Linguistic Fieldwork: A Student Guide

DRAFT: 03-April-2006

Daniel L. Everett
Illinois State University
(formerly of
University of Manchester)
Linguistic Fieldwork Daniel L. Everett
page 2
Preface
I have spent a large portion of my adult life in the Amazon, visiting over two
dozen American Indian communities along the way. There have been times when I was
so depressed, or so tired, or just so ill, that I wished I had chosen another area of
linguistics, or even another profession (pizza delivery comes to mind). But only a few
weeks after I return to my campus office, I am planning for the next field trip. Why? I
don't like bugs, heat, humidity, cold, hunger, loneliness, or ridicule any better than you
do. So my motivation is not masochism. I have two motives for doing fieldwork,
people and science.
The greatest privileges of my life have been to get to know people like the Pirahã,
the Kinsedji, the Banawá, the Wari, the Satere-Mawe, the Tzeltal, the Yarawara, and
others I have visited in the course of my field research. These people's character,
wisdom, humour and lack of pretension have challenged and enriched me deeply (and
their demands and occasional impatience with me have helped me to 'grow' as a
person). The many times I have had malaria, amoebic dysentery, infections, wounds,
bruises, headaches, typhoid fever, even the half-dozen times my life has been
threatened, have all been more than compensated for by the experience of getting to
know these people. And from what I hear talking to field researchers around the world,
this is a common conclusion.
The major reason I love fieldwork, though, is science. The excitement of
discovery and analysis of facts that no other linguist has ever worked on before me is
nearly addictive. As I review my curriculum vitae built up over these past couple of
decades, each publication, each funded research project, each invited course and lecture
on my fieldwork, recalls a fork in the road of my life, a place where I learned
something new which altered my view of language in one way or another.
In this book, I hope to communicate some of this awe that fieldwork inspires. But
I also plan to help the reader detour around the pits I have fallen into. My goal is that
those who read this book will be better equipped for research than they would have
been without it. Certainly, they will be better prepared by this book than I was when I
first went to the field in 1977.
Suggestions on how to use this book are found in the introduction.
Linguistic Fieldwork Daniel L. Everett
page 3
Acknowledgements
Many people have read portions of this manuscript and provided me with excellent
comments, not all of which I used, but all of which I learned from. These include Peter
Ladefoged, Bernard Comrie, Nigel Vincent, Jeanette Sakel, Martina Faller, Steve
Sheldon, Alan Vogel, Brent Berlin, Johanna Nichols, Nikolaus Himmelmann, Bob
Ladd, Keren Rice, Paul Postal, Geoffrey Pullum, Ljuba Veselinova, Cilene Camptela,
Miguel Oliveira, Joan Baart, Carlota Rosa, Stuart McGill, Monica Macaulay, Sarah
Grey Thomason, and Terrence Kaufman.
Two people in particular deserve special mention for without their support, this
book would never have come to be. They are Prof. Steve Parker, Head of the School of
Languages, Linguistics, and Culture at the University of Manchester, and Prof. Bernard
Comrie, Director of the Department of Linguistics at the Max Planck Institute. Steve
generously allowed me to take an early sabbatical leave, and Bernard generously
offered me space and financial support to spend the bulk of the 2005-2006 academic
year as a Visiting Scientist at the Max Planck Institute for Evolutionary Anthropology
in Leipzig, Germany. The personnel at the Max Planck are incredibly helpful. In
particular, Claudia Büchel and Julia Cissewski helped me make the transition to life in
Leipzig and work at the MPI. I don't know what I would have done without their help.
And Julia allowed me to join the MPI Choir, which gave me much-needed breaks from
the strain of writing. My project secretary, Anabella Niculescu has improved the
quality of my life by her initiative and competence in handling most administrative
responsibilities of my research projects, allowing me more time and energy to write.
Linguistic Fieldwork Daniel L. Everett
page 4
DEDICATION
This book is dedicated to the greatest linguist I have ever known, Peter Ladefoged.
Perhaps no single linguist has ever done more for the documentation of the sounds of
endangered languages than Peter. And no linguist was ever more caring, considerate,
and respectful of their colleagues than Peter. He made many comments on this book as
it was taking shape and I regret deeply that he could not see the final version.
Linguistic Fieldwork Daniel L. Everett
page 5
CHAPTER ONE: INTRODUCTION
1.1.ORGANIZATION OF THE VOLUME
This guide emerges from more than 28 years of field research in over two dozen
languages of the Brazilian Amazon, as well as from teaching field methods courses in
Brazil, the US, and the UK, and the enjoyable experience of taking many graduate
students and seasoned linguists from Brazil, the US, Israel, and the Netherlands with
me to train them in fieldwork methods. I have always enjoyed the isolation and intense
challenge of field research and the presentation of the lessons learned thereby to the
international linguistics community. Many of the suggestions found here are those I
wish someone had made to me before I began (or that I wish had followed in) my field
research career.
This guide has ___ chapters and ___ appendices, covering the topics I consider
to be most useful for field research, including personal preparation for field work,
designing 'lab sessions' (my suggested term for what are sometimes known as
'informant sessions'), the ethics of field work, selection of native speaker teachers,
processing data, design of web-based data presentations, and suggestions on fieldwork
(including suggestions on writing up results for publication) for specific linguistics
subfields, e.g. phonology, morphology, and syntax, as well as the history and
philosophy of fieldwork and suggestions on writing grant proposals to fund your
research.
I will believe that this book is a success if it helps researchers collect and
analyze linguistic data in a way that is helpful to other researchers.
1.2. WHAT THIS BOOK IS ABOUT
This book is intended for upper-division undergraduate students and above. It
presumes basic knowledge of most areas of linguistics. Some parts are harder than
others to understand at first blush. To lighten up the reading and make it more personal,
I have added anecdotes from my and others' experience in highlighted discussion
boxes. However, I believe that all chapters will be useful to the fieldworker. A special
feature of this book is its detailed suggestions for phonological fieldwork, including
chapters on segmental phonology and a very important guest-authored chapter by
Robert Ladd and Nikolaus Himmelmann on prosodic fieldwork. It also includes a
comprehensive phonological questionnaire, developed by the author, to aid in the
development of professional phonological studies of a field language.
The major theme to be developed in this book is how to do field work,
independent of any particular theoretical perspective. The book's major thesis is that
linguistic fieldwork can be successful with proper preparation and execution, bringing
deep personal and professional satisfaction for the researcher and her native-speaker
teachers. The book's purpose is to help linguists do, enjoy, and succeed at field
research. This guide is mainly step-by-step, detailed advice on how to go about the
business of fieldwork, from prefield preparation to field and post-field phases. All the
chapters will include many personal examples from my own and others' fieldwork
experience.
The history of research in general and field research in particular, is the history of
evolved creatures struggling to understand nearly agonizingly complex facts in an alien
environment. No one person is fully up to the demands of fieldwork. So the outputs of
our fieldwork will necessarily be incomplete records of each individual's progress in
understanding parts of wholes that exceed any one person's abilities. Thus, our research
reports, whether grammars or articles or talks or webpages are never more nor less than
our efforts to communicate with interested interlocutors about the beliefs we have come
Linguistic Fieldwork Daniel L. Everett
page 6
to form and hold, based on our experiences and how these beliefs affect our actions in
science and in life. This is our canopy of epistemic humility. No one can do more,
though there is still tremendous potential for qualitative variation in the effort.1
And we each need to approach the field well-armed with such humility.
Arrogance is ignorance, especially in field research, where one's limitations appear in
stark relief at all times. I am more and more convinced that the beliefs we have come to
hold about a particular language or grammar are constrained and shaped by the totality
of our experiences, not merely our linguistic training. If this is correct, an immediate
consequence for fieldwork that emerges is that compartmentalization of knowledge and
the isolation of knowledge from application should be avoided. Field research is
holistic – it involves every bit of the researcher's personality in every bit of the
language and culture under study, whether overtly or covertly, consciously or
unconsciously. This thesis underlies this entire book.

1.3. BECOMING AN ALIEN – MIND THE GAP

When you cross the boundary from your language and culture community to
another, you need to realize that you are transmogrifying yourself from someone
familiar into an alien. You could become a 'freak' instead of an attractive person; an
incompetent, instead of a respected professional; ugly instead of lovely; fat instead of
average; stinky instead of normal-smelling; and on and on. You may go from being
articulate and witty in conversation to being perceived as an infantile dullard who can
barely function in conversation. You will go from having many friends to having none.
From enjoying good company, to stark loneliness. From having your personal space
respected and being treated with dignity, to being seen as outside the normal politeness
conventions and treated like a fruit tree (people get things from you whenever it is
opportune for them). Familiar cues and clues are missing. You will be frequently
disoriented, feeling despair on the worst days.
Most disturbing, perhaps, if you are part of the majority group in your home
country, is that you will become a minority. It may or may not bring you prestige to
come from a wealthier background than people in the community who will be teaching
you their language. In my case, it occasionally means going from being called 'Dan' to
being called 'gringo', from being trusted, to being distrusted. Occasionally, traveling in
some parts of Brazil, I feel like if I say 'It's a nice day, isn't it?', my interlocutor is
likely to think 'Hmm. Why is he saying that? What is up his sleeve?'
People will almost certainly misclassify you. They may think you are an
anthropologist, not understanding what a linguist is (actually, then I think they would
be right, but that is another matter). They may think you're a missionary. They may
claim you're are with the CIA. They may think you're a different nationality even when
you're not. When they do find out your nationality, they are likely to have a different
view of it than you do yourself, assuming they are of a different nationality from you.
(So, for example, if you're German they may say 'Heil Hitler'. If you're an American
they may say 'Hooray for Osama' (which I have actually heard), and so on.)
The prospective field researcher must give careful thought, therefore, to the many
barriers they are crossing by entering another community to study their language. In the
next section, I briefly consider some historical examples of barrier-crossing from the
Americas.

1
Much of this section is taken directly from Everett (2004).
Linguistic Fieldwork Daniel L. Everett
page 7

FIELDWORK IN HISTORICAL PERSPECTIVE

Linguistic fieldwork has as many histories as there are countries in which it has
been carried out. And there is no global history of fieldwork (and virtually no local
histories either). Nor is this the place to write one. Nevertheless, some historical issues
are relevant to understanding the nature of field research and to preparing to do this
work. Therefore, I want to offer just a very brief discussion of field research in Brazil
and the USA, as examples of the kinds of issues, problems, and solutions faced by both
individual field researchers and the general enterprise of field linguistics.
Arguably, field linguistics in the Americas, as field linguistics most places, began
as an extension of colonial activity, specifically, missionary work. Let's first consider
the case of Brazil, then move on to consider the USA.2

Fieldwork in Brazil
In the Colonial Era (1500-1822)3
On April 22, 1500, a flotilla of ships commanded by Pedro Álvares Cabral
appeared off the coast of what is today the city of Porto Seguro, in the current-day
state of Bahia. Almost immediately, the sea-weary sailors of Cabral's ships spotted
men and women on the shore, looking out at the ships. A group of sailors rowed to
shore and were greeted warmly by those people bold enough to remain and not flee into
the jungle. Thus occurred one of the first contacts between Europeans and South
American Indians, in this case the Tupinambá. Cabral eventually sailed off towards his
intended destination of India, around the Cape of Good Hope, finally arriving back in
Portugal, with news of the new land, to be called 'Brasil' (for the pau brasil, a tropical
redwood that came to be highly valued in Europe). As it had begun with Ignatius of
Loyola (1491-1533), the founder of the Jesuits and the modern missionary movement,
the Church recruited missionaries to take the gospel to the newly-discovered heathens
of Brasil. One of the earliest missionaries to reach Brazil was the Jesuit Padre José de
Anchieta (1533-1597). Anchieta turned out to be a brilliant linguist (and administrator –
he was co-founder of both the cities of São Paulo and Rio de Janeiro). Anchieta began
his work near what is today the city of São Vicente between Rio de Janeiro and São
Paulo. The original people contacted by the Portuguese explorers were the Tupinambá,
a language of the Tupi-Guarani family.
Along with the very closely related language, Guarani, spoken to the south, in
what is today southern Brazil and Paraguay, Tupinambá was spoken along a sizeable
portion of the Brazilian coast, from São Vicente to what is today the city of São Luis
do Mararanhão. Wherever the Portuguese landed their ships north of São Vicente they
encountered the Tupinambá, eventually coming to refer to their language as the
'Brazilian language'. It was to this language and people that Anchieta gave the majority
of his attention during his missionary career in Brazil. Anchieta produced a grammar, a
2
Cite my SOAS paper, etc.
3
Much of the introduction to this chapter closely follows my article, 'Coherent
Fieldwork', Everett (2004).
Linguistic Fieldwork Daniel L. Everett
page 8
dictionary, and translations of catechisms. His grammar and dictionary still rank among
the best ever produced of a Brazilian language, nearly 500 years later. Although his
missionary activity was partially responsible for the complete extinction of the
Tupinambá people (largely because the Jesuits increased the size of Tupinambá
villages, thus increasing mortality rates when European diseases infected local
populations), Anchieta was a dedicated linguist whose work can be considered the
beginning of Amazonian linguistics (indeed, it would not be stretching matters too far to
call his work the beginning of linguistics in the Americas).
In addition to Anchieta, Tupinambá was also the object of some study by the
French Calvinist Jean de Lery (1534-1613), who originally went to Brazil to establish a
French Protestant colony. Lery's principal contribution was to record in written form
some naturally-occurring Tupinambá conversations. These enhance the picture of the
language presented in Anchieta's grammar and reinforce the importance of
conversational data in the documentation of endangered languages, since Lery's data is
now the only record we have of the living form of this language in use.
Several decades after Anchieta and Lery, another Jesuit, Padre Antonio Ruiz de
Montoya (1585-1652) arrived in what is today the border region between Brazil and
Paraguay to work among the Guarani people, speakers of a Tupi-Guarani language very
closely related to Tupinambá. Like Anchieta, Montoya was a brilliantly talented and
dedicated linguist, also producing a grammar and dictionary of the language (Montoya is
a partial model for the composite character of the priest played by Jeremy Irons in the
movie, The Mission).
After these few examples of precocious linguistic studies of endangered
languages (though Guarani has managed to survive this early troubled history), the field
of Amazonian studies was to lay fallow for the next several hundred years, aside from
reports and word lists from a succession of European explorers, mainly from Germany,
under the influence and example of Alexander von Humboldt (1769-1859).
So field research in Brazil began as a colonial activity. As such, its initial
purposes were utilitarian, to serve the Church, to get catechisms and the gospel into
indigenous languages. This story was repeated in country after country, around the
world. Native speakers were not valued for their knowledge and language but rather for
their role as objects in the colonial (and personal) goals of the missionary linguist. They
certainly played no active role in shaping the goals of the studies of Anchieta,
Montoya, and others, at least not that we have any record of or any reason to believe.
In modern days, however, missionary efforts have been very important in the
development of field research programs and traditions in different countries, though the
attitudes have remained very similar, in the sense that the native speaker community
plays no or very little role in shaping the missionary's objectives and activities among
them. To see this, let us consider the modern history of field research in Brazil.4

4
Spratt (2004) is a fascinating and largely convincing study of the influence of Native
Americans on new world philosophy, which has been both profound and uncredited.
The book, 1491, by Mann (2005) also demonstrates, very convincingly, the intellectual
richness of pre-Columbian Native American populations.
Linguistic Fieldwork Daniel L. Everett
page 9

Fieldwork in the contemporary era

Brazilian linguistics in the modern sense arguably begins with Joaquim Mattoso
Câmara Jr. (1904-1970), who dedicated a significant portion of his life to the
introduction of modern linguistics into Brazilian university (and pre-university)
training. Câmara did not spend much of his illustrious career on the study of Brazilian
indigenous languages, but he did encourage their study as part of the development of
Brazilian linguistics. In terms of the study of Amazonian languages qua endangered
languages, the pioneer in Brazil surely is Darcy Ribeiro (1922-1997), perhaps the first
government official of the Americas to invest government resources specifically
earmarked for the documentation and description (and for him, the 'preservation') of
endangered languages. During his tenure as Chefe da Casa Civil for Brazilian President
Jânio Quadros in the early 60s, Ribeiro invited the Summer Institute of Linguistics to
Brazil in the late 1950s. Ribeiro states his motive in inviting SIL to Brazil as (my
translation, DLE):

"My objective was to save for linguists of the future, who possibly will know how
to study them, the languages as crystallizations of the human spirit, in order that
we might learn more about mankind." (Ribeiro 1997, ---)
Ribeiro's adminstrative and anthropological concern for the indigenous peoples
in Brazil's survival and welfare was admirable and extremely forward-looking. We return
to the mixed results of his initiatives below.
In terms of personally-conducted research, the modern pioneer of the
documentation of Amazonian languages was Kurt Unkel (1883-1945) a German, later
naturalized Brazilian. This famous explorer, linguist, 'indigenista' , and anthropologist,
known to most Brazilians as Nimuendaju – the Guarani name he was given in 1906 and
used until his death in 1945 (partially) documented and identified a very large number
of Amazonian languages. Amazonian languages are still difficult to access physically,
culturally, and linguistically today. They were far more so in Nimuendaju's day. Yet he
managed to visit the majority of Brazilian Amazonian languages personally, taking
competent word lists from the many groups he visited, which have been extremely
valuable in the linguistic classification of these languages. Nimuendaju is today perhaps
the most revered figure in the history of the study of indigenous languages in Brazil,
making tremendous personal sacrifices to both study and support these languages and
their peoples. Stories of his life are currently only available in Portuguese to my
knowledge and even these are fairly superficial in their coverage. One hopes that one
day Nimuendaju's life and contribution to the study of Amazonian languages will
receive the attention it deserves. His concern for endangered languages and peoples
motivated not only his professional career but his entire life, from about 1906 until his
death. Nimuendaju was not motivated by the desire to change the people he studied, so
in this sense his work was an ethical improvement over earlier missionary efforts. He
wanted to provide a record of the peoples' languages and cultures. But his activities still
represent an intermediate level of ethical relationships with the communities, because
they still fall far short of engaging the native speakers as co-shapers of the records
Linguistic Fieldwork Daniel L. Everett
page 10
about themselves. Indians did not sit with Nimuendaju, for example and guide his
studies in any significant way, at least all records indicate otherwise, namely, that he
approached his studies with pre-determined objectives that were not negotiated in the
local context.
To most linguists, however, the true beginning of modern linguistic studies of
Amazonian languages in Brazil, entailing historical and comparative research, emphasis
on extensive grammars and dictionaries, begins with Aryon Rodrigues (1925 - ) – who
published his first articles on these languages before he was thirteen, as an eighth-grade
student in his native city of Curitiba, Paraná. Later Rodrigues was a friend and colleague
of Darcy Ribeiro at the University of Brasilia when Ribeiro served as the University's
first Rector (Rodrigues currently is a Professor Emeritus at the University of Brasilia).
Rodrigues combines most of the positive characteristics of previous figures
mentioned above. Administratively, he has founded linguistics programs, with strong
emphases on Amazonian studies, at the University of Brasilia, the Federal University
of Rio de Janeiro, the National Museum in Rio de Janeiro, and the State University of
Campinas (UNICAMP). Although Rodrigues has done little fieldwork of his own, he
has supervised countless graduate students' research (including my own MA thesis).
Rodrigues is a beloved advisor, from my own experience, but he, as a result of his vast
experience, has strong opinions (as most advisors will) about what the student should
be doing and how she should be thinking about the data she has collected. This raises
another issue, however, namely the role of the advisor in shaping the the field record of
the student field researcher. Students working under dominating advisors, or insecure
students working with a revered advisor, etc., can unconsciously or consciously allow
the advisor to tell them what to look for, how to think about it, how to relate to the
people, what conclusions make most sense, etc. This influence can be very helpful and
the modern academic system is partially predicated on the assumption that it will be.
However, it can also put the native speaker even farther away from the decision-making
and goal-planning stages of research (see section ___below for more on the role of the
native speaker). The student thus has an even more delicate and difficult task in field
research. She must engage the native speakers as conscious, willing, and active shapers
of the record, at the same time that she develops her own intellectual goals for the
research, while simultaneously satisfying an advisor that may be impatient or at least
skeptical of her decisions. These are natural tensions in life, of course, i.e. balancing
multiple demands of various people, but they are pervasive in field research and
students are particularly vulnerable. Therefore, the advice to the advisor is to give the
student as much freedom as is possible to work out her own field program, while at the
same time not relinquising the responsibility to ensure quality control.

Fieldwork in the USA

In the US, the Jesuits and other missionaries played a similar role to Anchieta
and Montoya in beginning studies of indigenous languages.5 However, professional
linguistic and anthropological fieldwork began with Franz Boas (1858-1942), who
trained a core of linguistically-aware anthropologists (Ruth Benedict (1887-1948),

5
MARY DORIA RUSSELL's THE SPARROW
Linguistic Fieldwork Daniel L. Everett
page 11
Edward Sapir (1884-1939), and in some classes and via Sapir, Mary Haas (1910-1996),
among others) responsible for the birth and growth of North American linguistics.
During the years of Boas's influence, roughly during his life and following his death
until the 50s, North American linguistics was concerned with describing specific
languages in detail, producing integrated studies of texts keyed to cultural studies,
grammars, and dictionaries, providing exactly the kind of pragmatist study that has
proven to be so important to knowledge of little-studied peoples and their languages
throughout the intervening years. In fact, though this is not the place to attempt a more
detailed intellectual history, a case can be made that this earlier descriptive linguists
were heavily influenced by the pragmatist philosophy underlying much American
intellectual endeavor until at the least the death of John Dewey (1859-1952), itself
arguably influenced by Native American philosophy (Pratt (2004)). Thus in a
roundabout way, Native American thought influenced the way that Native American
languages were studied and documented, at least until the 1950s. Consider some
remarks of Boas in his 1917 introduction to the first volume of the new International
Journal of American Linguistics (IJAL). According to Boas one of the principal goals
of the new journal was to provide what I would call a 'coherent' report of languages.
For example, he (1917, 201) laments the fact that "… the available material gives a
one-sided presentation of linguistic data, because we have hardly any records of daily
occurrences, everyday conversation, descriptions of industries, customs, and the like.
For these reasons the vocabularies yielded by texts are one-sided and incomplete." That
is, Boas felt that a full 'picture' of a given language was only possible by looking at the
language in the cultural context. Or consider Sapir's (1915, 186) assertion that more
studies are needed of cultural 'modalities of attitude' and consonantal alternations (I
discuss this further in Chapter Seven below), thus explicitly connecting grammar with
culture.
Boas (1911, 63-67), in his introduction to the handbook of American Indians,
provides perhaps the best statement of the relationship between language and culture
ever given. His discussion of this relationship was directly related for him to the
connection between fieldwork and theoretical research on the nature of language and
the nature of culture.
"If ethnology is understood as the science dealing with the mental phenomena
of the life of the peoples of the world, human language, one of the most
important manifestations of mental life, would seem to belong naturally to the
field of work of ethnology, unless special reasons can be adduced why it should
not be so considered."

In the same passage, Boas proceeds to consider and reject several proposed
'special reasons'. He goes on in this section to consider ways in which culture may
affect a language's morphology, lexicon, and grammar, concluding this section by
stating (p67)

"It does not seem likely, therefore, that there is any direct relation between the
culture of a tribe and the language they speak, except in so far as the form of
the language will be moulded by the state of culture, but not in so far as a
certain state of culture is conditioned by morphological traits of the language."6

6
This passage is particularly interesting in that it seems to contradict the linguistic
relativity hypothesis often associated with Boas. I think it is fair to say that Boas was
Linguistic Fieldwork Daniel L. Everett
page 12

Sapir carried on the Boasian tradition describing languages as manifestations of

culture, human psychology, and local conditions. And he influenced more than any
other person the course of North American linguistics until the 1950s, with the possible
exception of Bloomfield.7
Thus, for the first half of the Twentieth Century, the normal North American
conception of the linguist's 'job' was to study little- or un-studied languages in the field
and to produce coherent bodies of data on the interaction of culture, lexicon, texts, and
grammar. But by the 60s this had changed radically, with field research given more or
less the intellectual status of butterfly collecting. Postal (1968, ---) referred to previous
linguistic theories as 'taxonomic' while even today some refer to fieldworkers as 'data
fetishists' or 'hunters and gatherers'). What are the forces that changed the attitudes to
field research in North America (and eventually the world) so drammatically? It is what
we may call (with no perjorative intent), the 'Chomsky factor'. The twentieth-century
withering of fieldwork began innocuously enough, in the restlessness of a graduate
student at the University of Pennsylvania with his MA research:

"Harris suggested that I undertake a systematic structural grammar of some

language. I chose Hebrew, which I knew fairly well. For a time, I worked with
an informant and applied methods of structural linguistics as I was then coming
to understand them. The results, however, seemed to me rather dull and
unsatisfying. Having no very clear idea as to how to proceed further, I
abandoned these efforts and did what seemed natural; namely, I tried to
construct a system of rules for generating the phonetic forms of sentences, that
is, what is now called a generative grammar." (Chomsky 1975,25).

Chomsky's intellectual frustration with (an extremely easy version of) standard
fieldwork led indirectly to some of the most important developments in the 2000 + year
history of the study of language, so I am hardly complaining about the direction

here concerned with something else, namely, the classification of languages by culture,
which he rightly attacked as quite erroneous. Nevertheless, Boas does us a service here
by showing how the language-culture connection is bidirectional. This has obvious and
important consequences for field research, discussed in more detail in chapter six
below.
7
One influential linguist deeply impacted by Sapir was Kenneth Pike, who was both a
professor of linguistics at the University of Michigan and the President of the Summer
Institute of Linguistics (SIL) for over a quarter of a century. The rise of SIL in the
second half of the twentieth century and its Sapirian influence gave a huge impetus to
the study of American (and other) indigenous languages, as well as to the general
enterprise of field research. In the initial period of SIL's growth, there was little overt
questioning of the missionary enterprise. To my mind at least, SIL's nearly worldwide
acceptance and expansion in the late 40s and early 50s owes a great deal to the fact that
developing countries were, for obvious political reasons, welcoming citizens of the
country that defeated the Axis powers, with at least overt enthusiasm. The postwar
period is a period of the expansion of US influence with parallels to the 16th century
expansion of European colonies.
Linguistic Fieldwork Daniel L. Everett
page 13
Chomsky decided to take. Nevertheless, the very intellectual vigor and power of
Chomsky's subsequent work sufficed to pull most linguistics students and departments
away from the traditional emphasis on fieldresearch to theoretical work on, for the most
part, the linguist's native language. Though there is nothing inherently anti-fieldwork in
Chomsky's research programme, his attitude, as expressed in the passage just cited, and
his rejection of the intellectual priorities of Boasian linguistics led to an abandonment
of fieldwork in the US and a nearly five-decade neglect of the study of indigenous
languages and fieldwork throughout the linguistics world, as his influence soon became
massive and international. Over the past decade as the spotlight has begun to shift to
fieldwork once again, it has been primarily concerned with the study of endangered
languages (see ___ below) and has not yet recovered the 'Boasian imperative' of
coherent, integrated fieldwork. This is unfortunate and one hopes that we will continue
to make our way 'back to Boas'. This guide is meant partially as an aid to that journey.
In that sense, this guide is theory-situated.
As I say, the resurgence of interest in linguistic fieldwork (or at least in talk about
it) from the late 20th century is largely linked to the concern for documenting and
describing endangered languages. The interest in language endangerment itself had
been an important motive for early field research (see the quote, for example, from
Darcy Ribeiro above), especially among Boas and his students, but went out of vogue
for decades, making a comeback in the early 1990s. It is perhaps best exemplified
institutionally today by the Hans Rausing Endangered Languages Documentation
Project at the School of Oriental and African Languages in London and the Centre for
Linguistic Typology at Latrobe University in Australia, both established since the
'endangered languages' movement began, as well as through several funding (e.g. an
NSF program and the DOBES (http://www.mpi.nl/DOBES/) project for the study of
endangered languages) and technological initiatives (e.g. ELAN, part of the DOBES
project, and EMELD (Electronic Metastructure for Endangered Languages Data,
http://emeld.org), University linguistics departments and the general public have begun
to appreciate the fact that languages are dying daily and that with them die millennia of
accumulated knowledge and ways of talking about and experiencing the world and
examples of different linguistic evolutionary paths (or so Everett (2005) argues). From
a hard-nosed linguistic viewpoint, however, all languages need to be better described
and documented and the most important criteria for determining which languages
should be studied, to me at least, have to do with ensuring that the sampling of
languages we document is sufficient to warrant linguistic claims about theoretical
principles and typological universals of human grammars and languages. Although one
can accept the claim that endangered languages are the most urgent priority, the long-
term view of linguistics research must be to produce the best science it is able to do and
this means that we need diverse and robust data to better understand whatever it is
about Homo sapiens that ultimately underwrites their ability to have grammars and
language and use them. This entails more fieldwork, since so many areas of the world
are under-represented in linguistic research and because certain types of linguistic
phenomena are under-represented in the documentation of languages (e.g. intonation,
information-structuring, the phonetics underlying the phonology of a given language,
etc.).
In spite of the general belief in the scientific equality of all languages, there is a
sense in which workers on little-studied languages need a guide more than those who
study better-known languages. To see what I mean, consider that if someone makes a
claim about, say, English syntax or French phonology, there are hundreds of scholars
Linguistic Fieldwork Daniel L. Everett
page 14
and millions of native speakers that are in a position to challenge analytical assertions
they disagree with. But in work on little-studied languages it is often the case that very
few people, if any, will be in a position to seriously test the actual data used by the
linguist, unless the linguist has followed careful procedures that encourage, facilitate,
and promote as much replicability and soundness of presentation and analysis and data-
preservation as possible.
Unfortunately, there are few guides available to help the linguist go about the
business of fieldwork in the twenty-first century, especially for the linguist with the
goal of documenting and analyzing a large portion of grammar or language, while
working within a community of speakers of the language, away from their base
institution and confronted with the massive novelty of language in the real world. This
is because nearly all of the extant linguistics field guides were written decades ago and
fail therefore to respond to extensive developments in linguistic theory, methodology
and technology over the years (e.g. personal computers, lap-top software for acoustic
analysis, well-developed theories and notations for the study of intonation, advances in
morphological theory, discourse theory, functional and formal theoretical
developments, and on and on). To take one example, the field guide that I found most
useful in my early fieldwork, beginning in the late 1970s, was William Samarin's 1967
Field Linguistics, but this was already going out of date when I used it and many
portions of it are simply no longer applicable. And the very small number of field
guides that have appeared in recent years are for the most part orientated to special field
areas or otherwise limited in their general applicability.

A DEFINITION OF FIELDWORK
Let's begin our discussion of field research by hazarding a definition, from
Everett (2001, __):

(1.1) Fieldwork describes the activity of a researcher systematically analyzing parts

of a language other than one's native language (usually one the researcher did
not speak prior to beginning fieldwork), within a community of speakers of that
language, prototypically in their native land, living out their existence in the
milieu and mental currency of their native culture.

I think it is useful to consider this definition in more detail. Again, by dwelling

on my own definition, I am not claiming that it is 'right' in some absolute sense. But it
does raise issues worth considering, however one ultimately comes to understand the
essence of field research.
'Systematically analyzing' should be clear. We go into field research with a
system of ideas that guide our research. How does this system guide, then, what we are
going to study? What are the subparts of the system? How do the different subparts of
the system, projected onto the language of study, interact? For example, perhaps I am
conducting research to test a specific claim in the literature, e.g. 'language x lacks
embedding.' What system could there be to my investigation? Well, first, if a claim has
been made to this effect, I want to check the data that was adduced on its behalf with
native speakers. Do native speakers agree with all the grammaticality judgements
offered to support the claims being tested? Are there discrepancies across speakers?
And so on. To check data requires a plan. How many speakers should one check the
data with? How should one subcategorize and test discrepancies in speaker judgments?
How can one design and test alternative hypotheses? Second, if a language lacks
Linguistic Fieldwork Daniel L. Everett
page 15
embedding, it should be reflected several places in the grammar, not merely, say, in the
absence of complement clauses. Does the language have disjunction? Coordination?
Verb phrases? Complex noun phrases? Adjectival phrases? And so forth.
Or let's assume a different kind of study, e.g. Topicalisation in narrative
discourses of language y. How does one study this systematically? Another way of
putting this is 'what is my specific hypothesis and what are its component parts? How
do I order them for study?' Well, first it is crucial to separate out the various
components claimed to be part of Topicalisation, based on theories of Topicalisation in
the literature – things like the manner of use and frequency of proper names,
constituent orders for different sections of discourse, types of anaphora or cataphora at
different points in the discourse, and so forth. A study of each of these, independent of
the global hypothesis, to determine their functions and structures within clause and
sentences, should be undertaken, so as to minimise bias caused by looking at them only
in light of the hypothesis being tested. These superficial examples give some idea of
what I mean by 'systematic' in () above.
In other words, field research is like any other large, complex task. It requires
planning, administration, progress checks, self-evaluation, and reports (at least to
oneself). In Chapter Two we discuss the application of some basic management
principles to field research.
Continuing on with a discussion of (1), why does it refer to 'parts of a language',
rather than an 'whole language'? First of all, it is impossible to study a whole language.
Just consider the thousands of studies of English language and the fact that there is no
sign that research on English is coming to an end. A language is vast and beyond any
single researcher's ability to study in a human lifetime. Language is everything:
semantics, sociolinguistics, phonetics, phonology, syntax, morphology, ethnography of
communication, and on and on. Second, Language as an object of study is unclear,
unfocussed – there are no boundaries to identify either a coherent beginning or end of
the study if its object is 'documentation of English' or some such. The goal of 'parts of
language' requires a lowering of the of the sights from Language to selected
components. Their selection requires hypothesis management, and a notion of the
'systematic' just discussed. It also requires a coherent vision of how the parts fit
together, assuming that the study is to fit together at the end, that it is not strategically
opportunistic.8
Next, why does my definition include the qualification 'other than one's native
language'? First I am a bit of a purist. I think that being a 'field linguist' should be
different from being a 'linguist' (i.e. that hyponyms are not the same as hypernyms), and
that its difference is exactly the crossing of linguistic and cultural boundaries. But also,
I think that it is important to separate out different research foci, based on their peculiar
requirements on the linguist of training for and execution of the research. Not everyone
will accept this, I realize. But without some such qualification, field linguistics loses its
distinctiveness as a term.
Now comes one of the most vital components of my definition and the
component that sets it apart from many other conceptions of field research, namely, that
field research should be conducted 'within a community of speakers of that language.'

8
It very legitimately might be if one finds oneself with access to a language one hadn't
planned to work on, but which presents itself in a set of circumstances, with interesting
properties or is endangered, etc.
Linguistic Fieldwork Daniel L. Everett
page 16
This differs radically from others' conceptions of field research. So consider what
Samarin (1967, 1ff) says in his classic Field Linguistics:

'Field linguistics can be carried on anywhere, not just in the field, as its name
implies. [emphasis Samarin's, D.L.E.] A "field archaeologist" must go out to
where he expects to collect his data, but a linguist can bring his data to himself.
Thus, some fieldwork is done by bringing jungle dwellers to a city and is
conducted in an office instead of a lean-to.'9

Before I say why I disagree with Samarin and all the linguists who agree with
him (e.g. Hyman ()), let me hasten to say that circumstances have often forced me to
work with native speakers in offices, hotel rooms, and missionary compounds. If there
is a language I need or want to study, I will go where I have to go, even if the
circumstances are not ideal in some way.10 It will become clear in this book that we
must always be prepared to improvise. But my position is based on the view that
language and culture are inextricably intertwined. You cannot understand one without
the other (Everett (2005)). This is discussed further, with several examples, in ___
below, and it is implicit in Boas's statements above. By taking speakers out of their
communities or studying parts of languages outside their cultural contexts, I believe
that vital pieces of understanding go missing and that the resultant grammar or study
can be seriously flawed. The rest of the definition in (1) is there for the same reason,
namely, because grammar and culture affect and, to some degree, effect one another.

POSTAL'S MAXIMS
One of the books I learned from early in my career was Longacre's Grammar
Discovery Procedures (Longacre ()). The title seems like a tongue-in-cheek poke at
Chomsky's assertions to the effect that there are no such things as 'discovery
procedures', either in linguistics or science more generally (Chomsky 1957; 1975
[1955]). Interpreting the book charitably, Longacre lays out a general list of heuristic
procedures for developing hypotheses on the grammar of the language under
investigation. Although the book is still useful in many respects, however, it doesn't
really get at the core components of building a theory of aspects of a grammar or
language, the essence of field research.
Paul Postal noticed these shortcomings in his review of the book and made the
following, extremely useful, observations:

"... I would strongly suspect that the two most important 'discovery procedures'
[are to] learn the language of study as well as possible and attempt to

9
In an interesting paper on fieldwork as a state of mind, Larry Hyman suggests that
fieldwork can be done any place, so long as the linguist is properly prepared. To some
degree I agree with this. On the other hand, if I am correct about the connection
between language and culture, a language must be studied in a culturally robust
community of speakers.
10
George Carlin said in a routine at the Beacon Theatre in Boston that 'You know me.
If it's got a zip code, I'll f___in' be there.' Well, I am like that, substituting, say,
'grammar I am interested in' for 'zip code'.
Linguistic Fieldwork Daniel L. Everett
page 17
formulate an explicit account of the rules which generate the full syntactic
structures of its sentences." Postal (1966, 98)

I will refer to Postal's proposals here as Postal's Maxims. I restate them to better
suit my objectives here in (1.2):

(1.2) a. Learn the language.

b. Generate and test hypotheses about the language.

To rephrase, since I believe that this is crucial: The most important thing to
remember here and in all areas of linguistic analysis in the field that the most important
two methods are: learn the language and formulate and test new hypotheses daily. All
analysis is tentative. Nothing is ever certain. It is unreasonable for people to think that a
grammar must 'stand the test of time', for example. Grammars are temporary
documents. Ideally, they represent the best a given linguist can do in a particular slice
of the space-time continuum. New evidence comes to light to force a change in
analysis. New perspectives. New theories. There is no algorithm for discovering a
grammar. Only hard work, ideas, alternative ideas, and testing.
That said, let us consider suggestions for succesful field research by turning our
attention in Chapter Two to prefield preparation.
Linguistic Fieldwork Daniel L. Everett
page 18
CHAPTER 2. PREFIELD PREPARATION AND NON-LINGUISTIC CONSIDERATIONS FOR
FIELDWORK
2.1. Prefield preparation
You are going to the 'field' to study aspects of some language. You probably
have a general idea of what you plan to study. But it is always useful to bring your
research questions into ever sharper focus as your journey into the unknown takes
shape. So, before you apply for a research grant, before you buy your ticket, ask
yourself this: Am I clear on what I want to study? Can I explain it to others without
being too wordy or abstruse? Can I explain it convincingly to the mirror? 'What is the
exact object of my investigation?'
Part of getting an answer to these questions is to consider whether the object is
to study something directly observable or something only inferable. An example of the
former would be the measurements of formant frequencies of consonants and vowels
across all speakers in a single village. An example of the latter would be, e.g. constraint
rankings proposed to account for the morphological structure of the verb of language
'x'. This type of object clarification will affect your preparation, including your budget,
your need for skills (e.g. in sound analysis software or constraint ranking evaluation,
etc.), etc.
Another issue to consider in this regard is whether you are ultimately more
interested in the explanation of similarities between languages or in the documentation
of their differences. (Contrary to some opinions, the former is not necessarily a better
goal than the latter!) Are you interested in corpus-based studies or speaker intuitions?
Are you interested more in qualitative studies or quantitative studies, or a combination
thereof? Is your principal objective connected in any way to ethnography of
communication? And so forth.
A less obvious, but perhaps equally important, way to clarify your research
objectives, once you have decided on your main question, is to carefully consider the
ancillary questions implied by your 'big' research question(s). For example, if you read
Cowart (), you may decide that your big theoretical research objective could benefit
from some statistical analysis. How much time should you therefore give to the study
of statistics? Or to learning about questionnaire preparation?
It is common for PhD students (in particular) to specialize, to deliberately focus
on a narrower range of questions, to the exclusion of many other interesting, but not
directly relevant issues. This is quite reasonable in most contexts. But it can be
unreasonable and counter-productive in fieldwork, at least in an extreme form. A
fieldworker not only needs to know more, because they will be faced with more
information that requires knowledge to sort through, but they need more reflection
because they cannot leave the field to get additional training if ancillary issues require
it. This limitation has the corollary that very careful and detailed thought needs to go
into the formulation of research questions to be asked in the field and that these
questions and related issues should influence prefield training.
I suggest the following as a potential method of prefield research preparation.
First, develop a list of the research questions you want to ask. The first versions of this
list should be done hastily, just jotting down questions as they occur to you, things you
might be interested in researching in the field. These questions should be formulated
before, during, and after reading all you can about the field language and theoretical
issues you expect to research. Second, narrow this list down to those questions that are
most vital to your research and career objectives. Third, organize the questions (e.g.
what are the main vs. ancillary questions? Which ancillary questions accompany which
Linguistic Fieldwork Daniel L. Everett
page 19
main questions? And so forth). Fourth, operationalize each question – how can it be
made 'behavioural', i.e. into something you need to do in order to investigate it? Never
stray from the empirical core of your research at any stage of planning or execution.
Refine and add to this list as you feel necessary. Next, build an initial plan. How might
you ask and answer these questions in the field? (By the end of this book, I hope that
operationalizing your questions will be easier.)
Let's move now to another prefield question of importance, namely, selecting a
location for your field research. Here is a partial list of selection criteria you might find
useful:

(2.1) a. Language endangerment

b. Family history
c. Typological interest
d. Good things to eat
e. Geography
f. Funding agency priorities
g. Suggestion of advisor

For what follows, I suggest that each heading be thought of as a type of

constraint. The individual field worker can then use them to guide her decision as to
where to work by ranking them and trying to satisfy as many of them as possible, from
most highly ranked down.

Language endangerment
If you are interested in the documentation and description of endangered
languages, this will somewhat narrow down the range of language communities you
have to choose from. You won't, for example, work in monolingual communities of
tens of thousands of speakers undergoing no external pressure to switch languages or
any obvious health or other external threat. Your concern will lead you most naturally
to communities where there is a threat either to the physical survival of the people or
economic or social pressure on them to switch to another language. To determine
whether a given language is endangered, you will have to read on the socioeconomic
conditions of the region, sociolinguistic relationships between languages in the area,
speakers' attitudes towards themselves and their language, government policies on
minority languages, likelihood of other linguistic studies of the language in question,
and so on.
This is quite a worthy criterion. After all, if you study an endangered language,
you could be contributing not 'merely' to linguistics but to history and the naturalistic
record of Homo sapiens. Still, it is likely that the fieldworker will need more than these
altruistic intellectual goals to see him or her through the long, lonely spells of
frustration and ignorance that mark all initial periods of field research. So let's consider
some other possible motivations.

Family history
Alex Haley's () Roots alerted and excited many people about the possibility of
knowing about their family history. Haley's attention was initially caught by linguistic
evidence of the similarity between words he heard from his grandmother and those of
contemporary African languages overheard by him from fellow university students. If
Linguistic Fieldwork Daniel L. Everett
page 20
he had been a linguist, he might have very legitimately chosen to conduct research on
one of his identifiable ancestral languages from present-day Gambia.
The desire to research your family's linguistic history is quite a legitimate
motivation for selecting a field language and location. This can further have the side
benefit of creating family interest and support for your research and career choice,
something lacking in most linguistic research projects, where family members, like the
general public, often fail to appreciate linguistics.

Typological interest
If your goal is to advance our understanding of the linguistic possibilities of
Homo sapiens, then a powerful motivation for field research could be the desire to
advance typological knowledge, i.e. statistically valid clusterings of linguistic
properties and their explanation. So, for example, you might go to Northern Brazil to
study object-initial languages, which apparently only exist there. Or you might choose
to work in Africa if you are interested in properties clustering in the phonologies of so-
called 'click languages'. This selection criterion has the benefit of inserting your project
from the outset into the current concerns of linguistic theory. And this is a very
important advantage for the field researcher. As has been stated earlier, field research is
a risky enterprise professionally. Your career options will be maximally enhanced if
you can use your research results to challenge, refine, or advance current theorizing
about grammar or methodology.
How then might one use typological considerations to select a field location?
First, you should read widely in the typological literature. What are the main issues that
stand out to you? Which ones are you most interested in? Discuss these with a
typologist on your faculty or in another institution to make sure that you understand
them and that they are indeed issues of current debate. Next, select an area of typology
for concentrated learning. You need to master this area, reading, ideally, everything on
the subject. Next, investigate regionally-focused journals and surveys (e.g.
International Journal of American Linguistics, Oceanic Linguistics, Journal of African
Languages, etc.). In this part of your preparation, you are looking for information
discussed directly or simply mentioned and illustrated in some other context, that might
bear on your research interests. What you would like to find, again ideally, is a region
that you are attracted to and that seems to have languages in which the typological issue
you wish to investigate is potentially quite relevant and widespread. Finally, and this
advice goes for ALL field workers, read grammars from the linguistic/geographical
area where you hope to work. Reading grammars is a vital component of the field
researcher's healthy diet. Without knowledge of the intricacies of languages in the area,
you enter that area under-prepared. This doesn't mean, of course, that you must agree
with those grammars. But you must know them well. Form opinions, even become
opinionated, but know the material and the analyses proposed by your predecessors in
the area well.

Good things to eat

Some time ago, I attended a lecture by Professor Steven Anderson of Yale
University where he suggested that an important criterion in selecting a field location
would be 'where there are good things to eat'. I had never thought of that before. But it
strikes me now as great advice. And it extends to more than gustatory attraction. Insofar
as is possible, choose a place where there are things you enjoying seeing, eating, and
doing in general. Feel free to choose a climate you enjoy. There is just no reason for
Linguistic Fieldwork Daniel L. Everett
page 21
anyone to avoid their own comfort and pleasure in selecting a field location. It isn't
always possible, of course, to satisfy these additional constraints, but it often may be
and yet I suspect that many field linguists, if I am any measure, do not think about their
own comfort, as though this were somehow 'wimping out' or anti-scientific.
In my own case, although I didn't use the 'pleasure criteria' in selecting my field
location, I was fortunate nonetheless. Although the Pirahãs do not eat much I consider
tasty, the surrounding Brazilian culture certainly does. The weather, scenery, rivers,
beaches, music, people, and culture of Brazil have often inspired me to continue on
with my work on Amazonian languages, even when I was linguistically discouraged,
undergoing malaria treatment, or otherwise feeling tired of Amazonian field research.
We all need such inspiration. This can come if we choose a place to work where, even
if the research hits snags, we will still be able to enjoy where we are.

Funding agency priorities

Another important selection criterion for deciding where to work is the set of
research priorities of the funding agencies most likely to support your research. Chapin
(2004) is a useful and nearly comprehensive guide to research funding, based on
Chapin's many years as Director of the Linguistics Program of the National Science
Foundation of the USA. His book should be read by all younger scholars looking for
funds to support their research. A vital part of field research is money. Know where to
get it – and in large enough quantities to do your research well. Find out which funding
agencies or programs within larger agencies fund research for the area of the world you
propose to work in. Your university may have 'seed funds' to help newer scholars or
students conduct pilot studies. Agencies like the NSF often have subprograms that set
aside special funds for types of research, e.g. social sciences in specific regions of the
world. In the United Kingdom funding agencies such as the Economics and Social
Research Council and the Arts and Humanities Research Council have 'ring-fenced
awards' that give special priority to specific types of research (e.g. studies in Modern
European languages – great if you want to do field research on Basque or Northern
Italian languages, for example). The success rate for most grant proposals hovers
around 20-30%, depending on the size of the award, the research area, the funding
agency, etc. A younger scholar will usually have to learn the hard way by competing
for funds directly against senior scholars with plenty of fund-getting experience. So the
newer scholar must invest time researching funding possibilities and agencies. Contact
the relevant administrative staff at the agencies you are interested in. Most of them
prefer to discuss potential proposals in advance, rather than merely getting them
without prior discussion. Discussing the proposal in advance will usually make it more
competitive (or help you discover that this agency is inappropriate for the project you
have in mind).

A good programme officer at a major funding agency can be of inestimable benefit in

helping the researcher prepare a successful research proposal. One of my earlier
proposals to the National Science Foundation was rejected, though the comments from
the reviewers were not so negative, mainly saying that my project was too ambitious.
The Director of the Linguistics Program at the NSF in those days was Paul Chapin. He
recommended that I bring the referee comments and my proposal to his office hours at
the next Linguistic Society of America meeting. I did. He suggested that I keep the
research objectives but triple the time and the budget (well, he did not suggest the
latter, but that was entailed in the time suggestion). I did. I resubmitted. I got the
Linguistic Fieldwork Daniel L. Everett
page 22
money. If it hadn't been for Chapin, I might have given up. He encouraged and helped
me and thus is responsible for much of the research I have gone on to do over the past
25 years, almost all of it funded by the NSF, as I learned to become a more successful
grant-writer.

Suggestion of advisor
For many linguistics graduate students, the choice of a language for fieldwork
will be largely made for them by their advisor. However, even if the advisor is the
driving force behind the selection of the language for study, the student should give
some consideration to the factors mentioned here to better evaluate their chances of
success before entering the project.

Self-Evaluation
In the movie, Unforgiven, Clint Eastwood reminds us that 'A man's got to know
his limitations'. Fieldwork may sound like a great idea in the air-conditioned, bug-free
library at Hometown University, but it may produce a different reaction when little
children are pressing their hands against private body parts to determine your gender
(or, as I sometimes suspect, your species), even as you are exhausted from a long trip
and surprised by the smells and sounds of your new work environment. So be hard on
yourself before you travel. A very important part of prefield preparation is self-
evaluation: Do you have what it takes? Some of the components of the successful
fieldworker include the following question, arguably the most important of all: do you
have the talent and training for the job? And I suggest that the former outweighs the
latter (as Boas himself emphasized to his students (Darnell ()). This talent and training
will most clearly manifest themselves in the linguist's five senses (mental/physical data
input devices) and her ability to interpret the results she gathers (mental data-
processing). Do they think and read regularly about other languages? Do they have a
well-developed ability to distinguish segments and prosodies? A talent for language-
learning? An enjoyment of the exotic? A strong constitution? Ability to learn and teach
with patience and clarity? Can they make friends easily and defuse tense situations?
Can they tolerate lots of noise, successfully concentrate in a 'busy' environment; accept
criticism for his or her government's politics, his or her skin colour (and many other
things they can't change); tolerate lack of privacy and being laughed at every
day?These are by far the most important toolkit the fieldlinguist will possess.
With regard to training, did the potential fieldworker's training include work
with native speakers (ideally in their own environment)? Did they have teachers with
field experience? Do they have cross-cultural experience of any kind?
There is a picture of me somewhere, where my back is covered with wasps and
I am holding a microphone in one hand and a metal plate in the other, the latter in the
futile attempt to ward the wasps away. I was stung about four times a day that summer.
My right elbow was swollen to at least double size. My problem is that I sweat a lot in
the tropics and the wasps apparently love the salt in my sweat (my theory). Anyway, I
uttered many a foul word that summer. My sweat would trickle down my arm to my
elbow, attracting wasps. Then I would rest my elbow on the table, without thinking, and
that would anger the wasps and they would sting me.
Another time I was working just under a low thatched roof near the river's edge.
River frogs would get into the thatch and croak. That is bad enough, because it affects
the quality of your data recording. But what was worse was that snakes, some of them
poisonous, would then come up into the thatch to eat the yummy frogs. This became so
Linguistic Fieldwork Daniel L. Everett
page 23
common that I kept a foot-long, hardwood club by my feet. I would hear a rustling in
the thatch. Then a frog would jump out. Then a snake would come slithering out behind
it, often right above my desk. WHOMP! I killed many a reptile that field session. I even
enjoyed it. In fact, if I could not figure out a particular construction, morpheme, etc. I
looked forward to killing the next snake. WHOMP, indeed.
Now, if someone had told me these two stories before I went to the field I
would probably have had two reactions. First, this is unlikely to happen to me. Second,
I would freak out if it did. I was wrong on both accounts. But as I think about my
background, growing up in a rural area of Southern California, spending a lot of time
on my grandparents' farm, with chickens, cattle, pigs, and all the smells, sights, and
accompanying creatures and bugs that go along with cow dung, chicken entrails, and
pig blood (we ate these creatures, you see, and even while alive, they were unsanitary),
I can see that this was all good preparation for field linguistics.
If you come from such a background, some aspects of field research will be less
difficult for you. However, most academic researchers probably are not raised on
farms. Most come from cities I would think. If you are from a city and intending to
spend a serious period of time doing field research away from a major city, then you
will face similar things. What you face in this regard will, of course, vary by climate.
Exercise is important, though not always because fieldwork is so physically
demanding. There is no question, of course, that if a linguist moves from an urban
university environment to a rural field environment, that daily routine will differ and
will almost certainly require more lifting, walking, climbing, and general labor, as well
as more direct exposure to the elements. Exercise will help prepare anyone for such a
change. But exercise is valuable too because it can give the fieldworker a break from
the mental strain of both prefield preparation and fieldwork itself. Doing linguistics in
the field involves all of the work and pressure of doing linguistics in the city but adds to
this the strain of responsibility, novelty, culture shock, and change in diet, comfort, and
physical labor, etc. unique to the field. Trying to exercise or work in novel, less
comfortable surroundings, altering one's diet for a while, etc. are useful training for the
field, for relatively little personal cost.

2.2. Specific knowledge and skills preparation

In this section, I want to discuss the various general ways in which you can
prepare yourself for a more successful field experience, where the latter includes
scientific objectives as well as personal well-being, and contributions to the local
community.

Preparation of research questions

Consider again the list of research questions suggested in the previous section.
You need to learn as much as you can about each of those areas before you leave for
the field. Some additional suggestions to prepare for linguistic research in the field are
the following:

(2.2) Identify an advisor or mentor

You need an advisor or a mentor – a professor or a colleague whom you trust to
ask you useful questions and offer useful suggestions. The exact role of this person will
depend on your career stage. For graduate and undergraduate students, of course, an
advisor is vital. You need someone who is willing to answer your formal, informal, and
personal questions, including what we in the United Kingdom know as 'pastoral care' –
Linguistic Fieldwork Daniel L. Everett
page 24
personal counseling along the lines clearly expected from Boas by Mead in the quote
from her letter above. There may not be such a person at your home institution. It could
well be the case that your mentor is across the hall. But it is equally possible that he or
she is in another country, especially if your advisor and your mentor are different
people (the former playing a formal, institutional role, the latter a more personal,
intellectual role). With the internet and ever cheaper rates for international phone calls
(especially with voice-over IP technology), there is no overwhelming reason why the
latter couldn't be the case.
Ideally, you want someone with field experience, preferably in the geographical
and linguistic area where you will be working. You want someone successful in getting
data and publishing. You want someone who can put you in contact with local
community of scholars where you'll be doing research and give you the perspective of
the wider community of scholars who will be 'consumers' of your research. You should
have someone who can advise you on the bureaucratic, mundane aspects of fieldwork
(e.g. how to fill out forms in another language and culture, whether to buy a pressure
cooker, how many kilos of books are worth taking, when the best time to be in a
particular field location is, what the best local transport is, how to stay in touch with the
'outside world', etc. You also want someone who can read your work in progress and
tell you how to get data, ask questions, and read more to make that work better. In fact,
it is not inconceivable that you could benefit from multiple mentors. And these need
not, again, correspond to an actual academic advisor if your research is for a thesis. But
keep the number to less than three. Even with two, you will find opinions and advice
will grow exponentially.

Prefield Literature Review

Read on the country and world region you'll be going to. Find out about its
politics (especially its relations with and its populace's attitude towards both your home
country and the minority group whose language you will be studying), history,
geography, foods, language, culture, and laws, especially those governing foreign
researchers (e.g. the constraints on research authorizations and visas).
Next, read all the linguistics you can find on the region you plan to visit. Even if
the specific language you hope to work on is an isolate or part of an unstudied family,
read all you can on languages of that region. Areal characteristics of languages are
likely to be relevant and useful for you. As you read – THINK. How could this study
have been improved? What kind of data is necessary to write a paper like this? How
much time did the author spend with the people to collect the data upon which his or
her studies are based (unfortunately, grammars and papers often omit this information
or are imprecise in their description of it. For example, if an author says merely that 'I
have worked on this language for twenty years, what does that mean? Is the twenty
years in question eight hours a day in the field collecting data or three months in the
field every two years, followed by writing and reflection)? What has been written
already about the structure of the language you plan to study? How much has been
written about the ways in which that structure is put to use or acquired? How much on
the history of the language?
Take time to compile an exhaustive bibliography on the language, the family,
and the area. Then do your best to read it all (this is relative – for some languages there
will be very little, hence this assignment is easy, for others you won't be able to read it
all or even list it all, but will need to be selective, exhaustive reading and compiling for
some categories, very little for others). Your own research questions will have to guide
Linguistic Fieldwork Daniel L. Everett
page 25
you in your efforts. Classify the works you have read. Know which are most useful.
Often, you will want to take these with you to the field.
Three websites I recommend that supply tools for field research are the sites of
the Max Planck Institute for Evolutionary Anthropology in Leipzig
(http://lingweb.eva.mpg.de/fieldtools/tools.htm), The Electronic Metadata for
Endangered Languages Data (EMELD, http://emeld.org), and that of the Max Planck
Institute for Psycholinguistics in Nijmegen (http://www.mpi.nl/tools/).

Anthropology
Early in the history of North American linguistics, linguistic studies were seen as
a branch of anthropology. Today, however, most linguists would likely not think of
themselves as anthropologists, nor would most anthropologists identify even
descriptive linguists as a subfield of anthropology. Nevertheless, because doing field
linguistics is doing linguistics in a natural cultural setting, the field linguist cannot
avoid culture. They can approach the cross-cultural linguistic experience ignorantly or
informed – that is the only choice. Read a general text, e.g. Foley () or Duranti (), and
then do follow-up reading on topics of personal interest via the references to these
texts. Or, if you already have a background in anthropological linguistics, you can read
in the major journals, eg. the Journal of Linguistic Anthropology, Journal of
Anthropological Linguistics, and 'four-discipline' anthropology journals, e.g. Current
Anthropology. I also recommend that all field linguists read Sapir (1921) and work by
Lucy () and others on the neo-Whorfian approach to the language-culture interface.
I also recommend corresponding with anthropologists who have studied the
people whose language you plan to study or with anthropologists who have worked
nearby.
Other books looking at the connection between culture and grammar (language
structure) should also be read and carefully considered (e.g. Enfield's 2003,
Ethnosyntax). In chapter ___ below I give a series of examples of the interaction
between language and culture. That chapter reinforces the importance of
anthropological knowledge for the average field linguist.

Computers
Every researcher must use information technology in their research. It simply is
no longer acceptable to go to the field without good technological support for
collecting, recording, and analysing data. Residual Luddites that do attempt to do so,
however, should at least recognize that they cannot document a language nearly as well
without modern technological aids as they can with this technology.
Documenting a language involves creation of a multimedia record of the language
in use by native speakers. Describing/analysing a language also benefits tremendously
from and often requires technological support. Software and hardware for sound
analysis, video-editing, transcription, and preparation of data for long-term storage is
essential to field research.11

11
Several researchers have suggested that it is important to keep old versions of
software for processing field data because there can be incompatibilities with newer
versions and the old versions may be necessary, so long as one still has hardware that
will run them, to access the data properly. But care should be given to data-storage and
the software used for this purpose. Proprietary software such as Microsoft Word should
be avoided and, instead, use of XML software should be the standard.
Linguistic Fieldwork Daniel L. Everett
page 26
Time must therefore be invested to acquire at least an intermediate level of
technological skills prior to departure for the field. You should familiarize yourself
with your computer and its operating system and all relevant programs. Understand
how and where application files and documents are stored. Design an effective (for
you) filing system on your computer for your work. Treat it as a portable office. Make
sure you have hardware to back up all your files and programs (fast transfer portable
hard disks are very convenient for this).
Have calendars with alarms to keep you to your time schedule and goals.

Documentation
In the development of linguistic fieldwork, notions of documentation and
description have perhaps not been as carefully distinguished as they ought to be. In
earlier years of my career, they seemed to be used nearly interchangeably. To write a
grammar of a language, for example, was to document a part of the language. Likewise
a dictionary was a form of documentation. More recently, technological advances allow
us to create interactive databases for long-term storage and usage of primary data on
languages, i.e. audio and video files. Such data bases refine our concept of
documentation (though of course in the selection of data for such data bases the
researcher intrudes and obscures). In my opinion, 'primary documentation' is the
recording of audio and visual data. Secondary, and perhaps tertiary, documentation may
be thought of as data in increasingly interpretative matrices (e.g. grammars, theoretical
articles, and so on). The more interpretative the documents produced, the farther
removed from primary documentation in the view advanced here. Again, in times past,
description doubled for documentation as primary sources were not made available to
general linguists and all that we have/had on many languages were data as selected and
interpreted by linguists, explorers, anthropologists, missionaries, and others.
So now let us consider documentation a bit more (though audio documentation
is discussed in chapter __ , section __ on phonetics).
If you want to share the pleasure of your birthday party with your friends, you
could simply tell them about the party. Or you could show them photos and videos of
the party. Your friends may not want to see all your photos, but at least they can judge
for themselves whether Sally's dress was divine or the cake was lovely, etc. Still photos
can isolate moments. It is very important, therefore, that a fieldworker be familiar with
photography or at least that someone on the team have such knowledge. The quality of
your equipment (see ___ below as well for general considerations on equipment) will
depend on the knowledge available to you to use it, your budget, your goals, etc. But at
least a five pixel digital camera with an array of focus options should be part of your
toolkit. Read an introductory text to Visual Anthropology, e.g. those listed at
http://www.visualanthropology.net/. This is not absolutely crucial for field research in
linguistics but it can be very important for documenting certain kinds of claims on
meaning where facial expressions, gestures, and other visual cues can be crucial to
understanding the pragmatics of the utterances in question.

Medical/first-aid training
Fieldworkers should have basic training in first-aid, treatment of diseases
common to their chosen area of fieldwork. Ideally, if they are going to be working in
extremely isolated situations, they consider some training in suturing and bone-setting.
Linguistic Fieldwork Daniel L. Everett
page 27
They should have access to some basic bibliographic sources (e.g. Where there is no
doctor and Where there is no dentist) and have emergency numbers to call locally
and internationally (by satellite) for consultant help if necessary.
One afternoon among the Pirahãs, I was pursuing my never-ending quest to
understand the structure of the Pirahã verb. Suddenly, I heard yelling at the river. When
I looked, the Pirahãs were running towards the river, talking loudly. Someone came to
tell me that a man from the village, /abagi 'Toucan', was hurt. Sure enough. As they
brought him up into our house, his left arm was beet red, amazingly swollen, and
oozing pus. He had had an accident in his canoe and an arrow had entered his forearm
just above the wrist and emerged on the opposite side, below the elbow. He had a fever
and was in obvious pain, something the Pirahãs only admit to in extreme
circumstances. He freely admitted that the pain was nearly unbearable.
What was I supposed to do about this? It was clear that every Pirahã there
expected me, the outsider, to have some western medicine and to know what to do. So I
did what any courageous, knowledgeable, and resourceful field researcher might do – I
called my wife, Keren.
Keren was able to treat Toucan and he fully recovered. How? Well, before we
ever set foot in the Pirahã village we both took courses in first-aid. We also asked
various people – doctors, nurses, missionaries, and others – what kinds of health
problems we were most likely to encounter. We then purchased medicines accordingly.
In Brazil, as in many countries, a much wider and more potent range of medicines can
be purchased over the counter than in the USA. So we purchased several hundreds of
dollars worth (this in 1978) malarial medicines, analgesics, snake anti-venom
(antiophidic serum), local anaesthetics, syringes, sutures, and so forth. During our first
couple of days among the Pirahãs, we organized our medical equipment and supplies
on shelves, with our most useful medical manual, Where there is no doctor, by David
Werner, in the front.12
There are many field locations where the fieldworkers would not be expected to
provide health care. There are many places where unlicensed people dispensing
medicines would be in violation of local laws. But in many isolated communities, a
linguist or anthropologist may be the only hope for health care. Certainly the linguist
may need to care for his or her own health, or their partner's or children's, depending on
where they are. Therefore, training, reading, equipment, and medicines are all crucial
components of any fieldworker's kit.

Survey
Finally, if there is no extant sociolinguistic survey of dialects, language attitudes,
demographics, geographical distribution of the language, etc., this should be undertaken
at some point during the first field research trip. There are books that offer crash
courses in this kind of survey work. One such is Blair (1991).

2.3. NON-LINGUISTIC CONSIDERATIONS FOR TIME IN THE FIELD

2.3.1. Paperwork and bureaucracy

12
All fieldworkers should have a copy of this book. It can be ordered via the internet:
http://www.hesperian.org This website also contains a wealth of information and
material on health care in rural environments, etc.
Linguistic Fieldwork Daniel L. Everett
page 28
Going to 'the field' means crossing political boundaries. And this almost always
entails getting two broad types of authorization: authorization to enter the country and
authorization to do research. These permissions usually require the fieldworker to apply
for special visas, to get medical exams, criminal checks, official translations of
diplomas, etc. It will take time. And nobody enjoys the process. This book would be
less honest or helpful than it should be if it ignored this unpleasant aspect of fieldwork,
so I will try to highlight some of the bureaucratic processes involved and suggest ways
to help the process run more smoothly.
Most countries will not allow fieldwork on a tourist visa. And in some countries
tourist visas restrict the tourist so that they cannot visit minority communities outside of
major cities. I know some linguists who have done fieldwork on a three-six month
tourist visa, some even multiple times. Many an important linguistic study has been
done without proper authorization. But not only is this failure to secure the proper visa
unethical and illegal; if discovered it could bar the linguist from current or future
funding for the research (most funding agencies require evidence, in advance, that the
researcher has secured or will secure the proper legal documents for his or her research
from the local government).
So you will need to get a visa. But you are likely also to need permission from a
government department, ministry, etc. responsible for minority affairs. And your
scientific project is likely to need authorization from the national research or science
foundation. The latter may require that you have a national partner, i.e. a linguist or
other appropriate specialist who is personally supportive of your research and is willing
to be your academic sponsor. Often these various permissions seem to produce
'ordering paradoxes', 'infinite regresses', or 'Catch 22' situations, e.g. one source tells
you that need government authorization before you can get scientific authorization,
while another source tells you that you need scientific authorization before government
authorization. I have seen foreign researchers spend a year or more getting
authorizations in some countries.
In most countries these processes are all made easier and faster if you know the
people responsible for the authorizations. It is almost never possible for you to simply
handle this all with a phone call or over the internet. So how do you get to know the
right people? You will need money, time, and a willingness to bite your tongue and
keep silent at times.
First, where bureaucracy is especially labyrinthine, the prospective fieldworker
should attempt to find 'seed money' for a trip, on a tourist visa, to the country where
they would like to work. Before leaving, they should find out who the local linguists
doing field research are. They are quite likely to be the people who will be asked,
eventually, to evaluate the research proposal proper. If the linguist is lucky, they might
be able to establish initial contact with someone who could help, a graduate student or a
professor from the target country would could tell him/her who they need to see, where
to go, etc. This person might even be able to make the initial contacts for the
fieldworker, provide him/her with a letter of introduction, etc.
It would be useful if the fieldworker could speak the national language before
venturing to the target country, but this is not necessary. Academic contacts will speak
English and if all goes well they may be willing to help, advise, and even speak to the
authorities on behalf of the fieldworker. There is, to be sure, occasionally some distrust
of Western Europeans or North Americans, but by and large the international scientific
community is interested in promoting high-quality research and the fieldworker is
likely to be received warmly and make life-long good friends. So, although these initial
Linguistic Fieldwork Daniel L. Everett
page 29
contacts can be tense and humbling, they often end up pleasant and rewarding, both
professionally and personally. One should look at these initial steps as part of the
fieldwork. They are.
On this first visit, enjoy the food, the nightlife, and the people of the country.
Make the most of the experience. Life is short and you only go around once (so far as
any one can prove). So make the most of it and have a good time. That would be
neither unscientific nor unprofessional.
Expect the entire process of authorizations, once again, to take from 3-18 months
and expect to travel at least once at the beginning of the process (and perhaps again
after the process is well underway) to the target country to get help in speeding the
process along. Plan ahead! Also, keep careful written records of your contacts, their
expressions of support, and their willingness to help, etc. Get letters from them
whenever possible. Funding agencies will want to know that you are aware of the need
for authorization, etc. and that you have it well in hand.
Now I want to consider some additional nonlinguistic factors for time in the field.

2.3.2. Entertainment
When I took my first anthropology class, at Grossmont Junior College in El
Cajon, California, in 1970, the teacher (who influenced me tremendously, but whose
name I cannot remember), mentioned that what you read in the isolation of the field
will affect you more than it would in your home community. That is, it will reach your
emotions and mind more deeply. I have found that to be correct in my experience.
General suggestions I have found useful in my own field experience include the
following:

(2.3) Have a time for yourself: you need a time away from the language teachers, away
from as much of the hustle and bustle of the language community as possible (if it is
appropriate – it will not always be). Some time where you can simply reflect, relax, eat
one of the chocolate bars you brought from the city, write letters to friends, etc. Time to
just 'chill'.
(2.4) Read: take books to read in the evening or whenever the best time turns out to be
in your field situation. I take a mix of novels, history, philosophy, and biographies. I
find it necessary for myself to mix light and heavy reading. I can't take too much of
either without a break.
(2.5) Movies: Taking movies to the field is easy these days. Take DVDs that you are
willing to watch with members of the community. It will be possible at times to watch
DVDs alone, with no one around, but assume that someone will want to watch DVDs
with you. Choose programming that will entertain and relax you, but not offend the
community's sensitivities. Also, it is important to remember that movies and other
programs can be great educational tools for the community. So take a selection of
DVDs just for showing the community. The Pirahãs, for example, tremendously enjoy
material about other indigenous communities and their daily lives, as well as about
animals of all kinds. National Geographic movies are very useful. On the other hand,
the most popular movie among the Pirahãs is the old John Wayne movie, Hatari, about
capturing wild animals in Africa.
(2.6) If your community is accessible to mobile phones, I am sure that you will take
one. If the community is not accessible to mobile phones, then I suggest that you save
up to purchase a satellite phone. There are several options on the market and most of
them, though quite expensive with high per minute charges, can allow you to maintain
Linguistic Fieldwork Daniel L. Everett
page 30
contact with friends, colleagues, and others in the most remote locations. Some people
find this very useful. For example, with a satellite phone you can call for linguistic
help, phoning your home institution, etc. to get expert advice on how to analyse or
collect data on a particular subject (not to mention emergency help). Some satellite
phones (e.g. the Nera World Phone that connects to Inmarsat) also can enable you to
send and receive email, which can also be useful for getting advice or sending
examples of constructions you have collected, along with your ideas on them, to
another member of the research team, or a colleague or mentor, to ask for help, advice,
etc. On the other hand, satellite phone communication is expensive so will not be a way
to chat freely with your friends on the weekends.

2.3.3. Journaling
For the past three decades, I have, off and on at least, kept diaries, some of which
have evolved into journals. I have to admit that my journals have by and large had no
'theory of journaling' behind them. They are unsystematic records of my emotions and
activities for the most part. But they could be more useful to me in my research in my
ability to perform well in fieldwork. I want to offer some advice that I wish someone
had given me on how to keep a journal.

Journals
The first thing to get clear on is that a journal is not a diary. A journal is not
simply a chronological record of what you have done during the day, though it might
include that. So what is a journal for? What are the advantages of keeping one? Here is
a partial list of ideas:

(2.7) Reasons to keep a journal

a. To maintain a personal record of your time in the community, e.g. what has
been hard for you, what has provided new insights on language, life, culture, etc. what
you are thinking about on a given day, thoughts about family and friends from your
field site vantage point, etc.
b. To develop a record of your scientific findings, your incipient ideas about
analysis, suggestions to yourself about what to collect the next day or week, ideas that
did not work and ways that they might be retested, etc.
c. To have someone to talk to. The field gets very lonely. There are times when
no one around speaks your native language. You need to talk to someone in your native
language. Talk to your journal. I find that this is very useful. I can rant to my heart's
content about something and get it out of my system. I do not write these parts of my
journal for any present or future audience except myself, and would never want anyone
else to see some of my 'rants'. But it is good therapy for me nonetheless and I highly
recommend it. In fact, days or weeks later you may return to read your 'rant' and learn
how wrong you were about the matter that so exercised you at the time, building up a
valuable store of lessons on your growing understanding of the culture and yourself.
Also record what pleases you about the community, the location, the research. Record
positive observations too!
d. To have a place to focus and organize. Journals can provide you a space for
focusing your thoughts and organizing your thoughts and future course of action in
research, relationships, etc. A journal can be where you set and revise your goals,
redefine success (as the saying goes 'If at first you don't succeed, redefine success'), and
make notes of new things to read, new methods to try, etc.
Linguistic Fieldwork Daniel L. Everett
page 31
e. To track your intellectual and linguistic development. Writing down what you
think you know and what your attitudes are towards the work, the community, the
general aspects of field research can, as you read your own entries over the months of
fieldwork, provide an excellent source of data for you to study yourself and your own
development as a fieldworker. This can be useful for setting further goals, but it can
also be useful as an encouragement that you are not as bad as you thought you were at
this job (or alternatively, this can be a source of humility if you had too high an opinion
of yourself).
f. To record memories for post-field write ups of your experiences, your
developing analyses, etc. Journals also and naturally provide a historical record of the
linguist's time, but also of the community. If you are careful to record observations
daily about the people (see below) then your journal, as a document that scientists
(including you) might look at years hence can provide a record of cultural and historical
change of the community under study and of the role and relationship of the
fieldworker in these changes.
g. To develop your writing style and ability. Journals provide an excellent
opportunity to become a better writer. Let yourself go. Be flowery. Try new ways of
writing. Develop your writing style so that you can express your observations in such a
way that your own personality comes through. This makes for better field writing.

Approaches to journal keeping

One approach, which I learned from Peter Ladefoged, is where the journal takes
the form of a scrapbook, with photos, small bits of memorabilia (such as a feather, a
flower, a fish hook, etc. – small things that are evidence of important or memorable
events in the experience), as well as annotations. Keeping pictures of your language
teachers can be very useful and provide a personal link with your field research for
years to come.
Another approach is less interesting, perhaps, but it is my own, perhaps lazy,
approach. I simply write in my journal, nothing else. My journals are simply written
records that include the information suggested above and below.

Things to include in a journal

Here are some suggestions for what to include in a journal that I believe are useful
for a scientific journal. (i) the date and time of the entry; (ii) the location where the
entry is being recorded; (iii) the weather at the time the entry is recorded; (iv) your
mental and physical health (i.e. how are you feeling?); (v) the village activities going
on at the time of the journal entry and others planned for that day, i.e. before the next
journal entry. An analysis of observed activities should also be included; (vi) track an
individual, a family, or other cohesive group during the day and record their activities
(gists of conversations, who did what during the day, where they went, how far it was
from the community center, etc.); (vii) village health – is anyone ill as you record your
entry? What is the illness? Is it contagious? etc; (viii) your ideas on the language,
culture, nature, etc. (Theoretical, descriptive, musings, etc.)

Additional suggestions journal writing

Write freely. Allow yourself opportunities to release a 'stream of consciousness',
without worrying about form or even much about content.
Linguistic Fieldwork Daniel L. Everett
page 32
Describe scenes and events that catch your attention, rich in details. These
details can help you write up your field experience later. Clifford Geertz () talks about
'thick' versus 'thin' descriptions. By this he means that:
“Cultural analysis is intrinsically incomplete. And, worse than that, the more
deeply it goes the less complete it is… There are a number of ways of escaping this—
turning culture into folklore and collecting it, turning it into traits and counting it,
turning it into institutions and classifying it, turning it into structures and toying with it.
But they are escapes. The fact is that to commit oneself to a semiotic concept of culture
and an interpretive approach to the study of it is to commit oneself to a view of
ethnographic assertion as… ‘essentially contestable.’ Anthropology, or at least
interpretive anthropology, is a science whose progress is marked less by a perfection of
the consensus than by a refinement of debate. What gets better is the precision with
which we vex each other.” (29)13
In Geertz’s understanding, ethnography is by definition “thick description”—
“an elaborate venture in.” Using the action of “winking,” Geertz examines how—in
order to distinguish the winking from a social gesture, a twitch, etc.)—we must move
beyond the action to both the particular social understanding of the “winking” as a
gesture, the mens reaction, the state of mind of the winkers, their audience, and how
they construe the meaning of the winking action itself. “Thin description” is the
winking. “Thick” is the meaning behind it and its symbolic import in society or
between communicators. The fieldworker should attempt to develop this kind of
descriptive style in their journal. Among the cultural values that should be included are:
community recipes, aphorisms, folk wisdom, myths, etc. It is also useful to provide
pictures and samples of plants, insects, etc. whether you are interested in studying
ethnobiology, or simply generally interested in these topics.

The form of the journal

First, you will need material to write with. You will need to decide whether you
want to keep your journal on your computer or in hard copy. If the latter, are you going
to keep your journal in a notebook or a hardbound book? I suggest that you use
indelible, water-proof and acid-free ink and paper for your journal, and that the pen-
paper combination you use doesn't result in the ink being absorbed in the paper over the
years in such a way as to obscure what you have written. Let me close this section by a
consideration of an alternative type of journal, the blog.

Blogs (Weblog)
Blogs are different from journals primarily in that they are written for an
potentially large audience and located on a public or semi-public space on the internet.
They can be useful for that very reason, by letting others see, in real time (as you post
to the internet, perhaps from a satellite phone), what you are going through. They can
enlist help and advice, they can later serve as journal records, and are otherwise every
bit as useful as a journal. Therefore, what I have to say about journals also applies to
blogs. In fact, there is no reason why a journal could not also serve as a source of blogs,
leaving the private kinds of observations out.

13
This comes from: http://academic.csuohio.edu/as227/spring2003/geertz.htm
Linguistic Fieldwork Daniel L. Everett
page 33
2.3.4. EQUIPMENT AND SUPPLIES
2.3.4.1. General issues
Any discussion of equipment or technology will date a study quickly. So I do not
intend to spend much time on this. However, there are a couple of things to say in this
regard that will be somewhat impervious to time.
First, technology is vital in field research. Even though I believe that I have very
good 'ears', in my experience machines have been invaluable in helping me to notice
sounds and patterns which my unaided ears had missed. And technology provides a
record for the future, however outdated it eventually becomes. Consider, for example,
the significance of the portable cassette tape-recorder for the history of field research. It
is true, trivially, that early fieldworkers got by without this, now outdated, device, just
as everyone gets by without inventions yet to come. But wouldn't it now be priceless to
listen to audio tapes or watch video tapes made by Sapir, Boas, Newman, and others,
checking their facts and interpretations more carefully, or possessing a more complete
record of the languages they studied? As we recognize the need to study, for example,
endangered languages, technology capable of accurately preserving and measuring the
sights and sounds of these languages becomes ever more important.
Some questions to ask with regard to field-equipment are:
(i) Who will be able to use the output of your equipment now and in the future?
(ii) Is the equipment portable?
(iii) Does the equipment provide state-of-the-art accuracy, or as close to it as the
fieldworker can afford?
(iv) Will the equipment help record both the grammar and its cultural matrix?
(v) Does the equipment use a practical power source for the location in which it will be
used (such as solar power)?
(vi) Does the fieldworker's equipment include satellite capability, for email and phone
contact from the field site to any part of the world?
(vii) Do you have backup equipment for crucial items, e.g. extra microphones,
computer(s), recorders, etc.?
Point (iv) may seem strange, but it can be taken as a reason for using, in today's
terms, high-quality camcorders in the field, rather than relying exclusively on audio
recordings. It is also a reason to use portable computers in the field which have state-of-
the-art video and audio editing capabilities (e.g. the Mac G4 laptop in 2003). In
purchasing and planning, remember that quality is not something to be overly
economical with – pay top prices if necessary to get top equipment. There are other
areas to be frugal in, if that is necessary (and of course it always is).

2.3.4.2. Portable power systems

There are a number of small, relatively inexpensive (but fossil-fuel consuming)
generators available in most countries. These are extremely useful and convenient
tools. In fact, I have a lightweight generator (approximately 20 pounds) that will supply
all my energy needs in the field (lights, battery charging, etc. – it will even power a
refrigerator). But generators have severe disadvantages, e.g. (i) they are
environmentally harsh (use of fossil fuels); (ii) they require significant weight for fuel;
(iii) they require maintenance and careful storage and they break down. Therefore, I
suggest that the field linguist use a small solar-powered system when practical.
A solar power system would not have to be large if used only for charging a
camcorder, computer, and audio recorder, and for small light bulbs. I recommend two
32-watt solar panels (roll-up rather than rigid, for easier transport). These are then
Linguistic Fieldwork Daniel L. Everett
page 34
connected by cables to a 12 volt, deep-cycle battery (e.g. the kind used on many small
boats in the US. Deep-cycle batteries charge better and have a wider voltage tolerance).
This type of battery can be quite heavy, however, and needs to be replaced every one-
two years. Smaller batteries (e.g. a motorcycle battery) can be used, but these are
unable to provide as much power. You will then need a voltage converter to change the
power coming out of the battery to 110 (or 220) from 12 volt. A voltage regulator
would be useful because it can keep your batteries from overcharging or running while
undercharged. It will shut off the relevant input or output to protect the battery. If you
choose not to purchase and connect a voltage regulator, you should get a voltage meter
to check the power in the battery. It should never go about 13-14 volts and never below
11 or you will be without it soon.
Finally, wherever possible, native speakers should be trained in the use of the
equipment. This is important training for them and can be very helpful for the linguist,
even avoiding the need for the linguist to return to the community for small samples of
data that can now be collected by the native speakers.

2.3.4.4. Consumable study supplies

You will also need to take pens, papers, and other office supplies, even though
you should rely more on the computer than on paper. Consummable study supplies I
take to the field include:

(2.8) Consumables that might be useful:

a. Flashlight batteries and AA-size batteries. I only take alkaline batteries. You
can buy fewer of these than would be necessary for non-alkaline batteries since they
last so much longer. Therefore, they require less space.
b. Blank DVDs are important for backing up data or making copies of lab
sessions, songs, etc.
c. Indelible, acid-free, waterproof ink pens. Buy enough for one a week,
minimum.
d. Other pens: take various colors of pens for color-coding data and a range of
other uses.
e. At least 500 sheets of paper for a six-week stay, with side holes for including
in notebooks (or take a paper-punch).
f. Separate notebooks (appropriate for the paper just mentioned). You will need
one notebook for every major division you want to make (e.g. 'Verbs', 'Nouns',
'Transitive clauses', 'Subordinate clauses', and so on).
g. Paper clips, staples, stapler, clear adhesive tape, and plastic bags, e.g. rubbish
bags of different sizes, sandwich bags, etc. The plastic bags will be vital, especially in
rainy and humid areas, in preserving your notes and many other supplies (e.g. matches)
from the effects of humidity and leaky roofs.
h. Rice for dessicant and tupperware.
i. Back-up computer battery: computer batteries do fail. I always take an extra
computer battery to the field.
j. Cottonettes and rubbing alcohol – these are useful for for cleaning playing
heads (on older model cassette recorders. But if possible avoid these and purchase a
solid-state recorder). These are useful for personal hygiene as well, to help cut down on
orifice fungus. (I recommend that you put alcohol mixed with vinegar (1:3 ratio) in
your ears after every swim in the river, or hydrogen peroxide) .
k. Recording tapes or memory cards (for solid state recorders).
Linguistic Fieldwork Daniel L. Everett
page 35
l. Plastic binders/covers for loose papers. Papers which are not yet in particular
notebooks should also be carefully stored and plastic binders or some other way to
render your notes water-resistant are extremely useful.

2.3.5. Tools
Take a multiuse knife or two. I use both a Leatherman and a Swiss Army knife
(since the two types have different tools). A hammer is also useful. The tools you
choose will depend, obviously, on your particular field situation. If you use a boat
motor, you will need to know how to repair it, at least basic maintenance and simple
repairs, and you will need tools for that. I also recommend that if you use a boat, that
you have replacement parts, especially ignition modules, waterpump, fuelpump, and
spark plugs.
No fieldworker should travel without two essential aids: duct tape and durepox.
The former can fix most broken things. Durepox (which comes in two clay-like
sections, to be mixed together) forms a chemical bond which is strong enough to repair
holes in boats and cars. It is very valuable in the field. A flashlight, e.g. one that can be
attached to your head for hands-free lighting when trying to read or repair something is
also useful. The list can always be added to, of course, but each item is more to carry,
more to lose, more to potentially come between you and the community.

2.3.6. Care of equipment

The field is usually rough on equipment. Whatever you take with you is likely to
be exposed to some combination of the following: high humidity, sand and dust, bugs,
rain, and temperature extremes. As you travel and pack and repack, heavy bags and
other objects are likely to be put on top of delicate equipment, your equipment will be
dropped, perhaps even in a river, and will otherwise be treated very differently than it
might, say, at your home institution.
There are special carrying cases with adjustable foam linings and quality rubber
seals for laptops and other sensitive electronic equipment. These cases, when
maximally useful, are of metal or hardened plastic (yet relatively lightweight),
waterproof, and able to hold equipment tightly in place and secure, without rattling
around inside. There must be room in the equipment case for a dessicant of some sort if
you will be in a wet or humid environment. One popular dessicant is silicon gel.
Cheaper alternatives are available, as discussed in ___.
We now turn to consider the last, but certainly not least, issue in non-linguistic
fieldwork issues, having your partner and/or family accompany you to the field.

2.4. FAMILY IN THE FIELD14

2.4.1. Your partner
If you have a partner who enjoys the field experience, you are most fortunate. My
former partner was raised in the Amazon by missionary parents. Living among
Amazonian peoples turned out to be the most enjoyable activity in the world for her.
She can tolerate the bugs, heat, and other hardships much better than most people can.
And she is extremely organized, (which I am not), which helped tremendously in many

14
In 2005 I was interviewed on the BBC 4's Excess Baggage programme about raising
children in the Amazon. This is available at:
http://www.bbc.co.uk/radio4/excessbaggage/index_20050917.shtml
Linguistic Fieldwork Daniel L. Everett
page 36
ways, from shopping and packing, to organizing each day in the village. Her presence
always enhanced my own ability to work and thrive in the field. When both you and
your partner enjoy the field, this is a tremendous personal and professional help.15
I only went to tribal areas a couple of times without my family. In each case my
productivity fell by more than half and I swore I would never do it again. People are
very different. Whereas some linguists might relish the solitude of individual fieldwork,
I have never found it easy to be alone in the field, though when my family has gone
with me, i.e. most of the time, it was usually an enjoyable experience.
Even with the family it can be hard. My first visit to the Pirahãs, my oldest
daughter and my wife got very severely infected with falciparum malaria. There were
rarely any field visits where one of the family didn't get very ill.16

The first time that my family went with me to visit the Pirahã, 1979, was only the
second time I had ever gone to the Pirahã area. Things were apparently going very well.
One day I was lying in my hammock, swaying in the breeze off the river and
memorizing Pirahã words. I remarked to my wife that 'this life is really rough', with lots
of sarcasm. But I spoke too soon. It can be rough. Within 24 hours she and my oldest
daughter, Shannon (8 years old at the time) had fevers of 104 degrees and severe
headaches and backaches. I had taken introductory health classes during my SIL
training, and I had several medical resource books. Since neither Keren nor Shannon
were suffering from chills, I assumed that they could not be suffering from malaria.
Also, I didn't think (why I have no idea) that malaria would begin in two people
simultaneously, like any other infectious disease. (They both had falciparum malaria,
the most virulent South American form.) So since their symptoms matched something I
had suffered from on fieldwork in Mexico, typhoid fever, I began antibiotic treatments.
They got worse by the day. I had no radio to call for help or advice because these were
then prohibited for foreigners in tribal areas, and this was before the invention of
satellite phones. A plane was due to arrive in a week, but Keren had now drifted into a
coma, after being delerious for about 36 hours. Shannon was drifting in and out of
consciousness. Neither one had eaten or had more than a liter to drink in four days. I
had to do something or they might both die. And I was exhausted, taking care of them
and watching my 5 year old daughter and 2 year old son. The Pirahãs seemed to me
uncaring and cold during this time, though later I came to realize that this was part of
their stoic philosophy of life, not a reflection of how much they cared for someone, nor
a lack of concern. But I only had an old 9 horsepower motor, with very little gasoline,
and no knowledge of the way out to the nearest doctor. I had done no geographical
research, just assuming I would fly in and out of the village, as though I were going to a
US regional airport. I am not sure what I would have done, but a Catholic lay
missionary, an Italian man named Vicenzo, now deceased, came up to visit the Pirahãs

15
However, it is important, especially if you are both linguists, that you are not overly
competitive with one another and that you are not defensive about your language-
learning, linguistics, ability to cope with the environment, etc.
16
However, this is not standard for all fieldworkers! Desmond Derbyshire, a pioneering
Amazonianist, told me that he had an excellent language teacher within his first thirty
minutes among the Hixkaryana, who worked with him regularly for the next couple of
decades, and that he could never recall having been sick in the village.
Linguistic Fieldwork Daniel L. Everett
page 37
on behalf of a priest he knew. Seeing our plight he told me to take his boat, a small
aluminum canoe, barely big enough for my family of five, with a 6.5 HP outboard
motor. He told me how to get out. However, because he gave me all the place names in
Italian and because I didn't yet speak Portuguese well, no one I encountered farther
down the river knew what I was asking about. We left the village with Keren and
Shannon lying down in the middle of the canoe and Kris and Caleb sitting near the back
with me. The water was nearly up to the edge of the canoe. We were moving about 8
miles an hour and I had only 20 litres of gasoline, with no idea, in spite of Vicenzo's
help, of how long it would take us to get to the next river, the huge Madeira, where
boats could be hailed to take us to the nearest town. After several false stops and turns,
almost out of gasoline, we arrived at a small settlement and several men hung
hammocks on poles and carried Shannon and Keren through a jungle path to the
Madeira river, about a one-hour hike. I followed along with supplies, carrying Caleb
and Kris. A local family, Godofredo and Cesaria Monteiro, poor as all the others, took
us in and fed us (I had taken no money – we were a complete liability for that family).
After waiting for three days, with Keren ever nearer death, a boat came. Men from the
community carried Keren and Shannon down the steep banks of the Madeira, at 3AM,
to catch the big boat to the nearest town. Loaded on the boat, I was impatient to get to
the hospital. The next day, I offered to pay the boat owner any price he wanted to take
us in the motorboat he was towing to the nearest town, which would have saved us
some 12 hours. He said 'Comrade, if it is your wife's time to die, she will die. Rushing
will not save her.' Then he stopped the boat and his crew played the men of a local
settlement a two-hour game of soccer, with me beside myself with rage and fear. Keren
and Shannon were both losing weight and had terrible cases of diarrhea. My two year
old and five year old were running about the boat and I kept trying to keep them still.
Finally we reached the town of Porto Velho. The doctor examined the two and said that
Shannon might live but that Keren would die that night and that I should call her
family, which I did. She weighed 76 pounds and had lost a lot of blood (malaria causes
you to lose blood through urine). But she lived. And we have been back many times.
That was more excitement than I had bargained for. But it is, for better or worse, part of
what fieldwork can be like if you are unprepared. Don't be. I have had malaria many
times since then, as all of the family, except our son, Caleb, have. But being prepared
has meant that none of the subsequent cases were all that serious (except for the time
that I came down with malaria after just arriving from Brazil, while on a trip to
Disneyland with my family. I came close to dying. But that is another story).

2.4.2. Children and fieldwork

Your children's mental, physical, and social health will be of utmost concern to
you in the field. There are some dangers and things to worry about, but overall, the
field experience can contribute to raising aware, resourceful, and independent children,
respectful of other cultures and races. I highly recommend taking your family to the
field, in spite of occasional danger and adversity.
In 1977 my family went to Chiapas, Mexico for 'jungle training' with the then Summer
Institute of Linguistics. During our first few weeks in Chiapas, my six-week old son,
Caleb, came down with meningitis. We gave him six adult does of antibiotics per day
for days (I forget the exact treatment). He was very, very ill. And his first set of teeth
were all weakened and bad because of the antibiotics. And, as related above, in Brazil,
my daughter Shannon nearly died with malaria. Another daughter, Kristene, came
down with hepatitis. Health concerns should not be taken lightly. On the other hand,
Linguistic Fieldwork Daniel L. Everett
page 38
they should not be overly dramatized. Kids are tough. And cuts, scrapes, sickness, and
environments demanding adaptation are good for them. But it is important that their
parents react well to and enjoy the same environments. If your children sense fear or
guilt on your part for taking them, they will respond with fear and a sense that your
guilt is well-deserved. I noticed early on that whenever something unusual or
potentially unpleasant occurred, my children would immediately look at my wife or me
to see our facial expression. If we showed fear, then they did too. If we seemed
unperturbed, then so did they. You must be well-adjusted to the environment, secure,
and confident in your ability to cope with most things that come your way before you
take your children to the field. But if those conditions are met, they should thrive.17
Act like they are fortunate to have these experiences and you all will have a
great time too! That is the secret.

Part of well-being for your children is to ensure that they have an adequate social
life. This cannot be based around you and/or your partner. They need people their age.
Therefore it is essential that they make friends in the community under study.

My daughters, Shannon and Kristene, went with me to the live among the Tzeltal
people when they were 6 and 3, respectively. They went with me to the Pirahãs when
they were just a year older. I remember among the Tzeltales, waking up about 630AM
on most days. That was late for the Tzeltales. Our 'door' was a set of poles placed
upright in the doorway. Most mornings, a few poles were moved out of the way
because, Kris, three years old, had left the house. I would find her in front of a Tzeltal's
home, a huge cup of sweet coffee in one hand and a fistful of corn tortillas in the other
hand. Shannon, then 6, would often spend the day visiting different Tzeltal villages
with Tzeltal girls. Sometimes the girls would dress Caleb up in Tzeltal baby clothes (he
was about 2 months old) and take him around with them to other villages (though not
far because he was still nursing). Among the Pirahãs, Shannon and Kris would often
disappear in a canoe with Pirahã girls in the morning and return in the afternoon, after a
day of fishing, picking berries, and giggling. They seemed to enjoy themselves and tell
me that they have good memories (they are now both in their 30s). On the other hand,
the Pirahãs (like many peoples in the world) are sexually active very early and the
entire culture sees sex at most ages as a natural and innocuous pastime. So my
daughters were receiving sexual advances early on. But this need not be any big deal so
long as all are prepared and they know that they can say no. This could be a problem in
some societies, but the Pirahãs never expected us to go along with everything they did,
because we were obviously different from them.
Children also need physical exercise and conditioning. In the normal course of
events, fieldwork toughens them up physically, emotionally, and mentally. Let them do
the normal village chores (gathering firewood, hauling water, collecting fruits, etc.) and
they will almost certainly grow up fit. And they will be far from fastfoods and other
'junk' food. They will often mainly eat fish, wild fruits, and drink water. Not a bad died
for them, as a matter of fact.

17
I remember my first LaMaze (childbirth preparation) class, for my second child,
Kristene, when I was 21. The instructor said, "Folks, there is no secret to raising
children. If you show them love and respect, they will grow up healthy even if you raise
them in the dark eating raw meat." Pretty close.
Linguistic Fieldwork Daniel L. Everett
page 39
Of course, children's education in the field must be good enough to ensure that
when they return to their home country they will be able to fit in (academically at least)
with no serious problems. Both formal and informal education of children in the field
will take fieldworker's time. Nevertheless, it is such a rich experience for all involved
that it is, in my opinion, easily worth the time involved. Formal education of children in
the field can be carried out as part of an accredited correspondence school, e.g. the
University of Nebraska's course (http://eeohawk.unl.edu/ishs/) or the Calvert School
(http://www.calvertschool.org/engine/content.do). These were great experiences for my
children and they generally found normal US schools easy, and always were placed in
the most advanced sections. Of course, children in the field can also learn about culture,
linguistics, environment, nature, and a host of other things that are very hard for other
children to learn as well by first-hand experience. Discussing cultural values of the
Pirahãs, linguistic issues of the Tzeltal language, Banawa puberty rites, etc. with my
children was enriching for all of us. And my children often asked better questions about
culture than I did, helping me refine and improve my research in various ways. Buying
a celestial map and showing your children constellations, stars, and other celestial
objects is very enjoyable. (In the Pirahã village, we have even seen man-made
satellites.) Your children can learn about local flora and fauna, about nature
preservation and so on. There are limitless educational opportunities for children in
fieldwork.
Children can also contribute to the work by noticing things that the linguist fails
to notice, by learning the language quickly and better in many cases than the linguist,
and by learning cultural values more directly than the linguist. They really can be an
integral part of the research team. And I have found that having my children with me
helps me fit into the community better and that the community trusts me much more if
it sees me as a 'family man' than as a lone researcher.

When we first went to the Pirahãs' village, my children were upset. They thought that
the Pirahãs were ugly, dirty, and weird. I tried to teach them about valuing differences,
but my words were ineffectual. After nearly 8 months in the village at one stay,
however, my children seemed to be making good friends. We were then visited by a
Brazilian army major who had come to check out rumors of a gringo in the jungle
exploiting Indians. The major said that the Pirahãs were the ugliest people he had ever
seen. My children were angered immediately and opened up on the major, telling him
that he didn't know what he was talking about, that the Pirahãs were beautiful, kind,
and fun and that anyone who thought they were ugly needed to have their head
examined. Not only did my children react in fluent Portuguese (because of their time on
the field), but they showed that they had learned a beautiful cultural lesson, something
no other situation could have taught them so well.

But there is no question that taking one's children to the field will have a huge
logistic impact on your fieldwork, quadrupling perhaps the amount of equipment,
books, and supplies you might otherwise take. You will need school books, candy, toys,
presents, and so on. I have carried sets of encyclopedia through the jungle on my pack,
for miles, so that my children could have reference works for their correspondence
courses.
Linguistic Fieldwork Daniel L. Everett
page 40
CHAPTER 3. GETTING STARTED
3.1. The first field session
Somewhere, sometime, somehow you bit the bullet and decided that you wanted
to be a fieldworker. And now you are planning your first session in the field, wherever
that happens to be. From the moment you first enter the field, you will never be the
same. You will experience a professional and personal rite of passage. How do you
plan for such an experience? Well, it involves both the mundane and the conceptually
involved. Let's begin with the mundane by considering first the kinds of things the
fieldworker might take to the field and then move on to more abstract aspects of
planning for that first experience.

3.1.1. What do you take?

The things a fieldworker takes to the field, especially on the first field trip, can
play a significant role in the success of the project and the well-being of the field
linguist. Things to be taken fall into two broad categories: things for the work and
things for the soul. The greater the cultural gap between the linguist's home community
and the language community, the more the linguist will need to take things that enable
him or her to feel a sense of 'continuous identity'. As Wengle (1988, 20ff) makes clear,
'identity maintenance' in the field is crucial for successful fieldwork (or, at least, it is
crucial for a healthy fieldworker). The isolation of the field can be extremely disruptive
psychologically. Wengle (1988, 7ff) advocates the idea that '... the stability of an
individual's sense of identity depends directly on the "innumerable identifications" he
has established with the familiar, personal and impersonal, concrete and abstract,
animate and inanimate objects of his past and present existence.' Judicious selection of
mementos, photos, books, and music, for example, can help the linguist to maintain a
connection with her or his life outside of fieldwork.
Linguistic equipment is discussed in chapter ___. A list of the other kinds of
things I usually take with me to the field includes the items in ():

(3.1) List of personal items commonly taken by D.L. Everett to field sites in Amazonia:
a. Three changes of clothes (one to wear, one to wash, and one drying), hat, flip-
flops, closed shoes, and gym shorts for bathing. This would obviously not be an
adequate wardrobe for fieldwork among the Inuit. What you take varies on
where you are going. But the principle is to travel lightly.
b. Medicines (for first-aid and treatment of serious village health problems).
c. Multiple vitamins (my diet in some villages is largely starch and meat, with
few if any fresh vegetables or fruit).
d. Food (the amount and kind depends on whether the group I will be working
with has surplus food to sell or trade to me – not all do), including popcorn
(people have different 'treats'. For me, the day ends better if I have popcorn).
e. Hammock and mosquito net.
f. DVDs of movies and concerts.
g. Pictures and letters from people I care for.
h. Satellite telephone.
i. Merchandise for paying language teachers (when money is not wanted. Cloth,
machetes, files, hoes, axes, munition, fish hooks, sewing needles, flashlights,
batteries, etc.).
Linguistic Fieldwork Daniel L. Everett
page 41
j. Toiletries and bathroom supplies (I recommend, strongly, deodorant –
wherever you work, people are likely to think you smell bad, because you smell
different).
k. My entire music library (I have about all my music on iTUNES and nothing
cheers me up more during a dull point in the day than ZZ Top or Lynyrd
Skynrd, etc. For you it could be Mozart).
l. Books (linguistics, philosophy, nonfiction, and fiction – especially Henry
James and PD James).
m. Folding table and chairs.
n. Backpacking stove.
o. Kerosene.
p. Batteries (12 volt and flashlight size).
q. Solar panels and connections to charge 12 volt battery. (see ___)

If there are special treats (like popcorn for me), that you like to eat to cheer
yourself up, be sure to take an adequate supply. However, there is an important caveat
to consider before you decide what to take: things you take, no matter how innocent-
looking to you, could and probably will, be desired by people in the language
community. This can lead to theft, hard-feelings, jealousy, demand for equal treatment.
Plan to take things to give as presents, to share your coffee and sugar (for instance), to
give away most of what you have taken before leaving the village, and to make sure
that nothing you take is too important to lose, be destroyed, or be taken. Never let
'things' come between you and the community. If things appear to create interpersonal
problems, go without them. Never argue about them. And that entails avoiding taking
anything you would be prepared to argue about.

3.1.2. How long to stay

The length of the first stay in the community should be based not only on the
fieldworker's available time, but also on a realistic assessment of his or her ability to
withstand the separation, newness, anxiety, and work responsibilities of the initial visit.
The general rule is to underestimate your staying power, rather than to overestimate it.
I suggest less than one month for the length of the initial visit. There is a
psychological/cultural issue involved in this recommendation. In general, when the
fieldworker gives a date that they plan to leave the community, whether only to
himself/herself or to others, this date is taken as a sort of promise, a kind of moral
commitment.18 So if a fieldworker overestimates his or her staying ability for the initial
visit, then they can feel that they are breaking a promise or 'wimping out', that is, that
they have failed in some way. In fact, anyone who goes to the field and stays for a
reasonable period of time should feel pride in their accomplishment, not guilt. If the
fieldworker leaves after the first visit feeling guilty, this would be an unnecessarily
inauspicious beginning to her or his field program. Therefore, I recommend that the
first session be planned for three or four weeks. The fieldworker can stay longer if
things work out well or return quickly if conditions permit.

3.1.3. Estimating the length of the entire project

18
This no doubt says more about Western Culture's view of language than about
fieldwork.
Linguistic Fieldwork Daniel L. Everett
page 42
Samarin (1967, p71):

"Criticizing linguists because they have not had enough time with informants is
like condemning shipwrecked sailors for not having provided themselves with
food and water. What we can more wisely do is evaluate how much was
accomplished in the time a linguist had at his disposal."

The estimate of the length of the entire project clearly depends on several factors,
including the nature of the project (e.g. to collect data for a grammar, dictionary, text
collection, partial study of phonetics, phonology, constituent order typology, etc.), the
degree of physical isolation of the community (from the fieldworker's perspective), the
degree of bilingualism in the speech community, the state of documentation and
description of the language of study, the state of documentation and description of
related languages or languages of the area, the goodwill and cooperation of the
community, and so forth. And, as the quote from Samarin says, all linguists would like
more field time than they have had before writing up their results. Therefore, a rough
time estimate for a comprehensive grammar (no dictionary, no carefully compiled text
collection) of an Amazonian language lacking previous research on it or related
languages would be, after all permissions are received (see __ below), in the
neighborhood of two full years of near-daily work and contact with the community.
The writing up of the grammar would take at least an additional 6-12 months, including
follow-up visits to gather, verify, or clarify data (see ___). For a language of North
American, where at least some in the community speak English and related languages
are well-described, the time for data-gathering might be somewhere between 6-10
months, again of regular, daily work with language teachers. I base these estimates on
an assumption of three hours per day of good lab sessions, excellent language teacher
help for data-processing in the field, good health, and no undue interruptions in the
work routine. By this reckoning, a PhD program with the projected output of a
grammar as a dissertation (in my opinion the most challenging, intellectually
demanding, and personally rewarding type of dissertation), with courses, fieldwork, and
writing up, should be estimated to take six years (in the British system, since no coures
are required, four years are the maximum allowed by national regulations).

3.2. The first day

In early 1977, I began a four-month 'Field training course' offered by the

Summer Institute of Linguistics (for more on SIL, see ___ below) in Chiapas, Mexico.
To all participants, however, it was known informally as 'Jungle Camp'. Jungle Camp
had five phases. First, there was an orientation to Mexico, cross-cultural relations, and
equipment purchasing in Dallas, Texas. Next, there were several days of adaptation to
Mexico, including visits to the Ballet Folklórico, Tenocticlan, Basílica de San
Guadalupe, etc. After this general orientation to the country, we proceeded south to
Chiapas, to SIL's initial training base, Yaxmiquilpan, where we learned to butcher beef,
preserve meat, give injections, provide first aid care, travel in dugout canoes over
rapids, repair kerosene lamps, live under primitive conditions, learn how not to get lost
in the jungle, which jungle plants can be eaten, how to hang a hammock with mosquito
net, etc. and we began physical conditioning, involving multiple day-long hikes of up to
thirty miles through the jungle, in very hard conditions (90%+ humidity, knee-deep
Linguistic Fieldwork Daniel L. Everett
page 43
mud, mosquitoes, etc.), including one fifty-mile hike to the village of the Lacandon
people. The training's principal highlight was 'survival hike', when people were pulled
out of the group at random and taken into the jungle to spend 3-6 days alone, with no
food, no water, and only what they were wearing, no change of clothes. This meant that
everyone walked around with rope, matches, a machete, and a few other tools that
could be carried, in anticipation of the survival hike. The lesson learned was very
simple indeed – be prepared (and all that preparation did help when I was called out
and spent 5 days alone in the jungle, foraging and fishing and other things that linguists
don't normally expect to do).
The Lacandon village was the first completely new cultural experience for me.
All other trips I had made, I could at least identify a few points of similarity with my
previous experiences. Not with the Lacandones, however. Everything seemed so
different than anything I had ever experienced before, that I felt somewhat
overwhelmed. Unfortunately, I had not received any suggestions on what to observe or
how to observe in such circumstances, no background on the Lacandones, no linguistic
orientation to them, etc. I arrived in utter ignorance and left no less ignorant and
shocked at how different things were than I had expected to find. It is never excusable
for a student of culture or language to enter another group without preparing
themselves carefully first with good questions to answer.

3.2.1. The significance of the first day

No serious fieldworker should ever go as ignorantly to a location as I did to the
Lacandon village (see box). So many things could have commanded my attention. For
example, how was the village located in relation to the jungle, bodies of water, other
villages or settlements? Who lived in each house and what were their kinship or other
relations to people in other houses? What were the kinship constraints on spatial
distance or proximity of dwellings? What greetings were used? (By the people for
foreigners? Between men and women?, etc.) How were people dressed? (e.g.
differences by age? By gender? By village prestige?) What were the dimensions of the
houses? Of the open spaces between them? What were the terms for basic vocabulary
(e.g. the Swadesh word list, see ___ below)? And on and on. I measured nothing. My
photography had no real scientific objectives. My first day among the Lacandons was
largely a day lost, a precious opportunity squandered. The first day in a new
environment is the only time you will ever see the culture without previous experience
of it. At the same time, from that first day, the sensory meter starts running. And this
flow of sensory experience biases the mind, by accustoming it, by making the new old,
by dulling curiosity. This in turn means that it is unlikely that you will ever be as
acutely aware to the newness and strangeness of this culture and language as on the
first day. It is your responsibility, therefore, to arm yourself with questions and to begin
asking and trying to answer them within your first hour in the community.19
This first day is also the day when the community will form its opinions about
you. Are you a trustworthy person? A nice person? What are your purposes here? How

19
In my favorite movie of all time, Unforgiven, Clint Eastwood shoots the unarmed
owner of a brothel, in front of the local law officer, Gene Hackman. Hackman remarks
that it is very cowardly to shoot an unarmed man. Eastwood says 'Well, he shoulda
armed himself.' Opportunity does not always wait for one to arm oneself intellectually.
A linguist needs to be prepared.
Linguistic Fieldwork Daniel L. Everett
page 44
approachable are you? Will you give them things 'for free'? Are you someone people
want to know better? Linguistically, you are more likely to impress positively by
accurate mimicry and quick learning and use of phrases. As you unpack your bags, you
should listen for question-like intonation and try to mimic it. Don't worry about
mimicking things you do not understand. Of course you will embarrass yourself. Just
do not take yourself too seriously and you will be fine. Have a good time.
This is also when you begin to form your working and personal relationships
within the community. First impressions are very important. The consequences of a bad
first impression are difficult to gauge. So it is better to give a good first impression and
to make an effort to receive one. Yet, that can be tricky. If someone asks you for
something, will giving it to them make you seem like a pushover, forever dooming you
to nagging about giving away your possessions, or will it be seen as a sign of
generosity, not necessarily inviting more requests? You must watch others, ask others,
read, learn from others' experiences, etc.
On this initial day of 'National Geographic' newness, you should be learning
more than linguistics. For example, this is the time to draw maps of the community,
learn who the community leadership is, where they are located, etc. Talk to the
community leadership on this first day (and before ever entering the community if this
is possible, by radio, one-day visit, in a nearby city, etc.). Explain your objectives, your
aspirations, etc. Find out about theirs. Begin negotiating the understandings necessary
for all to benefit from the research (see ___ below). Learn where and how to get water,
take care of your rubbish and other waste. Find out about proper relations between
foreigners (you!) and the community. Photographs are best taken at this stage, before
jadedness sets in. This is also a vital time for gathering metalinguistic phrases, e.g.
'What is this?' 'What is she doing?' 'When are you going?', etc. Use this day to the
fullest. It will never come again. And no other days will be remotely similar. Another
important, crucial task of the first day or so in the community is the selection of
language teacher(s). We turn then to consider that vital part of the first visit.

3.3. Language teachers

3.3.1. Selecting the teacher(s)
Before discussing the selection of a language teacher(s), we should consider what
to call the person or people who work for the linguist to teach him or her the language.
The traditional term for these members of the language community is 'informant'. But
in today's world this sounds rather sinister. Even worse than 'sinister', however, the
term 'informant' is bad because it implies that the language community participants in
the linguistic research are passive, with no goals of their own, little more than inert
sources of information for the fieldworker. The crucial desideratum for any term
chosen in this regard is that the term reflect the active and pro-active role of the
language community in shaping the research. I use the term 'teacher' here. There may
be better terms, but 'teacher' seems to be simple and accurate. Other terms people might
find useful are 'consultant', 'co-linguist', 'John' (i.e. simply refer to the teachers by their
first names/preferred titles in the first instance), etc.
There may be a range of people with different backgrounds, linguistically,
socially, and economically, to choose from. Ideally, one should choose to work with as
representative a sample as possible, never being locked into just one language teacher.
There are various reasons for this. Linguistically, one wants diversity and statistical
significance (sometimes these conflict) in one's sample. But another reason for not
working with a single teacher is that the linguist should avoid becoming a patron (in
Linguistic Fieldwork Daniel L. Everett
page 45
the Latin sense), i.e. an employer responsible for the employee's overall well-being. I
recommend working with at least 6 women and 6 men over time, from a variety of
ages. (But at the same time the fieldworker must guard against confusing data from
different dialects and offer suggestions on how to keep dialects separate.)
There are of course situations in which, initially at least, you may have little or
no choice as to who your teacher is going to be. When I first began research on the
Kisedje (Suyá) language, I saw several potentially good teachers. The best it seemed to
me was a man who was the community's official translator. He was bilingual in
Portuguese and Kisedje. But when I approached him, he said that he had no time for
such work. Moreover, he said, it had already been decided by the chief that my teacher
was to be his daughter. Faced with a community-based decision of this type, it may be
acceptable to explain why you'd prefer someone else. But that can be risky at the outset
of field research, when you do not know the people or the culture well. By and large the
safest course of action is to make the best of the community's selected teacher(s) and
later, when you are better attuned to values, decision-making, persons, and
consequences in the cultural context, you can ask for a change in or additional teachers
if you think it would work better. Once again, however, linguistic research always
requires multiple teachers. This is because all conclusions are based on the assumption
that they have inter-speaker validity. You cannot make this assumption if you work
with only a single language teacher. (So don't worry too much about who you get as
your first teacher if this person is selected for you.)
Let's assume that the linguist can work with anyone they please. How do they
go about selecting a language teacher? Some qualities to look for in a language teacher
are given in (3.2), although none of these are ultimately necessary or sufficient to
guarantee a good teacher. But they are all helpful, I believe.

(3.2) Selecting a language teacher:

a. Good story teller
This is important because it will affect the quality of the texts you collect.
Especially when collecting oral literature, you should attempt to work with someone
respected in the community for telling stories of this sort well and appropriately.
b. Friendly
I have worked with good teachers who are rude or otherwise less than
friendly. Other things being equal, it would be nice to work with someone that you
have a good chemistry with, someone you consider friendly, for example.
c. Speaker of the correct dialect
It would be a disaster if you concluded a significant portion of your field
time only to discover that one of the speakers you had worked with in fact spoke
another dialect or had given you a very different register than what you thought you
were getting. So you must check out from the outset whether your teachers are
representative of the dialect you are studying.
d. Mother-tongue speaker
This sounds obvious, but if you have not worked on a language previously,
you might mistake someone for a native speaker of the language when in fact it is their
second or third language. You must find out what language their mother spoke to them
and what language they now speak in their homes.
e. Patient
The fieldworker is continually asking questions that a child should now,
from the perspective of the native speakers. And they ask questions about details that
Linguistic Fieldwork Daniel L. Everett
page 46
are extremely uninteresting, asking one to repeat the same word over and over. Thus,
working as a teacher for a linguist can be extremely tiring. A good teacher should be
patient.
f. Reliable
The linguist's time in the community and the success of the entire field
research program depends on the ability to plan. If a language teacher cannot be relied
on to show up sometime near the pre-arranged time and if this goes on for too long (the
'too' to be determined by the linguist), then the teacher is not reliable and should
therefore be switched, however delicately, to avoid offense, etc.
g. Unintimidated by linguists
Some language teachers are intimidated by the linguist's education,
nationality, money, big words, cushy job, and so forth. Some of these teachers therefore
do not like to correct the linguist's pronunciation, or his or her conclusions, data, etc.
h. No speech impediment
You need to be sure that the pronunciation you are hearing is a standard one
in the dialect/language under study. If someone has an obvious speech impediment, you
cannot be you are getting standard pronunciations. This can lead to faulty
generalizations about the language's phonetics or phonology. Hence, at least for sound-
system analysis, language teachers should be free from any speech impediment.
i. Shows ability to reflect on language as a formal system
For some language teachers, the linguist's structural investigations will have
little interest or meaning (moving stress around in the word, building and testing
sentence paradigms, i.e. avoiding questions on meaning for questions on structure. But
there are rare teachers who (like rare university students) turn out to be talented, natural
linguists and show an ability to reflect metalinguistically on their own grammar. Such
speakers can suggest paradigms, find regularities before the fieldworker sees them,
provide transcribed, translated data, etc. They are marvelous linguistic resources and
should be looked for, cultivated, and trained. Where appropriate, the fieldworker should
attempt to get linguistically talented teachers more formal training if they desire it.
Alan Vogel offers the following suggestions on training native speakers, based again on
his Jarawara experiences:
"I think two kinds of training may be relevant here. First, I trained Jarawaras to
transcribe tapes of recorded stories. These transcriptions are not perfect by any means,
but they are a huge help to me. For one thing, they hear a lot of things that I would not.
They gladly spend 8 hours a day transcribing tapes. Secondly, I made up a
questionnaire on plants with about ten questions. I gave the younger Jarawaras, who
write well, the questions, along with sheets of paper, each with a name of a plant that I
had gotten, and asked them to get the answers to the questions from the old men who
know all about the plants. It produced volumes of information on hundreds of plants.
It's important to realize, of course, that when we started with the Jarawaras, no one
knew how to read or write in Portuguese or their own language. We had to devise the
orthography (adapting the Jamamadi orthography), and after members of another
missionary organization made up a primer, my wife taught several young men how to
read and write Jarawara. She taught them how to teach others, and they did. The other
missionaries among the Jarawara did the same thing in Agua Branca, and today the
Jarawaras are basically literate in their own language."

j. Well-respected in the community (at least not marginalized)

Linguistic Fieldwork Daniel L. Everett
page 47
Occasionally someone available to work as a language teacher is ostracized
from the community or seen as marginal (in fact they may be trying to get work as a
teacher because they have no other employment). This can be detrimental to the
research if when you proceed to check examples with other language teachers they
reject all examples given to you by your 'marginal' teacher, simply because of that
person's social status rather than because of their linguistic ability. But you will not be
able to tell the difference easily, affecting the research in ways that might not be easy to
predict.
k. Expects to teach language, not translate the Bible, etc.
In some field situations, communities with a strong or new church, with a
history of close ties or cooperation with foreign missionaries, the community or
individual language teachers might confuse the fieldworker with a missionary and
expect that the linguistic work on the language is ultimately directed towards a Bible
translation. The linguist must make it clear at the outset that this is not what they are
there to do. Avoid creating false impressions. (See __ below on ethics for further
discussion.)

Regardless of how many or how few teachers you actually work with, it is
important to know how to select them. As you select your teachers, you should also
give thought to how you present your linguistic objectives, the nature of your job and
the nature of your teacher's job to the community. Perhaps the community will have
had previous experience with anthropologists (as with the Kisedje). In this case, it is
important that you distinguish your objectives, because the linguist will work quite
differently from the anthropologist, with objectives that are perhaps harder at times for
the people to grasp or sympathize with. Avoid claiming that your objective is to learn
their language unless you in fact intend to learn it. It can cause misunderstandings if
you are cheerfully working away but with no marked progress in your ability to speak
the language. If the teachers saw that as their primary goal yet you have not progresssed
much, this could lead them to conclude that they are bad teachers, you are a bad
student, or both. This can lead in turn to a lack of interest in helping you.20

3.4. The laboratory session

3.4.1. Linguistic planning of the session
The fieldworker must develop and maintain a high degree of professionalism in
preparing the space, time, language teacher, and their own mind for the lab session.
This section provides some suggestions for this preparation.
Have a point! Each lab session should begin with a well thought out, measurable
objective. For example, 'Today I want to record and translate a narrative of 15-20
minutes in length.' Do not waste your time or your language teacher's by scheduling
sessions for improvisation. As Peter Ladefoged told me, "...when in the field spend
more time thinking what you want to record than making recordings. I so often see
students and others wasting time recording masses of useless stuff, hoping they can use
it later when in fact it will simply delay their later work by their having to go through

20
My opinion is that you should do your utmost to learn to speak the language and to
demonstrate regular and obvious progress at each meeting with the language teacher. It
would be nice if others in the community complimented you and your teacher(s) for
your progress (if complimenting is something they do).
Linguistic Fieldwork Daniel L. Everett
page 48
it all. Think hard first, record later." [emphasis mine, DLE] Peter Ladefoged, email to
Dan Everett November 26, 2003

Additionally, you should have back-up plans. As Mick Jagger reminds us, 'You
can't always get what you want.' For example, if the speaker(s) do not seem to want to
give you a narrative text, you could have planned a series of questions, e.g. 'Tell me
about your hunting trip', 'Can you tell me how to weave a basket', etc. That is, open-
ended questions that likely will involve multi-sentence answers and can give you at
least something text-like. The secondary goals need not be related to the primary goal.
But time with the language teachers is precious, literally, and it should always be fully
and productively utilized. On the other hand, do not keep the teacher with you if you
have run out of planned material. There is no sense in paying someone to watch you
sputter and spin your wheels. In fact, you could give them evidence that you have no
idea what you are doing, which is not a good impression to make.
There are various types of elicitation that can be undertaken in the lab session
(and in this section I am heavily indebted to Samarin (1967, 112ff), where some of
these types are first spelled out).

(3.3) Types of elicitation

a. Questionnaire-based: I often find it useful to follow a questionnaire,
modifying it for intelligibility and local relevance, e.g. the Lingua Descriptive
Questionnaire. (I have included an appendix to this book with a phonology
questionnaire I developed, for the fieldworker thinking of doing more serious
phonological studies, though it is not necessarily a tool appropriate for
elicitation.)
b. Complementary: looking for further examples of material already collected.
c. Probing: asking open-ended questions, looking for explanations and
paradigms provided by the language teacher. (Answers in such elicitation are
simply one form of input and should never be taken as settling any issue.)
d. Hypotheticals: testing lists of hypothetical words generated by computer (see
William Poser's webpage (http://www.ling.upenn.edu/~wjposer/) for one such
program).
e. Translation: getting translations for texts and other material. Matthewson
(2004) discusses semantic elicitation (see also __) and some of the pitfalls of
eliciting translations. In spite of her well-taken warnings, though, getting
translations of the data collected is vital and lab sessions need to be dedicated to
this task. The fieldworker, however, must be careful to treat native speaker
translations as only one source of data about the meanings of texts, sentences,
etc.
f. Reverse translation: another way of getting at the meanings of texts, etc., is to
have one set of speakers translate a vernacular text (a text in the language under
study) into the national or local trade language and then have a different set of
speakers translate the translation back into the vernacular. The fieldworker can
then compare and contrast the different forms given and design further lab
sessions to probe and test the resultant hypotheses.
g. Corrective: this type of elicitation tests the linguist's understanding of a
certain set of constructions, expressions, texts, constraints, rules, etc. by
discussing examples that have been devised based on the linguist's current
understanding of the grammar. This is likely to have a low and unreliable rate of
Linguistic Fieldwork Daniel L. Everett
page 49
return but, so long as the fieldworker bears this in mind, it can play a useful role
in analyzing the relevant structures.
h. Filling in gaps: looking for missing paradigm forms, gaps in phonetic charts,
sentence types, different participant lists, etc.
i. Experiments: It is important, where expertise is available, that some questions
of structure, meaning, and understanding be tested by standard psycholinguistic
experimentation. Although I have nothing to say about experimental design
here, fieldworkers should either acquire the skills necessary to design, carry out,
and interpret psycholinguistic experiments or they should recruit expert help, to
either help them design experiments before leaving for the field or to
accompany them to the field to carry out the experiments en loco.

3.4.2. Non-linguistic planning for the lab session

An early decision is where the lab session will be held. It would be nice of course
to have a sound-protected location, comfortable, bug-free, dry, and free from
distractions. However, this is not usually possible. But it is vital that the site chosen be
as close to ideal as the fieldworker is able to get in the community. This could mean
little more than taking a couple of folding chairs and table out of the village, busy
neighborhood, etc. to a relatively calm, isolated area. Peter Ladefoged and I once
commandeered a hut in a Pirahã village to carry out phonetic investigations and I
simply asked people to keep the children and dogs away and to keep the general noise
level as low as possible. They kindly accommodated us. Care in the choice of location
will directly improve the research. And it can indirectly improve the quality of research
by showing that this is serious business.

Of course, like any field activity, selecting and developing a location for lab
sessions can have unintended consequences. For example, among the Pirahãs, I built a
small wooden structure about 100 meters outside the village, raised above the ground,
and screened in, with a lockable door. I discouraged people from looking in during
sessions and tried to allow in only the teachers working in a given session, rather than
them and all their friends. The Pirahãs do not mind this. But a Brazilian government
agency investigating the activities of a 'gringo' in the area asked the Pirahãs about this
small structure when I was absent from the village. The Pirahãs, in their nearly non-
existent Portuguese, were able to communicate that I spent a lot of time in there and
that few people were allowed in. The agency representatives wondered what sort of
fiendish experiments I might be running out there. Finding out about this, I quickly
visited the agency headquarters and gave them a full explanation.

I cannot imagine a successful lab session without the following equipment:

headset microphone, digital and analog recorders, computer, pen and paper, chairs,
table, and protection from the most intense features of the elements. In ___ below I
discuss the linguistic equipment and its use in more detail.
The time of the lab session must take into account primarily the constraints of the
community and the language teacher. When is it best for the community and the teacher
for the lab session to take place? If the linguist can choose the time to work freely, then
I suggest that the session take place while the community is quietest and calm. The
linguist must keep lab sessions appointments and show that they take these times
seriously. This could be in some cases part of the training of the language teachers as
well – all participants must respect the research time. If language teachers show
Linguistic Fieldwork Daniel L. Everett
page 50
themselves unable (for whatever reason) or unwilling to show up at agreed times, then
they and the linguist should discuss possible solutions. If such discussions still fail to
provide a satisfactory solution, the linguist may need to take the issue up with the
community leadership, asking for advice.

3.4.3. During the session

Fieldwork is about getting high-quality data in the appropriate quantity (see ___).
Nowhere is the data more likely to be gathered than the lab session. And the success of
these sessions depends largely on the rapport, trust, and mutual understanding of the
linguist and the language teacher. Therefore, the lab session must be a meeting of
equals. The language teacher must be fully confident that they are respected by the
linguist, and vice-versa. Care must be taken to ensure that the physical arrangement of
the space (chair, table, etc.) reflects equality rather than hierarchy. The linguist's facial
expressions, tone of voice, and body language must at once show professionalism,
respect, and enjoyment of the task at hand.
Frequently in the sessions the linguist will find a particular response puzzling,
apparently irrelevant, or otherwise unhelpful. Receive any 'unhelpful' response with the
same gratitude that should mark the linguist's attitude towards every response and help
from the language teacher. Respect means allowing the teacher to offer information that
does not always coincide with the linguist's request.

As I have listened to some of my more than twenty-five year old tapes of the
Pirahãs, I have, with my hard-earned ability to speak the language, realized that some
of these 'unhelpful' responses were in fact attempts to correct some of my mistaken
impressions, offering, instead of what I had ignorantly asked for, a response that was
much more helpful, had I only recognized the Pirahãs' ability to teach me, rather than to
simply answer my questions. That is, had I realized that they were teachers rather than
mere passive 'informants'.

Further, the language teacher will look for evidence that the linguist has
understood or at least heard correctly their response. The best way to show this (and
that the linguist is paying attention) is to repeat every example back to the teacher.
Make it clear that you want to be corrected.
One source of potential pressure on the language teacher that can lead to less
useful responses is for the linguist to reveal too much about his or her predictions and
analysis before it has been carefully verified. this could bias the results, leading the
teacher to, in a friendly way, look for examples to confirm what you are saying or
otherwise bias the nature of the results obtained. (This excellent bit of advice is taken
from Boas (1911, 59)).
Finally, the linguist and teacher must know when the session should conclude.
There are various criteria. First, conclude at the agreed time, unless there is mutual
agreement that the session should continue for some reason. Second, conclude when the
linguist runs out of material for the session. Third, the session should be concluded if
this is a bad day for the teacher and they cannot get focussed. Fourth, conclude the
session if there are too many distractions (e.g. a hunter has just brought back game and
the language teacher is concerned about missing his or her share).

3.4.4. Third-party interpreters

Linguistic Fieldwork Daniel L. Everett
page 51
In some field locations, it may not be possible to engage in either monolingual or
bilingual work directly with the language teacher. In such cases, it may be necessary to
employ a third-party interpreter (see also Samarin ()). Such a situation arises when the
linguist and the language teacher fail to share a common language and the monolingual
method is impractical for some reasons (e.g. when the time available to the linguist is
too short, the teacher is one of a few survivors of a moribund language, etc.).
There are some special precautions to take in working with interpreters. First, the
linguist must be clear in all instructions given. It can be safely assumed that all
questions asked through an interpreter will be distorted to one degree or another.
Ambiguity or vagueness in questions will create more severe problems when asked
through a translator. Second, the linguist should be suspicious of the quality of the
question, the translation, or both if short questions to the interpreter become very long
questions or conversations between the interpreter and the language teacher. Answer in
such circumstances should be viewed as highly suspect. Third, be equally suspicious of
short answers by the teacher which require apparent circumlocutious translations to the
linguist or whose translations are hesitant, unclear, etc. Finally, since the interpreter has
a closer linguistic connection than the language teacher to the linguist and, perhaps,
more formal schooling, there may be a tendency for the language teacher to perceive
that they have lower status. Avoid this. Neither the linguist (!) nor the interpreter should
communicate any attitude of superiority to the language teacher.

Lab sessions can involve culture shock and problems that the linguist did not
foresee. For example, I was working once with a couple of Pirahã men who had
contracted colds from visiting Brazilians on the TransAmazon Highway. As we were
working, they would sneeze and then take the mucous and fling it away from them,
often landing on my notebook paper or on me. Since the Pirahãs see absolutely nothing
unhygienic about this, they could not understand my request that they not do this. 'We
have colds', they remonstrated. 'This stuff comes out your nose when you have a cold.
Don't you get colds, Dan?' Or another time I was working with a group that had
apparently eaten something hard to digest. There were sounds and smells coming out of
their nether regions that are hard to describe. As I told a friend later, they passed more
gas than the Alaska pipeline. One time a female language teacher decided to flirt with
me. She wasn't very subtle. She took her dress off and asked if I had ever had sex with
women other than my wife. This was apparently no big deal to her, but it was, um,
distracting in the midst of transcription to think of a way to say no politely. And then,
of course, I offend others. I am hairy to Amazonian Indians and smell quite bad
(Westerners perspire much more than Native Americans and thus do smell much worse
in hot, humid climates). Be prepared to accept things with a smile that you might not
have anticipated.

Monica Macaulay has written one of the best short pieces on culture shock (or culture
strain) in the field that I have read. Here are some of her experiences:
"Let me describe some of the things I was dealing with:
First, I had trouble finding food. Eventually, I was able to establish a routine where
I had a noon meal at a restaurant, and otherwise ate bananas and tomato sandwiches.
But bananas, tomatoes and bread were all things I could only buy twice a week, on the
two market days. If I didn’t buy enough, and ran out of food, sometimes I could buy
little packaged pastries at one of the stores, but that was about it. I lost a lot of weight.
Linguistic Fieldwork Daniel L. Everett
page 52
Second, it was the rainy season, and I was at an altitude of about 9,000 feet. It was
extremely cold, and I was not prepared for it. I wore layers and layers of clothes, and
froze when I had to wash something. We did have electricity in the house I lived in—
most houses in the village had it. But of course there was no heat, and often when it
rained the lights would go out.
I almost immediately became covered in little red bites from invisible insects. No
bug spray or lotion helped, and this continued for the entire time I was there...
I did manage to find consultants fairly easily, but they stood me up all the time. Of
course the notion of scheduling appointments was not quite the same to the people I
was dealing with as it was to me, with my Midwestern expectations of promptness and
politeness. If the consultant wasn’t available, the day was shot for me—it was very
hard to get people to agree to work without at least a day’s notice.
Then there was my work itself: the more I worked on the language, the more
incompetent I felt. I had terrible fears that I was putting myself through this torture for
nothing—that there would be no results to show for it when I returned home.
One thing that I expected that I actually did not find was unwanted attention from
men. I had been in enough big cities in Mexico to know what to expect along those
lines, but it did not happen in Chalcatongo, at least not from the locals. They watched
me, certainly, but it was more like being an animal in a zoo than a woman being ogled
by men.
Unfortunately, there were a few men there from bigger cities, and they did give me
some trouble. One in particular was a real problem. He would get drunk and pound on
my door, and say strange things to me. At times I felt completely confined to my
room—that it simply was not safe to go out. This was especially problematic since of
course the bathroom was across the courtyard.
Most of the time, though, I was simply a curiosity. Occasionally there was ridicule
that I was aware of, but usually they just stared at me...
... how alien I felt, and how alien they regarded me as being."

3.4.6. Serial teaching, group teaching, and individual teachers

3.4.6.1. Group sessions
In any single laboratory session, the linguist will work with at least one
language teacher. But they could also work with multiple teachers simultaneously. It is
often useful for the linguist to work with small groups as well as with individuals.
Group sessions (3-4 language teachers) have various advantages: (i) they can allow the
linguist to put the language to use right away, by checking immediately what one
speaker gives with another speaker in the same session; (ii) they immediately provide
the linguist with alternative phrasings and pronunciations of the data and alternative
ways of telling and/or interpreting a short text; (iii) they provide speakers a chance to
discuss their language (by discussing their answers and the linguist's questions),
developing linguistic sophistication and awareness of the linguist's objectives, at the
same time that it also helps the linguist to learn more of the language and meta-
language by listening attentively to the native speakers converse.

3.4.6.2. Serial teaching in lab sessions

The Pirahãs have never developed a great deal of patience for sitting and teaching
me their language. So rather than have one person sit for long sessions, I schedule
different language teachers for every 20-30 minutes for the entire morning, giving
myself a coffee break or two in between sessions. This breaks up my day, keeps
Linguistic Fieldwork Daniel L. Everett
page 53
language teachers from getting bored, and provides many opportunities for clarifying
and checking other teachers' data. Serial sessions also have another significant
advantage over working long periods with a single teacher. Speakers who only work
15-20 minutes per day will not usually see the linguist as substantially in debt to them
for their time and services. They can see their participation as a break, as 'fun', as a
change, etc. rather than serious work. And this lightened attitude and environment can
translate into better exchanges and more openness between the linguist and the teacher.
The upshot of this is that the fieldworker should remember that there are various
options that should be explored for working with language teachers beyond working
with a single person for long periods of time.

It is not so much what you do, but what you do next

This may all make the lab session look like a traumatic experience. But in fact
these sessions will almost certainly be rewarding and exciting experiences. A general
rule of thumb is the old advice 'It's not what you do that matters, it is what you do next.'
You will make lots of embarrassing, side-splittingly stupid and funny mistakes. You
will murder the language, utter profanities unintentionally, and do any number of other
things that will embarrass you and confuse the language teacher terribly. This is all
unavoidable. But what the linguist does next, what you do next, that is, after the
embarrassing mistake, is what really matters. Laugh it off. Go back to the drawing
board and start again. Do that and you will succeed. Don't do it and you will fail.
Period.

3.5. Training language teachers

3.5.1. Why train teachers
Why would the linguist, whose time is already far too limited, want to take any
extra time away from his or her linguistic investigations to train language teachers in
any skill not directly related to the linguist's own goals? The answer is primarily that
this is an ethical responsibility. Just as the linguistic research is ultimately intended to
enhance the linguist's career and improve the linguist's quality of life, so the research
should enhance the teacher's quality of life. This is empowering fieldwork and it is not
an option. Linguists and anthropologists, like missionaries, have the skeleton in the
early histories of their disciplines of following in the wake of and contributing to
attitudes of colonialism (every fieldworker should read Said's () Orientalism). Passing
on skills and knowledge in mutual exchange and respect should be the hallmark of
modern fieldwork. The linguist takes and must give back. And money is not enough.
When teachers are trained and interested in the research on their language, they
can become pro-active co-researchers. As they come to understand and identify with
the research on their language, they can gather a good deal of the data, record it, and
transcribe it. They can also translate it themselves. The linguist can then review all the
data they have collected and verify it against the developing analysis and verify it in
light of the developing analysis and check it with other speakers for relevance,
naturalness, accuracy, etc. I found that this method worked reasonably well with the
Banawá. Alan Vogel (p.c.) tells me that this works well with the Jarawara people of
Brazil.

3.5.2. What kind of training?

The linguist has an obligation to at least train the teachers and others in the
community in technology, mechanics, medicine, linguistics, general world knowledge,
Linguistic Fieldwork Daniel L. Everett
page 54
and any other kind of knowledge that the community sees as important that the linguist
is able to help them with.

I have trained Pirahãs in the operation of my motor boat. We have had lots of
fun driving this National Science Foundation-purchased boat and 40 horsepower engine
around the Maici river. Pirahã men have piloted us right into jungle growth on the river
bank and onto beaches. But we had a blast. And now when I am in the village, I know
that if I am injured or ill, the Pirahãs themselves can pilot me out. So I get a direct
return on this investment of time. And the Pirahãs now have a skill. They do not have
motorboats, but they enjoy showing outsiders how they can pilot mine. The Kisedje
people of the Xingu regions of Brazil already know how to drive cars, pilot motor
boats, and handle other types of Western equipment. However, they need computers for
some of the activities of their tribal association and want their people trained to use
these computers. So my project provided them with a laptop computer and training in
its use. We are on regular email contact now, no matter where I am in the world. I have
also trained people in giving injections of anti-venom and in the use of anti-malarial
pills. I regularly take in National Geographic and other movies about other regions of
the world so that they can learn about other peoples, regions, animals, etc. These
discussions are enjoyable activities for all of us.

How does one go about training language teachers? This will depend
tremendously on the level of familiarity of the people with the things that the linguist
brings to the field, e.g. education, equipment, medicine, etc. The more familiar they are,
the more likely they are to have specific objectives in training in one or more of these
areas. For linguistics, the linguist should discuss his or her objectives and the
methodology of linguistics. And they must always be careful to explain global and local
goals. Why did they come to the community? What are they trying to learn about in
today's session. Why is the linguist asking this or that question? What are they trying to
learn today? Training can also be facilitated by asking the teacher for his or her advice
and insights. This type of reflection helps the teacher to think like a linguist about the
language. Different approaches are required by different cultural contexts.

3.6. Styles of work

Some field linguists work with language teachers eight hours a day, seven days a
week, the entire time that they are in the community. Others, myself included, find that
a less intensive lab schedule works better. I recommend two-four hours per day of lab
sessions, including verification of data collected with multiple teachers. The remainder
of the day should be used for data-processings, analysis, hypothesis formation,
language practice, and planning the next session. (See also section __ on how I spend a
typical day in the village.)

3.7. More on data-collection

3.7.1. Processing the data
Data that are not processed in the field, with the intensive participation of native
speakers, are relatively useless. The fieldworker will waste much effort if texts and
other data are collected to be processed when they return to her home institution. Lab
sessions must be dedicated to transcribing, translating, and testing all material.
Part of a the data processing, the easy part, is careful annotation of the data for
storage (this is a form of 'metadata', though I avoid this term here because of its largely
Linguistic Fieldwork Daniel L. Everett
page 55
computational connotations, taken up briefly in 5.8.). Suggestions for annotation are
given in (3.4) below (which borrows from Samarin ((1967); see also the EMELD
website, http://emeld.org).

(3.4) Data annotation

a. Code: This gives the genre of the text (e.g. N(arrative), E(xpository),
H(ortatory), P(rocedural).
b. Sound file number, i.e. where the text is found in the audio database.
c. Topic: What is the text about ('fishing story', 'story about jungle spirits', etc.)
d. Date: When the text was collected and when the tape was transcribed (thus
there will be two dates).
e. Language teacher information: Who produced the text; who helped transcribe
and translate the text; sociocultural information on language teachers (e.g. age,
gender, marital status, occupation, birthplace, knowledge of other languages,
religion, language of parents, societal status, etc.)
f. Team member information: Which linguists worked on the text and what was
the contribution of each?
g. Place: where was the text recorded (village, city, university, etc.).
h. Any notable characteristics of the text (e.g. this ritual text includes prosodic
phrasing that appears quite unusual; the language teacher uses expression x a
great deal, etc.).
i. Other comments the linguist believes are relevant.
j. Each line of the text should be numbered and cross-referenced to its location
in the audio database (how can a reader find it to listen to it, quickly?).

3.7.2. Transcription
As a rule of thumb, never estimate less than a 4:1 ratio between transcription
and recordings. That is, for each hour of recording, it will take the linguist at least three
hours to transcribe the data, if they are already very familar with the language and its
sounds. If you are just beginning in a new language with unfamiliar sounds, the ratio is
more likely to be 5:1 or 6:1, or even more, until you get used to transcribing this
particular language.
I usually proceed as follows. First, I transcribe the text (or other data) on my
own. Then I read my transcription to a native speaker. With a different color of ink, I
write in that speaker's correction of my transcription, as I pronounced it. Where doubts
and confusion arise, I play the original recording of the text and ask the teacher to tell
me what was said and what it means. I use my own pronunciation initially because it is
the best and most immedate check of my ability to pronounce and transcribe the
language. The native speaker will almost always easily understand the original text,
since it is by another native speaker. But I want to see if I have it right. I indicate
whether I think any correction is of the actual phonetics or whether it is a 'prescriptive'
correction, i.e. that it reflects what the second speaker believes that the first speaker
should have said if speaking correctly (it is very common for speakers to make fun of
one another and to think that the other's use or knowledge of the language is inferior to
their own – common to anyone who has much experience around universities).
I then continue with this same language teacher to get a translation of the text as
a whole, the individual lines or sentences of the text, and, to the degree possible,
individual words and morphemes. Following this section, I study the text and
translation, making notes of doubts I still have and structures that are unusual in
Linguistic Fieldwork Daniel L. Everett
page 56
content, form, linguistic complexity, or cultural information. I often repeat this entire
process with one or two other language teachers. Finally, I often go through the entire
text, when it is particularly complex, with the original giver of the text.
In cases of unresolved doubt on the sounds (segments or prosodies) of the text, I
look at wave forms and spectrograms. In cases of continuing doubts on meaning, I may
continue to investigate with other language teachers or I may simply set the text aside
for a while (perhaps even years).
I recommend too that indelible ink (waterproof at a minimum) be used for all
transcriptions and that corrections be done only by strikethrough, never completely
obscuring the corrected data, because that could, on subsequent reflection or
investigation turn out to be the more accurate transcription after all.

3.7.3. Fieldwork without language teachers

David Gil (email, 14-October-2005) gave me an interesting answer to a question
I put to him on working with language teachers:

'To be honest, for the last decade or so I've actually done very little informant
work myself. Reasons are, I find other sources of evidence generally more
reliable, and using other such sources more fun and more rewarding. In fact, I
would go so far as to say that I haven't done any informant work at all in many
years on any of the issues that REALLY interest me; the little that I have done
sporadically is mostly to check things for other people, or for very mechanical
things like collecting word lists... As a result, my main source of evidence is
naturalistic speech (via either eavesdropping or recording and then transcribing
longer stretches), with various types of experiments as an alternative.'

Eavesdropping can indeed often be a useful resource, even when the linguist is
not in a situation similar to Gil's. One can often find what they need by spending
unstructured time with native speakers and engaging in directed, selective
eavesdropping. That is, the linguist can have a specific question in mind and listen for it
to come up (obvious examples would be greetings and leavetakings, verb forms,
pragmatic conventions, phonological features of natural speech as opposed to elicited
speech, etc.). The linguist can also use eavesdropping for probing, i.e. looking for new
structures that they have not encountered before (or doesn't remember encountering).
But it is important then that the linguist always have pen and paper or, where practical,
a small taperecorder, to hand at all times to jot things down. It is of no use to try to
remember at the end of the day useful sentences overhead earlier.

3.7.4. Coaxing texts out of speakers

Here are some of my thoughts on ideas on collecting texts. It is not always easy
to get a speaker to give you a natural text. I recommend Longacre () as a general
overview of discourse analysis. The reader is also referred to ___ for more on discourse
analysis.

Alan Vogel, who has worked for many years on the Jarawara language,
recommends the following:
"One thing I do is always carry a list in my pocket of stories I want to record. When I
am talking to people, and I hear a reference to some experience or story that sounds
worth recording, I make a note to myself to record the story later, jotting down the
Linguistic Fieldwork Daniel L. Everett
page 57
person's name and something about the subject matter. And I ask the person if he would
be willing to record the story another day. Then later, when we both have time, I record
it. This is a good way of avoiding the problem of sitting down with someone and asking
them to tell you a story, and they don't tell a good story, because you don't know what
stories they know (or maybe they are not a good story teller)."

NARRATIVES: Some speakers will not want to give informal narratives. They will
want to provide the linguist with formalized/stylized texts of cultural significance, i.e.
recognized oral literature. Such material is of course wonderful and the linguist
certainly must collect a good deal of it. However, the linguist also needs informal
narratives of the type 'What I did for my summer vacation', 'How my fishing trip went
today', 'What I told my son to do to avoid jaguars', 'How my little girl made her dress',
etc. Different speakers will have different abilities and preferences for different kinds
of narrative texts. There is no magic answer. One simply has to experiment with
different methods and speakers. But one should begin with the simple stories and work
up to traditional stories, which are longer, often use archaic language, and are much
more difficult to analyze. Moreover, in many communities only certain elders are
allowed to provide, translate, or otherwise work on traditional stories with the linguist.
PROCEDURAL: This kind of discourse is easier to collect, because it involves a
speaker telling the linguist how to make something, following a specific order of
activities, e.g. a recipe, how to make a bow and arrow, etc. These were the first texts I
ever collected among the Pirahãs. And they are among the easiest to understand, since
you can pick out individual parts of the process and figure out fairly easily how the
overall structure of the text works. It is an excellent genre for collecting imperatives,
temporal connectives, and other natural features of recipes, etc.
EXPOSITORY: These are explanations, ideally of culturally important information.
So a language teacher might explain how a man becomes a shaman, what the village
headman was talking about last night in the text you could not understand, etc. Again,
these are usually not too difficult to collect. They have characteristic aspects, tenses,
and other features which set them off.
HORTATORY: These texts are often quite difficult to get in elicitation, though they
are usually much easier in eavesdropping (and for this reason the effective
eavesdropper should always be armed with a recorder and good microphone, ready to
be activated at a second's notice).21
CONVERSATIONS: Conversations are at once the most important genre of text to
collect (because they contain natural, everyday use of the grammar) and the most
difficult. To set two or three speakers down together and then say to them 'Converse',
is not conducive to natural discourse. But conversations can be recorded. In July 2004,
for example, I was able to record a natural, long (30 minutes), and linguistically rich
conversation from two Banawá men, Sabatao and Bido. They each wore headsets with
high-quality unidirectional microphones and were recorded onto separate channels on a
digital taperecorder. The conversation was then transcribed, glossed, and filed. It is the
best data I have ever collected from the Banawá.

21
Of course, one must never record anyone without their permission. So the
community should be asked if it is OK for the linguist to record things spontaneously.
Assuming it grants permission, then for any recording made in eavesdropping
conditions, the linguist must then get the specific and explicit permission of all those
recorded for that recording to be kept, processed, and used in linguistic analysis.
Linguistic Fieldwork Daniel L. Everett
page 58
At the same time, it proved much harder to collect a similarly natural
conversation from the Pirahãs, whom I know much better. I tried various strategies
after it became clear that just asking two people to converse was not going to give
useful results. I have just set with people in their huts, with the recorder running, often
getting natural conversation. However, since they are obviously not wearing headsets in
such circumstances, the sound quality of such recordings is far inferior. Nevertheless, it
is good enough for prosodic and morphological analyses in most cases.

3.8. Digitization of data and the internet

Before travelling to the field, the field researcher and all members of her team
(which may include no one else but the lone fieldworker) should be trained in the field-
collection of audio, video and pictorial primary data and determine the form of the
notes and metadata to be associated with the collected data in the data base, e.g. Name,
Tribe, Dialect; Gender; Age; etc. Any less-experienced members of the field team
should also practice data-collection and field-analysis prior to the fieldwork seasons.
Once in the field, members of the team should follow their pre-agreed upon plan to
collect, transcribe, and conduct preliminary analysis of captured data. Again,
unprocessed data is almost always useless away from the field situation. It is also vital
that the team ensure that all collected data are secured and backed up to alternate
media, e.g. DVDs and copied onto each team member's computer in the field (i.e. each
team member's laptop should contain a full copy of the entire team's data). Returning to
the home institution the data should be further backed-up, preferably using institution-
wide resources that enjoy a long-term commitment of the institution's resources and
administration.
In today's world, it is important that the fieldresearcher attempt to make her
research results (at some stage of development, not necessarily including raw data)
web-accessible, assuming that the speech community agrees. The planning process for
this should include at least the following: (i) a list of desiderata for a preliminary
website (e.g. data-retrieval, video-audio coordination on the website); (ii) a plan for the
participation of a web-programmer in the site's design; (iii) a discussion for answers to
questions like the following: (a) how can data be optimally accessible to linguists and
other users via the Internet; (b) how the underlying data-base is best structured and
constructed; (c) how will web-based users interact with the site; (d) what is the efficacy
of different kinds of material to be made available for retrieval from the web data base;
(e) how best to link video, audio, and labelled files to the orthographic and phonetic
transcriptions of the texts collected. etc.
Finally, in designing the final site, the fieldresearcher/team should take a 'story-
board' approach to experiment with design ideas, also inviting other university staff,
colleagues, and students to participate. Presentations and discussions should be used to
discover usability errors or conceptual design flaws before the latter implementation
stages. At each stage of development of ideas on web-preservation expected outcomes
should be clear to the field team.
These planning stages will then be followed by actual site construction. During
this phase the research team, in conjunction with the web-programmer, will: (a)
construct the underlying database; (b) populate the database from analysed field data;
(c) construct the website front-end working from the agreed story-board; (d) program
the agreed methodologies for the extraction and viewing of the underlying data; (e)
conduct usability trials on the prototype system with interested parties (e.g. linguists
experienced in web design and/or field linguistics); (f) consult with a wide-range of
Linguistic Fieldwork Daniel L. Everett
page 59
experienced experts to ensure best-practice and comply with emerging international
standards in web-archiving of field data, e.g. the Open Languages Archive Community
(OLAC); Electronic Metastructure for Electronic Languages Data (EMELD); Oxford
Text Archive (OTA) and others.
It is always useful for field researchers to handle and process the data multiple
times to better control it. But fieldnotes in longhand also provide for creating symbols
not available on computers initially, and can be done when not near one's computer, as
often arises in field situations). All data should be copied to a separate server, e.g. the
server of the linguist's home department. DVD copies are also made and then ideally
would be made available to the appropriate institutions of the country of research, e.g.
for Brazil, the Museu Nacional in Rio de Janeiro. At the home institution, initial field
transcriptions are expanded to include as much of the phonetic nuances as possible, up
to the point that members of the team agree that they have reached the 'point of
diminishing returns' relevant to the project. Sound files are also analyzed in more detail,
again using an advanced acoustic measurement program. Video files should be studied
and discussed, e.g. as to how hand gestures, facial expressions, and other features
(might) correlate with grammatical and cultural features of the texts being told.
Tagging of texts, audio files, and video files will include any special/additional
information agreed upon by the team.
Attention must also be given to then assessing the final product and ensuring
that it continues to serve the agreed-upon purposes of the particular language project,
e.g. (i) transcriptions of written texts downloadable for each text, both separately and
along with their supporting audio (and/or) files. Readers might also have the options of
downloading with the transcriptions other information, e.g. ToBI labellings (i.e. any
combination of files, including downloading only audio or video files); (ii) each
individual sentence of a transcribed text should ideally be downloadable individually,
along with its supporting sound file and various annotations/labellings; (iii) some
thought should be given to a general constraint to serve all interested parties with
Internet access, including many in the third world with slow, dial-up connections.
Therefore, during the pilot-webpage development phase, the fieldresearcher should be
seeking feedback from all users to design a future site that can facilitate use by both
high-end and low-end technology users. These are of course just a few ideas and their
implementation and nature will vary dramatically from project to project.
Webpages shuld use platforms, software, and equipment backed by long-term
institutional commitments.
Linguistic Fieldwork Daniel L. Everett
page 60
CHAPTER 4. SEGMENTAL PHONOLOGY AND PHONETICS
4.1. Introduction
What are you trying to figure out when you begin to analyse the segments of a
language's sound structure? Not everyone who sets out to study segmental phonology
in a field situation has the same objectives. One person might study the segmental
phonology so that they can have a consistent system for representing language data for
research on other aspects of the grammar. Another might study the segmental
phonology in order to help develop a literacy program with a particular community. Or
one might study the segmental phonology because they are a phonologist or otherwise
interested in theoretical issues of sound structure. No matter what one's overall
objectives are, it is vital that they do a good job and that results be backed-up by high-
quality digital sound files so that the analysis can in principle be replicated, without the
need, insofar as possible, for additional field trips. The analysis of the segmental
phonology of the fieldwork language should accomplish the following:

I. Provide an explicit, clear account of the phonetics (contrast, variation, and make-up)
and distribution of all sounds of the language.

II. Offer a clear statement of the phonemes (or distinctive segments) of the language,
with supporting analysis and data.

III. Discuss or point out any typologically unusual or theoretically noteworthy

characteristics

In my experience, phonology is usually the least developed portion of descriptive

grammars. This is due to several reasons. First and most significant, perhaps, is that
most linguists are interested in morphosyntax and semantics and many (wrongly) view
phonology as unconnected to those goals or as inherently less interesting (see the next
chapter on prosody for an effective refutation of that idea). Another reason is that field
phonology requires a well-developed skill, namely, phonetic ability, i.e. ability to both
produce and recognize sounds of languages in field situations, in particular well-
developed auditory discernment. Field phonologists must have good ears. Instruments,
while necessary, are not sufficient. Ears are necessary and nearly sufficient (but not
quite).
But all linguists can do high-quality field phonology. They can develop their ears
and use acoustic analysis as an aid (and much more!). The purpose of this chapter is to
help the field linguist provide his or her colleagues with a useful, convincing, and
testable analysis of the segmental phonology of the language under investigation.

4.2. THE PHONEME

A phoneme is a set of sounds that speakers hear as the same sound for reasons of
grammar. Because it is label for a set of sounds, a phoneme is, properly speaking, never
heard.22

22
In the late 1980s I was asked to visit a newly arrived linguist who was working on a
language in the state of Amazonas, Brazil, to help him get started on analysis of the
language he wanted to work on. So I flew to the nearest airstrip and then, with an
Indian guide, hiked about 8 hours through swamps (walking through water up to my
chest), over logs, up and down the banks of many streams, with 'cut grass', thorns, bugs,
Linguistic Fieldwork Daniel L. Everett
page 61
At one time, many linguists likely thought that generative phonology (Chomsky
and Halle (1968, ---)) had done away with phonemes. But as researchers in Lexical
Phonology (see Mohanon ()) pointed out, the phoneme, or something like it, needs to
be retained in phonological theories for at least two reasons. First, native speakers have
intuitions of phonemes. Second, phonemes are theoretically significant (in some
theories) and practically important in orthography design. In Lexical Phonology
(Mohanan ()), for example, the output of the lexical rules is roughly equivalent to a pre-
generative phonemic representation. Also, discussing the segmental phonology in terms
of phonemic analyses is perhaps a more theory-neutral way of presenting research
(though still problematic, see __).
On the other hand, there is no standard terminology or theory that a field linguist
should feel obliged to conform to except that which produces quality and clarity of
analysis and presentation. Finding generalizations and understanding the system,
whether as rules, constraints, principles, or all three, is what the field linguist is after.
Phonemic analysis is just one way of getting at this goal.
The field linguist should, however, be aware that his or her phonemic analysis has
ethical and political, as well as scientific implications, especially if it turns out to be the
first graphic representation of the language. This is too often overlooked. Entering a
community as a linguist will, in many communities, mean that some see you as an
expert whose work should be taken seriously. So as you conduct your phonological
analysis many are likely to take your proposals more seriously than you might have
imagined. To give an example of a problem and how it might develop, let's say that in
your early efforts you confuse a morphophonemic rule with a phonemic rule. This
could have serious implications for the community. To see how, consider the
hypothetical analysis of English nasals, beginning with the facts in () and ():

(4.1) a. It is pink. [phiNk]

b. I want a pint. [phaint]
c. He is a pimple. [phImp;]

(4.2) a. It is inconceivable. [...Nk...]

b. It is intolerable. [...nt...]
c. It is impossible. [...mp...]
d. It is illegible. [...ll...]
e. It is inordinate. [...n...]
f. It is irrealis. [...rr...]

etc. plaguing me along the way (plus an impatient guide who could not believe how
slow I was or how frequently I fell down). Arriving in the village where the linguist
was beginning work, we began to discuss the language. 'How is the phonology
coming?' I asked. 'Nothing to it', came the answer. 'Why is it so easy?', I asked
skeptically. 'Oh, because you can hear the phonemes' came the answer. This person
was a friend so I did not strangle him and toss him out in the swamp for the wildlife. I
realized that this person who had made it into a good graduate program and had come
this far, was in need of remedial phonology instruction. You cannot hear analyses; you
cannot hear phonemes; you cannot even hear allophones until after you have done an
analysis. I may strangle the next person, though, who tells me this. I am older now and
less patient.
Linguistic Fieldwork Daniel L. Everett
page 62
The linguist will notice from examples like () that nasals and following obstruents
are homorganic, i.e. that they share the same place of articulation. And the same is
apparently true of the examples in ().
Now let us further assume that the linguist has found, say by minimal pairs or
near-minimal pairs (see ___) like those in (), that the nasals in question are separate
phonemes:

(4.3) a. sing [siN] (at least in my Southern California dialect)

b. sim [sIm] (as in 'my phone needs a new SIM card)
c. sin [sIn]

(4.4) a. mother
b. nother (as in, 'Nother?')
c. --- (no word-initial /N/ in English)

Based on (4.3) and (4.4), the linguist is reasonably sure that the nasals are each
separate phonemes. Then, based on these facts, they could, wrongly, conclude,
wrongly, that when immediately preceding an obstruent, the place of articulation
contrast between nasals is neutralized. One might proposal a rule along the lines of ():

(4.5)
+nasal → [βplace]/ ___ -sonorant
αplace βplace

But this would be wrong, problematic, and misleading. It would predict that there
is only one kind of underlying representation and only one phonological process
involved. To see why this is mistaken, compare the examples in () and ():

(4.6) a. unobtrusive
b. unreal (cf. *urreal)
c. unpopular (cf. ?/*umpopular)
d. untouchable
e. unkillable (cf. ?/*uNkillable)

If rule () is correct, why does it fail to apply to the examples in ()? Well, it doesn't
apply to () because it is incomplete. There are two 'neutralization' processes for nasals
in English. The first is a constraint on nasal + obstruent sequences within morphemes
and the other is a constraint on nasal + obstruent sequences across only a certain set of
morpheme boundaries (or alternatively, only affecting certain morphemes, e.g. {in-},
but not {un-}.
It is true that the linguist could fix the initial bad analysis on their next visit (see
___ on morphophonemic analysis). What is the big deal? Just how could this possible
become an ethical problem?
The latter could arise if the linguist presented his or her analysis to the community
too hastily and, if consequently, orthographic decisions were made on the basis of it
(e.g. if some more progressive elements of the community began to write their
language as the linguist represented it to them). If the linguist then says 'Whoops, I
made a mistake. We need to redo the writing system, folks.' This would lead to
Linguistic Fieldwork Daniel L. Everett
page 63
confusing representations or difficulty in teaching native speakers to read their own
language. To attempt to undo this harm would require telling the community that the
linguist was wrong and that they were going to need to retool and relearn how to write
their own language. This would likely seriously undermine community confidence in
the literacy process. It could even contribute to a feeling in the community that their
language is inferior to the national (etc.) language, since, as far as the community
knows, no such inconsistencies are ever found in teaching the larger or more
prestigious language. These are ethical problems. The political issue is that it is the
community, not the linguist, that must eventually agree on and adopt a particular
orthography. The linguist must avoid inadvertently or purposely bypassing the
community in introducing the orthography. In fact, they should have nothing to do with
such issues unless invited to do so by the community. And they should participate only
when phonological analysis has been carefully refereed by other linguists, whether or
not it is published. We want to do it right, as right as we can, and avoid such problems.
That is part of what this chapter is about.
The chapter is organized as follows. First, I present the standard methodology for
'phonemic analysis'. Then I consider potential problems and shortcomings of this
procedure, supplementing it with suggestions based on modern phonological theory.
This is followed by a section considering segmental phonology in the wider
grammatical context. Some suggestions are then made for phonetic fieldwork, though
this section is small and I refer the reader to the best source available anywhere on
phonetic fieldwork, Ladefoged ().

4.3. METHODOLOGY
4.3.1. Caveat
There is no guaranteed way of analysing anything in any language. You could in
principle come up with an excellent analysis without following anyone's suggestions,
due to luck or genius. Or you could fail to come up with a satisfying analysis after
trying all the methodological suggestion of this and other books on field methods. All
things being equal, however, it is best to plan to use a tested methodology rather than to
rely exclusively on being a genius.

4.3.2. Train the ears

I would recommend that the first step the linguist take before trying
phonological analysis in the field is to train their ears for transcribing sounds and
prosodies. One form of training is to listen to tapes of other languages and practice
imitating them and transcribing them, preferably having someone who knows the
language make the corrections. A related form of training is to find speakers of other
languages in one's hometown, university, etc. and practice fieldwork with them. You
can also train your ears if you play a musical instrument by learning new styles of
music 'by ear'.
Another training method I have found very useful is to memorize texts of the
language under study before learning what they mean. Then repeat these texts to native
speakers and see if or how well they understand you.
It is also useful to eavesdrop and then imitate what you hear (assuming that no
one minds and that you never record anyone without permission).
An additional means of hearing and pronunciation training is to compare
spectrograms of your pronunciations of words of the language under study with
spectrograms from native speakers' pronunciation of the same strings of words or
Linguistic Fieldwork Daniel L. Everett
page 64
sounds. The linguist then practices and works to make his or her spectrogram, wave
form, etc. look more and more like the native speaker's.
Last but not least, perhaps the best way to train your ears for phonological
analysis is to learn to speak the language (see chapter ___). This is not only useful ear-
training, it is of inestimable value for all your linguistic objectives and is important
socially and culturally.
However, one thing that is less than useful, except for some aspects of prosody, is
to rely for your practice of the language on analog recordings in the absence of native
speakers. The reason for this is that analog recordings simply are not good enough for
distinguishing many sounds, especially voiceless consonants.

4.3.3. Chart all sounds

It is essential that all sounds transcribed be charted in a standard phonetic chart.
The sounds of the phonetic chart can be referred to as 'surface segments' or, to use a
more traditional term, 'phones'. (Following analysis, the linguist produces another chart.
This on is of 'distinctive segments' (or 'underlying segments', or, depending on your
analysis, as 'phonemes'), which are those from which the linguist believes all related
allophones are derived or which they believe is a useful label for the particular set of
sounds.)
Once the linguist has charted the surface segments they have encountered,
analysis should begin. Do not wait. Analysis of the segmental phonology should begin
on the first day of linguistic research, though it might not be concluded for several
weeks or much longer. As one proposes analyses to oneself, one better equips oneself
to look for more and better data on the phonology. Your analysis feeds and enhances
your fieldwork. So it should begin immediately (and this is true for all areas of
analysis). The first thing I do in studying the phonology of a 'new' language is to
indicate on my chart of surface segments what I consider to be suspicious pairs and
suspicious segments. A 'suspicious pair' of sounds is any two sounds that differ in no
more than one manner or one place of articulation. A 'suspicious sound' is one that can
serve both as a syllable nucleus or a syllable margin (i.e. onset or coda). Once you have
these indicated to yourself you are ready for the next stage of analysis. Before going
into that, however, let's 'walk through' an example of what is suggested in this
paragraph.
So let us say that you are working on Pirahã. You will notice immediately that
there are very few surface segments in the language. They are shown in the charts
below (one for consonants, one for vowels).

Table One
Pirahã Surface Segments
Consonants

p t k ?
kW
pä tS
p÷
b g
b) l&7
m n
Linguistic Fieldwork Daniel L. Everett
page 65
s S h
h2 (with movement of tongue
towards s position)
w y

Vowels
i
I
E o
a

So what is suspicious in the charts above? First, all the voiceless bilabials are
suspect because they differ in no more than one manner of articulation. The voiced
bilabials likewise. The weird lateral, [l&7] (see Everett (1982)), and [n], the two [h]s,
the two sibilants, [s] and [S], and the front vowels should all be tested as groups to see
if they are separate distinctive segments or members of the same phoneme
(allophones). Second, the segments [w] and [y] are both suspect because they could
simply be the phonetic realization of nonlow vowels ([i] and [o]) in onset position
(Pirahã has no codas). How do we check these? Well, according to standard
structuralist methodology, we first look for 'minimal pairs' or Contrast in Identical
Environments (CIE). Herbert Landar (1980 IJAL 46:228) summarizes the structuralist
viewpoint well when he says, “as with vowels and consonants, so with stress, pitch, and
juncture: one minimal pair certifies phonemic integrity.” CIE is defined in (4.7):

(4.7) CONTRAST IN IDENTICAL ENVIRONMENTS: Two sounds contrast in identical

environments if and only if they are found in at least one set of two words that differ in
meaning, where the only sound difference (segments and prosodies) distinguishing the
words of the set are the sounds in question, and where the latter appear in exactly
corresponding positions in each of the words in the set. The set is referred to as a
minimal pair.
An example from English:

(4.8) a. bit [bIt]

b. beet [bit]

We will return to problems with this notion directly. But what if the linguist
cannot find a minimal pair or CIE for a pair of sounds? Then they should look for
Contrast in Analogous Environments (CAE) which can be defined as in ():

(4.9) CONTRAST IN ANALOGOUS ENVIRONMENTS: Two sounds contrast in analogous

environments if and only if they are found in at least one set of two words that differ in
meaning, where the sound differences (segments and prosodies) distinguishing the
words of the set are the sounds in question plus some other sound(s), where the former
appear in nearly the same corresponding positions in each of the words in the set and
any other different sound(s) is(are) unlikely to be responsible for the differences in the
manner or place of articulation of the sounds said to contrast.

(4.10) a. spit [spIt]

Linguistic Fieldwork Daniel L. Everett
page 66
b. Pete [phit]

Example (4.10) is CAE because the words compared differ in meaning and for the
sounds in question (the vowels [i] and [I]), as well as the additional [s] at the beginning
of (a). This [s], however, is unlikely to be responsible for the fact that the vowels in
question differ as they do, i.e. by the feature, [Advanced Tongue Root] ([ATR]).
If the linguist finds neither CIE nor CAE, then they should turn to consider that
the two sounds in question are in Complementary Distribution or, in other words, in a
relationship of allophony. Complementary distribution is the Superman vs. Clark Kent
relationship (The reason Lois Lane suspected Clark Kent of being Superman was that
he was similar to Superman and he was never in the same place at the same time as
Superman):

(4.11) COMPLEMENTARY DISTRIBUTION: Two sounds are in complementary

distribution if one is found in a specific phonological environment and the other one is
never found there.

Let's examine this notion in more detail. Recall from introductory phonology
courses that there are two forms of complementary distribution, namely, ambient and
positional.
Ambient conditioning refers to changes in the sound under study (or,
alternatively, selection of an allophone of the phoneme under study) effected by the
'contamination' from surrounding sounds. Consider the English example in ():

(4.12) can't [khQ)nt]

In this case, the oral vowel, /Q/, has been contaminated by the nasalization of the
following nasal consonant. This is known as regressive assimilation (i.e. it spreads
'regressively', from right to left).
A further, well-known type of ambient conditioning is illustrated in Turkish, i.e.
Vowel Harmony. Cases of Vowel Harmony can be very interesting because they show
that conditioning can take place from non-adjacent items. Consider the following
examples, which are standardly understood to illustrate the features of vocalic backness
and lip rounding:

(4.13) Turkish Vowel Harmony: ip−in ’rope’; kız−ın ’girl’; yüz−ün ’face’; pul−un
’stamp’, etc.

Epenthetic vowels (vowels inserted to satisfy constraints on syllable structure

usually) also undergo Vowel Harmony, receiving rounding and backness specification
from neighboring vowels
according to the same rules, as shown in (3).
(4.14) a. /koyn/ → koyun 'bosom'
b. /metn/ → metin 'text'
c. /sabr/ → sabır 'patience'
Linguistic Fieldwork Daniel L. Everett
page 67
In this case, assimilation is progressive, the vowel to the left triggering harmony
on the vowel to its right (though not necessarily its immediate right, just the first vowel
to the right).
Assimilation may in some cases produce a slightly different kind of result. An
alternative to simple harmonic assimilation, a more complete version, if you will, is
Vowel Coalescence, i.e. where two vowels merge to form a single vowel. Consider the
following examples from Wari' Vowel Coalescence (Everett and Kern (1997, 331)):

(4.15) a. i+i=i
b. xi' + -in = xin [tSin]
1pincl:rf -3sn

(4.16) a. i+e=i
b. xi' + -em = xim [tSim]

(4.17) a. i+o=u
b. 'iri' + -on = urun [Y'RYn]
1pincl:rp/p -3sm clitic cluster

(4.18) a. i+u= u
b. xiri- + -u xuru [tSY'RY]
house- 1s 'my house'

(4.19) a. e + i = ei
b. je + -in jein [/y)e)i)]
2p:rf -3n clitic cluster

(4.20) a. e+o=u
b. hwe + -ocon huhun [hY'hYn]
2p:rp/p -3pm clitic cluster

(4.21) a. e+a=e
b. hwe + -am = hwem [hwIm]23
2p:rp/p -3sf clitic cluster

(4.22) a. a + i = ai
b. wita- + -in -= witain [wi'tãi)]
mat -3n 'its mat'

(4.23) a. a+e=e
b. 'ina + -em = 'inem [/i'nIm]
1s:rp/p -2s

23
The phonetic form differs from the phonological form, /e/, here, but is a natural
realization of /e/ before /m/.
Linguistic Fieldwork Daniel L. Everett
page 68
(4.24) a. a+o=o
b. ta + -on = ton [ton]
1s:rf -3sm

(4.25) a. o+u=u
b. toco- -um tucum [tY'kYm]
eye- -2s 'your eye'

In general, what is happening is that the lower vowel is deleted and the higher
vowel remains, except with [ei] and [ai]. This example is like vowel harmony except
that that 'harmony' results in a blending of the vowels into one.
Positional conditioning is change produced by the location of the sound in
question in the phonological or morphosyntactic constituent structure of the utterance,
rather than by other sounds per se. Consider aspiration in English:

(4.26) a. paper [pheip´]

b. tart [thart]
c. call [khaL]
d. cheap [tShip]

Or consider the change in vowel quality effected by the vowel's location relative
to stress placement in the word:

(4.27) a. photo ['foto]

b. photography [fo'tçgr´fi]
c. photograph [fot´'grQf]

These are the standard methodological teachings that linguists have received over
the years. And this methodology is extremely useful, which is why I have given space
to it here. All field linguists should familiarize themselves with it. In the next section,
however, I consider some problems with this methodology and offer additional
suggestions for phonological analysis.

4.4. The Structuralist Methodology Reconsidered

4.4.1. Reconsidering minimal pairs
In his important monograph of 1968, Postal points out that a major weakness of
'autonomous phonemics' (the source of the methodology just introduced) is its inability
to allow exceptions. This observation has several applications, but the situation I am
referring to here is where linguistic generalizations are either not observed or not
described because the linguist has found apparent minimal pair contrast. Specifically
referring to minimal pairs, Chomsky puts his finger on a crucial problem:

'In general it should be observed that ‘minimal pair’ is not an elementary

notion. It cannot be defined in phonetic terms but only in terms of a completed
phonemic analysis …' (Chomsky 1964:97).

And Postal (p28) paraphrases this to say, '… contrary to almost every
introductory exposition of autonomous phonemic theory or practice, the discovery of
Linguistic Fieldwork Daniel L. Everett
page 69
phonetically minimal pairs does not necessarily [emphasis Postal; D.L.E.] permit an
immediate conclusion about underlying phonological contrast'.
Postal goes on to argue, to me successfully, that analyses should be based on
systematic regularities rather than 'static' exceptions. We do not thereby eliminate
minimal pairs from analyses, but rather we bring the principle of their application into
proper perspective.
Chomsky and Halle develop this notion further, as seen in statements such as:
“Clearly, we must design our linguistic theory in such a way that the existence of
exceptions does not prevent the systematic formulation of those regularities that
remain” (1968:172).
The structuralist methodology from which CIE, for example, emerges has no
elegant way of recognizing exceptions. What is clear is that regularities are the basis for
linguistic analysis. Minimal pairs might, in fact, illustrate such regularities, in which
case by all means use them. Or they might be useful predicting subregularities or
creating doubt. Doubts refine hypotheses. But doubts caused by a single minimal pair,
for example do not require, say, complete redesign of an orthography on the basis of
two words.
Contrary to the Landar quote above, we do not 'certify' results by a
methodological 'proof'. We do not do this in phonology, fieldwork, or science
generally. Let's consider a couple of minimal pair problems to drive this point home.

Empirical and epistemological extensions

The misuse of minimal pairs has various empirical consequences. Consider the
following examples.

Pirahã
Previous analyses of Pirahã tone (see Heinrichs () and Sheldon ()) based
themselves primarily on series of minimal pairs. The following are some examples used
to support the previous analysis proposing three phonemic tone levels.

(4.28) [/a#òí] MLH ‘hand’ H, '= high tone

(4.29) [/a#o#i##] MMM ‘ear’ M, - = mid tone
(4.30) [/àòi#] LLM ‘foreigner’ L, ` = low tone
(4.31) [/àòí] LLH ‘skin’
(4.32) [/a#òì] MLL ‘Brazil-nut shell’
(4.33) [/a#ói#] MHM ‘basket’

In a theoretical framework which allows “minimal pair shortcuts,” it is clear

that (4.28)–(4.29) argue strongly in favor of a three-tone analysis. But this is why
minimal pairs can be dangerous.
Closer examination of this series, however, reveals that between the [o] and [i],
in each of these examples a semivocalic glide [w] occurs. This is rather uninteresting
phonetically since such a glide is pretty much expected in this position for obvious
reasons. However, as any first-year linguistics student knows, such a glide has (at least)
three possible interpretations: (i) this could be perceived as simply a trivial transition
effect, needing nothing more than a footnote; (ii) the glide could be functioning as a
consonant /w/ in the language; (iii) the glide could in fact be a vowel in underlying
representation.
Linguistic Fieldwork Daniel L. Everett
page 70
The determination of which option a–c is correct will have serious implications
for the analysis of Pirahã. In my analysis, option c was selected. This not only allows
for prediction of stress placement (Everett 1988), but notice the change in the tonal
patterns of (4.34)–(4.39) (where [w] carries tone because it is an underlying vowel.
Compare Everett, Forthcoming, for details.)

(4.34) [/a#o$w@í] MLHH ‘hand’

(4.35) [/a#o#w#i #] MMMM ‘ear’
(4.36) [/àòw#i#] LLMM ‘foreigner’
(4.37) [/àòw@í] LLHH ‘skin’
(4.38) [/a#òw$ì] MLLL ‘Brazil-nut shell’
(4.39) [/a#ów@ì] MHHL ‘basket’

These representations then allow us to write rules to derive the mid-tone from
perturbations affecting both high and low tones. These rules are given below. This
phonological decision completely alters our initial perspective of the minimal pairs in
()-(), showing that the concept of what is 'minimal' depends on our analysis.

Pirahã Tone Rules

TONE LOWERING:

(4.40) +high → [-high]/ [+low] ___ [+low]

-low #

TONE RAISING:

(4.41) [+low] → [-low]/ [-low] ___ [-low]

#

Asyllabification:

(4.41) +syl [-syl]/ +syl ___ [+syl]

+rd +rd

And these rules are important for Pirahã since they also explain the rarity of
tone 2, a mere 'allotone' or surface variant, according to this analysis.

Tonal displacement
A further example of empirical problems for a minimal pair analysis may be
seen in languages which manifest what has been termed “tonal displacement”:
Richardson (1971) discusses a phenomenon which he calls displacement,
whereby tonal contrasts are realized several syllables to the right of their original
position. The words [ný-kòlò] 'sheep' and [ný-kòlò] 'heart' in Sukama differ, in that
'heart' etymologically carried a high tone on the last syllable. Both are pronounced
identically in isolation. However, consider the following forms:

(4.42) a. ný-kòlò ný-tàalè ‘big sheep’

b. ný-kòlò ný-tàalé ‘big heart’
Linguistic Fieldwork Daniel L. Everett
page 71

The original tonal contrast is realized on the adjective big (Hyman and Schuh
1974:103).
Hyman and Schuh proceed to give a feasible account of this phenomenon, with
'sheep' and 'heart' contrasting in underlying form. Note that such an analysis would
have been 'messy' or very difficult to state in a structural framework wherein
preanalytical minimal pairs are phonetic units 'certifying phonemic integrity'.

Portmanteau
A final example in support of the thesis expressed in this paper is the
phenomenon known as Portmanteau (or Coalescence, as seen in the Wari' examples in
()-()). A discussion of this is found in E. V. Pike (1974a:24):
A portmanteau phone is one surface sound which is produced by the 'fusion' of
two underlying or distinctive segments. A unique phone, the one not part of a
symmetrical pattern may turn out to be a portmanteau phone. When the units which
make up the portmanteau phone are recognized, their occurrence should help to make
symmetrical one of the nonsymmetrical patterns.
Pike gives the following examples (among others) of Portmanteau:
'Harris (1951:92) describes a flapped nasal which occurs in some environments in
some dialects of American English (as in, for example, painting) as actualizing the
sequence /nt/.
In Quiotepec Chinantec, the sequence /mï/ is actualized as a syllabic bilabial nasal
(Robbins 1961:245).
In Ayutla Mixtec, the sequence /ae/ is actualized as [æ] when following an
alveopalatal consonant (Pankratz and Pike, E. V. 1967:289)….'

Discussion
One might possibly respond to this data by saying, “Well, the initial minimal
pairs were incorrect. As a matter of fact, once the proper forms were defined the
contrast became obvious.” But this reasoning is fallacious. The 'correct form' of these
minimal pairs was determined phonologically, not phonetically. That is, no phonetician
could have told us that the semivocalic glide from [o] to [i] in the Pirahã data was an
underlying /o/ (which carries tone). Nor is the phonetic data sufficient to determine the
presence of a displaced tone in the Sukama examples. Portmanteau is even more
effective in revealing the shortcomings of CIE/minimal pairs because sounds are 'lost'
on the surface.
Another way of thinking about this is in terms of the more general notions of
dependent vs. independent variable. These are defined in () and (), respectively:

(4.43) DEPENDENT VARIABLE: The item in an experiment or study whose changes

are effected by one or more independent variables.

(4.44) INDEPENDENT VARIABLE: An item manipulated in an experiment or study

that effects change in the dependent variable.

To give an example, let us assume that we are studying the nasalization of

vowels. We might ask whether the nasalization on the vowel, the dependent variable, is
caused by the nasalization of the consonant, the independent variable according to this
hypothesis.
Linguistic Fieldwork Daniel L. Everett
page 72

(4,45) a. can't [khQ)nt]

b. cat [khQt]

In (), the [n] is the independent variable because we are manipulating it to see if it
causes the nasalation of the vowel [Q], which is our dependent variable. From this we
can draw the modest conclusion that it is ambitious, to put it mildly, for the fieldworker
to believe they have controlled and distinguished are variables in any pair of words at
the beginning of field research. Rather, the understanding and recognition of such
variables comes, if ever, only after a significant amount of rigorous analysis. Therefore,
CIE, CAE, and CD, like other methodological notions, do not provide us with
analytical algorithms but with heuristic procedures, rules of thumb.
To summarize what we learn from the cases above, the lesson is simply that if
the fieldworker had stopped with the superficial phonetic form he could, according to
structuralist methodology, have considered that:

a. Pirahã has three tones and (by further data) unpredictable stress placement.
b. Languages like Sukama have arbitrary allomorphs in grammatical sequences.
c. Languages with portmanteau phenomena have arbitrary patterning in their
phonemic inventories.

This type of data further illustrates the difficulties and dangers of minimal pair
analysis for field linguists. Carrying the conclusion a step further, we might even say
that a good analysis should determine minimal pairs rather than vice-versa. We need to
remind ourselves that methodological suggestions are just that, suggestions. Ultimately,
they are the 'icing' on a 'cake' of theoretically and typologically-informed analysis.24

24
As an epistemological aside, it might be noted that minimal pairs represent the
effects of empiricist philosophy in linguistics. That is, they are the vestiges of the naive
notion that “proofs” exist in science, in general. As Chomsky has frequently observed,
data by itself is not sufficient for criticism of a given theory. Rather, one must say
something about the data, which, by its very nature, is a theoretical activity. To criticize
analysis a, for example, it is not enough to merely present contrary data. It is also
necessary to (1) show how an analysis b would treat the data more effectively, and (2)
how analysis a cannot be extended to handle this “extra information.” Pure inductivism
is a dead-end road.
In astronomy, a researcher might criticize a colleague’s theory by noting that
light rays and planetary motion in a particular section of the galaxy do not conform to
this colleague’s theory. Then the colleague may simply respond by saying, “Well,
there’s this thing called a ‘black hole’ up there which, although invisible, exercises an
effect.”
So, let’s get some money from NASA and send up a rocket to check out the story. No
black hole! Now we’ve got him! But, when presented with this new evidence, the
shameless fellow replies, “You didn’t find evidence of a black hole because your
instruments were fouled up by magnetic clouds in the area (this example is largely from
Marcelo Dascal, verbal communication).
This type of thing can go on and on unless colleague b gets fed up and says,
“Listen—I have had it with your old fairytales. I have developed a theory which
Linguistic Fieldwork Daniel L. Everett
page 73

4.4.2. Effect of change

Fieldworkers can also be confused by historical change that has left
synchronically strange effects in the grammar. As an example, consider once again the
chart in () of Pirahã surface segments. Because the pairs [g] and [l&7] and [g] and [n]
differ by multiple points and modes of articulation, they are not suspicious pairs so the
fieldworker is unlikely to suspect them of being potential allophones/surface variants.
And yet, they are surface variants of a single distinctive sound, as per the rules in ():

(4.46) a. -continuant → +nasal /#___

+velar +coronal
+voiced

b. -continuant → +continuant /+syll __

-syll +lateral
+velar +high
+round
+voiced

c. +consonantal → [+posterior]/___
-continuant
+voiced
-labial

This strange rule is, according to Everett (1979), the result of a historical process
in which a diachronically prior sound [d] shifted to [g], while the remaining two sounds
of the original phoneme remained as [+coronal]. This may be due to pressure to
'disperse' sounds more effectively to aid speaker perception (see X (), Y () for further
development of the concept of 'dispersion'). But it is not expected from a purely
synchronic view and, therefore, would pass beneath the radar of just about any
methodology.
Another interesting case comes from the Ge language, Suyá (Kisedje). Consider
the process of lenition in this language (Everett () and Foresti ()):

1) /t/ → [R] / _____ phrase boundary

2) /p/ → [w] /_____ phrase boundary
3) /k/ → [g] / _____ phrase boundary

[tEp] → ['tEwE] 'fish'

[khet] → [kheRe] 'no' (negation)
[s√k] → [s√g] ~ [s√k] ~ [s√kh] ~ [s√g√] 'bird'

explains all of these phenomena, simply and satisfactorily without black holes,
magnetic clouds, and so forth.”
So it is with minimal pairs. They are only acceptable as evidence within a
theory. The lack of a theory rules all of the data in the world irrelevant.
Linguistic Fieldwork Daniel L. Everett
page 74
Notice that although the coronal, /t/, and bilabial, /p/, stops lenite to continuants,
the velar stop, /k/, is alternatively realized as [g], [k], [kh], and [g√]. As Everett () points
out, there is no continuant realization of /k/ in the phrase-final position, contrary to
realizations for /t/ and /p/ in this position, because there is no distinctive segment that is
velar and continuant. According to Everett (), this process is structure-preserving, that
is, it can only subsititute one distinctive segment or phoneme for another. Since /R/ and
/w/ are both distinctive segments in the language, they may be substituted for /t/ and
/p/, respectively, in phrase-final position. But since there is no continuant velar
distinctive segment, /k/ behaves differently, as already seen. This is interesting because
it shows that some constraints (Structure Preservation in this case, see Kiparsky ()) may
prevent rules from being maximally general, contrary to the standard fieldwork
methodology, and it also shows a possible case of a change in progress. In other words,
the behavior of /k/ could be seen as asystematic by new generations of Kisedje/Suyá
and this could lead to a new phoneme (e.g. [F], which is continuant and velar) or to a
removal of the Structure Preservation constraint). In any case, it is a case that is not
completely compatible with the simple methodology in ---. So, once again, although
that methodology is very helpful, one must be very careful to analyze and argue
carefully for conclusions. And careful argumentation and analysis require knowledge of
the literature, typological, theoretical, and descriptive.
For another caveat on the use of the standard methodology, see section ___ of
chapter __ on phonology and culture.

4.5. Orthography design

Phonological analysis is a necessary condition for orthography design, but it is not
a sufficient condition. As is stated in ___ the choices involved in designing the visual
representation of a language involve community values and discussion. And various
other issues, such as non-trivial, but conflicting cultural values.
For example, do the people want to read in the first place? Do they want to read
their language or the national language or both (each of these questions has
implications for the literacy programme). Do they want their language to look like the
national language or would they prefer that it look different, unique to them? Do they
want the linguist's advice or not? Should the linguist offer advice even if it is not asked
for (the linguist will have to answer this based on their understanding of the culture in
question).

In 1980, Keren and I dedicated every night, six nights a week, to literacy and
math classes among the Pirahãs, at the Pirahãs' request. Near the end of this time, we
finally succeeded in getting the Pirahãs to read a word. We wrote the word bigí on the
black board we were using. Everyone read the word out loud. Then they all started to
laugh (about 30 people were present). I asked 'What's so funny?' They answered 'That
sounds like our word for 'ground'. 'It IS your word for 'ground'', I replied. 'Oh, no' they
said, 'We don't write our language. Is that what you are trying to teach us? Oh, we don't
want that.' And they never returned to literacy classes.

Many agraphic societies will recognize the value of reading and writing their
language, though certainly not all will. If you have an opportunity to contribute to
literacy or other educational goals of the community, especially as these golas implicate
your linguistic skills, then you, qua linguist, are making a valued contribution to
community life. This is both personally rewarding and community-empowering.
Linguistic Fieldwork Daniel L. Everett
page 75
Therefore, the fieldworker should feel privileged to contribute in this way. But at the
same time, the fieldworker must take this privilege very seriously, recognizing the
responsibility that it entails.

4.6. Back up your analysis with sound files

Until very recently it was impractical for the average reader of a grammar or
phonological analysis to have the option of listening to the sounds upon which an
analysis was based. Audio recordings would have been made on analog tapes stored
exclusively at the linguist's home institution (by and large). However, digital files,
servers, and the internet make it both possible and desirable for all phonological
analysis to be backed up empirically such that any reader can replicate the
measurements and analysis. This is a significant improvement over 'impressionistic',
take-the-linguist's-word-for-it analysis. Sound files should be stored in .wav format (see
the E-MELD 'best practice' page at: http://emeld.org/school/). For any phonological
data archive or published study, each datum or analytical point should be accompanied
by an address where the sound file which supports it can be downloaded.

4.7. Phonology in the wider context of the grammar

The procedures for segmental phonological analysis sketched above are
incomplete because they treat phonology as an autonomous module unconnected with
the rest of the grammar. However, as the examples from English in ()-() illustrate,
morphosyntax can be causally implicated in phonological analysis. Although culture is
also occasionally implicated in phonological analysis, this is treated in chapter ___.
Here I want to consider ways in which morphology is crucial to understanding
phonology. We begin by considering what Everett () refers to as encliticization, also
discussed in Sheldon ().

Pirahã Encliticization

(4.47) tíì /ísìtòí /óògàbàgàí → tíì /ísìto&ògàbàgàí

1psg egg want
'I want an egg.'

(4.48) kàhàí /áàgàhá → kàhàa&àgàhá

arrow is
'It is an arrow.'

In these examples, we see that the final vowel of the first word, the verbal
object, deletes, as does the initial glottal stop of the following verb. However, the tone
does not delete, as shown by the rising tone on the vowel to the right of the deletion site
in both examples above. This means that the only way to get at the correct analysis of
Pirahã tones, i.e. that they are underlyingly level and not contour, is by taking into
consideration the effect of 'cliticization' on the surface forms of the tones. If the forms
to the right of the arrows were analyzed without regard to the forms to the left of the
arrows, the tonal analysis would have severe problems.
Moreover, as Pike and Fries () observe in their pioneering study, loan words can
affect the phonological system of a language. So, for example, consider the phoneme
/Z/ in English. This sound only occurs in loan words, e.g. azure, garage (some
dialects), fromage, etc.
Linguistic Fieldwork Daniel L. Everett
page 76

4.8. PHONETICS FIELDWORK

This section is a brief and humble introduction to how to begin to answer
questions of phonetic theory. Maddieson () and Ladefoged () provide excellent
introductions to phonetic field research, the latter a book-length survey and this section
is meant as little more than an annotated reference to those works.
There are two broad domains of phonetic research: the nature of contrast, i.e. how
the sounds of a language are perceived as different by native speakers, and variation,
i.e. how sounds change diachronically and synchronically. Phonetics fieldwork in some
senses underwrites or warrants all other fieldwork. If the phonetics (articulatory or
acoustic) is bad, the entire documentary-descriptive effort is bad. But phonetics is not,
as some people might think, the first type of fieldwork undertaken by the fieldworker.
Before you begin serious phonetic analysis, you should have other analyses well under
way.
Moreover, with phonetic fieldwork, as with all types of research, you must first
be clear on your methodology, purpose, and technology. As an example, let's say that
you want to study the contrast between two closely related sounds. You will have
already established that these sounds contrast phonologically. You should next look to
your dictionary or word list in order to find systematic sets of examples to test. If you
do not have a sufficiently complete dictionary to arrange the necessary examples,
William Poser has written a programme that will generate lists of hypothetical words,
based on an input of the language's syllable structure and segmental phonology.
What kind of word list do you need? What are the guiding principles in devising
a word list for phonetics field work? The basic idea is quite simple: the linguist wants
to construct examples for testing in which the dependent and independent variables are
properly controlled. That is, you the linguist want to be sure that you are studying what
you think you are studying. Consider how you might go about studying the difference
between a voiced stop and a voiceless stop. First, assemble words with the sounds in
minimal or near-minmal pairs (see ___ above). For example, suppose that you want to
study the contrast between /g/ and /k/. Word lists like those in () and () would give you
a reasonable start (where I assume that all words differ in meaning).

(4.49) gabi, bagi, gut, tug, grat

(4.50) karg, garg, gark, kark, kãg, gãk, etc.

(4.51) kabi, baki, kut, tuk, etc.

That is, each segment to be tested should be recorded preceding and following all
vowels and all consonants and in word-medial, initial, and final positions. Once this is
done, if the recordings are of adequate quality (see ___ below) and quantity, then you
have the basis for comparing spectrograms of the two segments. Likely, the most
interesting distinctive phonetic process you will find in this case is Voice Onset
Timing. However, perhaps you have reason to believe that the articulation of the
sounds is also different in some way. You may want to make palatograms of each of
them. (How to do this is described in ___ below.) You may want to film speakers or
use ultrasound, etc., depending on where you are at and what kind of research budget
you have. I turn now to consider technical aspects of phonetic and phonological studies.
Linguistic Fieldwork Daniel L. Everett
page 77
Recording
First, as mentioned in ___ above, the few the moving parts in your recording
equipment, especially your recorder, the less chance there is for motor hum and other
intrusive sounds to contaminate your recordings. Recording on solid-state digital
recorders or directly onto your computer is therefore recommended over even quality
digital recorders that use tapes, compact disks (CDs), or digital video disks (DVDs).
Second, your microphone should be a dynamic, uni-directional mic, attached to a
headset. (A capacitor microphone would be better if you were particularly interested in
studying low-frequency sounds.) The purpose of the headset is to ensure a constant
distance and angle between the microphone and the native speaker's mouth. Hand-held
or table-top microphones are notorious for registering false differences in loudness, for
example, as the speaker turns their head while recording or they hold the microphone at
different distances from their mouth while speaking.
Care must be taken too to ensure that the recording volume on your recording
device is set to maintain a clear signal to noise ratio (see __ for a discussion of this
ratio). That is, you do not want the recording volume so high that you get distortion at
the upper ranges of speaker volume. But you also want the volume high enough so that
the speaker's voice easily drowns out all ambient noise. The goal, visually, is to
produce high-quality spectrograms, such that formants, vocal chord vibrations,
nasalization traces, etc. are clearly visible. If the recording is well-controlled and in a
silent environment, words will be separated on the spectrogram by white, clear space.
This is almost never possible to achieve in the field environments, though
approximating this ideal should be the goal.

When recording among the Banawa in 2004, I was delighted to collect the best natural
conversation between two speakers that I had ever collected in all of my field research
in the Amazon. Two language teachers, Sabatao Banawa and Bido Banawa, sat in front
of me in a small study just out of the Banawa village (kindly provided for our use by
SIL member Ernest Buller). Each wore a headset with a high-quality microphone. I
recorded them onto a professional digital recorder (). I even recorded the entire session
on digital video, which provided additional high-quality audio back-up as well, since
most camcorders have superb microphones. Two things happened when we returned
from the field, however. First, someone broke into our project office in Manchester and
stole a valuable computer, and all of our video recordings from the entire first year. So
I lost the video record. Next, as I looked at the audio recording of the conversation, I
noticed a high-frequency band of noise going through the entire conversation. It turned
out to be a cicada that I had not heard while recording, because I was so focussed on
collecting the actual conversation. A bit disappointing. Nevertheless, the cicada and
other ambient noise were relatively easy to work around because the signal-noise ratio
was good.

An ideal addendum to your laboratory sessions would be to video-record all of

them so as to capture at least all the words and expressions (facial, body posture, hand
gestures) of the teacher. It would also be useful to record yourself at the same time.
Your own actions, interactions, and reactions can be useful in determining the kind of
feedback you might inadvertently be giving the language teacher (e.g. frowns, smiles,
furrowed eyebrows, crossed arms, face in hands, etc.). All recordings should include a
preface of the linguist speaking, giving at least the following information: date and time
of recording, speaker name, location, objective of this recording, speaker gender and
Linguistic Fieldwork Daniel L. Everett
page 78
age, speaker dialect, and speaker's level of bilingualism. Ideally, this tape should be a
stand-alone document that any other linguist could use, e.g. students looking for thesis
projects. Remember: backup all data files immediately and keep copies in a different
building from the originals.

Palatography and linguography

Ladefoged () gives an excellent set of suggestions for field palatography. He
suggests painting tongues or palates with either purified charcoal or the scrapings from
burnt toast mixed with olive oil. These can be painted on the subject's palate (for
linguography 'writing on the tongue') or tongue (for palatography 'writing on the palate)
with a never before used or sterilized paintbrush, approximately 0.5 – 1.0 inches in
width.

In my fieldmethods class at the University of Manchester, a couple of undergraduate

students became enthusiastic about palatography and their fieldwork papers showed
excellent palatograms. One of the students decided, however, that olive oil + burnt toast
scrapings tasted 'yucky' so she used softened chocolate and butter. She got reasonable
palatograms doing this. However, the disadvantage to tasty 'paint' is that it produces
more salivation, smearing the palatogram and thus generally producing less useful
results. So bad-tasting palatographic 'paint' may be better for science.

Linguography marks the palate and then photographs the subsequent markings
on the tongue after the word or syllable has been pronounced. Palatography marks the
tongue and then photographs the palate.
To photograph subjects' tongues and palates, use a high-resolution digital
camera. Have the subject stick their painted tongue out and use a mirror to photograph
the palate.
Airflow and air pressure (these are not the same thing) studies can also be useful
in the field for more carefully examining supraglottalic pressure and nasal airflow. The
data collected from such studies can be essential in distinguishing stops from
continuants, nasal sounds from oral sounds, and any other contrast involving the source
and direction of airflow (e.g. pharyngeal vs. pulmonary air, etc.) or air pressure. For
example, variations in air pressure can indicate increased or decreased energy in
production of sounds and can be used, for example, to distinguish stressed vs.
unstressed syllables. Greater airflow with lesser air pressure can, to give another
example, indicate difference in phonation types (e.g. normal vs. 'breathy' voice).
Ladefoged (2004, 55ff) has an excellent discussion of methodology in recording and
measuring airflow and air pressure. There is a cost in undertaking such studies and that
is the purchase and transport to the field of the equipment necessary to do the
recordings. Such equipment these days is greatly reduced in bulk from what earlier
field phoneticians had to contend with, but it still entails greater expense and
inconvenience. Nevertheless, the more urgent the documentation of a particular
language, the greater the argument for taking a range of phonetic equipment from the
outset of field research on that language. Otherwise, my suggestion is that high-quality
digital recordings and phonological analysis should take place for the first couple of
sessions, to be followed later by phonetic field analysis. For most of what the average
field linguist wants to do they can learn to do it themselves, and they should. On the
other hand, if there is money in the budget and you can find a willing phonetician, it
Linguistic Fieldwork Daniel L. Everett
page 79
could be useful to have an expert along to help you with your recordings and
measurements for phonetic analysis.

In my own case, as I have mentioned before, I had the extreme privilege of having
perhaps the world's greatest phonetician, Peter Ladefoged, accompany me to the
Pirahãs, the Banawas, the Wari's, and the Oro Wins. But Peter wanted to come for two
reasons. The first was to document the phonetics of endangered Amazonian languages
for which there were at least initial phonological analyses. But the second was to check
out some of the things I had been saying in the literature about Pirahã and Banawa
stress. As I picked Peter up at the Porto Velho airport in Rondonia, Brazil, I felt like I
was about to be audited by the US's Internal Revenue Service. I felt sure that my
analyses were correct, yet at the same time, I knew that Peter Ladefoged would be
concerned about the facts and not about what disproving everything I had said would
do to my reputation! On the one hand this was a good feeling, as a scientist. But as a
person I was tense. When Peter got off the plane and I was helping him to the car with
his bags (and portable phonetics lab), he said that two of his colleagues had '... asked
me to check out very carefully what you say about Pirahã stress because they are
skeptical of your analysis.' And yet at the same time that I was concerned about
whether my earlier statements would really stand up (paranoia), I was pleased that
someone was taking this research seriously enough to check it out. Ultimately, the
stress and tone that Peter identified in all the relevant words was what Everett &
Everett (1984) had predicted. Thus Pirahã stress became a more widely accepted and
important part of crosslinguistic stress studies. But the point is that replicability is a
crucial part of science and having your results checked independently is a service to
your own research and to the scientific community as a whole, even if your 'auditor' is
perhaps not quite as distinguished as Peter Ladefoged.
The analysis of the entire grammar rests on the foundation of phonetics. If the
phonetic quality of your data is poor, then everything from the morphology to the
semantics is suspect.
Linguistic Fieldwork Daniel L. Everett
page 80
Chapter 5. Prosodic fieldwork

by Nikolaus P. Himmelmann and D. Robert Ladd

This chapter is a little different from the preceding and following chapters, because it
devotes a lot of space to basic issues that have nothing directly to do with fieldwork.
The reason for this is simple. There’s a lot of basic agreement about how segmental
phonology and morphosyntax work, and most linguists know what kinds of things they
will be looking for when they begin work on a new language. So in chapter 5 and 7 the
focus is not on what to look for, but how to look. But with prosodic features – the
kinds of things that often don’t show up in a segmental transcription – many linguists
feel that they are on shaky ground. They are insecure about hearing prosodic
distinctions and unclear about the way these distinctions might be used in different
languages. This means that an important goal of this chapter is to make you aware of
what you might be looking for – not just how to look.

There are two fundamental ways that prosodic features differ from more familiar
segmental features. One is that they are relevant at different levels of structure: there
are both word-level or lexical prosodic features and sentence-level or “post-lexical”
ones. Probably the best known typological difference based on this distinction is the
one between “tone languages” like Chinese, where pitch serves to distinguish otherwise
identical lexical items, and non-tonal languages like English, where pitch only serves to
signal sentence-level differences of “intonation”. However, the use of prosodic
features at different levels applies more widely as well: in English we can use stress at
the lexical level to distinguish one word from another (e.g. PERmit [noun] and perMIT
[verb]), but also at the post-lexical level to distinguish one sentence meaning from
another (e.g. I only put salt in the STEW and I only put SALT in the stew.)

The other important property that sets prosodic features apart from familiar segmental
features is that their sentence-level functions – like intonation and sentence-stress – are
often broadly similar even in completely unrelated languages. For example, it is very
common cross-linguistically to signal questions by the use of sustained high or rising
pitch at the end of an utterance, even in languages that also have lexical tone. Because
it works at different levels and because it has both universal and language-specific
aspects, prosody is likely to seem mysterious and difficult. Speakers of a language that
uses a given feature in one way are likely to find using it in a different way strange and
exotic and (more practically) hard to hear: this is a common reaction of speakers of
non-tonal languages when they encounter a tone language. Furthermore, sentence-level
distinctions are probably inherently more difficult to think about than lexical
distinctions: the difference between a pin and a bin is instantly obvious and easy to
demonstrate, whereas the difference between the two versions of the sentence about the
salt and the stew in the preceding paragraph takes careful explaining.

Nevertheless, prosody is an essential ingredient of every spoken language, and a

description of prosody is an essential ingredient of every complete language
description. In the following sections we will sketch some of the key phonetic,
functional, and typological aspects of prosodic features, then go on to outline various
techniques for achieving a satisfactory analysis of prosodic features in the field.
Linguistic Fieldwork Daniel L. Everett
page 81

5.1 The phonetic fundamentals

We begin by briefly introducing four phonetic parameters which are relevant to

prosody: pitch, duration, voice quality, and stress.

5.1.1 Pitch

Pitch is the property that distinguishes one musical note from another. In speech, pitch
corresponds roughly to the fundamental frequency (F0) of the acoustic signal, which in
turn corresponds roughly to the rate of vibration of the vocal cords. It is physically
impossible to have voice without pitch – if the vocal cords are vibrating, they are
necessarily vibrating at some frequency. In English and many other European
languages we talk about pitch being “higher” or “lower” as the frequency of vibration
gets faster or slower, but other sensory metaphors are used in other languages and
cultures (“brighter/darker”, “sharper/duller”, etc.). Perhaps because pitch is a necessary
property of voice, all languages – so far as we know – exploit pitch for communicative
purposes.

The most striking thing about pitch is that it varies conspicuously from one speaker to
another – men generally have lower voices than women. This means that the phonetic
definition of pitch for linguistic purposes cannot be based on any absolute level of
fundamental frequency but must be considered relative to the speaker’s voice range.
Normalization for speaker differences must also deal with the fact that speakers can
“raise their voice” without affecting the linguistic identity of pitch features. The details
of how this normalization should be done are not fully clear but the basic principle is
not in doubt. Moreover, this seldom causes serious practical difficulties in fieldwork,
because we can usually hear whether a given pitch is relatively high or low in the
speaker’s voice.

However, even if we find it relatively easy to abstract away from differences of overall
pitch level, there are still major difficulties in the phonetic description of pitch. This is
reflected in the lack of any agreement on an IPA system for transcription of pitch
distinctions. One of the key issues for transcription is the relevance – or lack of
relevance – of the syllable. In many tone languages, terms like “rise” and “fall” must
be defined relative to the syllable: a sequence of a high-tone syllable and a low-tone
syllable can be lexically completely different from a sequence of a falling-tone syllable
and a low-tone syllable, even though both sequences involve an overall “fall” in pitch
over the two syllables. In such a tone language, the overall “fall” is not relevant for
phonetic description. In a language like English, on the other hand, a phonetic fall on a
monosyllabic utterance (e.g. John) and a phonetic high-to-low sequence on a disyllabic
one (e.g. Johnny) may be completely equivalent in the intonational system, which
suggests that the “fall” must be regarded as a phonetic event regardless of the number
of syllables it spans. This idea is strengthened by recent work showing that in
languages like English and German functionally equivalent pitch movements can be
“aligned” in different ways relative to syllables in different languages and language
varieties.
Linguistic Fieldwork Daniel L. Everett
page 82
In studying an unfamiliar language, in short, the fieldworker needs to be alert to the fact
that descriptive assumptions can be hidden even in an apparently neutral label like
“pitch fall”. For fieldwork, the most important thing to know about pitch is that a
useful phonetic description of pitch depends on the way pitch is used in the language.
More practically, the fieldworker whose native language works like English must be
prepared to detect the syllable-by-syllable phonetic chunking of the pitch contour that is
likely to be found in a language with lexical tone.

5.1.2. Segment duration

To the extent that we can divide an utterance into phonetic segments with clearly
defined boundaries, we can measure the duration of the segments. In many languages
duration is systematically manipulated for prosodic effect (e.g. distinctions between
long and short vowels), but in all languages segment duration is affected by a whole
range of other factors as well. These include some nearly universal allophonic effects
(e.g. vowels tend to be longer before voiced consonants than before voiceless
consonants; low vowels tend to be longer than high vowels; fricatives tend to be longer
than stops) and effects of speaking rate (faster rate means shorter segments, but vowels
are generally more compressible or expandable than consonants). Segment duration is
also affected by other prosodic factors: specifically, stressed vowels tend to be longer
than unstressed vowels; segments in phrase-final syllables tend to be longer than in
other syllables; and word-initial and phrase-initial consonants tend to be longer than
consonants in other positions.

For fieldwork, these differences mean that any suspected duration distinctions must
always be checked in similar sentence contexts. In particular, if you ask someone to
repeat two items that appear to be a duration-based minimal pair (like Stadt ‘city’ and
Staat ‘state’ in German), it is important to hear the two members of the pair in both
orders. That way you will not be misled by any lengthening (or occasionally,
shortening) of whichever item is pronounced second.

5.1.3. Voice quality

The phonetic description of voice quality is less well advanced than that of other
prosodic features. Many differences of voice quality – described by such
impressionistic terms as “harsh”, “breathy”, “creaky” and so on – are based on
different configurations of the glottis. As such they are difficult to observe directly,
either in ourselves or in others, except by the use of special equipment. The standard
work on the impressionistic description (and transcription) of voice quality is Laver
1980, which remains a useful reference for fieldwork. Much recent research has
focused on understanding the acoustic correlates of voice quality differences and/or the
glottal configurations that give rise to them. This work is not likely to be of much
direct relevance to descriptive fieldwork, but good fieldwork can provide the basis for
directing instrumental phonetic studies into fruitful areas of research.

5.1.4. Stress
Linguistic Fieldwork Daniel L. Everett
page 83
Roughly speaking, stress is the property that makes one syllable in a word more
prominent than its neighbours – for example, signaling the difference between the noun
PERmit and the verb perMIT. Perhaps surprisingly, it is extremely difficult to provide
a phonetic definition for this “greater prominence”. Impressionistically (for native
speakers of most European languages), the phonetic basis of stress is “loudness” – the
stressed syllable seems louder than neighbouring unstressed syllables – but perceived
loudness is psychophysically very complicated, not just in speech but in all auditory
stimuli. The most important correlate of loudness is intensity (sound energy), but
duration and fundamental frequency have also been shown to play a role – for the same
peak intensity, a longer or higher-pitched sound will sound louder than a shorter or
lower-pitched one.

A possibly more useful phonetic definition of stress is “force of articulation”, which

shows up less in effects on the overall energy in a segment or syllable and more in the
distribution of energy in the spectrum of the sound. Specifically, it has recently been
suggested that stressed vowels in Dutch have more energy at higher frequencies than
unstressed vowels (they have “shallower spectral tilt” [Sluijter and van Heuven 1996]).
There may also be effects of “force of articulation” on the relative duration of
consonant and vowel portions of a syllable, although the details are not at all clear.
Additionally, stressed syllables usually contain full (peripheral) vowels while
unaccented syllables may contain reduced (centralized) vowels such as schwa;
alternatively, a language may have only or mainly peripheral vowels, but stressed
syllables may allow for larger vowel inventories than unstressed syllables. For
example, Catalan distinguishes seven vowels /i e a o u/ in stressed syllables but
only three /i u/ in unstressed ones.

Whatever the phonetic basis of stress turns out to be, and however much it turns out to
differ from one language to another, it is clear that part of the problem defining the
phonetic basis of “stress” is the existence of conceptual and theoretical problems with
the classification and description of accentual systems generally. We return to this
issue in the next section, and in section 6.3.3.

5.2. Typical functions of prosodic features

5.2.1. Lexical and morphological functions

The lexical functions of prosody are, on the whole, like the function of most segmental
phonological distinctions: to distinguish between one lexical item and another. Just as
English pin and bin differ minimally phonologically but are two unrelated lexical items,
so pairs like Chinese niàn ‘study’ and nián ‘year’ or Dutch man ‘man’ and maan
‘moon’ or Greek [ ] ‘space’ and [  ] ‘dance’ involve unrelated lexical
items that are minimally different phonologically. Similarly, just as segmental
distinctions can be used to signal different morphological categories (for example,
English foot/feet for singular/plural or drink/drank for present/past), so prosodic
features can be used in the same way, as in the difference between Efik [  ] “I
buy” and [  ] “I would buy” or Dinka [dèk] ‘drink (infinitive)’ and [dè:k] ‘s/he
drinks’ or Italian [ ] ‘I speak’ and [  ] ‘s/he spoke’.
Linguistic Fieldwork Daniel L. Everett
page 84
The examples just given illustrate the three most commonly encountered types of
lexical prosodic distinctions: tone (as in the Chinese and Efik examples), quantity (as
in the Dutch and Dinka examples) and accent (as in the Greek and Italian examples). It
is common to treat the three of these together as “suprasegmental” features, and to
identify them with the phonetic parameters of pitch, duration and stress. A classic
statement of this view, still useful for the data it contains, is Lehiste’s book
Suprasegmentals (1970). However, this view is misleading in two distinct ways. First,
the linguistic categories of tone, quantity and accent are often cued in multiple phonetic
ways. Tone is primarily a matter of pitch, but may also involve accompanying
differences of segment duration and voice quality: for example, in Standard (Mandarin)
Chinese syllables with “tone 3” are not only low in pitch but tend to be longer in
duration and to have creaky or glottalised voice as well. Quantity distinctions are based
on segment duration, but often involve differences of vowel quality or (in the case of
consonants) manner of articulation as well: for example, German long vowels tend to
be higher and less central than their short counterparts, in addition to being longer in
duration. As for accent, there are so many different phonetic manifestations of things
that have been called “stress” or “accent” that there is very little agreement on what
these terms refer to. In short, it is at best a gross oversimplification to think of tone,
quantity and accent as the linguistic functions of the phonetic features pitch, duration
and stress.

The second reason for not treating tone, quantity and accent together is that they are
functionally quite different. Where they exist, distinctions of tone and quantity are often
functionally similar to segmental distinctions. Tone – especially in East Asia, West
Africa, and parts of the Americas – generally has a high functional load, and it is not at
all uncommon to find extensive minimal sets distinguished only by tone, for example
Yoruba igba ‘two hundred’, igbá ‘calabash’, ìgbá ‘[type of tree]’, ìgbà ‘time’.
Quantity systems are similar: in many languages with distinctive vowel or consonant
quantity, all or almost all the vowels or consonants can appear both long and short in
pairs of unrelated words, for example German bitten /bitn/ ‘request’ vs. bieten /bi:tn/
‘offer’; beten /be:tn/ ‘pray’ vs. Betten /betn/ ‘beds’, etc.

By contrast, accentual differences are often rather marginal in the lexicon of a language
as a whole, yielding few minimal pairs and/or involving some sort of morphological
relatedness. For example, in English the lexical accent in a word is certainly a
distinctive part of its phonological make-up, and a misplaced stress (e.g. in foreign
pronunciation) can make word identification very difficult. Yet there are very few
minimal pairs in English based on lexical accent, except for derivationally related
noun-verb pairs like OBject-obJECT and PERmit-perMIT. This difference is due to the
fact that accent involves a syntagmatic relation (the relative prominence of two
syllables) whereas tone and quantity, like most segmental features, are a matter of
paradigmatic contrasts between members of a set of possible phonological choices. It
is clearly meaningful to say of a monosyllabic utterance that is has a long vowel or a
high tone, because these terms can be defined without reference to other syllables. It is
often less clear what it means to say that a monosyllabic utterance is “stressed” or
“accented”.

Finally, we should mention lexical distinctions of voice quality, which are often not
considered under the heading of “prosody” at all. In some languages there are
Linguistic Fieldwork Daniel L. Everett
page 85
phonemic distinctions of voice quality which are associated with specific consonantal
contrasts: for example, in Gujarati voice quality distinctions are historically related to
the distinction between “voiced” and “voiced aspirated” stops in other languages of
Northern India, and are found only in the presence of certain specific stops. Similarly,
in many East Asian tone languages there are characteristic differences of voice quality
that accompany pitch differences in distinguishing between one tone phoneme and
another, and which are therefore generally described as part of the tonal system. (This
is the case with the glottalisation that often accompanies Mandarin “Tone 3”, as we just
saw above.) However, voice quality distinctions can be independent of both segmental
and tonal distinctions: for example, in Dinka, the two distinctive voice qualities (often
described as “creaky” and “breathy”) can cooccur with any of the tone phonemes, any
of the distinctive quantity categories, and most of the vowel and consonant phonemes.

5.2.2 Sentence-level functions

At sentence level, prosodic features typically play a role in marking three general
functions: sentence modality and speaker attitude, phrasing and discourse
segmentation, and information structure and focus. However, there is nothing
intrinsically “prosodic” about any of these functions: all of them may also be marked in
a non-prosodic way in addition to, or instead of, a prosodic marking. Thus, for
example, while sentence modality and focus are often marked by intonational means in
many European languages, many other languages employ particles or affixes in the
same functions (e.g. focus particles in Cushitic languages, question marking clitics in
western Austronesian languages).

An important problem in studying the prosodic signaling of these functions is that

many pitch-related phenomena are quasi-universal, which reflects their link to
prelinguistic ways of communicating that we share with other species. As noted in
section 6.1.1, women have higher-pitched voices than men, and individuals can “raise”
and “lower” their voices for various expressive purposes. These “paralinguistic”
functions of pitch and voice quality are broadly similar the world around, though there
are big differences between cultures in the way the paralinguistic functions are
evaluated. For example, a voice raised in anger sounds much the same in any language,
but raising the voice in that way may be dramatically less acceptable in one culture than
in another. Similarly, in some cultures it is highly valued for males to have very low
voices and/or for females to have very high voices, and speakers tend to exaggerate the
biologically based differences, whereas in other cultures little importance is attached to
such differences.

5.2.2.1 Sentence modality and speaker attitude

The prosodic expression of modality and attitude is most closely identified with speech
melody and voice quality. Together, these are the characteristics we are most likely to
think of as the “intonation” of an utterance. Typical examples include the use of overall
falling pitch in statements, overall rising pitch in yes-no questions, or the use of overall
high pitch in polite utterances.

These examples are also typical examples of the difficulty of distinguishing linguistic
and paralinguistic functions of pitch. For example, there have been disagreements about
Linguistic Fieldwork Daniel L. Everett
page 86
whether overall rising pitch in “question intonation” is part of a language-specific
intonational phonology or merely based on the universal use of high pitch to signal
tentativeness or incompleteness. Our view is that it is necessary and appropriate to talk
of “intonational phonology” for at least some sentence-level uses of pitch (see further
section 6.3.1 below). It is important to remember that languages may diverge
considerably from the quasi-universal tendencies mentioned above: there are languages
such as Hungarian or some dialects of Italian, where question intonation includes the
kind of final fall which is typical of statements in other western European languages.
Nevertheless, we acknowledge that there is genuine empirical uncertainty about how to
distinguish phonologized uses of pitch from universal patterns of human paralinguistic
communication.

5.2.2.2 Phrasing and discourse segmentation

In all languages, so far as we know, longer stretches of speech are divided up into
prosodically defined chunks often called intonation units (IUs) or intonation(al) phrases
(IPs). To some extent this division is determined by the need for speakers to breathe in
order to continue speaking, and in the literature the term breath group may also be
found for what we are here calling IU. However, it is important not to think of IUs
purely as units of speech production, because they almost certainly have a role in
higher-level linguistic processing as well, both for the speaker and the hearer. That is,
intonation units are also basic units of information (e.g. Halliday 1967, Chafe 1994,
Croft 1995) or of syntax (e.g. Selkirk 1984, Steedman 2000). Closely related to the
issue of segmentation into IUs are the prosodic cues that help control the smooth flow
of conversation (e.g. signals of the end of one speaker’s turn) and the cues that signal
hierarchical topic structure in longer monologues such as narratives (e.g. “paragraph”
cues). An eventual theory of prosodic phrasing will cover all these phenomena.

The phonetic manifestations of phrasing and discourse chunking are extremely varied.
The clearest phonetic marker of a boundary between two prosodic chunks is a silent
pause, but boundaries can be unambiguously signaled without any silent pauses, and
not all silent pauses occur at a boundary. Other cues to the presence of a boundary
include various changes in voice quality and/or intensity (for example, change to
creaky voice at the end of a unit), substantial pitch change over the last few syllables
preceding the boundary (such as an utterance-final fall), pitch discontinuities across a
boundary (in particular, “resetting” the overall pitch to a higher level at the beginning
of a new unit), and marked changes in segment duration (especially longer segments
just preceding a major boundary). However, it is also important to note that there are
extensive segmental cues to phrasing as well, especially different applications of
segmental sandhi rules. For example, in French, “liaison” – the pronunciation of word-
final consonants before a following vowel – is largely restricted to small phrases and
does not occur across phrase boundaries: allons-y ‘let’s go’ (lit. ‘let’s go there’) is
pronounced [al zi] but allons à la plage ‘let’s go to the beach’ is normally pronounced
[al alapla ], signalling the presence of a boundary between allons and à la plage.

An important conceptual problem in discussing phrasing and discourse segmentation is

that we need to recognize different levels of prosodic structure, and there is no
agreement on how to do this. In corpora of ordinary spontaneous speech it will often
be easy enough to distinguish a basic level of IU, perhaps 6-10 syllables long, set off
Linguistic Fieldwork Daniel L. Everett
page 87
by relatively clear boundaries signaled by silent pauses and other cues. However,
merely dividing texts into a single level of IUs tells us nothing either about the smaller
units that distinguish one syntactic structure from another, nor about the larger units
(often called “episodes” or “paragraphs”) that signal higher-level textual organization
in monologues. This important topic is unfortunately beyond the scope of this chapter.

5.2.2.3 Information structure and focus

Related to the marking of boundaries and cohesion is the use of prosody to signal
semantic and pragmatic features often collectively known as “information structure”.
This includes notions like “contrast”, “focus” and “topic”, and refers to the way new
entities and new information are introduced into a discourse and to the way in which
entities and information already present in a discourse are signaled as such. One
important means of conveying this kind of information is to put specific words or
phrases in prosodically prominent or non-prominent positions. In some languages word
order can be extensively manipulated in order to achieve this, whereas in other
languages the same string of words can have different prosodic structures. Both
strategies are exemplified in English constructions involving direct and indirect objects:
we can say either I gave the driver a dollar or I gave a dollar to the driver, putting
either the amount of money or the recipient in the prosodically prominent final position.
Other things being equal, the first construction is used when the amount of money is
more informative in the discourse context and the second when the point of the
sentence is to convey something about the recipient. However, we can achieve similar
effects by restructuring the prosody so that the major sentence-level prosodic
prominence occurs on a non-final word: I gave the DRIVER a dollar (… not the waiter)
or, somewhat less naturally, I gave a DOLLAR to the driver (…not a euro).

There is an extensive literature on these matters, especially in the European languages;

the reader is referred to Lambrecht (1994) and Ladd (1996) for useful summaries.
Fieldworkers should probably be wary of expecting to find close analogues of
European phenomena in languages in other parts of the world.

5.3. Phonology of tone, intonation and accent

From the foregoing sections it will be clear that “prosodic” features – defined on the
basis of phonetic properties that are not normally indicated in a segmental transcription
– do not form a linguistically coherent set. Among other things, this means that there is
no way of knowing ahead of time how the phonetic features loosely referred to as
“prosodic” – pitch, duration, and so on – are going to be put to phonological use in any
given language. Speakers of all languages produce and perceive differences in pitch,
duration, voice quality, and probably relative prominence, but they may interpret these
differences in radically different ways. There is no unique relation between a given
phonetic feature and its phonological function.

As we suggested earlier, some “prosodic” distinctions turn out to work in ways that are
no surprise to any linguist, while others – sometimes involving the same phonetic raw
material – are still in need of extensive new theoretical understanding before we can be
sure that our descriptions make sense. What seems fairly clear is that the
“unsurprising” prosodic features (like lexical tone and quantity) involve linguistic
Linguistic Fieldwork Daniel L. Everett
page 88
elements that are grouped into strings and contrast paradigmatically with other
elements, like most segmental phonemes. The “problematical” prosodic features (like
accent and phrasing) are somehow involved in signaling phonological structure, the
grouping of linguistic elements into larger chunks. In this section of the chapter we
provide a little more detail on two problematical topics: the tonal structure of
intonation, and the nature of “accent”.

5.3.1 Tone and intonation

As we’ve already seen, pitch provides the main phonetic basis for prosodic distinctions
both at the word level (“tone”) and at the sentence level (“intonation”). Tone languages
are extremely varied, and it would be possible to devote this entire chapter just to
describing the many varied phenomena of lexical and grammatical tone. However,
since there are good descriptions of numerous prototypical tone languages from around
the world and a substantial body of literature discussing various aspects of their
analysis, it would be pointless to attempt a mere summary here. The textbook by Yip
(2002) provides a comprehensive survey, and is a useful guide to various descriptive
and theoretical problems. Anyone embarking on the study of a language known or
suspected to have lexical and/or grammatical tone should be well acquainted with this
literature before leaving for the field.

We focus here instead on intonation. We use the term here in a strict sense, to refer to
sentence-level uses of pitch that convey distinctions related to sentence modality and
speaker attitude, phrasing and discourse grouping, and information structure. The
phonological structure of intonation is better understood now than it was a few decades
ago, but there are undoubtedly plenty of intonational phenomena waiting to be
discovered in undocumented languages, and plenty of things that we will understand
better once we have a fuller idea of the range of possibilities. What we present here is a
minimal framework for investigating intonation in a new language. Our discussion is
based on the now widely accepted “autosegmental-metrical” theory of intonation (for a
review see Ladd 1996).

The most important phonological distinction to be drawn is the one between

intonational features at major prominent syllables and intonational features at
boundaries: in current terminology, the distinction is between “pitch accents” and
“boundary tones”. The existence of such a distinction has been recognized by some
investigators since the 1940s, and is made explicit in current autosegmental-metrical
transcription systems for numerous (mostly European) languages. The difference
between the two can be readily appreciated in English when we apply the same
intonational tune to sentences with markedly different numbers of syllables and/or
markedly different accent patterns. For example, imagine two different possible
astonished questions in response to the sentence I hear Sue’s taking a course to become
a driving instructor. One might respond Sue?! or one might respond A driving
instructor?! In the first case, the pitch of the astonished question rises and then falls
and then rises again, all on the vowel of the single syllable Sue. In the second case, the
pitch is briefly fairly level at the beginning, then there is a steep rise in pitch on the
lexically stressed syllable dri-, immediately followed by a fall, then a level low-pitched
stretch until the very end of the utterance, at which point there is an abrupt rise. [Really
need a sound file and a picture here!///] At a minimum, therefore, the contour consists
Linguistic Fieldwork Daniel L. Everett
page 89
of two separable parts: a rising-falling movement at the main stressed syllable and a
rise at the very end. On the monosyllabic utterance Sue these two parts are compressed
onto the single available syllable, which is both the main stressed syllable and the end
of the utterance. But with a somewhat longer phrase the separateness of the two
prosodic events becomes clear.

One important clue to the correctness of the distinction between pitch accents and
boundary tones is the fact that in some lexical tone languages, where pitch primarily
conveys lexical information, there are nevertheless intonational pitch effects at the ends
of phrases or sentences. These effects typically involve modifications of the lexically-
specified pitch contour on the pre-boundary syllable. Early descriptions of this effect
were given by Chuang 1958 for Szechuan Mandarin and by Abramson 1962 for Thai.
This coexistence of lexical and intonational pitch can be described easily if we
recognize boundary tones: in these languages the pitch contour of an utterance is
principally determined by the lexical tones of the words that happen to make it up, but
at the edges of phrases it is possible to add an additional tonal specification - a
boundary tone.

However, it should be emphasized that not all lexical tone languages use intonational
boundary tones; for example, some West African tone languages appear not to have
them, so that in these languages the pitch contour of an utterance is almost completely
determined by the string of lexical tones. Conversely, there appear to be languages
with intonational boundary tones that have neither pitch accents nor lexical tonal
specifications. In these languages, all intonational effects are conveyed by pitch
movements at the edges of phrases, and “nothing happens” phonologically in between.
Obviously, there is phonetic pitch wherever there is voicing, but the linguistically
significant pitch effects are restricted to phrase edges, and the pitch in between is
determined by simple interpolation. Clear descriptions of such systems are given by
Robert and Rialland (2001) for Wolof and Jun (1998) for Korean.

Current transcription systems for pitch accents and boundary tones, based largely on
the ToBI system first designed for English in the early 1990s, analyze these pitch
movements further: the astonished question contour just discussed would probably be
transcribed as a L+H* pitch accent, an immediately following L- “phrase accent”, and a
H% or L+H% boundary tone. The details are well beyond the scope of this chapter, but
the reader who expects to deal with an unfamiliar intonation system in a language
without lexical tone should consult the ToBI web site (URL http://www.ling.ohio-
state.edu/~tobi/) and its extensive series of links to ToBI systems that have been
designed for a number of other languages.

Before we leave the subject of intonation, we must note that in addition to pitch accents
and boundary tones, intonation can make crucial use of what we might call “register
effects”. Recall that the phonetic realization of pitch distinctions is somehow relative
to the speaker’s pitch range: “high” does not refer to some absolute fundamental
frequency level, but a level that is high for a given speaker in a given context. This
even applies within a single utterance: as a result of the widespread phenomenon of
“declination” – a gradual lowering of pitch across a phrase or utterance – the pitch of a
“high” tone at the end of an utterance may be lower than that of a “low” tone at the
beginning. That is, the phonological interpretation of pitch level is somehow relative to
Linguistic Fieldwork Daniel L. Everett
page 90
a frame of reference that varies not only from speaker to speaker and from context to
context but also from one part of an utterance to another. Such changes of the frame of
reference during the course of an utterance can be exploited for communicative
purposes in various ways, and these are what we are calling “register effects”. The
clearest examples of such effects involve the interaction of lexical tone and overall
pitch level to signal questions. In Chinese, for example, it is possible (though not very
usual) to distinguish yes-no questions from statements in this way.

5.3.2 Lexical accent systems

The existence of tone languages is such a remarkable fact from the point of view of
speakers of non-tonal languages that there are at least two typological schemes –
devised by speakers of non-tonal languages – that attempt to accommodate
lexical/grammatical tone in a larger theoretical understanding. One of these is based on
the “domain” of pitch distinctions, while the other is based on a typology of “word
prosody”. Looking at the domain of pitch, languages have been divided into “tone
languages” (where the domain of pitch distinctions is the syllable), “melodic accent
languages” (where the domain of pitch distinctions is the word), and “intonation
languages” (where the domain of pitch distinctions is the phrase or utterance). This
typology goes back at least to Pike (1945) and is found in work as recent as Cruttenden
(1997). Looking instead at the lexical uses to which “prosodic” features are put, we
can divide languages into “tone languages” (in which each syllable has different tonal
possibilities), “melodic accent languages” (in which one syllable in a word or similar
domain is marked by pitch in some way), and “dynamic accent languages” (in which
one syllable in a word or similar domain is marked by stress in some way). This
typology is suggested by Jun 2005. Both typologies have obvious problems (e.g. the
existence of intonational distinctions in tone languages, the existence of languages like
Swedish with both dynamic accent and lexically specified melodic accent), and neither
commands wide acceptance.

In our view, the problems with these typologies result from trying to incorporate tone
and accent in the same scheme. As we pointed out earlier, tone often functions like
segmental distinctions: it involves a choice of categories from a paradigmatic set, and it
is meaningful to talk about e.g. a high tone on one syllable without reference to the tone
on any other syllable. Accentual distinctions, on the other hand, are syntagmatic
distinctions: they involve contrast with immediately adjacent syllables in a string.
Consequently, we believe that it is quite misleading to see, as in Pike’s typology, a
continuum from tone to melodic accent to intonation, and equally misleading, but in a
different way, to take “tone” and “stress” as different kinds of “word prosody” that a
language may have. Rather, we think it will be useful to discuss the ways in which
accentual systems can differ without necessarily trying to place them into a typological
scheme that places them in the same dimension as intonation and tone (i.e. the typology
of prosodic systems should involve three, at least partially independent dimensions:
tone, accent, and intonation).

A general and possibly universally valid definition of lexical accent is the singling out
of a specific syllable in a word or similar domain (such as the “foot”) for some sort of
prominence or other special prosodic treatment. Lexical accent, as conceived of this
way, is an abstract structural notion, and says nothing about how exactly the “special
Linguistic Fieldwork Daniel L. Everett
page 91
prosodic treatment” is manifested in the acoustic signal. In some languages, the special
status of the accented syllable is based entirely on association with a specific pitch
feature; in other languages, the accented syllable is distinguished from other syllables
by phonetic “stress” – greater force of articulation leading to some combination of
longer duration, greater intensity, more peripheral vowel quality, shallower spectral tilt,
etc. (cf. section 6.1.4). This suggests a distinction between “melodic” and “dynamic”
accent, a traditional distinction recently reestablished by Beckman (1986).

The distinction between melodic and dynamic accent is a phonetic one. Other
typological dimensions on which accentual systems appear to differ involve structural
properties. These include obligatoriness, culminativity, recursivity, transitivity,
intonational anchoring, and lexical distinctiveness. We briefly outline these six
properties here:

Obligatoriness: In some accentual systems, an accent must occur within each domain
of the specified size: if the “prosodic word” is the domain of accent, then each prosodic
word must have an accent. In other systems, the accent may or may not occur in a
given domain. For example, in Japanese, words can be accented or unaccented,
whereas in English any word of more than one syllable must have at least one syllable
that stands out as more prominent when the word is pronounced in isolation.

Culminativity: In some systems, for every accent domain there is a single major
prominence peak. This does not preclude the possibility that other syllables in the same
domain may also be prominent relative to surrounding syllables (see further below
under RECURSIVITY), but there is only one which is the most prominent one of them all.
In a non-culminative system, there may be two prominences within the same domain
without either of them being more prominent than the other one (in some languages,
e.g. Chinese, accentuation in compounds appears to be non-culminative).

It is a matter of debate whether it is useful to distinguish obligatoriness and

culminativity. The alternative is to operate with a single parameter, usually also called
simply culminativity, defined as the property where every lexical accent domain has a
single major accentuation. If one separates culminativity (in a narrow sense) and
obligatoriness, languages such as Japanese have a non-obligatory, but culminative
accent-system (i.e. not every word has to have an accent, but those that have an accent
have only one). If one operates with a single parameter culminative (in a broad sense ),
then Japanese is non-culminative since not every word has an accent.

Recursivity: In some languages, it is possible and useful to distinguish different levels

of lexical accentuation. Thus for English, for example, one commonly distinguishes at
least three different levels of syllable prominence: primary accent, secondary accent,
and unaccented. Primary accent is assigned to the most prominent syllable in a word (as
the English accent system is culminative, there can be only one such syllable).
Secondary accents are assigned to syllables which are also somewhat prominent and in
certain contexts can actually become carriers for the primary accent. There can be
several of these in an English word, as in èxtramètricálity (using grave accents to mark
secondary accents). However, in some languages there is no evidence – or at best very
weak evidence – for anything resembling secondary accent: a single accent is assigned
to a word domain, and all the other syllables are simply “unaccented”.
Linguistic Fieldwork Daniel L. Everett
page 92

One widely-adopted analysis of such secondary accents in languages that have them is
in terms of sub-word domains called (metrical) feet. In a word with secondary accent,
the word domain consists of two or more feet, each with its own most prominent
syllable, and one foot is singled out as the most prominent foot of the word. The
prominent syllable of the prominent foot is the primary accent; the prominent syllables
of the other feet are secondary accents. In languages without secondary accent, we may
say either that there is no level of structure corresponding to the foot, or that the feet are
“unbounded”, i.e. that they are coextensive with the word. See Ewen & van der Hulst
(2001) for a comprehensive introduction to metrical structure.

Transitivity: Just as accentual prominence may apply within domains smaller than the
word, so we may also find accentual prominence relations at the phrasal level when
words are joined together to form phrases. Within a phrase such as yellow paper one
word (normally paper) is more prominent than the other word, which entails that its
most prominent syllable is more prominent than the most prominent syllable of the
other word. That is, the most prominent syllable of the most prominent word becomes
the most prominent syllable of the phrase, often called phrasal prominence or sentence
stress. However, not all accent systems have this feature of transitivity, and then it is
not possible to single out one accented word as the most prominent in its phrase.

Phrasal prominence can be analysed in the same way as lexical secondary accent, in
terms of nested domains each with its own most prominent constituent. However, not
everyone accepts this point of view. In some analyses, phrasal prominence is treated as
being qualitatively different from lexical prominence: on this view, lexical prominence
is usually described as “stress”, and phrasal prominence is described in terms of
intonational “pitch accent” (see e.g. Selkirk 1984 or Shattuck-Hufnagel and Turk
1996). For this reason it is extremely difficult to make reliable and generally
acceptable typological statements about these matters.
Intonational anchoring: In many languages, as we saw in sec. 6.3.1., a lexically
accented syllable serves as the ‘anchor’ for the pitch accents that make up the
intonational tune. This means that in e.g. English and German the lexically most
prominent syllable of the most prominent word in an utterance also carries an
intonational pitch accent. This is the basis for the view of transitivity sketched in the
preceding paragraph: according to this view, lexical accent is phonetically “stress”,
while phrasal prominence is “pitch accent”. We prefer to see this as a fact about the
relation between the accentual system and the intonational system of a given language;
lexical accents may or may not serve the role of intonational anchors. In Japanese and
many other languages with melodic accent, for example, there is no additional
intonational feature that targets accented syllables. But this is not a function of having
a melodic rather than a dynamic lexical accent: in Swedish and Basque, syllables
marked with a melodic lexical accent may additionally also serve as anchors for an
intonational pitch accent. Conversely, recent work on the Papuan language Kuot
(Lindström & Remijsen 2005) suggests that it has dynamic lexical accent (phonetic
stress) but that the intonational pitch accents do not have to occur on a stressed syllable.

Lexical distinctiveness: Finally, another commonly drawn typological distinction

among accentual systems is that between fixed or predictable accent and lexically
distinctive accent. In both Greek and Japanese, despite the fact that the former uses
Linguistic Fieldwork Daniel L. Everett
page 93
dynamic accent and the latter melodic accent, the location of accent can be used to
signal differences between one lexical item or another. In other languages, the position
of stress is either completely fixed (as on the initial syllable in Hungarian or Czech) or
entirely predictable (e.g. Latin, where the penultimate syllable is accented if heavy,
otherwise the antepenultimate).

The dimensions of accentual typology just discussed are probably not completely
independent. Accentual systems with dynamic accent (or phonetic stress) typically
have obligatory and culminative lexical accent, exhibit recursivity and transitivity, and
involve intonational anchoring, and in fact it is widely assumed that all dynamic accent
systems exhibit these properties more or less by default. Although there is no doubt
that the dynamic accent systems of Europe typically show this cluster of features, for
fieldworking purposes we strongly advise you not to take this as given. Kuot and Wolof
are examples of languages with phonetic stress which show that one should be prepared
to encounter unusual combinations and to try to provide substantial evidence for each
of the parameters.

Finally, since melodic accents are realized primarily by pitch changes, they are
sometimes difficult to distinguish from tonal distinctions, and in a number of cases
there is an ongoing discussion whether a given language is better analysed as a tone
language or a melodic accent language. This problem typically arises when there are
only two distinct pitch patterns (high/low or marked/unmarked) and when the pitch
pattern changes only once per lexical item. This type of accent system is widely
attested African and Papuan languages and often discussed under the heading of ‘word
melody’ (see Donohue 1997, Hyman 2001, Gussenhoven 2004, for examples and
discussion). The core issue in analyzing these languages is whether tonal marking has
essentially a paradigmatic function, distinguishing one lexical item from the other, or
rather a syntagmatic (or organizational) function, rendering the marked syllable(s)
prominent in comparison to the neighbouring syllables. While this distinction is
reasonably clear on the conceptual level, there are many borderline cases in actually
attested systems which may be quite difficult to assign to either category (Hyman
(2001) uses a set of parameters similar to the ones above for distinguishing typical tone
and accentual systems). The existence of such borderline cases is not surprising given
the fact that prototypical lexical tone systems may change into melodic accent systems
and vice versa.

In concluding this section, a note on the ambiguity of the term “pitch accent” as used in
much of the literature is in order. This term is now regularly used in two distinct ways:
on the one hand, it refers to the sentence-level (intonational) pitch features that may
accompany prominent syllables in an utterance in a language like English; on the other
hand it refers to the word-level – lexically specified – pitch features that accompany
accented syllables in a language like Japanese. In this chapter, we have opted to use the
term pitch accent only for intonational pitch features and use melodic accent for
lexically specified accentual pitch features.

5.4 Working on prosody in the field

Linguistic Fieldwork Daniel L. Everett
page 94

In approaching the analysis of segmental phonology or morphosyntax in an unfamiliar

language, there are various well-tested techniques for determining the elements and
structures one is dealing with. These include things like minimal pair tests and
permutation tests, which are described in more detail in Chapters 5 and 7. For certain
purposes, these are also relevant for prosody – for example, we have already described
the existence of lexical minimal pairs that differ only in tone, and once you have
determined that you are dealing with a lexical tone language it may be both possible
and appropriate to elicit minimal pairs for tone in exactly the same way that you would
for segmental differences. However, to the extent that prosodic features are not
organised like ordinary segmental phonological and morphosyntactic features, different
techniques are required.

The most important problems in studying prosody in the field are the fact that prosody
is pervasive – you can’t have an utterance (even a single elicited word) without prosody
– and the fact that it is influenced by both lexical and sentence-level factors and may
thus be contextually variable in ways that are difficult to anticipate, or to notice. For
example, if you were asked out of context to give the name of the famous park in the
middle of London where people come to make speeches to anyone who happens to
want to listen, you would say Hyde Park, with the two words about equally prominent.
However, if you were in a conversation about great urban parks – like Grant Park in
Chicago or Central Park in New York or Stanley Park in Vancouver – you would
probably say HYDE Park, with the main prominence on Hyde. (In fact, if you read the
previous sentence aloud you will find it is very difficult to say the list of park names
without putting the main prominence in each on the proper name and de-emphasising
Park in each case.) If you were doing fieldwork on English and knew nothing about
the language, you would have to become aware of this contextual effect before you
could accurately describe the prosody of expressions like Eiffel Tower or Princes Street
or Van Diemen’s Land that consist of a proper noun and a common noun.

In this section, therefore, we will discuss research procedures which are particularly
useful in prosodic research but rarely used in working on other aspects of the grammar
of a given language. We begin by describing some useful “first steps” to take in the
prosodic analysis of a previously undescribed language.

5.4.1 First steps

It is important to establish early what sort of lexical prosodic features are found in the
language you are working on. The literature on neighbouring and related languages
may provide important pointers in this regard, but it is obviously necessary to remain
open to all possibilities until clear language-internal evidence points in one direction or
the other. If you are working on a language with distinctions of quantity or of lexical
accent (whether dynamic accent or melodic accent), it may take some time to become
aware of the distinctions, because as we noted earlier the functional load of such
distinctions may be relatively low. If you are working on a prototypical lexical tone
language, it is likely to become evident quite quickly, because native speakers will
usually point out to you that items that you appear to consider homophonous are not
homophonous but clearly distinct for them. However, unless you are working with
speakers who are also familiar with a well-described tone language, they will not
Linguistic Fieldwork Daniel L. Everett
page 95
necessarily make reference to tone (or pitch) in pointing out these differences. They
may simply assert that the items in question sound very different, sometimes perhaps
even claiming that the vowels are different.

Although there may be some languages with no lexical prosodic features whatever, in
general it will be a useful starting hypothesis that in any given utterance some prosodic
features will be lexically determined and some determined at the phrase or sentence
level. Both levels are inextricably intertwined; there is nothing in the signal to tell you
whether a given pitch movement is lexically motivated (e.g. lexical tone),
intonationally motivated (e.g. sentence accent), or even both (e.g. the combinations of
lexical and intonation tone commonly found on sentence-final syllables in Chinese or
Thai). This problem is of central importance when analyzing pitch, but sometimes
affects the analysis of quantity and accent as well. Perhaps the most important lesson
to begin with is that recording and analyzing words in isolation does not in any way
provide direct, untarnished access to lexical features. This is a classic mistake,
unfortunately widely attested in the literature. A single word elicited in isolation is an
utterance, and consequently cannot be produced without utterance-level prosodic
features. For example, if you compare ordinary citation forms of the English words
PERmit (noun) and perMIT (verb) [link to sound files], you might conclude that high
pitch, followed by a fall, is a feature of lexical stress in English. However, high pitch
associated with the stressed syllable is actually a feature of declarative statement
intonation in short utterances. If you utter the same words as surprised questions, the
stressed syllables will be low, followed by a rise in pitch to the end. If you utter the
same words as part of a long and boring list, the stressed syllables may be high, but
followed by a rise in pitch (rather than the fall seen in the isolated citation form). In
short, even for single word utterances it is not a straightforward matter to distinguish
between lexical and intonational prosodic features. There is no intonationally
unmarked ‘citation form’; every utterance has intonation.

In order to separate the two levels, lexical items have to be observed in a number of
different syntactic and semantic-pragmatic contexts. Whatever prosodic features remain
constant across these contexts most likely pertain to the lexical level; features which
change may relate to the sentence level. But especially in dealing with lexical tone
languages, even this statement needs qualifying, because in many such languages there
are complex locally-conditioned variations in tonal pattern, sometimes called tone-
sandhi (see Yip 2002 for examples and discussion).

In order to elicit target words in different contexts, one can construct short clauses or
phrases where the target words may occur in different positions (i.e. initial, medial,
final). A particularly useful variant of this technique is to record short (3-5 word) lists
of target words with the words in different positions in the list. If speakers produce a
coherent list rather than a sequence of minimal utterances, the result is likely to be a
contrast between list intonation and minimal declarative utterance intonation. This may
allow you to distinguish word-level prosodic effects. More generally, list intonation
may be particularly useful in the initial stages of such an analysis for three reasons.
First, it is relatively easy to elicit naturally: the act of listing elicited items does not
differ in principle from listing items as part of a procedural description, whereas
enacting a question is quite different from actually asking a question. Second, list
intonation tends to be fairly simple in the sense that there is usually only an opposition
Linguistic Fieldwork Daniel L. Everett
page 96
between non-final and final members, or sometimes a three-way distinction between
non-final, penultimate and final. In particular, there are no differences of information
structure (focus, topic) in lists, which often complicate the interpretation of prosodic
features in other types of examples (see also section 6.2.2.3 above). Third, list
intonation may be more consistent across speakers, which makes it easier to recognize
the same intonational targets across speakers and at the same time provide an indication
of inter-speaker variability.

5.4.2 Elicitation

All modern descriptive and documentary fieldwork includes the recording of a

substantial corpus of (more or less) spontaneous ‘texts’ (where ‘text’ subsumes all
kinds of communicative events including conversations, narratives, oratories, etc.). If
these recordings are done with reasonable quality, they can form the basis for
subsequent auditory and instrumental analysis of many prosodic features of connected
speech, features that may be difficult to observe in structured interview sessions and
difficult for most native speakers to be aware of. However, just as you would not
expect to study phonology or syntax solely on the basis of a recorded corpus, so in the
case of prosody it is important to complement recorded texts with elicited data.

In eliciting data for prosodic analysis it is important to keep various factors in mind that
are of only secondary importance for eliciting many other kinds of data. First and most
important, it is essential to keep in mind the kind of effects that context may have, and
to adjust elicitation procedures accordingly. For example, in English it is common for
WH-questions to be pronounced with an overall falling contour in neutral contexts
(Where is he going?), a relatively high level followed by a low rise at the end in polite
contexts (Where would I find Dr. Anderson?), and an overall rising contour in
repetition or reminder contexts (Where did you say you were from?). Eliciting such
distinctions may require you to get native speakers to put themselves mentally in
different contexts, which is not necessarily easy to do. We return to this topic at some
length in the next section.

Second, it is important to record several speakers rather than relying on one or two
primary consultants. One reason for this is the conspicuous difference of voice pitch
between males and females; another is that many prosodic features vary more between
individuals and between socially defined groups than do centrally “linguistic” features.
Fieldwork situations will usually put severe limits on how many speakers you can work
with, but if at all possible it will be valuable to record elicited material from at least 4
and as many as 8 or 10 speakers. Gender balance is an important concern in putting
together a set of speakers. In situations where it is impossible to find several speakers
for the same task, it may be useful to record the same material with the same speaker a
few days or weeks apart. There is little use in recording the same example set twice as
part of the same session because this will almost certainly produce repetition effects.

Third and finally, it is important to keep in mind that instrumental acoustic analysis is
increasingly regarded as an essential part of reliable descriptions of prosody, and that
preliminary instrumental work in the field may be invaluable for guiding your work.
This means that elicitation must be done in such a way that the resulting recordings are
usable for instrumental analysis. In devising test examples for prosodic features, it is
Linguistic Fieldwork Daniel L. Everett
page 97
important to pay attention to the segmental make-up of the example in order to
minimize microprosodic effects (see sec. 6.4.5). However, it is often not possible to
come up with materials that perfectly control for microprosody; either the phonotactics
of the language may prohibit certain sequences that would be useful to include in your
materials, or the only lexical exemplars of a particular sequence may create
meaningless, obscene or ridiculous sentences that native speakers may refuse to say or
will be unable to say naturally. As usual in experimental work, there is a trade-off
between naturalness and the control of interfering variables.

5.4.3 Problems in prompting speakers

As the example of English WH-question intonation makes clear, eliciting example

sentences for prosodic research requires attention to various factors that are not usually
of concern to fieldworkers, and makes demands on speakers that ordinary phonological
and syntactic fieldwork may not. Suppose you carefully construct a question-answer
pair, paying attention to both pragmatic plausibility and segmental make-up. It is not
enough to get native speakers to produce the segments of which the example sequence
consists; they have to produce the first part as a question, the second as an answer. Do
not underestimate the problems involved in explaining the idea of pretending to pose a
question or give an answer. Moreover, be aware that some speakers may be unable to
do things like this naturally, even if they understand the idea. This is one of the reasons
why it is important to record multiple speakers wherever possible: without being able to
compare across a sample there is no way of knowing who is acting reasonably well and
who is doing something else.

We just spoke of carefully constructing question-answer pairs for native speakers to

produce, but there is a significant problem of how to present tokens for prosodic
research without unduly influencing the speakers. It is of little use to have a speaker
repeat what the fieldworker is saying, since there may be direct effects of repetition on
the speaker’s production, or the speaker may in some way imitate the researcher’s
model. If you are working in a literate community, reading can be a good method for
eliciting intonational data, provided that the speakers understand the need to vocally
enact the illocutionary force of the example sentences. Unfortunately, it often happens
that even literate speakers are unable to read fluently in their native language; it is
common to find speakers who are literate in a majority or national language but have
little practice or experience reading their native language. One technique that has been
successfully used with such speakers is to present them with material written in the
language they are comfortable reading, and ask them to give equivalents in their own
language. But only some speakers will produce natural-sounding utterances under such
conditions. It is also known from work on major European languages that the intonation
patterns found in reading do not perfectly match those found in spontaneous
conversation. Here the influence of the standard norm may be a major issue.

If reading is not feasible, various role-playing and experimental tasks may be useful.
For example, rather than constructing question-answer sequences in advance and
asking speakers to ‘enact’ them as naturally as possible, one may try to involve
speakers in some kind of game or role play which requires them to ask questions. A
technique widely used for this purpose involves matching tasks where one speaker
instructs another speaker in reconstructing an arrangement of figures, pictures or points
Linguistic Fieldwork Daniel L. Everett
page 98
on a map which is only visible to the instructing speaker (see map task, space games,
etc. /REF/ and chapter ??.??[needs input by Dan]). Another technique is to have
speakers look at a picture sequence or watch video clips and then to describe these or
comment on them (pear film, frog story, etc. /REF/ ?? [needs input by Dan]). The big
advantage of these techniques is that speakers are prompted with non-linguistic
materials, and relatively spontaneously produce naturalistic speech. Moreover, unlike
completely open-ended tasks such as recounting narratives or engaging in free
conversation, these tasks permit a certain degree of control over what speakers will do,
which makes it possible to collect comparable data from several different speakers.
While it is rare that speakers produce completely identical utterances in these
circumstances, a well-devised task usually requires them to use particular words,
phrases or constructions and to engage in specific linguistic routines such as asking
questions or giving directions.

Such tasks are not without their problems, however. The major problem is that speakers
in small and remote communities are generally not familiar with the idea of role-
playing or experiment and may be unable or unwilling to participate. It is not unknown,
for example, that speakers who are asked to retell a video clip they just watched
comment on the colors of the main participant’s clothes or the nature of the setting
rather than the action depicted in the clip. Considerable time and ingenuity may thus be
required in adapting the experimental set-up to the specific circumstances found in a
given speech community and in explaining the task.

5.4.4 Perception experiments

For prosodic analyses it may also be desirable to obtain some perceptual data in
addition to the production data generated with experimental tasks or documented in
narratives and conversations. Perceptual data are needed to answer questions such as:
Do native speakers actually perceive prominences at those locations where they appear
in the acoustic data (or where they are perceived by the fieldworker)? Which of the
various factors contributing to a given prominence (intensity, duration, vowel quality,
change and height of pitch) is the one actually of major importance in the language at
hand? Which parts of a pitch contour are actually perceived as major cues for question
intonation? Such questions can generally only be answered with some degree of
certainty by devising perceptual tests, i.e. manipulating the prosodies of example
clauses or phrases and testing speakers’ reactions to them. For example, one may
reduce the duration of putatively stressed syllables and ask speakers to identify stressed
syllables in tokens computationally modified in this way, comparing the results with
results obtained when identifying stressed syllables in naturalistic (unmodified) tokens.
See van Zanten et al. (2003) and Connell 2004 for detailed descriptions of such
experiments.

Once again, however, it has to be pointed out that administering such experiments is
not a straightforward matter and will not necessarily produce satisfactory results. Apart
from problems involved in getting speakers to participate at all in a listening
experiment (in some instances, putting on a headset may already be a problem), the
main problem pertains to defining a task which speakers are able to perform and which
also generates relevant data. In most non-literate societies, it will be impossible to use
concepts such as syllable or prominence in explaining a task. Task types that may work
Linguistic Fieldwork Daniel L. Everett
page 99
- to a certain degree at least - are: a) asking speakers to comment in a general way on
prosodically modified examples (which produces very heterogenous and non-specific
results but may still be useful in providing pointers to relevant parameters); b) tasks
which involve the comparison or ranking of similar tokens (Which of these two items
sounds ‘better’/’foreign’? Which token would you use when speaking to your mother?
etc.).

5.4.5 Computer-aided acoustic analysis

Perception experiments of the kind just mentioned presuppose the use of programs for
acoustic analysis such as praat, emu, wave surfer or speech analyzer (see chapter ??.??).
Use of such programs is strongly recommended for all kinds of prosodic analyses. The
main reason for using them is that they may be of help in overcoming biases in one’s
own perception of prosodic data and in detecting phenomena one has not been listening
for. As further discussed shortly, acoustic data are always in need of interpretation and
auditory crosschecking. Nevertheless, they provide the only objective source of
prosodic data, and an analysis which goes against major acoustic evidence is almost
certainly false.

The programs just mentioned provide fairly reliable acoustic analyses of duration,
intensity and F0. These can be done on a laptop in relatively short time and hence are
feasible also in field situations provided that laptops can be used at all. Handling the
programs can be learned in a few hours (in particular in the case of speech analyzer or
wave surfer). Hence, it would be most inefficient not to use these tools when tackling
the prosodic analysis of a previously undescribed language.
The current section briefly reviews the most important parameters to keep in mind
when interpreting F0 extraction. For effective fieldwork it is not necessary to
understand the mathematical and engineering aspects of F0 extraction. However, it is
necessary to know something about the factors that affect F0 in order to interpret pitch
contour displays appropriately and to select speech materials for phonetic analysis. It is
easy to be misled by what you see on the screen, and easy to make instrumental
measurements that are nearly worthless.

The rate of vibration of the vocal cords can be briefly but substantially affected by
supraglottal activity – that is, by the fact that specific vowels or consonants are being
articulated at the same time as the vocal cords are vibrating. Such effects are often
collectively referred to as microprosody. Fig. /// shows instrumental displays of two
English utterances, pronounced with pitch patterns that are impressionistically the same
(listen to sound files ///). However, it can be
seen that the two pitch contours look very different.

a Is that Betty’s kitten?

b Are you Molly Newman?

The most obvious difference is that in (b) the contour is continuous, whereas in (a)
there are many interruptions. This makes sense if we recall that we must have voice to
have pitch: voiceless sounds have no periodic vibration and therefore no F0. As
listeners we are scarcely aware of these interruptions, but on the screen they are very
Linguistic Fieldwork Daniel L. Everett
page 100
conspicuous. Even more conspicuous is the fact that the F0 in the immediate vicinty of
the interruptions jumps around a lot. These so-called “obstruent perturbations” are
caused in part by irregular phonation as the voicing is suspended for the duration of an
obstruent, or (in the case of voiced obstruents) by changes in airflow and glottis
position as the speaker maintains phonation during partial or complete supraglottal
closure. Even an alveolar tap (as in Betty) often causes a brief local dip in F0; a glottal
stop (as in kitten) often causes a much greater local dip. The consequence of such
obstruent perturbations is often that the pitch contour on a vowel flanked by obstruents
(like the first syllable of kitten) looks like an abrupt fall on the visual display, even
though perceptually and linguistically there is no significant pitch change on the
syllable at all.

Methodologically, the existence of obstruent perturbations means that great care must
be taken in interpreting visual displays of F0. Beginners tend to overinterpret what they
see on the screen. In case of a conflict between what you see on the screen and what
you hear, trust your ears! Obstruent perturbations also mean that the best samples of
speech for making instrumental measurements of pitch are stretches containing as few
obstruents as possible.

The other type of microprosodic effect that it is important to be aware of is “intrinsic

pitch” or “intrinsic F0” of vowels. The phenomenon here is very simply stated: vowel
quality affects pitch. Other things being equal, a high vowel like [i] or [u] will have
higher F0 than a low vowel like [a]. If you say to Lima and a llama using the same
intonation pattern and being careful not to raise or lower your voice between the two,
the F0 peak on to Lima will be higher than that on a llama even though they sound
exactly the same. This effect appears to have some biomechanical basis, although it is
not entirely clear what that basis is. No language has ever been discovered to be
without intrinsic F0 effects, although in some languages with more than two lexically
distinct level tones the effect may be smaller than in other languages.

The methodological significance of intrinsic F0 is that if you want to measure F0 level

instrumentally, you need to control vowel quality. Don’t try to compare measurements
of mid tones and high tones if all the mid tones occur on [i] and all the high tones occur
on [o]. Be sure to compare like with like.

5.5. A final thought

In addition to being a central part of any language description, prosody is relevant to

the fieldworker in a very different way, because it may affect communication with
native speakers and local authorities. It has frequently been suggested that
misunderstandings in cross-cultural communication can be caused by misinterpreting
prosodic cues. Although there are certainly generalizations about the sentence-level
uses of prosody that are valid in language after language, the details may differ in
crucial ways. What sounds rude and aggressive to one party may just be the normal
way of marking emphasis for the other. A noticeable fall in pitch at the end of a unit
may signal a simple assertion to the non-native hearer, but the speaker actually intended
to pose a polite question. And misunderstandings may occur even if the fieldworker
and the community members use a contact language to communicate, because both
Linguistic Fieldwork Daniel L. Everett
page 101
parties will tend to bring their native prosodic systems to the contact language. So an
appreciation of the ways in which prosody can differ from language to language is in
itself an essential tool for successful fieldwork.

References

Abramson 1962
Beckman, Mary E., 1986, Stress and non-stress accent, Dordrecht: Foris
Chafe, Wallace L., 1994, Discourse, Consciousness, and Time, Chicago: The
University of Chicago Press
Chuang 1958
Croft, William, 1995, “Intonation units and grammatical structure”, Linguistics 33:839-
882
Cruttenden, Alan, 1997, Intonation, 2nd ed., Cambridge: Cambridge University Press
Donohue, Mark, 1997, “Tone system in New Guinea”, Linguistic Typology 1:347-386
Ewen, Colin J. & Hulst, Harry van der, 2001, The phonological structure of words,
Cambridge: Cambridge University Press
Gussenhoven, Carlos, 2004, The phonology of tone, Cambridge: Cambridge University
Press
Hyman, Larry M., 2001, "Tone systems", in: Martin Haspelmath, Ekkehard König,
Wulf Oesterreicher & Wolfgang Raible (eds), Language Typology and Language
Universals, Berlin: de Gruyter, 1367-1380
Jun, Sun-Ah, 1998, "The Accentual Phrase in theKorean prosodic hierarchy",
Phonology Phonology 15: 189-226
Ladd, D. Robert, 1996, Intonational phonology, Cambridge: Cambridge University
Press
Lambrecht, Knud, 1994, Information structure and sentence form : topic, focus, and the
mental representations of discourse referents, Cambridge: Cambridge Univ.
Press
Laver, John. 1980. The phonetic description of voice quality. CUP
Lehiste, 1970
Lindström, Eva & Bert Remijsen, 2005, "Aspects of the prosody of Kuot, a language
where intonation ignores stress", Linguistics 43(4), 839-870
Pike, Kenneth L., 19
Rialland, Annie & Stéphane Robert, 2001, "The intonational system of Wolof",
Linguistics 39: 893–939
Selkirk, Elisabeth O., 1984, Phonology and Syntax: The relation between sound and
structure. Cambridge, Mass.: The MIT Press
Sluijter and van Heuven 1996
Steedman, Mark, 2000, The syntactic process, Cambridge, MA: The MIT Press.
Linguistic Fieldwork Daniel L. Everett
page 102
Van Zanten, Ellen, Robert W.N. Goedemans and J.J. Pacilly, 2003, "The status of word
stress in Indonesian", in: Van de Weijer, Jeroen M., Vincent J.J.P. van Heuven
and Harry G. van der Hulst (eds), 2003, The Phonological Spectrum II:
Suprasegmental Structure, Amsterdam: Benjamins, 151-175.
Yip, Moira, 2002, Tone, Cambridge: Cambridge University Press
Linguistic Fieldwork Daniel L. Everett
page 103
"One of the first tasks which confront the linguist in examining a new language with a
view to discovering and describing its structure is the identification of the minimal
meaningful units of which the language is composed." Nida (1974 [1946], p6)

CHAPTER 6. MORPHOLOGICAL FIELDWORK

6.1 INTRODUCTION
In my early days of research among the Pirahãs, every single utterance was
nearly impossible to understand. Since the Pirahãs do not speak Portuguese, there was
no way for me to get a translation of anything they said. And, for reasons I was not to
understand for many years, they didn't seem to have simple verbs. Where I expected a
simple verb for, say, smelling, I would get long multisyllabic utterances. Each word I
was able to isolate and understand was a costly accomplishment. Learning what a
particular morpheme meant was cause for celebration.
In polysynthetic languages, e.g. many American Indian languages, the hardest nut
to crack is the morphology. And the difficulty of understanding the meaning of
individual morphemes is not likely to be significantly easier in any language, even in
languages that lack inflectional morphology (e.g. Wari', Everett (2005a, 2005b) or even
have no morphology (if there are in fact such languages). If a language has little in the
way of morphology proper, this will simply move the problem of understanding the
meanings and functions normally conveyed by affixes one level higher, to the word or,
more likely, to the phrasal level. In other words, no fieldworker can avoid analyses of
words and how they fit into the grammar, a matter that always entails morphological
principles.
Therefore, what I want to do in this chapter is to help equip the fieldworker for
successful morphological field research. Of course, a single chapter, or two or three, is
not really enough space to provide significant detail about all aspects of morphological
analysis. There are, after all, entire books dedicated to these matters (e.g. Nida () and
Pike (). Nevertheless, this chapter aims to provide at least a 'check list', with examples
to remind the fieldworker of the minimum required for data-collection and
morphological analysis in the field.
One might think we could begin by simply answering the question 'What is
morphology?' and then get on with a methodology for it. Morphology can be usefully
defined as the study of words and how they are formed (see Everett (in progress) for
various approaches to the study of word formation).
And yet it is not always clear where the dividing line falls between morphology
and syntax, a problem that can confound both the fieldworker and the theoretician.
Syntax is often defined (at least informally) as the study of how words get put together
to form sentences. But in practice, the fieldworker can often only understand word
structures if they understand the sentential contexts (structural and semantic) in which
the various word-forms may appear and vice-versa. So one needs to study syntax and
morphology together, even though one might believe that morphology and syntax are
ultimately distinct components of grammar (e.g. that they are organized in different
ways and/or have different principles – see Ackema and Neeleman (2004) for useful
discussion).
Although the morphology and syntax of a language need to be studied together,
there are still principles and methodological procedures unique to word structure. So I
focus on those here. But where necessary, syntax is brought into the discussion to
Linguistic Fieldwork Daniel L. Everett
page 104
explicate morphological concepts and their implications.25 I want to offer three caveats
on morphological fieldwork before beginning with a consideration of morphological
fieldwork proper.

6.2. CAVEATS
6.2.1. PHONOLOGY IS NOT MORPHOLOGY
You will hear a cacaphony of confusion when you first enter an area where you
do not speak the language (e.g. as I do now as I write this in Leipzig and ich spreche
nur ein bisschen Deutsch.) And it can depress you if you think too long about that fact
that it is your responsibility to bring order to your confusion, to find euphony in the
apparent cacaphony.
In my first linguistics course, Professor Kenneth L. Pike asked the class how we
might analyze the form in (1), which he claimed to be common in English (class
lectures, University of Oklahoma, 1976):

(6.1) squeat

I did not know what he was talking about. What does this mean? Pretending for
the moment that we do not speak English, let's approach this problem like fieldworkers.
We first need to consider the context in which () is uttered. It turns out that it is usually
uttered by a member of a group preparing to dine together. The linguist must observe
this context him/herself. They cannot ask speakers directly what an utterance means
without supplying them with a context. In fact, native speakers should only be asked to
opine about data – and then only when they fully understand the relevant context of the
items under discussion in very carefully controlled circumstances. After providing a
context, on the other hand, if we then asked speakers of this dialect what squeat means,
some speakers might understand it and might be coaxed to repeat it more carefully. If
they do, they are likely to say something like:

(6.2) Oh, you mean 'let's go eat'.

With this more careful repetition, it becomes clearer that the phonology of ()
obscures its morphology. This is an important point. To see this perhaps more clearly,
let us go deeper into the phrase. How might a field linguist analyze the s at the end of
[lEts], for example? They might be tempted to conclude that this s is part of the word
[lEt], basing this morphological decision on the phonetics of the word form. 26 But this

25
Interestingly, in my experience when a language has a complex morphology, that
morphology is more difficult to analyze than the syntax of the language, however
complex the latter.
26
This is not a textbook on morphology, but some terminology would be useful here
nonetheless:

word form: a phonological 'word', without regard to its morphological composition, a

constituent which fits the phonological definition of word in the language. So, for
example, let's is a word form in English. As is am, is, were, gone, going, etc.
Linguistic Fieldwork Daniel L. Everett
page 105
would be wrong, clearly. Although s is indeed part of the phonological unit that
precedes it, it is best analyzed as an independent word syntactically, a shortened form
of the pronoun, [Us]. This is what Zwicky (1977, page?) labels a 'simple clitic'. In other
words, the phonological data in this (simple) case once again conflicts with the
morphosyntactic data. This leads me to offer the following related warning:

(6.3) Be careful not to misanalyze clitics, affixes, and function words because of their
phonology.

A subtype of (3) is (4):

(6.4) Cliticization is often confused with affixation.

Consider, for example, how a novice fieldworker might analyze the English
examples in (5) and (6):

(6.5) a. John's hat.

b. The man's hat.
c. The man who was running's hat.
d. The man you spoke to's hat.

(6.6) I'll come tomorrow.

In (5) there is an enclitic which attaches to the final word of the noun phrase. This
enclitic, as it turns out, is indeed morphologically, syntactically, and phonologically
part of the preceding phrase. By analogy, the fieldworker might analyze the 'll of (6) in
the same way, assuming that it too 'goes with' the word on its left. But semantically and
syntactically, 'll forms a unit with the verb that follows it, not with the word that
precedes it. That is, its phonological host is unrelated to it syntactically. This
phenomenon is often labelled 'wrong-way' cliticization in the literature, a case where
the phonology and morphosyntax conflict. An additional example of wrong-way
cliticization, further underscoring the importance of keeping phonology and
morphology separate in fieldwork, comes from Yagua, a language spoken in Peru, from
(Everett 1989, 343, who takes it from Payne & Payne ()):

(6.7) Sai -pu7@u7@chi(*-níí) Pauroi rooriy(*-níí) viimú -nííj Anitaj.

3sg.cl -carry Paul house into 3sg.cl Anita
'Paul carries Anita into the house'

word: an abstract morphosyntactic unit that is not heard, but posited to account for
morphosyntactically relevant word behaviors.

lexeme: the minimal unit of the lexicon.

Similar (though not identical) definitions for these are offered in Matthews (1991,
24ff).
Linguistic Fieldwork Daniel L. Everett
page 106
Everett (1989) attempts to explain, among other things, the observations of Payne
() and Payne and Payne () that Yagua object clitics, like English auxiliaries, undergo
'wrong-way' cliticization. The clitic –níí in () must attach to the immediate left of the
object, in this case, Anita. It cannot attach directly to the object, but it must
nevertheless be immediately to its left. So although the clitic and the object do not form
a constituent phonologically, they do syntactically (at least according to Everett
(1989)).
To sum up these points aphoristically: What you hear is not what you get. Or
more scientifically, articulation obscures morphosyntactic boundaries in normal
speech. You must think, analyze, and argue for all your conclusions. Nothing is
obvious in linguistic analysis, especially not in morphological analysis. Do not be
fooled by confusing phonology with morphology! Let us consider now another caveat,
namely, that:

6.2.2. SEMANTICS IS NOT MORPHOLOGY

The fieldworker must also be careful not to assume a direct or transparent
relationship between semantics and morphology. A meaning that is expressed as a
single morpheme in one language may not be expressed at all in another language or it
may be expressed by multiple words or morphemes. Although semantic arguments, like
phonological and phonetic arguments, can be useful in identifying and classifying
words, word classes, morphemes, etc. in a language, there is no algorithm for
translating from one to the other or any necessarily direct relationship.
To understand this a bit better consider the category of tense. Consider the
question, "Does English have a future tense?" in light of the examples in ():

(6.8) a. I will go tomorrow.

b. I am going to go tomorrow.

Are either of these the future tense in English? Well, if you mean semantics, then
the answer is 'yes', they both express (among other things) a future meaning. But if you
mean morphology, then the answer is 'no', neither of these examples contains a
morphological future tense. In fact, morphological and semantic categories are often
confused in just this way in the professional literature. Everett (1993), for example,
argues that Pirahã has no morphological tenses, though it does have 'absolute tenses'
(Comrie (1985)) semantically. But I argue there as well that Pirahã has no relative
tenses, either morphologically or semantically.
It is common as well for novice field linguists to find it difficult to keep semantics
and syntax separated as they prepare a grammar of an unstudied language. For
example, linguists I have worked with in following some sort of questionnaire e.g. the
Lingua Descriptive Series Questionnaire (LDSQ), often initially answer affirmatively
to some of the structural or constructional questions of the LDSQ (of the type 'Does the
language have relative clauses?') even when the language in question lacks such
structures, just because there is a way to express that meaning in the language under
investigation.27 So, for example, does English have 'evidentials' (see section ___ below)
– morphological markers that indicate the source/type/quality of evidence for a given
assertion? Morphologically, no, English does not have such markers. But of course an

27
The LDSQ can be found at: http://lingweb.eva.mpg.de/fieldtools/linguaQ.html
Linguistic Fieldwork Daniel L. Everett
page 107
English speaker can always add to a statement, 'I saw that with my own eyes' or 'So
they say', using these expressions like evidentials semantically. So English can express
evidentiality, even though it has no evidentials. This is a very important point that is not
nearly as easy to recognize in fieldwork as it is in the linguistic classroom. In other
words, when linguists investigate structure, they are not thereby investigating the
circumlocutions that might be used to express the meaning or function of the structure.
They are concerned with understanding both form and meaning, of course, recognizing
however that they should be studied separately (initially at least) in order to better
understand each. An excellent discussion of these issues is found in Comrie, et. al.
(1993).
A final caveat before beginning our discussion of morphological fieldwork proper
is that:

6.2.3. SYNTAX IS NOT MORPHOLOGY

Although, as I said, syntax and morphology need to be analyzed nearly
simultaneously due to their mutual influence, it is possible to tease them apart. They are
not the same thing, at least not operationally in the field (see Ackema and Neeleman
(2004) for interesting discussion of how these might be distinguished in the grammar
(or not) and the consequences of different decisions). They differ most strongly, in my
opinion, in that word formation rules are paradigmatically constrained (see Everett
(2005) and ___ below).
And yet, morphology and syntax often appear to overlap – which can and will
confuse the fieldworker unprepared for this. As Ackema and Neeleman (2004, 48ff)
show, morphology and syntax can often 'compete' with one another. So, for example,
consider the contrast in (9) and (10) (again, see Everett (in progress) for a survey of
approaches to morphological analysis):

(6.9) a. *felicitouser
b. more felicitous

(6.10) a. bigger
b. *more big/more bigger

If an adjective cannot take the comparative suffix, -er, then it is made

comparative syntactically, or periphrastically. The suffix might be said to block the use
of the syntactic form. That is, the syntactic or periphrastic form and the affixal form are
not in free variation in most dialects (though this is an empirical question that has not
been well-investigated to my knowledge, even though its consequences for
morphosyntactic theory are potentially important). We will see more evidence for this
in section ___ on periphrasis. Such interactions reinforce both suggestions made here,
namely, that syntax and morphology must be analyzed together by the fieldworker, yet
they are not the same thing. Understanding English comparatives, in other words,
requires us to understand the different roles of syntax and morphology in constructing
comparatives, as it also requires that we analyze both together to see the full set of
patterns emerge from the data. These are the main caveats. Now we are ready to begin
our discussion of morphology fieldwork proper. I prefer to discuss morphological field
analysis in terms of semantic distinctions and how these are marked, since as a
fieldworker, I found this to be the most natural way to think about the issue.
Linguistic Fieldwork Daniel L. Everett
page 108
6.3 SEMANTIC AND GRAMMATICAL DISTINCTIONS
6.3.1. NOMINAL CATEGORIES
The two most important questions for morphosyntactic fieldwork are (i) What
semantic and grammatical distinctions are marked on words? and (ii) How are these
distinctions marked? That is really all there is to morphology. There is a large, but
fortunately, not too large, number of answers to each of these questions and many
combinations thereof. But we can in fact offer a couple of simplified lists that, I
believe, will to better conceptualize the issues.
I am using 'semantic distinctions' to refer to properties of the lexicon (as I see it)
that are relevant for the syntax, e.g. aspect, number, class, case, tense, etc. Another way
of putting this is as in (11):

(6.11) A morphologically relevant semantic distinction is any meaning contrast of a

language x that is reflected in the words of x.

The terms we use at this point are not crucial. What we are after here is a
counterpart to the philosopher J.L. Austin's (1962) famous How to do things with
words, namely 'how to do things to words'. The most important rule of thumb in
morphosyntactic field research is to be alert and be creative in your thinking, following
Postal's Maxims (see ____).
We begin with an overview of the major semantic distinctions encountered on
nouns, grouped under my headings:

6.3.1.1. REFERENCE: specific vs. referential vs. definite

This is a fundamental distinction that is perhaps a semantic universal, though its
morphological reflex (marking on word forms) is less so.

Consider the following sentences:

(6.12) a. John bought a book about Vietnam.

b. John bought the book about Vietnam.

These examples are, respectively, indefinite and definite. In general (but see
Lyons (1999)) the definite article indicates shared knowledge between speaker and
hearer, or old information from the discourse. The indefinite is often used to introduce
new participants into a discourse, or to offer a different type of quantificational reading.
But there is more to them than this. In both of these examples the speaker can either
have a very specific book in mind or not. For example, both can be continued as in (13)
and (14):

(6.13) a. John bought a book about Vietnam, in fact he bought Fire in the Lake, by
Frances FitzGerald.
a. John bought a book about Vietnam, but I don't know which one.

(6.14) a. John bought the book about Vietnam, or at least that is how he was
describing it, I really don't know anything else about it.
b. John bought the book about Vietnam, you know, the one in the window
we saw yesterday?
Linguistic Fieldwork Daniel L. Everett
page 109
What is going on in these examples? Well, they show that in addition to
definiteness, examples can be crossclassified by the notion of referentiality, that is,
whether they refer to a specific individual, i.e. one known to the speaker. In some
languages, this can be marked morphologically or syntactically. So consider the
following example from Persian (from the LDSQ):

(6.15) a. Hasan yek kita:b-ra: xarid

Hasan bought a specific book
b. Hasan yek kita:b xarid
Hasan bought some book or other

Both of these examples are indefinite. But the special marking, -ra on kita:b
'book' in (15a) is there to indicate that the example is referential, that is, that it refers to
a specific individual. Such distinctions, though 'only' semantic in English, can be vital
to understanding the morphology of some languages.
Another semantic distinction vital for morphosyntactic analysis is between major
participants of the designated verbal event or state:

6.3.1.2. GRAMMATICAL ROLES

GRAMMATICAL RELATIONS
Perlmutter & Postal () launched an extremely interesting and very influential
research program on grammatical relations, Relational Grammar, by explicitly adopting
the hypothesis that grammatical relations are syntactic primitives which are causally
implicated in numerous syntactic operations, especially those based on VOICE (see ___)
and INFORMATION STRUCTURE (see __). One of the appeals of Relational Grammar was
that its concept of grammatical relations and their relation to syntactic operations was
intuitively very appealing. Initially, at least, the theory also seemed to enjoy a great
deal of empirical success and conceptual simplification in handling numerous
operations that were much more clumsily handled in standard Transformational
Grammar, for example. Many linguists still find it convenient to talk about notions like
subject, direct object, indirect object, and so on. To see how such notions can be useful,
consider that the passive in English and other languages can be generalized to a rule
like ():

(6.16) Passive: Promote the direct object to subject position and demote the subject to
an oblique relation.

(6.17) a. John hit Bill.

b. Bill was hit by John (John has been 'demoted' to the oblique object of
by).

Unlike transformations that operate on syntactic categories, the use of

grammatical relations rather than categories such as NP, PP, etc. as the basis of
grammatical rules offers clearer and more satisfying cross-linguistic accounts of voice
alternations. The reader is urged to consult Perlmutter & Postal (), and the large
literature on Relational Grammar for further discussion of the usefulness of
grammatical relations in understanding syntactic processes.

MACROROLES
Linguistic Fieldwork Daniel L. Everett
page 110
An alternative to grammatical relations is proposed in Role and Reference
Grammar (Van Valin and La Polla ()), under the label of MACROROLES. Macroroles are
different from Grammatical Relations in that, among other things, they are linked
simultaneously to the semantic roles specified in a verb's lexical entry and the ways in
which the specific nature of these lexical roles are 'neutralized' in the syntax.
Macroroles overlap with Grammatical Relations in many simple cases, but they are by
no means isomorphic to Grammatical Relations.
For example, ACTOR is a generic term for all semantic roles that are associated
with causing or controlling, etc., an action in a specific clause. That is, it is the label for
the set of roles whose specific semantics are neutralized in a particular construction.
UNDERGOER Is the label for the neutralization of roles associated with arguments that
suffer the consequences of the action in some sense. Consider the following examples,
from Van Valin (), with the actor vs. undergoer roles labeled:

(6.18) a. The teacher is reading the words. Actor of transitive

V
b. The teacher is singing. Actor of intransitive
V
c. The teacher fainted. Undergoer of
intransitive
V

Example (6.18c) is interesting because it shows that subjects of simple clauses in

which there is no obvious dislocation can also be Undergoers, i.e. that subjects do not
map directly to Actors.

(6.19)
a.*The teacher are reading the words. *Undergoer of transitive V
b. The words are being read by the teacher. Undergoer of passive V

Example (6.19a) shows that Macroroles do not directly govern agreement in

English. In Relational Grammar this is because subjects, not macroroles govern
agreement in English. However, Van Valin and La Polla (1997) argue that Grammatical
Relations are really not what is needed to account for agreement and other apparently
Grammatical Relation-sensitive processes in syntax. Consider other ways in which
Macroroles may be manifested in the syntax.

(6.20) a. Chris wants to drink a beer. Actor of transitive V

b. Chris wants to sing in the park. Actor of intransitive V
c. Chris wants to be stronger. Undergoer of intransitive V
d.*Chrisi doesn’t want the journalist to interview ___i.*Undergoer of
transitive V
e. Chris doesn't want to be interviewed by the journalist. Undergoer of
passive V

This example is interesting because it shows that Macroroles can only govern
verb agreement or appear in a dislocated position if certain conditions on Voice are
met. The next examples show that English is quite liberal in the semantic roles that may
Linguistic Fieldwork Daniel L. Everett
page 111
head relative clauses. In many languages only the Actor or the Actor plus the
Undergoer roles can head relatives (all such cases are discussed in detail in Van Valin
(2005)).

(6.21) Mary talked to the man (a) who [A] bought the house down the street.
(b) who [U] the dog bit.
(c) to whom [recipient] Bill showed the house.
Mary looked at the box (d) in which [locative] the jewelry was kept.
(e) out of which [source] the jewelry had been taken.

Verbs have a semantic structure that, among other things, specifies for a given
verb how many arguments it needs and what the semantic relationship of those
arguments is to the verb.28 Languages rarely, if ever, mark these various lexical
relations directly in the morphology or the syntax (though cf. VVLP's treatment of
Acehnese (pp255ff)). More commonly, languages choose to 'neutralize' (VVLP
pp251ff) some number of these distinctions by grouping the more finely-grained lexical
argument distinctions under larger categories. Tagmemics (Pike & Pike (1976)) and
Role and Reference Grammar (VVLP), suggest the labels 'Actor' and 'Undergoer' as the
two basic 'macroroles' (VVLP, 139ff), under which the more specific semantic roles are
neutralized. These macroroles can have important consequences for a language's
mrophology, as seen in examples like (16)-(19), from Kulina, an Arawan language
(Wright 1988):

(6.22) Kodzo tshite i -na -hari.

lizard.M shoot 3on3 -AUX -COMP.M
'S/he shot the lizard.'

(6.23) Aoi dzoho i -na -haro.

tapir.F carry 3on3 -AUX -COMP.F
S/he carried the tapir.'

(6.24) Kodzo tshite ∅ -na -hari.

lizard.M shoot 3 -AUX -COMP.M
'He (*She) shot the lizard.'

(6.25) Aoi dzoho ∅ -na -hari.

tapir.F carry 3 -AUX -COMP.M
'He (*She) carried the tapir.'

Wright (1988) takes the examples with the i- '3on3' prefix (third person Actor
and third person Undergoer) ((16) and (17)) to be the basic/underlying forms. If one
makes this decision, then in the basic form there is agreement with the object. In the
'derived' form, the agreement is with the subject and the object no longer triggers
agreement. From this perspective, the i-/∅ alternation's effect on gender agreement
looks (somewhat) like an ergative/antipassive alternation (see __ below on ergativity
and antipassive). On the other hand, if one takes the ∅ -forms to be basic, and the i-

28
Arguments are expressions required to saturate a verb's valence (see ___ below).
Linguistic Fieldwork Daniel L. Everett
page 112
forms to be derived, then the structure will look more like a passive (this is the analysis
of Campbell and Campbell 1990 for Yamamadi (a language closely related to Kulina),
where hi- is the cognate of i-. It should be noted that that the hi-/i- prefixes only appear
when both subject and object are third person (any number). When either or both is
nonthird person, there is no special agreement prefix and only the gender alternation is
seen. Cf. (20) and (21):

(6.26) Kodzo tshite o -na -haro.

lizard.M shoot 1 -AUX -COMP.F
'I shot the lizard.'29

(6.27) Kodzo tshite o -na -hari.

lizard.M shoot 1-AUX-COMP.M
'I shot the lizard.'

In these examples the only visible difference is the gender of the verb. In (20) it
is governed by the subject and in (21) by the object.
For our purposes, the crucial observation is that Arawan has morphological
strategies for distinguishing different pragmatic functions of Actors vs. Undergoers. It
is not necessary to label them as either 'passive' or 'antipassive'. One way to think of the
alternations in Arawan, as well as related, but nonidentical, alternations in other
languages, e.g. passive and antipassive, is that a language can select either the
Undergoer or the Actor as the unmarked morphological (and pragmatic as well, in most
cases) argument. When the topicality of that unmarked argument is altered (by
becoming more or less topical), then the morphology will usually 'kick in' to signal
this.30

6.3.1.3. POSSESSION: alienable, inalienable, intermediate

Languages often distinguish between alienable vs. inalienable possession, i.e.
nouns that need not be marked as possessed in the morphosyntax and nouns which must
be so marked (or, alternatively, must have a special form to signal that they are
violating the normal expectation that they be possessed). Below I illustrate all of these
points from Banawá and Wari.
As in other Arawan languages, there are two main syntactic types of possession
in Banawá, alienable and inalienable. The former is subdivided into kinship vs.
nonkinship. All nonkinship alienably possessed NPs include the clitic ka 'possession'
occurring between the possessor and possessum. Kinship terms do not always take the
ka clitic and in such cases are similar to inalienably possessed forms. They are unlike

29
First and second persons and all plural forms are feminine throughout the Arawan
languages, regardless of the sex of the real-world referent.
30
In addition, Actors, Undergoers, and other semantic relations, can be further
neutralized by interposing another level of organization on the verb's arguments,
namely, grammatical relations, e.g. Subject, Object, etc. But not all theories recognize
these grammatical relations as having any usefulness to morphosyntactic analysis. Even
theories as otherwise distinct as Role and Reference Grammar and Minimalism reject
any role for grammatical relations in their view of grammar.
Linguistic Fieldwork Daniel L. Everett
page 113
inalienably possessed forms, however, in that the gender of a possessed kinship term is
determined by the kinship term rather than its possessor.

Alienable possession:
(6.28) Inaso ka yomai rabikei no, bede ka -wei.31
Ignacio ka dog.M sick.M past, run comit -cont.M
'Ignacio's dog got sick and ran away.'

Inalienable Possession
(6.29) kosiba efe wiri to -witi -ei
babassu:palm leaf strip away -end.point -M
The babassu palm leaf was stripped off to the end.'

In example (6.29), the gender agreement on the verb can be controlled by the
possessor (kosiba 'babassu palm'), but not the possessed item (efe 'leaf') and there is no
ka possession marker allowed. These are two of the criteria for recognizing inalienable
possession in Banawá. Banawá also shows a somewhat intermediate category of
possession for kinship:

Kinship possession:
(6.30) oda ka ami -rawa oda da -daba -bisa towa -maro
1P ka mother -pl.F 1P red -accomp -also be -past
'We stayed with our mothers also.'

(6.31) ami da -daba -bisa towa -maro

mother -pl.F red -accomp -also be -past
'You stayed with your mother also.'

In example (25), the possessor of ami 'mother' is omitted, but trivially

recoverable because ami is the suppletive term for second person's mother. With third
person kinship terms (also marked by suppletion in Arawa) following Campbell (), the

31
In personal communication, Alan Vogel suggests that the crucial portions of the
examples in ()-() in the text should in fact be analysed as below:
() bede kawei
bede ka- na -wei
run COMIT- AUX -now+M

() wiri towitiei
wiri to- na -witiei
strip.off away- AUX -out+M

() towamaro
to- ha -maro
INCH- AUX -past

Vogel may well be right. But this relatively abstract analysis has not yet been
demonstrated to my satisfaction.
Linguistic Fieldwork Daniel L. Everett
page 114
possessor of the kinship term can be omitted if it is recoverable from the discourse,
unlike normal alienable possession, but like inalienable possession. Thus kinship can
occasionally behave intermediate between the two types. Everett (in progress) accounts
for this in terms of qualia (see ___ below).
Kinship terms and possession in Wari' are discussed in detail in Everett and
Kern (1997:434ff.). There are two formation classes of kinship terms in Wari’. The first
includes what Everett and Kern (1997) call -xi’ nouns, where -xi’ is the first-person
plural inclusive possession suffix. Only inalienably possessed nouns take possession
suffixes and these in turn can be inputs to other word-formation rules, as discussed in
Everett and Kern (1997:235ff.). In these forms, the citation or base for the paradigm is
first-person plural inclusive. The first-person singular inflection of six of the terms is
suppletive; otherwise they all inflect for possession with the –xi' suffix. The complete
list of –xi' nouns, all in their ‘1pl inclusive’ forms, is given below (first-person singular
alienably possessed forms are in parentheses).

Simplex:
(6.32) a. ’aramanaxi’ (‘aramana) ‘sister’ (lit., female of the species; male ego)
b. cainaxi’ (caina) ‘daughter’ (female ego)
c. cawinaxi’ (cawiji ) ‘son’ (female ego)
d. humajixi’ (humaju) ‘children’ (female ego)
e. japinaxi’ (wijapi) ‘wife’s mother’
f. jinaxi’ ( jina) ‘granddaughter’ (male ego or female ego’s daughter’s daughter)
g. manaxi’ (mana) ‘wife’ (‘hole’; mana ‘my wife/my hole’)
h. namorinaxi’ (namori) ‘wife’s sibling’
i. tamanaxi’ (tamana) ‘husband’s mother’
j. taramajixi’ (taramaju) ‘brother’ (male of the species; female ego)
k. taxixi’ (taxi) ‘husband’
l. wijinaxi’ (wiji ) ‘grandchild’ (female ego’s son’s child)
m. winaxi’ (wina) ‘grandson’ (male ego or female ego’s daughter’s son)
n. xerexi’ (xere) ‘sibling’
o. xinaxi’ (wixi) ‘sister’s son’ (male ego)
p. xi’ (na) ‘mother’

The only way to understand how '–xi' nouns' differ from other nouns semantically
and morphologically in Wari' is to recognize the (non-universal) semantic distinction of
inalienable possession (note that it is not necessary to understand the semantic basis per
se as to why a given noun is an inalienably possessed noun. This can be helpful, but for
diachronic reasons can be fairly asystematic).
Moreover, even languages that do not have inalienable possession
morphologically can show traces of it in their syntax, as the following examples from
Portuguese show:

(6.33) Joao operou a perna.

(i) Johni operated on the or hisi/j leg.
(ii) John's leg was operated on.

In this example, quite common in Brazilian Portuguese, the 'passive' reading in

(6.33ii) is available for part-whole relations, the kind that are often the basis for
inalienable possession. Consider also:
Linguistic Fieldwork Daniel L. Everett
page 115

(6.34) O carro furou o pneu.

'The car's tire was punctured' (literally, 'The car punctured the tire.'

This example, also common in Brazilian Portuguese, is possible because pneu

'tire' is part of carro 'car'. Everett () explores these constructions in detail. The
relevance of part-whole/inalienable possession relations to the syntax of such
constructions is shown by ungrammatical examples like ():

(6.35) *A casa roubou o carro.

'The house was robbed of the car.'

Although the syntax of this example is the same as (6.34) (perhaps coming from
an underlying form e.g. ___ roubou o carro da casa, again, see Everett () for details),
the example is ungrammatical because () does not involve a part-whole relation (cars
are not parts of houses).
Notice too that wherease inalienable possession shows up (mainly) in the
morphology of Wari' and Banawá, it appears in the syntax of Brazilian Portuguese.
This again shows how semantic distinctions should not be confused with their modes of
expression. Form is not function.
From the semantic category of possession, let's move to consider the category of
number.

6.3.1.4. NUMBER: singular, plural, intermediate

Apparently all languages of the world, except Pirahã (Corbett (2001)), mark
number morphologically. The most common distinction is singular (one entity) vs.
plural (more than entity). But there are many languages that make intermediate number
distinctions, e.g. 'dual' (two) and 'paucal' (3-5 or so), in addition to plural vs. singular.
Therefore, when eliciting number, be sure to test for all these categories, especially if
you already know that other languages of the area or the family make such distinctions.
You can do this fairly simply by eliciting phrases, e.g. 'One man came'; 'Two men
came'; 'Three men came'; etc. plus 'I saw one man'; I saw two men'; 'I saw three men';
up to five at least (be sure to collect numbers in at least the subject and object
positions). This will cover most of the possibilities. However, once again, always 'keep
your guard up' for distinctions that didn't show up in initial, simple tests, but which
might appear in more subtle ways, i.e. elsewhere in the phrase, as you progress in your
understanding of the language.
Another important, number-based semantic distinction made in most languages is
the distinction between count vs. mass nouns, i.e. nouns that can be counted and nouns
that cannot be, at least grammatically. So consider English examples like 'Some water'
vs. *'Two waters' (where 'water' refers strictly to the subtance H20, and not some
standardized forms for containing or channelling it, e.g. cups or rivers) or 'Two arrows'
vs. *'Some arrow'. Although linguists almost universally attribute these distinctions to
countability, in my view, they may arise from something even more basic than
counting, namely, visual perception. For example, the Pirahã language of Brazil
(Everett 2005) lacks number, numerals, and counting, yet it apparently has a 'count' vs.
'mass' distinction. This could be evidence for the hypothesis that 'count' vs. 'mass' is a
distinction of some 'universal grammar', so that it is found even if a language lacks
means of expression for the related cognitive faculty of counting. But the simpler
Linguistic Fieldwork Daniel L. Everett
page 116
hypothesis is that this derives from the fact that some entities are easier to perceive as
made up of smaller units (i.e. they can be 'individuated'), whereas others are not so
obviously decomposable visually. Thus it is possible that the distinction here has
nothing to do with universal properties of grammar but, rather, with visual perception.
The point is that although the fieldworker should be aware of distinctions drawn in a
huge percentage of the world's languages, they must always exercise creativity,
thinking 'laterally' to use a term from de Bono (1973).
The next important semantic distinction languages often make in one way or
another on nouns (or their containing phrases; see section ___ below on ways of
marking semantic distinctions) is class membership.

6.3.1.5. CLASS MEMBERSHIP: gender and more

The classes into which languages may divide their nouns are likely to have some
sort of semantic motivation, at least diachronically. But in general, this motivation is
obscured, perhaps completely, by normal historical change of the language.
The simplest class distinction is masculine vs. feminine. Languages that use this
simple class system, as well as those that include a third class of 'neuter', are said to
have gender systems. This is fine as a label, but the linguist should not forget that this is
nothing more than a simple class system and that it is absurd to think that speakers of a
language with 'gender' actually see every object in their universe as male or female. A
number of objects in each linguistic gender class will indeed have a link to male or
female in the real world (e.g. 'man', 'woman', 'girl', etc.) but certainly not all. For
example, in Brazilian Portuguese, one says A cola 'The (feminine) cola', but O guarana
'The (masculine) guarana', where in this case guarana and cola are just two forms of
softdrink. No Brazilian I have ever talked to thinks that cola is female while guarana is
male in the real world. More complex classification systems have less obvious semantic
unity in each class, by and large, and it is almost always a waste of time (i.e. a violation
of the 'law of diminishing returns') to look for semantic motivation for each member of
each class. What the field linguist really needs to understand is how these classes, such
as they are, are causally implicated in the morphosyntax of their containing clauses and
discourses. This is not to say that it will always be pointless to look for the semantic
motivation behind class membership, not at all. But it is the case that looking for such
motivation is less likely to be useful to initial fieldwork than carefully describing how
the classes work structurally.
Class is often marked as a separate morphological category. For example,
consider how gender can be marked in the Arawan language, Banawá. Banawá is
interesting in that person and number agreement are marked on the rightmost boundary
of the verb, i.e. as prefixes or proclitics, whereas gender is marked at the end of the
verb, as a suffix. Moreover, gender agreement is not directly governed by either the
subject or the object, but by the topic of the sentence which can be subject, object, or
the inalienable possessor of the subject or object (see Everett (in progress). Dixon
(2004) discusses similar facts in the related language, Jarawara).
We saw some examples of Arawan gender marking earlier from Kulina (()-()).
When the Actor is the topic (what Dixon () refers to as the A-construction), the Actor
governs gender agreement on the verb:

(6.36) a. kobaya mere waka -nei

pig wild:M 3P.OBJ break -AUX:M
'He killed pigs.'
Linguistic Fieldwork Daniel L. Everett
page 117

b. oi -mank wete ∅ j -ka -seij

1S.F -arm.M tie.up 3 -comitative -aspect.M
'He tied up my arms.'

When the Undergoer is the topic (Dixon's () O-construction), it controls gender

agreement:

(6.37) dee ka -maki yomai owa nabowa bone32

2P MOT -follow:F jaguar:M 1S.OBJ kill INTENT:F

awa -tee o -ka abi ati

maybe -RECENT.PST.QUOT+F 1S -POSS father say

-nei
-AUX:M

'You all come here. The jaguar will kill me, my father said.'

But, just as with Kulina, when the Actor or the Undergoer is inalienably
possessed, its possessor controls gender agreement:

(6.38) Mate, mate ime yama hi- kaba metemone

buttock:F buttock:F flesh:F thing:F 3- eat long.ago:F
mate
buttock:F

hodi wa -metemone, mate tone wa

hole:F stand -long.ago:F buttock:F bone:F stand -

-fara -metemone.
in.open.area -long.ago:F
'The flesh of her upper thigh was being consumed by something, it was open,
and her thigh bone was exposed.'

When the Undergoer is treated as topic and where the Undergoer is inalienably
possessed, the possessor of the Undergoer, rather than the Undergoer itself, governs
gender agreement:

Possessor (of Undergoer) controls gender agreement:

(6.39) oi -manoj wete hi -ka -sai.
1S.F -arm.M tie.up 3/3 -comitative -aspect.F
'My arms were tied up by him.'

32
Vogel (p.c.) also says that bone 'intent feminine' should in fact be bona 'intent
masculine'. But my source is fairly clearly bone. I suspect that Vogel is right here,
knowing far more than I about these languages, but I leave it this way just the same.
Linguistic Fieldwork Daniel L. Everett
page 118
These examples are very useful because they show the interaction of possession
types, voice alternations, and gender. Each of these semantic distinctions and the way
that each is marked must be carefully analyzed or their interactions will be mysterious.
Everett () argues that the simplest analysis of the facts is that gender agreement is
pragmatically governed, whereas person and number agreement are syntactically
governed. Pragmatic government of the process requires the whole rather than the part
to be the topic. That is, to consier (), you cannot tie up my arms without affecting me. It
is 'my' that is the topic, therefore, not 'arms'. The point relevant to this guide is that
CLASS is subtle, has effects throughout the clause, and care must be taken in its
analysis.

6.3.1.6. PRAGMATICS
Arguably, transitivity alternations in languages have the function, among others,
of tracking the pragmatic functions of nominals.
It is useful for languages to mark such alternations. We have already seen
examples of this from Salish (() – ()) and Arawan (() – ()). The point of this section is
simply to remind the fieldworker to look for a variety of possible correlations between
pragmatics and morphology. And the fieldworker must be aware that there are also
important morphology-pragmatics connections unrelated to information structure, e.g.
politeness markers and honorifics. So consider the examples below from Brazilian
Portuguese and Japanese:

(6.40) Tu vai(s) para o cinema hoje, rapaz?

'Are you (most informal) going to the cinema today, fellow?

(6.41) Voce vai para o cinema hoje, rapaz?

'Are you (informal) going to the cinema today, fellow?'

(6.42) O senhor vai para o cinema (, % rapaz)?33

'Are you (formal) going to the cinema?'

Generally dialects of Brazilian Portuguese differ as to whether the informal usage

is expressed by tu or voce. However, in some dialects, there is a three-way distinction,
with (34) least formal and (36) most formal. Although not all dialects of Brazilian
Portuguese that use tu mark the verb differently (hence the parentheses on the s in ()),
some do. These examples thus show that politeness forms can affect word-formation.
In Japanese, on the other hand, there is a much more elaborate pragmatic-based
system, with the forms referred to as honorifics (Potts & Kawahara (2004, 1)):

Japanese Honorifics
(6.43) a. Sam -ga warat -ta.
Sam -NOM laugh -PAST
‘Sam laughed.’

b. Sam -ga o -warai -ninat -ta

Sam -NOM subj.hon -laugh -subj.hon -PAST

33
% indicates that the addition of this word would be pragmatically strange.
Linguistic Fieldwork Daniel L. Everett
page 119
i. ‘Sam laughed.’
ii. ‘The speaker honors Sam.’ [subject honorific]

c. Sam -ga warai -yagat -ta

Sam -NOM laugh -antihon -PAST
i. ‘Sam laughed.’
ii. ‘The speaker views Sam negatively.’ [antihonorific]

The authors go on to illustrate how honorific marking can be used to mark

speaker attitude not only to other individuals but to the content of propositions as well
(p2ff):

(6.44) Kathyrn -wa Sam -o hai -ken -shi -nakat -ta

Kathryn -TOP Sam -ACC obj.hon -see -do -not PAST
i. ‘Kathryn did not see Sam.’
ii. ‘The speaker respects Sam.’ [object honorific]

"We divide our attention between these cases and the honorifics represented in
(39) and (40)[my numbers, DLE], which associate semantically with some aspect
of the propositional content rather than with the denotation of an argument
nominal."

(6.45) Mary -ga ringo -o tabe -mashi -ta

Mary -NOM apple -ACC eat -perf.hon -PAST
i. ‘Mary ate the apple.’
ii. ‘I am speaking nicely to you.’ [performative honorification]

(6.46) John -wa [Mary -ga nesugoshi -chimat -ta] -koto-o

John -TOP Mary -NOM oversleep -antihon -PAST –fact

shitteiru.
know
i. ‘John knows that Mary overslept.’
ii. ‘It sucks that Mary overslept.’ [antihonorification]

As they continue:

"The first is called performative honorification in Harada (1976) (certainly a

suggestive name). It goes by the name ‘polite speech’ as well. We call the second
kind antihonorification. Here, the morpheme chimau, which appears as chimat-,
signals that the speaker has contempt for the proposition expressed by the clause
in which it appears."

The Japanese case thus illustrates that pragmatic markers like this can be 'co-
opted' for other grammatical functions. Therefore, such categories are doubly important
to recognize, with potential ramifications that go beyond their original functions.

6.3.1.7. CASE
Linguistic Fieldwork Daniel L. Everett
page 120
Verbs are said to 'govern' cases, which means that in some languages verbs
partially determine the shape of the nouns that are specified in their lexical
representation. English has vestiges of this in its pronoun system:

(6.47) a. I saw him.

b. He saw me.
c. *Me saw he.

Of course, anyone who has studied Latin, Greek, Finnish, or any one of a huge
number of languages will have learned about case systems much more complex than
English. Case is an interesting feature because it is in a sense 'displaced'. It is perhaps
best understood as a verbal feature that is borne by nouns. That is, what the case will be
on a noun depends on what the verb is. But the verb itself does not bear case. In
Minimalism (Chomsky 1995), case is one of the 'uninterpretable' features. But these
features in Minimalism seem to be exactly those features that are displaced in this
specific sense. To understand case in a language requires a solid understanding of
verbal lexical structure.
Case is functionally straightforward. It is used to help interlocutors track actions
and their participants. Consider the template for a simple transitive clause (order
irrelevant):

(6.48) S V O

Now in a language like English, with rigid word order, the use of additional word
markers (case affixes or prepositions) to indicate that one NP is Subject or Actor and
the other is Object or Undergoer is unnecessary. A language could distinctively mark
these participants but it does not need to. On the other hand, if a language allows its
nominals to be ordered more freely than English, a morphological marker on one or
both arguments of the verb becomes very useful. Marking just one of them is more
efficient (a lesser effort to results ratio) than marking both, although all possible options
are found in natural languages. The three possibilities are shown in (43):

(6.49) a. Smark V O
b. Smark1 V Omark2
c. S V Omark

The system in (6.49a) is called 'ergative' or 'ergative-absolutive'. The case that is

marked on the subject is called the ergative case and the unmarked case on the object is
called the absolutive case. The case marked on the object in (6.49c) is called the
accusative case and the unmarked case on the subject is referred to as the nominative
case. The (6.49b) example is rarer, because less efficient, but can fall into either
ergative or accusative patterns or yet a separate kind of pattern. Part of knowing how to
interpret (6.49b) is learning how the language marks the single argument of an
intransitive clause:

(6.50) a. Smark V
b. S V
Linguistic Fieldwork Daniel L. Everett
page 121
Marking of the sole argument of an intransitive clause is unnecessary, so (b) is the
most common case. When coupled with (c) the null case on the S is called nominative.
When coupled with (a) it is called the absolutive. When coupled with (b), if the case of
the intransitive S matches the case of the transitive S, we have a nominative-accusative
system. If it matches the case of the O, then the system is ergative. If it matches neither,
the system is neither. (But again, labels are not all that crucial. Clearly presented,
accurate analyses and descriptions are.)Variations are possible, the two most common
being split-ergativity and active-marking.34
In a split-ergative system, the case marking is nominative-accusative for some
purposes and ergative-absolutive for others. From our definition in () below of
prototypical transitivity, it is not surprising to learn that the dividing line is often
punctiliar past and/or definiteness of the arguments. In general, in split-systems,
ergative marking is more common with the punctiliar past and nominative-accusative
with deviations from prototypical transitivity. But there are other ways to divide the
case-marking system. So beware. One final observation is that languages do not have to
exploit case-marking to show at least traces of ergativity. In Wari' for example (Everett
& Kern (1997, 331ff), many verbs take a suppletive form when the Object or the
intransitive Subject is plural. Since those two are grouped together by this process, the
process is ergative in a sense. Or consider (from Lyons ()) the vestiges of ergativity in
English:

(6.51) a. John ate breakfast.

b. John ate.
c. *Breakfast ate.

(6.52) a. John broke the window.

b. The window broke.
c. *John broke.

Verbs like break and eat in English, as in many other languages appear in
transitive and intransitive clauses. However, when they do, the eat verbs allow the
Actor to appear as their sole argument, never the Undergoer (c), while the break verbs
allow only the Undergoer as their sole object, never the Actor (c). Therefore, the break
verbs group together the Subject of the intransitive form with the Object of the
transitive form and, as Lyons () so rightly pointed out, they are ergative verbs in this
sense.
Another kind of distinction, the last we consider for now, is what some call an
active-stative case system. Consider the hypothetical examples in (47):

(6.53) a. Johnmark fell down (on purpose/*accidentally).

b. John fell down (*on purpose/accidentally).

What is going on in such examples? In this type of system, volitional Actors

(acting intentionally) are case-marked while non-volitional Actors (things happen to

34
The best crosslinguistic typological comparison is WALS, Haspelmath, Martin &
Dryer, Matthew & Gil, David & Comrie, Bernard (eds.) 2005. The World Atlas of
Language Structures. (Book with interactive CD-ROM) Oxford: Oxford University
Press. ISBN: 0-19-925591-1.
Linguistic Fieldwork Daniel L. Everett
page 122
them without their intention), are not case-marked. Such systems of case-marking are
called active-stative.
The key to case analysis is to be alert for ways that nominals are being grouped
together. This is rarely arbitrary (modulo vestiges of some historical process now
almost gone from the language). Labels are not that important initially. Recognizing
patterns is.
Let's move now to another important semantic domain with morphosyntactic
consequences for nominals, qualia.

6.3.1.8. QUALIA
Morphosyntactic research must also pay attention to a semantic set of categories
used in classifying nouns in subtle ways. Pustejovsky (1996, 76ff and 85-105) discusses
these in a detailed and helpful manner, labeling them qualia. As Pustejovksy (85)
explains these, the qualia are the four factors which 'drive our basic understanding of an
object or a relation in the world. They furthermore contribute to (or in fact determine)
our ability to name an object with a certain predication'.
What this means for the fieldworker is that until they at least roughly understand
the qualia of a particular lexical item, they cannot fully understand that item's
appearance in sentence structure. This is relevant for morphosyntactic field research
because subregularities in word behavior, both in verb and nominal or NP structures
(e.g. selection of determiners, numerals, and, occasionally, affixation), can be
determined by the qualia of a given item. There are four qualia, according to
Pustejovsky (85-86).

(6.54) "1. CONSTITUTIVE: The relation between an object and its constituents, or
proper parts.
i. Material
ii. Weight
iii. Parts and component elements

2. FORMAL: That which distinguishes the object within a larger domain.

i. Orientation
ii. Magnitude
iii. Shape
iv. Dimensionality
v. Color
vi. Position

3. TELIC: Purpose and function of the object.

i. Purpose that an agent has in performing an act.
ii. Built-in function or aim which specifies certain activities

4. AGENTIVE: Factors involved in the origin or 'bringing about' of an object.

i. Creator
ii. Artifact
iii. Natural Kind
iv. Causal Chain"
Linguistic Fieldwork Daniel L. Everett
page 123
For example, it is possible that qualia can be reflected in noun classification. It is
likely that qualia can cause some nouns to behave in otherwise confusing ways, so that
they would group with some nouns for some purposes according to qualia, but with
others, according to, say, gender/class. Consider, for example, the following examples
from Brazilian Portuguese:

(6.55) João usou uma faca nova no frango.

'John used a new knife on the (cooked) chicken.'

(6.56) João usou um terçado novo no leitão.

'John used a new machete on the (whole cooked) pig.'

Because of what faca 'knife' and terçado 'machete' mean at the level of qualia, we
can use a generic verb like usar 'to use' and still know that cutting/chopping is being
talked about. If a language morphosyntactically distinguishes, say, an aspectual
difference for cutting vs. other actions, then () and () could function identically with
regard to aspect, as a result of the qualia of the nouns involved, yet fall into separate
classes with respect to gender (as in this case, feminine vs. masculine, respectively).35
Or consider an analysis of the Banawá possession facts in terms of qualia. As
Pustejovksy (1995, 79) observes, "The CONSTITUTIVE (henceforth CONST) quale refers
not only to the parts or material of an object, but defines, for an object, what that object
is logically part of, if such a relation exists." (p98)

To illustrate let us use Pustejovky's (1995, 99) lexical structure for hand:
(6.57)
hand
ARGSTR = [ARG1 = x:limb ]
QUALIA = FORMAL = x
CONST = part_of(x,y:body)

"That is, the relation in the CONST allows for reference to what something is
constituted of as well as what it constitutes, in part; i.e., a hand is part of a
body, and a body has a hand..." (Pustejovksy (1995, 99))

My conclusion is that topic-agreement (gender) in Banawá picks out the index of

the whole (body in (51)) in the qualia structure for its referential index. For inalienable
possession this will be the possessor (that is the whole) but for alienable possession it

35
This has been useful to me in my on-going analysis of Banawá possession. Topic-
agreement (gender) picks out the index of the whole in the qualia structure for its
referential index. For inalienable possession this will be the possessor (that is the
whole) but for alienable possession it will be the possessed item (because that is the
whole). Moreover, this allows us to say why kinship is intermediate. It is like
inalienable morphologically (the cases without ka(a)) but like alienable syntactically (in
terms of agreement). The latter fact is because the possessor of a kinship term is not the
whole. The former fact is because it IS inalienable possession without ka(a). In other
words, there are two things happening: (i) inalienable vs. alienable possession and (ii)
part-whole relations. Each is important in different ways.
Linguistic Fieldwork Daniel L. Everett
page 124
will be the possessed item (because that is the whole). Moreover, this allows us to say
why kinship is intermediate. It is like inalienable possession morphologically (the cases
without ka) but like alienable syntactically (in terms of agreement). The latter fact is
because the possessor of a kinship term is not the whole. The former fact is because it
IS inalienable possession without ka. In other words, there are two things happening:
(i) inalienable vs. alienable possession and (ii) part-whole relations. Each is important
in different ways.
Other semantic distinctions may be less 'deep' or not inherent. One such property
is comparison.

6.3.1.9. COMPARISON: more, less, better, best, etc.

Comparison is a ranking of objects relative to one another and is thus not inherent
in a single object, by and large, as qualia are. Not all languages have comparison as a
morphosyntactic category, but most do. So consider English:

(6.58) a. A good man.

b. A better man.
c. The best man.

Not only is comparison reflected on the modifier of the English NP, choices of
comparison (e.g. (c)) can also affect other characteristics of the NP, e.g. marking for
definiteness. Comparison, like other semantic categories, may be marked on the N, NP,
Verb, or Clause (section ___ below). Be aware that regularities on one word may be
caused by or associated with another word. Therefore, it is vital once again that
phonology not be confused with morphology (see ___ above). Let's move now to
consider a different type of semantic categorization which can affect the morphology of
a given language, namely, possible distinctions for different classes and subclasses of
words, e.g. common nouns, proper nouns, pronouns, abstract nouns, concrete nouns,
quantifiers, number words, etc. All or any of these might play a significant role in the
morphosyntax of the language under study. We consider just a few possible subclasses
of nouns here.

6.3.1.10. NOUN SUBCLASSES

PROPER NOUNS:
Proper nouns are those which name specific individuals, e.g. John, Fido, Bank
of America, Brazil, and so on. Morphosyntactically they usually do not cooccur with
definiteness marking (since they are inherently definite), plurals, or much other
morphology usually associated with common nouns.36 And they behave differently in
discourse in some if not all languages. Recognizing them, however, is not always as
easy as it sounds. In Pirahã, for example, all proper nouns are morphologically complex
or derived from common nouns. Here are a few:

(6.59) a. /ahoaógií 'big night'

36
But consider Brazilian Portuguese examples like ():

(i) Vamos na casa do João.

'Let's go to the John's house.'
Linguistic Fieldwork Daniel L. Everett
page 125
b. Kohoibiíhai 'blood eater'
c. Toisií 'eagle'
d. Piihoábigí 'frog' (literally: 'water-side-dirt')
e. Piiopí 'capyvara' (literally: 'enter water')

The first proper noun in (53) is a modified noun, the second a nominalized verb +
object, the third comes from a common noun, the fourth from a phrase which is most
commonly used as a common noun, and the last from a verb + goal, and is used
commonly as the noun for capyvaras. In other words, there is little in Pirahã to
distinguish proper nouns from other nouns morphosyntactically, at least in isolation.
Only in context can proper nouns be distinguished (mainly by syntactic position and the
type of clitic used preverbally), as in (54):

(6.60) a. Piiopí hi /iaagá.

name masc. is hungry
'Piiopí is hungry.'

b. Piiopí /ísi /iaagá.

animal
'The capyvara is hungry.'

The proclitic on the verb tells us whether the noun is a proper noun (referring to a
person named capyvara), as in (54a) or a common noun (a capyvara) as in (54b).

PRONOUNS, CLITICS, AND AFFIXES

The first caveat offered at the beginning of this chapter is that the fieldworker
must be on guard to avoid confusing phonological facts with morphological facts. This
is perhaps nowhere truer than in the analysis of clitics, pronouns, and nominal affixes.
To get an idea of why this is hard, consider the following examples from Wari' (Everett
()):

(6.61) Quep na -in xirim te pane ta.

do 3s:rp/p -3n house father:1s rem:past emph
'My father made a house long ago.'

(6.62) Ten ta wao'.

weave pass:3s type of basket
'Baskets are woven.'

(6.63) Mi' non -on con hwam hwijima' mon tarama'.

give 3p:rp/p -3pm prep:3sm fish children coll man
'The men gave the children fish.'

The italicized material almost always is found to the immediate right of the verb,
as in the examples shown. Should this material be analyzed as a sequence of pronouns,
as clitics, or as suffixes on the verb? Such decisions are non-trivial and require
consideration of syntactic, morphological, phonological, and, occasionally, semantic
evidence. In the case of Wari', we conclude that these are clitics for three main reasons:
(i) Vowel Harmony operates within words, but is blocked from spreading from a verb
Linguistic Fieldwork Daniel L. Everett
page 126
to an agreement marker, i.e. across a clitic boundary; (ii) stress falls on the last syllable
of each word, but clitics are never stressed as part of any adjacent word, receiving
either no stress or an independent stress in clitic 'clusters' (as in ()-()); (iii) these
constituents may follow embedded sentences, which are not words, and thus can appear
without any potential word for them to affix themselves to (Everett (in progress)).
Perhaps the best summary of the differences between clitics and affixes is given in a
well-known article by Arnold Zwicky and Geoffrey Pullum (19--). Another useful
discussion on distinguishing clitics, affixes, and pronouns (or words more generally)
comes from the article by Anderson and Zwicky () in the second edition of the Oxford
International Encyclopedia of Linguistics:

"Since the unusual properties of clitics ... are bound up with their ambiguous
status between affixes and words, we may consider some criteria which distinguish
affixes (determined, bound, reduced) from words (undetermined, free, full):
(a) The typical word, but not the typical affix, has an independent accent.
(b) The phonological shape of a word must be listed in the lexicon, but the phonology
of an affix is described in general by saying how the shape of some stem is
altered (so that affixes can have ‘process’, as well as affix, realizations) [See
Affixation].
(c) Separate, language-specific restrictions can govern the possible phonological
shapes of words vs. affixes. In particular, affixes, but not words, are often non-
syllabic.
(d) Syntactically, words belong to (lexical) categories, i.e., word classes; but the
assignment of affixes to such categories is problematic...
(e) Syntactic rules introduce word classes as co-constituents with other syntactic
categories; but an affix is syntactically dependent, described by rules which
locate it by reference to syntactic elements (e.g., on Nouns, on the head of VP, in
the first word of S, at the right edge of NP).
(f) For each affix, morphological rules specify the class of words with which it can
occur, and the properties of the resultant combination; but the syntactic rules
distributing words typically make reference to phrasal categories rather than to
word classes. From this, it follows that affixes are typically very selective in the
word classes with which they occur, but words are unconstrained with respect to
the word classes that happen to occur adjacent to them.
(g) Syntactic rules cannot alter morphological structure. In particular, syntactic rules
cannot allow a word to interrupt a stem + affix combination; a word attached to
such a combination must have edge position.
(h) Syntactic rules which introduce a lexical category are blind to the morphology
and phonology of its co-constituents, but rules which introduce an affix may be
contingent on such properties of its stem. From this it follows that there can be
arbitrary gaps and morphophonological idiosyncrasies, including suppletion
[q.v.], in the set of stem + affix combinations, but not in the set of word + word
combinations.
(i) Alternative orders of words within a constituent are common, but the ordering of
an affix – with respect to its stem, and to other such affixes – is fixed (although
the same affixes may combine in different orders to express different meanings:
e.g., the passive of a causative is not the same as the causative of a passive, and
the affixers involved may reflect this in their ordering)."
Linguistic Fieldwork Daniel L. Everett
page 127
The question here, of course, is how do we formalize such observations and turn
them into rules of thumb for fieldworkers? First, consider the basic structures
distinguishing clitics from affixes proposed in Everett (1997, PAGE ) (in these examples
()s = phonological boundaries and []s = morphological boundaries):

(6.64) a. Y is a free word ([X]) ([Y])

b. Y is a clitic ([X] [Y])
c. Y is an affix ([X [Y]])

Affixes are morphologically parts of words. Clitics are phonologically dependent

(though their placement may also be syntactically constrained, e.g. Everett (1989)
argues for Yagua). Free words are independent in both senses. The problem is how to
determine the type of dependency, morphological or phonological. Thus, by the view
represented in (58), the terms 'clitic' and 'affix' are nothing more than labels for
morphophonological relationships. It is those relationships which are crucial for the
field researcher. Zwicky and Anderson's discussion offers very useful considerations
for this.
The three questions the fieldworker must answer, put in a very general way, for
affixes, clitics, and free words are:

(6.65) a. Does the item behave like a free word?

b. Or does it behave like an affix?
c. Or does it have mixed qualities, e.g. behaving like a word syntactically
but like an affix phonologically?

Put in this way, it is clear that correctly understanding a particular item as a clitic,
affix, or free word requires knowledge of linguistics, the language, and, especially, how
to define a word in the language under study. Let us turn now to consider another
subset of nominals/nominal modifiers (depending on the language).

NUMERALS AND QUANTIFIERS

There are also special classes of nouns that can vary a good deal from language to
language as to whether they have idiosyncratic or relatively normal noun behavior.
These include, but are not limited to, the following: number words, quantifiers, abstract
nouns, and color words.
To take number words as the first example, consider the following examples from
Russian (Rappaport (2002, 329)):

"Numeral phrases in Russian have long presented a puzzle to those attempting a

theory of the syntax of this language. The puzzle itself is not difficult to formulate.
When the phrase as a whole stands in a position to which the syntax assigns a
direct case (nominative or accusative), the numeral acts like a nominal head of
the phrase, assigning the genitive case to the quantified noun (and its modifiers)
as if it were a complement. When this same phrase stands in a position assigned
an oblique case, the quantified noun acts like the head of the phrase, with the
numeral falling into line with the other modifiers in agreeing with that quantified
noun. For example, contrast (60a,b)[my numbers, DLE]:
Linguistic Fieldwork Daniel L. Everett
page 128
(6.66) a. videt' pjat' krasivyx ptic&ek.
to see fiveACC=NOM beautifulGEN.PL birdsGEN.PL
‘to see five beautiful birds’
Linguistic Fieldwork Daniel L. Everett
page 129
b. vosxis&c&at'sja pjat'ju krasivymi ptic&kami.
to-be-enthralled-by fiveINST beautifulINST.PL birdsINST.PL
‘to be enthralled by five beautiful birds’"

Or consider the English quantifier all:

(6.67) a. All the children came today.

b. *The all children came today.
c. The children all came today.
d. *The children came all today.
e. *The children came today all.

The English quantifier, all, 'floats', that is, it can appear in different positions in
the phrase, unlike other words. In this case, it shows no morphological irregularity,
though its syntax is clearly linked to its morphological class. Therefore it is important
in classifying words, part of morphological analysis, to determine what special features
they have phonologically, morphologically, or syntactically.

6.3.1.11. NOMINAL TENSE

Fieldworkers will expect certain semantic categories to be marked on a closely
corresponding set of syntactic categories for the most part. For example, it is natural to
expect tense to be marked on verbs because a principal function of tense is to situate
events, expressed almost always by verbs, in time. On the other hand, tense is not marked
exclusively on verbs crosslinguistically. For example, a word like ex-wife in English
arguably marks the tense of 'wife' as past by the prefix ex-. Nordlinger & Sadler () have
studied the distribution and typology of nominal tense extensively. They say that:
"Although the categories of tense, aspect and (to a lesser extent) mood are
traditionally considered to be properties of verbs, the morphological expression of tense,
aspect and/or mood (henceforth TAM) on nominals is attested across a range of
languages. We distinguish two major subcases of nominal tense, that is, of cases in
which TAM marking occurs on a nominal or other constituent of NP/DP. In some cases,
a dependent nominal or nominal phrase (of whatever grammatical function) bears some
TAM marking which serves to temporally, aspectually or modally specify the clausal
predicate which is itself distinct from the nominal argument. This is illustrated by
examples (62)-(63)[my numbers, DLE]) in which the case marking of the dependent NPs
changes to reflect the tense (future vs. nonfuture) of the clausal predicate:
(6.68) Ngamari -lu ngunytyi -ka
mother -ERG give -PAST

ngali -nha mangarni -marru -nga -nha

we.DU -ACC bone -having -GEN -ACC

kathi -nha.
meat -ACC.

'Mother gave us the doctor’s meat.' (Pitta Pitta (Australia), Blake 1987:60, ex.
4.12)
(6.69) Ngamari -ngu ngunytyi ngali -ku
mother -NOM.FUT give we.DU -ACC.FUT
mangarni -marru -nga -ku
bone -having -GEN -ACC.FUT
Linguistic Fieldwork Daniel L. Everett
page 130

kathi -ku.
meat -ACC.FUT.
'Mother will give us the doctor’s meat.' (op. cit., ex. 4.14)"

Thus, initial expectations based on training notwithstanding, we must be careful in

fieldwork to look for tense on categories other than the verb. In general we must look for
the possibility that many of our semantic categories will not match up as expected with
our expected morphosyntactic categories.

6.3.2. VERB CATEGORIES

We now turn to consider the hardest nut to crack in morphological fieldwork, the
verb. As one example, verbs in Pirahã have sixteen suffix positions (giving 216 possible
verb forms, modulo semantic incompatibilities), none of which include tense or number
or person. Pirahã affixes encode notions like evidentiality, aspect, directionality,
negation, and mood, among others (see Everett (1986, ---)). So although Pirahã's verbal
morphology is very complex, one of the most complex systems I have encountered, its
morphological categories overlap surpisingly little with those of English.
Nevertheless, many categories are commonly marked in verbal morphology
cross-linguistically and I want to survey those very briefly here.

6.3.2.1. TENSE
Tense is a deictic ('pointing') category that locates an action or event with
respect to the moment of utterance (absolute tenses) or another tense referenced on the
verb (relative tenses). (See Comrie (1985)).) Consider the examples from Portuguese:

(6.70) a. Ele fala. 'He speaks.'

b. Ele falou. 'He spoke.'
c. Ele falará. 'He will speak.'
d. Ele falara/tinha falado. 'He had spoken.'
e. Ele terá falado. 'He will have spoken.'

The first three tenses are absolute. This is because they are defined with respect to
the moment of utterance, i.e. when they are spoken. So each means 'he speaks as I am
speaking' or 'he spoke relative to my current speaking' or 'he will speak after I am done
talking now', respectively. The last two examples are relative tenses. They refer to
events situated in relation to other events. So you might use (64d), for example, in a
sentence like (6.71):

(6.71) Ele tinha falado quando eu cheguei na sala.

'He had spoken when I arrived in the room.'

And you might use (64e) in a context like:

(6.72) Ele terá falado quando voce terminar de comer.

'He will have spoken when you finish eating.'

In (6.71) and (6.72) the events in the initial clauses are situated temporally in
relation to the events of the second, embedded clauses, rather than directly to the
moment of utterance.
Linguistic Fieldwork Daniel L. Everett
page 131
Not all languages necessarily encode tense as a linear category. So for European
languages, which do seem to categorize temporality as principally linear, we may
represent tense by a line extending metaphorically into the past and into the future,
generally representing the future as leading off to the right and the past as to the left.

(6.73) Metaphorical representation of tense as temporal linearity:

<---------------------------------------->
| | |
Past Moment of Utterance Future
|
Present

By this line, Past, Present and Future are all absolute tenses because they are
defined directly with respect to the moment of utterance (simultaneous = present;
preceding = past; following = future). Relative tenses are defined on the line with
respect to the temporal placement of other events. So, for an example like 'When you
arrived I had eaten', our linear representation would look like:

(6.74)
<------------------------------------------------------------------->
| | |
My eating → Your arrival → Moment of Utterance

So here the 'eating' event of 'I had eaten' is located before the 'arrival' even of
'when you arrived'. Thus the former is not situated in relation to the Moment of
Utterance but in relation to the arrival event.
On the other hand, the linear conception of time is not the only one imagineable
nor even the only one implemented in natural languages. For example, in Piraha,
according to Everett (1986 – an analysis I would not swear by anymore. I need to do
some more thinking about this) there are two tense-like suffixes, -a and –i, marking
'remote' and 'proximate', respectively. Remote in this analysis refers to events either
remote in time from the present (moment of utterance) whether future or past, or out of
the speaker's control. Proximate on the other hand refers to the converse, something
close to the moment of utterance temporally or within the speaker's control. If this
analysis is correct, then rather than a linear conception of time, the best way to express
the Piraha conception would be as 'concentric' layers:

(6.75) Possible metaphorical representation of Piraha tense:

Remote

Proximate Moment of Utterance

6.3.2.2. ASPECT
Linguistic Fieldwork Daniel L. Everett
page 132
Understanding a report about an event or state can be made much easier if the
speaker gives his or her perspective on how to classify the temporal states being
reported. Such temporal classification is known as 'aspect'. So consider the examples
below from Brazilian Portuguese:

(6.76) a. Eu estive com ele ontem.

I was with him yesterday.

b. Eu estava com ele ontem.

I was with him yesterday.

These examples are distinguished not by tense but by aspect, a speaker's

perspective on the classification of the 'measuring out' (Tenny (1989)) of an event or
the non-deictic temporal properties of an event or state. Both are past tense. The
punctiliar aspect in (67a), i.e. a reference to a single being with him at a point in time, is
contrasted with the non-punctiliar 'smearing' across time in (67b).
Languages can draw many distinctions based on aspect. A few of these are:

(6.77) a. Perfective and Imperfective

b. Habitual
c. Telic and Atelic
d. Progressive
e. Punctual and Durative
f. State and Dynamic
g. Iterative
h. Simultaneous
i. Terminative
j. Ingressive
k. Semelfactive

Perfective aspect marks the verb to represent that the speaker is viewing the event
as an indivisible whole (Comrie 1976, 16ff). Imperfectivity, on the other hand, is used
to make "explicit reference to the internal temporal structure of a situation" (Comrie
1976, 24ff). Comrie's important introductory work goes on to break down a number of
aspectual relationships into a single hierarchical relationship, as in ():
Linguistic Fieldwork Daniel L. Everett
page 133
(6.78) Classification of aspectual oppositions

Perfective Imperfective

Habitual Continuous

Nonprogressive Progressive

This chart represents the broader set of classifications likely to be found in field
research. There are other possibilities, however, as languages introduce 'fine-tuning' or
specific twists into this classification. Like other linguistic categories, the semantics
and formal nature of aspect can only be gotten at more or less reliably by (i) studying
the distributions of forms in the grammar and (ii) following a semantic methodology
based on conditions of felicity and truth-conditions (see ___ below).
Some other aspectual distinctions the fieldworker may encounter include things
like those in the remained of the list in (). Aspectual distinctions are fairly standard
across theories, though different theories may emphasize different features of the
oppositions. Just to get a better feel for these oppositions, let us consider the one that I
consider the most important distinction for the syntax, the telic vs. atelic opposition.
This is relevant for the syntax because, as Tenny (1990) shows, it is crucial for
understanding the relationship of the verb + undergoer to the actor.

Telic vs. Atelic

Telic refers to an end point, what Tenny () refers to in relation to verbs as the
'measuring out' of the event, in particular, marking when the event has reached its
lexically specified ending point or goal. This need not be a morphological category, but
is often reflected in the language independent of morphology, e.g. in the syntax of
English:

(6.79) a. John ate spaghetti. (Atelic)

b. John ate the spaghetti. (Telic)

In (6.79a) 'spaghetti' is treated like an unspecified mass and so there is no claim

that 'John' finished eating the spaghetti, merely that he engaged in the activity of eating
it. In (6.79b), on the other hand, the definite article draws a boundary around the
spaghetti and indicates that there is an identifiable whole of spaghetti, such that when
'John' eats that he can in fact be said to have completed the goal of his eating. Hence
this is a telic clause. Telicity (as Van Valin and La Polla (1997, ---) show) has
implications for the syntactic behavior of clauses. But in many languages, e.g. Piraha,
telicity is also a morphological category:

(6.80) Tí so/óá /ísigíhií kohoáip -á.

1sg. already meat eat -telic
'I already ate (the) meat.'

(6.81) Tí so/óá /ísigíhií kohoáip -i.

1sg. already meat eat -atelic
'I already ate meat.'
Linguistic Fieldwork Daniel L. Everett
page 134
Note that in the examples given, the presence of the telic suffix favors an
interpretation of the direct object as definite, even though Piraha otherwise lacks a
definite/nondefinite distinction morphosyntactically. This is a natural correlate of
telicity, however, as we saw in the English examples in ().
There are various other possibilities that involve differential marking (syntactic
vs. morphological) vs. semantics of aspect. Although the fieldworker should be familiar
with the basic categories mentioned here, they must be careful to arrive at conclusions
about 'their' specific language by distributional studies specific to the language in
question.

6.4.3. VOICE, VALENCE, AND TRANSITIVTY

Voice is, roughly, is the relationship between the action or state expressed by a
verb, and its arguments (nominals to which it assigns a semantic role, e.g. agent,
patient, goal, etc.). It is a rather sloppy term that is often used to refer to either Valence
(also referred to as Valency) or Transitivity. When applied to Transitivity, different
voices increase or decrease the degree of transitivity of a clause, as in the Montana
Salish –(e)m morpheme discussed below. When applied to Valence, these mark
changes from avalent to trivalent:
There is a great deal of confusion in the literature, even in books which ought
otherwise to clear up this confusion, between VALENCE and TRANSITIVITY. The former
refers to the lexically specified number of arguments of a verb. The latter refers to the
way that these arguments are 'mapped' to the syntax. Let's consider a simple example
from English and then consider some more complex examples from Montana Salish.
This issue is not all that complex, but it is very important and often overlooked (though
Hopper & Thompson (1980) and VanValin and LaPolla (1997, 147ff) are both
noteworthy for their care in distinguishing these concepts. Consider the following
examples from English which illustrate the differences between the two:

(6.82) a. The captain sank the ship in order to collect the insurance.
b. The ship was sunk (by the captain) in order to collect the insurance.
c. The ship sank.
d. *The ship sank in order to collect the insurance.

(6.83) a. The captain scuttled the ship in order to collect the insurance.
b. The ship was scuttled (by the captain) in order to collect the insurance.
c. *The ship scuttled.

The a. and b. examples in (6.82) and (6.83) are bivalent. That is there are two
arguments lexically required by the verb 'to sink'. In the a. examples and in the
parenthetical material in the b. examples, each of the two arguments is overt. But when
the parenthetical information in the b. examples is missing, the two arguments are still
both there, one explicit and one implicit (i.e. understood to be present in the meaning
but not heard, like a well-behaved child). This can be seen by the fact that the purpose
clause 'in order to collect the insurance' is still grammatical, whether or not the Actor is
overt. But purpose clauses can only be used when an agent is either present or
understood as part of the meaning of the clause. When the Actor is implicity, it is often
nevertheless indicated by voice morphology, e.g. the English passive, which signals the
presence of a less topical, potentially implicit Actor (if it appears in the 'by phrase' of
English, e.g. 'It was seen by John' then it is overt). The c. examples are what some refer
Linguistic Fieldwork Daniel L. Everett
page 135
to as the 'middle voice'. With middle voice, the agent is not even implicit. Therefore,
with not even an implicit agent allowed with the middle, no purpose clause is permitted
in (d).
It is interesting that there is no middle voice for the verb 'to scuttle', (c). The
reason is that the middle voice eliminates the agent altogether, i.e. it can be thought of
as altering the lexical representation of the verb/deriving from a separate lexical
representation without an agent. But 'to scuttle' will lack a middle voice altogether
because it has a very specific, non-alterable meaning of 'an agent sinking/destroying' a
vessel. Therefore, (c) is ungrammatical. With regard to Valence vs. Transitivity, we can
say that the a. examples are bivalent and transitive; the b. examples are bivalent and
intransitive; the c. examples are monovalent and instransitive.
Another example might be helpful. In English, as for so many semantic
categories, the distinction between causative vs. non-causative is periphrastic or
syntactic:

(6.84) a. John cried. (non-causative)

b. Mary made John cry. (periphrastic causative)

A mark on a word that increases the number of arguments required for a lexical
item, as in () or periphrastic representation of an added argument, as in English, is said
to increase the valency of the lexical item. The causative is one of the most common
valency-increasing operations. (Others include the applicative, 'dative shift', or,
arguably, simple object-addition, as in English 'Bill ate' vs. 'Bill ate eggs'.) In Piraha,
causative is a morphological category:

(6.85) a. Baóhoipaíi /i koabáipí

cat it died
'The cat died.'

b. Baóhoipaíi /i koabá -a -ip

cat it die -causative -completive

-í
-proximate
'Someone killed the cat.'

There are several types of voice alternation in natural language, including (with
my rough descriptions) Antipassive (the detopicalization of an Undergoer); Passive
(detopicalization of an Actor), Applicative (increased topicalization of a benefactive or
similar prepositional object), Dative Shift (increased topicalization of a GOAL
argument. To illustrate some of these, let's consider the extremely interesting case of
Montana Salish (from work by Thomason, Thomason, and Everett ()). In this language,
as in most Salish languages, there is a morpheme, –(e)m, that marks a number of
apparently unrelated types of meaning. However, when looked at more closely in
relation to the Valence vs. Transitivity distinction, a simple analysis emerges. Here are
some of the facts.
First, –(e)m can mark Antipassives, that is a voice alternation whereby the
Undergoer of an action is marked (and is in many languages further indicated by
Linguistic Fieldwork Daniel L. Everett
page 136
marking the Undergoer with an oblique case) as less topical in the current discourse
context. In the (73a) and (74a) examples, we have the antipassive, compared to the
normal transitive in the (73b) and (74b) examples. In these examples, the morpheme
appears in the position normally reserved for agreement with the Undergoer.

(6.86) a. tSn tS'n'eyn -m t o'ol'ó

1sgSItr trap -ANTIPASSIVE OBLIQUE marten (small weasel)
'I trapped some/a marten'

b. tS'n'ey -s -t -∅ -n lu o'ol'ó
trap -TR -TRANS -3OTr -1sgSTr 2ndary marten
'I trapped the marten'

(6.87) a. ∅ -n -'úlxw -m t lúk'w

3SItr -LOC:in -bring -ANTIPASSIV OBLIQUE wood
'He brought in some wood'

b. n -'úlxw -st -∅ -s lu lúk'w

LOC:in -bring -TRANS -3OTr -3STr 2ndary wood
'He brought in the wood'

This same morpheme can also be used to mark the passive, an unusual fact. To
get a passive reading from –(e)m, simply put –(e)m in the position on the verb
normally reserved for agreement with the Actor:

(6.88) a. 'íln -∅ -∅ -m
eat -TRANS -3OTr -BCK.AG
'Someone ate it'

b. 'íln -∅ -∅ -m lu scnc'á t wílwlt

eat -TRANS -3OTr -BCK.AG 2ndary frybread OBL.
scoundrel
'The fry-bread got eaten by a scoundrel/scoundrels'

Compare:

(6.89) 'íli -∅ -∅ -s lu scnc'á t wílwlt

eat -TRANS -3OTr -3STr 2ndary frybread OBL scoundrel
'The scoundrel(s) ate the fry-bread'

(6.90) a. xwúy -st -∅ -m

go -TRANS -3OTr -BCK.AG
'Someone made him go'

b. xwúy -st -∅ -m lu Coní t Malí

go -TRANS -3OTr -BCK.AG 2ndary Johnny OBL Mary
'Johnny got made to go by Mary'
Linguistic Fieldwork Daniel L. Everett
page 137

Th e Salish –(e)m can also be used to mark what Thomason, Thomason, and
Everett () refer to as 'Derived Transitives'. The traditional Salishanist understanding is
that in Derived Transitives a verb not normally transitive (in fact monovalent) becomes
transitive (more accurately bivalent), i.e. the number of arguments increases. Or, as L.
Thomason () puts it, more rigorously:

"Atelic stems can be turned into telic stems through the addition of -m and a
transitive marker. The meaning of such derived stems is lexicalized to a greater or
lesser degree; locative prefixes frequently help to direct the newly-telic action
towards an object. Two examples of derived transitives are given in [...]:"

(6.91) a. tSn 'áyχ w-t

1sg SItr tire-STATIVE
'I am tired'

b. 'áyχ w -t -m -n -∅ -∅ -∅
tire -STATIVE -DER.TR -TR -TRANS -3OTr -1sgSTr
'I tire of him/it, I find him/it tiresome'

c. 'áyχ w -t -m -n -∅ -∅-∅ lu s-qwlllú-s

tire -STATIVE -DER.TR -TR -TRANS -3 OTr-1sg STr NOM-
story-3 Poss
'I tire of his story/stories'

Another use of this morpheme is to mark continuative aspect on transitive verbs,

as in (79) and (80), in which case it follows the object marker. This is referred to as the
'Transitive Continuative':

(6.92) es -λ'e' -∅- ém -s lu ululím

stative -seek -3OTr-CONT -3Poss 2ndary money

t Coní
OBL Johnny
'Johnny is searching for the money'

(6.93) kwu es -'ác'χ -m -s t Coní

1sgOTr stative -look.at -CONT -3Poss OBL Johnny
'Johnny is looking at me'

The reader may have also noticed a suffix –n(t) (e.g. examples (73), (74), and
(78) above) which marks the verb as being greater than monovalent. Although it is
refered to in the Salish literature (and it is roughly pan-Salishan) as a transitivizer or as
'the transitive apparatus', its real function is to mark valence (Thomason, Thomason,
and Everett ()), whereas the –(e)m suffix marks deviations with regard to prototypical
transitivity, (). There are two questions the Salish morphologist must answer with
respect to these examples. First, why does the language have both the valency marker, -
n(t), and, second, why does –(e)m have so many functions? If we distinguish, as we
Linguistic Fieldwork Daniel L. Everett
page 138
should, the lexical notion of Valence, a nongradient concept, from the syntactic notion
of Transitivity, which is gradient, then we predict that, as in Salish, a language can
mark variations in each with separate morphology (just as we predict that they can vary
independently in the first place). And the different functions of the transitivity
morpheme, –(e)m, can be understood as marking, in the first instance, deviations from
the prototypical notion of transitivity in (81) (which is meant not as a theoretical
statement or commitment to prototype theory at all, but merely as a very useful rule of
thumb):

(6.94) Prototypical Transitivity: The prototypical transitive situation is one in which a

definite Actor affects a definite Undergoer in punctiliar past.

In light of (81), consider the different uses of the transitive morpheme just looked
at. Why does the same morpheme mark both antipassive and passive, for example?
Because it indicates, by appearing in the relevant 'slot' on the verb, that the Actor or
Undergoer is indefinite, hence that there is a deviation from Prototypical Transitivity.
Why does it mark the continuative in transitive verbs? Same reason – a deviation from
(). How can it mark a Derived Transitive? In this instance, it marks a valence change
not a transitivity change. It is crucial to note that in the Derived Transitive, –(e)m
appears to the left (i.e. closer to the root) of the –n(t) morpheme. This relative
proximity to the root means that it (following Bybee (1985)) is likely to affect the
meaning of the root more severely. In this case that is the right prediction – it does
indeed affect the meaning more – it alters the valence (a lexical property of verb
meaning), not merely the transitivity (a property of clauses). In these cases it moves the
verb closer to Prototypical Transitivity, by signalling that the verb has a sufficient
number of arguments to be transitive, thus licensing the occurrence of the bivalence
marker/'transitive apparatus'. In other words, unless we carefully distinguish transitivity
from valence, we have no account of Salish verbal morphology.
Now why have I given so much attention to this particular analysis of Salish here
in a field methods guide? Because it is a very clear illustration of the vital distinction
between valence and transitivity that, if it is not controlled for, can mislead or puzzle
the field researcher for a long period of time. And because most linguists fail to
recognize the distinction.

6.4.4. MOOD (or MODALITY):

Mood is used to indicate the relationship of the reported event or state to reality.
There are various kinds. The simplest mood system with opposing moods is two: realis
(action or state is perceived as real) vs. irrealis (action or state perceived as potential,
hypothetical, future, etc). There are more complex mood systems, e.g. indicative,
subjunctive, imperative, hortatory, optative, etc. in Romance languages. Mood is often
marked on the verb directly, but not always. Consider English (82) vs. Portuguese (83):

(6.95) a. If I were you.

b. If I was you.
c. If it had rained, I would not have come.
d. If it rains tomorrow, I will not come.
e. ?It rains tomorrow.
Linguistic Fieldwork Daniel L. Everett
page 139
Although the English subjunctive is disappearing, a mood used to indicate
hypothetical or conditional statements, a vestige of it is seen in (82a). For dialects that
use (82a), this is a subjunctive form (it would be ungrammatical if used as an
indicative, as in *'I were too here yesterday'). And (82e) needs the help of might, as in
'It might rain tomorrow', with the indicative ending –s missing from the verb, to
indicate the subjunctive mood.

(6.96) a. Se eu fosse voce.

b. *Se eu sou voce.
c. Se tivesse chovido, nao teria vindo.
d. Se chouver amanha, nao virei.

In Portuguese, unlike English, as (83) shows, mood is an obligatory

morphological category.
Mood is also important (as are numerous other verbal categories) for syntactic
reasons as well, because it can indicate the degree of formal relatedness between
clauses, as the Portuguese examples below show:

(6.97) a. Se fosse você, não viria/vinha/*vim/*verei amanhã.

if went:pastsubj. you not would come/was coming/came/will
come tomorrow
'If I were you, I would not come/was not coming/*came/*will come
tomorrow.'

b. Se for amanhã, terá/tem/*tinha/*tive que levar um casaco.

if go:fut.subj. will have/have/*was having/had to take a coat.
'If you go tomorrow, you will need to take a coat.'

The correlation between the mood of the protasis (the embedded 'if clause') and
the tense and mood of the apodosis (the matrix 'then clause') indicates embedding in
Portuguese and can indicate different degrees of clausal interconnection in other
languages, so can serve as a useful diagnostic for embedding and other aspects of
analyses of interclausal relationships.

6.4.5. DIRECTIONALS
Verbs are further restricted in some languages by affixation of directionals. These
affixes may be used in addition to a similar range of prepositions or they may obviate
the need for most prepositions (so I advise the fieldworker to look with particular care
for verbal directionals if there are only one or two prepositions in the language!).
Directionals primarily function to fix the event in space. But they are often extended in
meaning. Intransitive prepositions in English often function as directionals, as seen in
():

(6.98) a. The plane takes off

b. Look up the tree; look down the street.
c. He threw up; he upchucked (from the stomach to the mouth is seen as
an upward motion, but the extension of the directional meaning is also present).
Linguistic Fieldwork Daniel L. Everett
page 140
But also (see Lakoff and Johnson for further extensions of verb + particle
constructions in English):

(6.99) a. Look him up vs. ?/*look up him.

b. Look Bill up vs. look up Bill.
c. He is moving up in the world.

To my mind the contrast in ( a) shows an interaction of constraints on English

cliticization (where the prepositions are simple clitics, merely phonologically reduced
forms needing a 'host' in Zwicky's (1977) terminology). A prominal clitic in English,
e.g. [m] 'him', attaches to what it is an argument of. If it is an argument of the verb, it
attaches to the verb, as in (). If it is an argument of a preposition, as in (), then it
attaches to the preposition, producing a very different reading. As with other
morphological categories, directionals have a fairly straightforward and literal basic
function with subtle semantic extensions that can take considerable research time to
understand.
So compare () above with (): (see Lakoff and Johnson () for a large number of
such extensions) consider:

(6.100) a. Look him up; */?Look up him.

b. Look Bill up; Look up Bill.
c. Take on city hall.
d. He is moving up in the world.

The contrast in (a) shows an interaction between the constraints on English

cliticization (simple pronoun clitics, i.e. those which are merely phonologically reduced
forms of full pronouns, but otherwise unexceptional syntactically, see Zwicky ()), such
that a pronominal clitic must attach to the verb before a prepositional clitic. A
pronominal clitic in English, e.g. him [Im] in (), attaches to what it is an argument of. If
it is an argument of the preposition it attaches to the preposition, as in (), but if it is an
argument of the verb it must attach to the verb. Thus () is bad but () is good. The
reading where it is an argument of the preposition up, (), is quite different, of course,
than where the pronoun is an argument of the verb.
Examples like (), and many more in Lakoff & Johnson, illustrate how the uses of
directionals can be extended to be interpreted in ways other than their etymological
source might indicate. As with other morphological categories, directionals have a
fairly straightforward and literal basic function with subtle semantic extensions that can
take considerable semantic research to understand.

6.4.6. AKTIONSART
Aktionsart, 'kind of action', is a semantic classification of verbs, based on the
inherent temporal characterization of a verb's specified action or event.37 In VanValin
and LaPolla (1997, __), as in many other works, the aktionsart of verbs is an important
component in the analysis of their morphological (e.g. the range of affixes they can

37
This section borrows heavily from the excellent introduction to verbal semantics in
Van Valin ().
Linguistic Fieldwork Daniel L. Everett
page 141
take) and syntactic (e.g. the range of constructions in which they may appear) behavior.
This notion is introduced clearly by Van Valin (---):
"Verbs can be subclassified in various ways. Some researchers have found some
ways of subclassifying them more useful than others. One system that has been widely
adopted is to subclassify verbs based on Aktionsart ('kind of action', from Vendler ()),
the inherent temporal characterizatoin of a verb's specified action or event. There are
four subtypes of verbs in the traditional Aktionsart classification proposed by Vendler:
(6.101) a. States: be sick, be tall, be dead, love, know, believe
b. Achievements: pop, explode, shatter (the intransitive versions)
c. Accomplishments: melt, freeze, dry (the intransitive versions); learn
d. Activities: march, walk, roll (the intransitive versions); swim, think, snow,
write, drink

States depict static situations which are inherently temporally unbounded

(atelic), and both achievements and accomplishments express changes of state, which
are inherently temporally bounded (telic): achievements are instantaneous, while
accomplishments are not. Activities are dynamic, inherently temporally unbounded
(atelic), states of affairs. Vendler proposed this taxonomy based solely on the analysis
of English verbs, and yet it has proved to be of great cross-linguistic validity.
It would be reasonable to hypothesize that these distinctions are the universal
basis of the organization of verbal systems in human language. There is one important
non-Vendlerian Aktionsart class, namely semelfactives (Smith 1997). Semelfactives are
punctual events which have no result state. Examples are given in

(6.102) a. The light flashed.

b. Chris coughed.
c. The tree branch tapped on the window.
d. Dana glimpsed Kim.

There is a derivational relation between two classes which is very important

crosslinguistically, namely that between activities and what are called ACTIVE
ACCOMPLISHMENTS, the telic use of activity verbs. This general pattern relates
activity verbs of motion (e.g. run), consumption (e.g. eat), and creation (e.g. paint) to
the corresponding active accomplishment verbs. This is illustrated in (87) for
English.[my numbers, DLE]

(6.103) a. The soldiers marched in the park. Activity

a´. The soldiers marched to the park. Active Accomplishment
b. Dana ate fish. Activity
b´. Dana ate the fish. Active Accomplishment2
c. Leslie painted (for several hours). Activity
c´. Leslie painted Mary's portrait. Active Accomplishment

There are a few verbs in English that are lexical active accomplishments, e.g.
devour and go; that is, they do not alternate with an activity counterpart like the verbs
in (6.104).
Most fundamental is the distinction between static and non-static verbs, which
distinguishes verbs which code a ‘happening’ from those which code a ‘non-
happening’. In other words, with reference to some state of affairs, one could ask,
Linguistic Fieldwork Daniel L. Everett
page 142
‘what happened?’ or ‘what is happening?’. If, for example, a sentence like Bob just ran
out the door could be the answer to this question, then the verb run is [– static]. On the
other hand, a sentence like John knows Bill well could not be the answer to this
question, because nothing is taking place. Hence know is a [+ static] verb. By this
criterion activities, achievements, semelfactives, accomplishments and active
accomplishments are [– static]. States, however, are [+ static]."

However, it would be a mistake, a serious one, for a fieldworker to attempt to

simply translate the diagnostics for Aktionsart from English. They will likely not be
useful in many other languages, at least without considerable adaptation. Finding useful
diagnostics for linguistic categories, especially diagnostics that give such (apparently at
least) clear-cut distinctions is not easy and in a particular field situation it is possible
that nothing corresponding to these will be found. Aktionsart is an important concept to
the fieldworker not because it will always be useful in every field situation, but because
it may be very useful in figuring out how the morphology works in some languages.
But beware, again, of simple solutions. It may be very difficult to come up with
satisfying, watertight classifications. Nevertheless, your morphological research will
benefit tremendously by investigating at least the possibility of verb or verb
combinations classified on the basis of Aktionsart.

6.4.7. FINITENESS
Verbs can be more or less finite. Finiteness is the degree to which the verb is
referentially anchored. The greater the amount of inflection on the verb, the more
precisely is it anchored referentially. Consider the following examples from English:

(6.104) He owns the place.

(6.105) a. He wants to own this place.

b. He thinks he can own this place.

(6.106) Owning is not always a good idea.

In (88) the verb is marked for present tense, indicative mood, third person,
singular number, and has an aspectual reading that indicates on-going ownership. This
verb thus provides the hearer with a good deal of information to help situate the
'owning' in time, reality, and relationship to identity of owner, among other things. But
in (89) the same verb is not marked for these semantic categories. This less finite form
can be overtly marked by an item such as syncategorematic to (Pullum ()), (89a) or by
no marking at all, (89b). But in these examples it is temporally-aspectually unanchored,
i.e. it refers to no specific 'owning' of 'this place'. Its gerundive form in (90) likewise
marks it as lacking anchored temporal-aspectual, etc. reference. And the gerundive
marking further indicates that it is 'less verby' (see __ ), becoming more like a thing (an
abstract thing, i.e. the act of ownership).
Finiteness can manifest itself in different ways in different languages. So consider
the examples below from Karitiana (Arikem family) and Pirahã:

(6.107) ÿn nãka paka -j ÿn pykyp

I affirmative clean -tense I clothes
'I will clean the clothes.'
Linguistic Fieldwork Daniel L. Everett
page 143

(6.108) ÿn i paka ÿn pykyp

I it clean I clothes
'I will not clean the clothes.'

(6.109) ti gái -sai kó'oí hi kaháp -ií

1 say -nom. name he leave -intention
'I said that kó'oí intends to leave.' (literally 'My saying kó'oí intend-
leaves')

(6.110) a. hi ob -áa'áí kahaí kai -sai

he see -attractive arrow make -nom.
'He knows how to make arrows well.' (lit: 'He sees attractively arrow-
making')

b. kahaí kai -sai hi ob -áa'áí

arrow make -nom. he see -attractive

c. *hi kahaí kai -sai ob -áa'áí

he arrow make -nom. see -attractive

Karitiana is interesting here because a negated verb takes a non-finite verb form.
Pirahã is interesting for other reasons. Let us consider first the Karitiana example.
Apparently in Karitiana, the negation of a verb removes, in standard cases, its verbal
referentiality. It no longer takes finite verbal morphology because a negated action or
state simply never occurred and thus can be left unanchored.
The Pirahã case is perhaps even more interesting. Everett (2005) argues that
Pirahã lacks recursion, at least in the syntax. Yet, at the same time, some functions of
clauses are less finite, corresponding partially, but by no means completely, to what in
other languages would be embedded clauses. On the other hand, the reduction of
finiteness in this language does not mark embedding. For example, in (), it is (what
most linguists would take to be) the matrix verb that receives the –sai suffix, where as
in () it is what might otherwise be taken to be the embedded clause which is so marked.
This difference in marking violates normal expectations of finiteness as an embedding
marker (see Koptjevskaja-Tamm (1993a and 1993b for detailed discussion of
nominalizations and finiteness more generally). Everett and VanValin (in progress)
analyse the –sai suffix as an information marker, indicating topical or old information.
Nonfiniteness of the verb can mark embedding. One reason that this is common is
because, as Cristofaro (2003) argues, subordinate clauses are not assertions. Therefore,
it is common for their verbs to be less referential in the sense of this section. However,
that is not the only function of nonfiniteness. It can also indicate that the verb has
'moved' in the sense of Ross's (1973) scale towards 'nouniness'. An analysis associating
nonfiniteness exclusively with embedding could not capture the generalization that
Pirahã otherwise lacks any evidence for embedding.

6.4.8. EVIDENTIALS
Many languages employ morphological marking to indicate the speaker's
epistemological stance with regard to their assertions, as an indication for the hearer(s)
of the reliability of or the warrant for the content of the speaker's assertion.
Linguistic Fieldwork Daniel L. Everett
page 144
For example, La Polla (2003) discusses the interesting case of the Tibeto-Burman
language, Qiang. This language has a suffix, -k, which is labelled as 'inferential/mirative',
depending partially on the Aktionsart of its verbal host. That is it is 'inferential' with
activities, but 'mirative' (i.e. marking unexpected or new information) with resultatives
and states. For example:

(6.111) a. the: ¸d¸yta: ˙a q´ -k

3sg Chengdu:loc or go inferential
'He went to Chengdu (I infer based on the evidence, e.g. bags are
missing and was told earlier about the trip, etc.)

And from Shipibo (Camacho, Elias-Ulloa ()), '...an inferential can be combined
with a reportative, yielding a potentially ambiguous sequence: either the speaker
reports someone else's inference, or s/he expresses an inference from reported
evidence."

(6.112) Ani -ronki i -bira [a] -i jawen jema.

large rep be -inferential -incl poss3 village:absolutive
'It appears the door is open/apparently the door is open.'

As Camacho and Elias-Ulloa go on to observe:

"Several studies have tried to explore whether propositional source-marking

obeys some kind of cross-linguistic tendency. Willett (1988, p. 57), for example,
¯nds in a survey of 38 languages, that evidentiality falls within three categories:
attested (which can include subdivisions that include visual, auditory or other
sensory information), reported, and inferred. Speas (2004, p. 4), on the other
hand, proposes four hierarchical categories: personal experience, direct
(sensory) evidence, indirect evidence and reported evidence (hearsay).
Aikhenvald & Dixon (2003, p. 3), suggest two broad types of evidential
systems: those that state the existence of a source of evidence without specifying
it, and those that specify the source. Within the second type, they discuss several
subtypes:

(6.113) Two-distinction systems:

1. Eyewitness and noneyewitness.
2. Nonfirsthand and everything else.
3. Reported and everything else.

(6.114) Three-distinction systems:

1. Visual, inferred, reported.
2. Visual, nonvisual sensory, inferred.

(6.115) Four-distinction systems:

1. Visual, nonvisual sensory, inferred, reported.
2. Visual, inferred (2), reported.
3. Nonvisual sensory, inferred (2), reported.
Linguistic Fieldwork Daniel L. Everett
page 145
4. Visual, inferred, reported (2).

It seems clear, particularly for three and four-distinction systems, that they are
built on Willett's three basic distinctions, with some possible subdivisions within
each of those categories. Speas, following Oswalt (1986), suggests an additional
category different from visual-sensory: personal experience. Willett, Oswalt and
Speas suggest that there is an underlying hierarchy corresponding to the degree
in which the source directly involves the speaker's evidence. This scale goes from
more direct experience to no experience at all (Speas 2004, pg.258):

(6.116) personal experience >> direct (e.g. sensory) evidence >> indirect
evidence >> hearsay."

Evidentials are interesting for what they have to teach us about the interplay of
semantic scope between different types of affixes, about what they reveal of different
epistemologies, about the nature of morphosyntactic structure, and about the cultures
that produce them.

6.4.9. DISLOCATION38
Modern syntactic theories focus (to an inordinate degree in my opinion) on just
one of the many items that we must 'memorize' about a language, namely, whether or
not it allows for constituents to be displaced or dislocated from their purportedly
'underlying' or (perhaps more accurately) their unmarked positions. The functions of
dislocation in a language (aside from the extremely theory-internal arguments of some
syntactic theories) include marking the information structure of the sentence (see
Lambrecht () and section ___ of chapter __ below) and scope alternations. But of
course dislocation is simply one among many different strategies for marking different
information structures and scope, inter alia. Otherways of marking these include
prosody, morphological marking, separate construction types, or a combination of some
subset of these options, among other strategies. Still, it is true that many languages
choose dislocation as perhaps the principal means of pragmatic highlighting, scope
alternations, or information structure changing. Occasionally, dislocated elements may
have consequences for the morphology of a language, which is why dislocation is
important to us in this section. So consider question-related dislocation and
morphological marking in Wari:

(6.117) Ma' co tomi' na?39

that:prox:hearer m/f:rp/p speak 3s:rp/p
'Who is speaking?'

(6.118) a. Ma' co tomi'ca?

38
See also the section on dislocation in chapter 8 below.
39
Example (4) is interesting because it illustrates that questioning the subject of the
sentence requires tense in second position, to the immediate right of the question word,
and also immediately to the right of the verb. WH-questions of subjects require that
tense be expressed twice in the sentence. This, as (5) – (7) show, is not true of any other
questioned constituent.
Linguistic Fieldwork Daniel L. Everett
page 146
that:prox:hearer m/f:rp/p speak 3sm
'Of whom is he speaking?'

b. Ma' carawa ca pa' caca mon

that:prox:hearer animal nrp/p kill 3pm coll

tarama'?
man
'What thing/animal did the men kill?'

(6.119) Ma' ca para 'aca ca pije ma'?

that:prox:hearer nrp/p why cry 3sm child that:prox:hearer
'Why is that child crying?'

(6.120) 'om ca mao ca.

not:exist nrp/p go(sg) 3sm
'He did not go.'

Dislocation is morphologically relevant in Wari' because the nature of the

grammatical dislocated (e.g. subject vs. direct object vs. adjunct) results in different
kinds of agreement patterns. First, any dislocation causes tense to appear before the
verb. This is due to a simple (and crosslinguistically very common) constraint that
requires tense to appear following the first constituent of the clause. But notice that in
(), when the subject appears in sentence-initial position, tense appears both in second
position and immediately following the verb (it's 'normal' position). Only one
occurrence of tense, however – in second position of the clause, is found when objects
and adjuncts are 'dislocated' to the beginning of the clause (() and ()). Whatever the
reason for this asymmetry, it is evidence that dislocation can have morphological and
not merely syntactic effects.
Another morphological configuration that some languages allow is Incorporation.

6.4.10. INCORPORATION
The first clear reference to Incorporation (though not by that name) I am aware of
is Montoya (1585-1652), Jesuit missionary and linguist who worked among the
Guarani of Paraguay. Like Dislocation, Incorporation is a common device for marking
changes in information structure. It can also be used for other functions, e.g. additional
specification of the verb. As Mithun (1986, --) says:

"Noun incorporation is perhaps the most nearly syntactic of all morphological

processses. Examination of the phenomenon across a large number of
geographically and genetically diverse languages indicates that, where syntax
and morphology diverge, incorporation is a solidly morphological device that
derives lexical items, not sentences. It is used for four different but related
purposes; these fall into an implicational hierarchy which in turn suggests a
path along which incorporation develops historically. Differences in its
productivity from language to language show that this development may be
arrested at any point – resulting either in the eventual disappearance of the
process, or in its resurgence as a productive system of affixation."
Linguistic Fieldwork Daniel L. Everett
page 147

An example from Yucatec Mayan

Linguistic Fieldwork Daniel L. Everett
page 148
(6.121) a. tinč’akah če’
t- in- č’ak- Ø- ah če’
COMP- I- chop- it ASP

b. č’akče’nahen
č’ak- če’- n- ah- en
tree chop- tree- APASS- ASP- I.ABS
“I chopped a tree.”

Noun-incorporation has various manifestations and a range of functions. In the

examples of Piraha encliticization repeated below, there is no real incorporation, but the
phonological divisions between object and verb have been reduced in a way that doesn't
happen with subjects:

(6.122) tíì /ísìtòí /óògàbàgàí → tíì /ísìto&ògàbàgàí

1psg egg want
'I want an egg.'

(6.123) kàhàí /áàgàhá → kàhàa&àgàhá

arrow is
'It is an arrow.'

There is something 'iconic' in this in that some structures show less phonology to
represent a tigher semantic relationship. What that semantic 'tightness' actually means
in any given language has to be determined by the linguist by careful analysis and
argumentation.

6.4.11. AGREEMENT
Agreement is a way of keeping track of verb participants by marking them
morphologically, either on the verb (as in English, John runs) or elsewhere in the
clause, with Wari's second-position verb agreement markers:
Most languages will have agreement in one form or another. As Everett (1996)
observes, it is imperative to analyze carefully the relationship between pronouns,
pronominal clitics, and agreement affixes. Everett (1996) considers a variety of
languages in which these categories interact to produce a type of 'agreement mosaic'.
This is also discussed in the section on clitics above and illustrated in many examples
throughout this book.
Let us turn now to strategies commonly employed to mark the distinctions
discussed above.

6.5. MORPHOLOGICAL REFLEXES OF GRAMMATICAL AND SEMANTIC DISTINCTIONS

6.5.1. CONCATENATION
Easily the most common morphological marking strategy is the addition of extra
material to the root (the base of a word from which all morphology has been stripped),
what morphologists know as concatentation. Material may be added to the root at the
right or left margins (prefixation and suffixation, respectively), at both simultaneously
(circumfixation), or within the root (infixation). Of these various strategies, suffixation
Linguistic Fieldwork Daniel L. Everett
page 149
is the most common, followed by prefixation, then perhaps infixation, then
circumfixation. These are illustrated in (6.124) – (6.126):

Suffixation: adding to the end of a particular morphological constituent (word, root, or

stem)

(6.124) John runs.

Prefixation: adding to the beginning of a particular morphological constituent (word,

root, or stem)

(6.125) John enlivened the party.

Infixation
(6.126) Tagalog Infixation (McCarthy & Prince 1993)

a. /um + /alis/ → //um-alis/ ‘leave’

b. /um + tawag/ → /t-um-awag/ ‘call’ pf., actor trigger
c. /um + gradwet/ → /gr-um-adwet/ ‘graduate’

Circumfixation is the simultaneous addition of material preceding and following

the root, stem or word to mark a single distinction. Circumfixation is perhaps best
known from the German past participle (ge- -t for regular verbs). The verb spielen, for
example, has the participle gespielt.

6.5.2. SUPPLETION
Another common strategy for marking distinctions like those discussed above
morphologically is suppletion. Suppletion is the label given to alternations where there
is no rule-based phonological relationship between the various allomorphs. It is quite
common in the world's languages. So consider the examples below from English and
Portuguese:

English to be
(6.127) I go, you go, he goes, they go, we go, I went, you went, etc.

Portuguese ir
(6.128) Eu fui; voce foi; eles foram; Eu vou; voce vai; nos vamos, eles vao.

It is common to find suppletion in more frequent verbs, e.g. 'to be' or 'to go'. This
is likely because it is easier to remember the forms when they are heard more often.
Another not uncommon type of suppletion is found in Wari', where the suppletion
follows an ergative pattern, the form of the verb governed by the object of the transitive
or the subject of the intransitive. Not all verbs participate in this process (and it would
be an interesting study to see if these verbs follow any particular pattern). But those that
do behave as in (), from Everett & Kern (1997, 331):

(6.129) a. Toc nana com.

drink:sg 3p:rp/p water
Linguistic Fieldwork Daniel L. Everett
page 150
'They drank water.'

b. Tototoc nana com.

drink:pl 3p:rp/p water
'They drank and drank water.'

This example is interesting, because although 'water' has no plural in Wari', the
suppletive plural of the verb can also indicate a plurality of events. This scope variation
of suppletion, i.e. marking either object plurality or event plurality, is common. Abdias,
Walker, and Everett (1984) discuss a similar case for the Mayan language, Huastec. But
there are many other examples where the object itself governs plurality:

c. Cao' 'ac cacama na.

eat:sg travel 3pf 3s:rp/p
'Then they ate.'

d. Mon cacacao' 'ac cacama -in capam' con

slowly eat:pl travel 3pf -3n cornbread prep:3sm

jowin ma' na.

monkey:species that:prox:hearer 3s:rp/p
'Then they ate cornbread and jowin monkey for a long time.'

The Wari' case is very interesting because the suppletion shows an ergative
pattern (see __ above). For a very useful and detailed study of morphological
suppletion, the reader should consult Veselinova (2004).
Let's consider one more, often overlooked strategy for morphological marking,
periphrasis.

6.5.3. PERIPHRASIS40
Consider the contrast between aspectual marking in English vs. Portuguese:

(6.130) a. Eu estava aqui ontem.

b. Eu estive aqui ontem.

(6.131) Eu tomava cafe, sim.

(6.132) a. I was here yesterday for a while.

b. I was here yesterday.
c. I used to eat breakfast.
d. I was eating breakfast.

Portuguese marks punctiliar, continuative, and habitual aspects morphologically.

English marks the latter two periphrastically. The fieldworker can easily mistake
periphrasis, a morphological strategy, for syntax. But periphrasis, as Börjars, Vincent,

40
This section assumes the paradigm to be a theoretically primitive of linguistic theory.
For eliciting lists of related lexical items commonly known as paradigms in a non-
theoretical sense, the reader is referred to __ of chapter __ on monolingual fieldwork.
Linguistic Fieldwork Daniel L. Everett
page 151
and Chapman (1997) show, is an important component of natural langauge
morphology. To see this more clearly, consider the case of periphrastic pronouns in
Wari', discussed in Everett ():

TABLE ONE
Paradigm of spatial demonstrative pronouns

Proximate to Proximate to Distal

Speaker Hearer
Masculine co cwa' co 'ma' co cwain
singular
Feminine cam cwa' cam 'ma' cam cwain
singular
Neuter 'i ca' 'i 'ma' 'i cain
Plural caram cwa' caram 'ma' caram cwain

TABLE TWO
Paradigm of temporal demonstrative pronouns

Heard/not seen Recently absent Long absent

Masculine co paca' co pacara ne co pacara pane
singular
Feminine cam paca' cam pacara ne cam pacara
singular pane
Neuter 'i cara 'i cara ne 'i cara pane
Plural caram paca' caram pacara ne caram pacara
pane

The constructions in these tables are not simply words. Phonologically, words in
Wari' manifest two important characteristics that these periphrastic forms fail to show.
First, Wari' words disallow internal consonant clusters where one of the consonants is a
glottal stop. Thus, the initial glottal stop on 'ma' [/ma/] 'that proximate to hearer' in
Table 1, as in co 'ma' 'that masculine one proximate to hearer' and 'i 'ma' 'that neuter
one proximate to hearer', should not occur, since word-medial glottal-consonant
clusters do not otherwise occur. By this otherwise inviolable criterion, in the cases of co
'ma' and 'i 'ma', co and 'i must be interpreted as morphologically independent and not,
say, as prefixes. We interpret them, as well as cam and caram,as clitics (as discussed in
Everett and Kern (1997, 413ff)). Second, in Wari' words prefixes always undergo
Vowel Harmony with their hosts. The kinship terms discussed in the penultimate
section of this paper, which are analyzed as words, rather than periphrastic forms, do
indeed show Vowel Harmony between the co and the root. The absence of Vowel
Harmony in the forms in Tables 1 and 2 is therefore additional strong evidence that
these are not words (or, at the very least, that they violate expectations on word
phonology).
Among the many arguments Everett (2005) gives that these pronouns are
periphrastic morphology, rather than syntax, is the following. Periphrastic pronouns in
Wari' mark all three Wari' genders. But neuter gender marking is accomplished by
Linguistic Fieldwork Daniel L. Everett
page 152
using the morpheme cara, just in case it it is in a periphrastic construction. Cara 'heard'
otherwise has no inherent gender association (Everett (2005))). This means that cara's
interpretation in the tables above is a function of the paradigm (an abstract constraint of
morphology but not syntax), rather than the syntax and is thus non-compositional and
morphological. Therefore, Everett (2005) concludes that periphrasis is an important
morphological strategy in Wari', a way of co-opting phrases to function as 'hybrid'
words.
Methodologically, therefore, it is always useful to try to discover unifying
principles of morphological groupings, assuming the paradigm as at least a
methodological convenience if not a theoretical primitive. This way it will be easier to
spot the use of syntax for morphological means as well as 'mixed categories'.

6.5.6. INTERCALATION
Intercalation is most often referred to in the literature as 'non-concatenative
morphology'. However, frm what we have seen in this section, there are various kinds
non-concatenative strategies. So I prefer to refer to this type of morphological marking
as Intercalation. This is illustrated in (110) from Arabic:

(6.133) TH: Gloss: Morphological analysis:

a. qātal 'he killed' 3 p. masc. sg. perfect
b. qittēl 'he murdered' 3 p. masc. sg. perfect intensive
c. qotēl 'killing' active participle (qal)
d. jiqtol 'he will kill' 3 p. masc. sg. imperfect/future
e. niqtal 'he killed himself' 3 p. masc. perfect reflexive
f. qe tālām 'he killed them' 3 p. masc. sg. perfect + pers. pron.

One way to conceie of this is that the morphology of Arabic accesses structures
for vowels separately from structures for consonants and then, following precise
constraints (e.g. those in McCarthy (1979)) intercalates them.

6.5.7. PROSODY
The marking of morphological distinctions can also be accomplished by prosodic
features such as tone or intonation, length, and stress. For example, consider the way
that Pirahã indicates an inherent vs. an accidental property of an individual via the tone
on the verb:

(6.134) Hi /ii /aagá.

'He hungry is.'

(6.135) Hi hiaitíihí /áagá.

'He is a Pirahã.'

Or consider the distinction between some nouns and verbs in English:

(6.136) permit (verb) vs. permit (noun); contract (verb) vs. contract (noun), etc.

6.5.8. REDUPLICATION – KAMAIURA

Another common morphological marking device is reduplication, where some
Linguistic Fieldwork Daniel L. Everett
page 153
part of the root or stem is duplicated in order to mark a semantic distinction. One well-
known example comes from Kamaiurá (Everett & Seki 1986, McCarthy & Prince
1993):

(6.137)
a. ohuka → ohuka-huka ‘he laughed/he kept laughing’
b. apot → apo-apo-t ‘I jump/I jump repeatedly’ (*apot-apot)
c. o-mo–tumuõ → omotumu–tumu–õ ‘he shook it/repeatedly’
d. je–umirik → jeumiri–miri–k ‘I tie up/repeatedly’
e. o–etun → oetu–etu–n ‘he smells/keeps on smelling’
f. o–ekvj → oekv–ekv–j ‘he pulls/repeatedly’

According to Everett & Seki (1986), these are examples of suffixation. According
to McCarthy & Prince (1993) they are best analyzed as infixation. Whatever the
analysis, it is important for the fieldworker to be familiar with this type of marking. A
more detailed study of reduplication (but to my mind one with the defect of predicting
the absence of the Tupi pattern just investigated) is Inkelas and Zoll ().

6.6. WHERE TO MARK DISTINCTIONS

Now let us consider where the morphological strategies discussed above may
appear. For many features they can differentially appear, crosslinguistically, on either
the head of the phrase or a dependent (daughter) of the head. This distinction is very
important, as Nichols (), the first one to make the distinction in these terms, points out:

"Morphological marking of grammatical relations may appear on either the head

or the dependent member of the constituent (or on both, or on neither). Grammatical
relations – and whole languages – may be classified according to their propensity for
using one of these types of marking. Implicational relations among various marking
patterns can be stated: languages display a tendency to use one type consistently
throughout their grammar. The difference in patterns provides a typological metric and
a functional explanation for certain word-order preferences. For historical linguistics,
it provides a diagnostically conservative feature and a clue to genetic relatedness.
Although the head-marked pattern is cross-linguistically favored, grammatical theory
is strongly biased toward the dependent-marked patterns that happen to dominate in
Indo-European."

Consider the two English examples in (115) – (117). The first one marks the
relevant morphological distinction on the head of the phrase (the verb in this case),
while the other marks it on a dependent of the head (the embedded NP in the possessor
position in (116) and the case on the pronoun in (117)):

(6.138) a. The man runs.

b. The men run.

(6.139) John's book is wet.

(6.140) a. I saw him.

b. He saw me.
Linguistic Fieldwork Daniel L. Everett
page 154

Likewise in Portuguese, the gender of the noun is marked on its dependents:

(6.141) a. O homen bonito.

b. A mulher bonita.

Or, as in Wari clausal morphology, the marking may appear on neither the head
nor one of its dependents, but in the clausal 'second position', also known as
Wackernagel's Position, from the work of the first person to draw attention to this
position for clitic placement, Wackernagel (1892).

6.7. DERIVATION VS. INFLECTION

A venerable distinction in morphological theory is the subclassification of
morphological processes which change the category of a root to form a stem as
derivational and morphology added to the stem, and usually agreeing with a semantic
distinction on another word in the phrase, as inflectional. Not all morphologists
attribute primitive theoretical status to this distinction. Here I want to briefly consider
some of the classical reasons for distinguishing these two subclasses of morphology –
not because I necessarily endorse the labels, but because recognizing the concepts
behind the labels is useful for any fieldworker. Consider the examples in ()-() from
English:

(6.142) a. farm
b. farms
c. farmer
d. farmers

(6.143) a. live
b. liven (as in Liven up the party!)
c. enliven (as in Enliven (*up) the party!)

The suffix -s is an inflectional morpheme. It is used to produce 'fit' between a

lexeme and its syntactic context (e.g. its phrase or clause), as well as to signal that a
meaning option (plurality) has been selected from the relevant inventory of semantic
distinctions (number in this case).
Morphology that changes the meaning of a root to form a stem (an internally-
complex morphological input to inflection) often is less productive, more idiosyncratic
in meaning and morphology, more constant in its phonological form, among other
things, than morphology that does not affect core lexical meaning and that applies to
either roots (from which all morphology is absent) or stems. Whatever you call it, this
distinction crops up in language after language and it is to the fieldworker's advantage
to be aware of it, look for it, and carefully subclassify this and any other distinct
mrophological processes that appear to be operative in a given language.
Up to this point, we have considered distinctions often made in natural
languages and how and where these may be marked. I now want to offer practical
suggestions for morphological fieldwork, i.e. how to collect, analyze, and process
morphological data.

6.8. METHODOLOGY FOR MORPHOLOGICAL ANALYSIS

Linguistic Fieldwork Daniel L. Everett
page 155
6.8.1. INTRODUCTION
Returning now to elicitation of morphosyntactic data, the researcher must first
answer the obvious question of what it means to 'describe', 'document', or 'analyze' the
morphology of a language. Simply put you have two goals in morphosyntactic
description: to identify the primary semantic distinctions in word meanings and to
determine how these distinctions are signalled. The task is therefore easy to
conceptualize.
Of course, the actual doing of it is much harder. Let's say you hear the two forms
in (6.144), for example:

(6.144) a. pandapa 'I ate'

b. pandapã 'You ate'

On the surface, it would look like 2nd person is marked by nasalization of the final
vowel, /a/. But here are some obvious questions: (i) are you sure there is no
nasalization on the final vowel of the first word? (ii) did you get this pair repeated from
multiple speakers? If so, did they all use nasalization in (b) and none in (a)? (iii) can (b)
also be uttered without nasalization on the final vowel without changing the meaning?
These questions may seem obvious now, but they are not quite so obvious in a field
situation. The field researcher will soon realize that there is an enormous difference
between analyzing printed data and analyzing spoken data. Having 'good ears' is crucial
to good morphosyntactic analysis. Acoustic analysis will help, of course, but without
good ears, you will be reduced to hours and hours of unnecessary spectrographic
labors.
Linguists disagree on how to collect data for morphosyntactic and semantic
analysis in the field. Some advocate a text-only approach to data collection. Some
claim that specific sentence elicitation can play a useful role in field data collection.
The problem to me is similar to the problem faced by any second language
learner. Do you learn a language from books and classes, or from natural conversations
in a community of native speakers? The choice is a false one, at least to me. One must
converse and develop a need to use the second language. But all other sources are also
of great potential usefulness and will almost always complement the information
otherwise learned, helping in fact to learn constructions and expressions that might
have otherwise gone unperceived, at least for a considerable period of time.
Likewise in morphosyntactic research, we need data from a variety of sources.
Some linguists are concerned about a 'data problem, arguing that elicited data (one
sentence at a time, for example) should be avoided in favor of conversational or text-
based data in which sentences appear in natural contexts. Only in such 'natural contexts'
we are told, are data reliable. However, I disagree. I believe that elicited data, natural
texts, conversations, corpora, and any other source of data can and should be used in
constructing a theory of a particular grammar. The 'data problem' is to me a judgment
problem. Gathering data requires experience, intuition, a lot of linguistic knowledge,
security in what you are trying to do, and the ability to distinguish contrived data from
data that could actually emerge in a natural way from the mouths of speakers. (We don't
know if they have it in their grammar until we have heard it from their mouth.)

6.8.2. Record, transcribe, annotate (the context and what you think it means)
As you perambulate about the speech community where you are working you will
hear words, both words you ask for and words that are offered to you, or just words and
Linguistic Fieldwork Daniel L. Everett
page 156
expressions in the environment. You probably will not know at the beginning of your
work whether these are words or sentences or phrases (consider the gavagai problem
discussed in section ___). And you will almost certainly not have a secure
understanding of their meaning. Many things you encounter, e.g. rare animal or rituals,
or descriptions of culturally sensitive activities you might not be welcome to see again
(e.g. something you just innocently stumbled across). So always be prepared to
transcribe and, ideally, record in audio and video what you hear. This is where a small,
analog recorder, with easy rewind and record functions is very useful. However, there
are often occasions that it would be useful to record an image along with the word. My
personal digital camera (not a camcorder) allows me to take a picture and a brief
recording associated with that picture. This is useful, for example, if I am collecting the
names for flora and fauna. I am not a zoologist. Thus whenever I see a new tree or
species of rodent, I usually do not know what it is called in English, much less in Latin
(i.e. its scientific name). Some fieldworkers do know this. But in my case, when I get
back to the village, I pull out one of the flora and fauna books I have taken (see ___
below) and look for what I have just photographed. Failing to find it, I can take the
photo with me to a specialist upon returning to my home institution.

6.8.3. Test with native speakers:

You haven't learned a morpheme until you have it down articulatorily and
acoustically. You need to test your pronunciation of it (or the words in which it occurs)
by practicing with native speakers. As you practice, you need to also listen to make
sure you understand how to use the morpheme appropriately in context and any
extensions that it has. This is partially pragmatic knowledge of the morpheme and
partially semantic. Semantic methodology is presented in __ below.

6.8.4. Get translations and paraphrases from a variety of speakers.

Never rely on the pronunciation, usage, or judgements of a single speaker when
analyzing a morpheme. You need to test your hypotheses and judgements with a variety
of speakers, at least 3-6 men and women before feeling that you have reached a
definitive conclusion.

6.8.6. Text tracking

One of the best examples I am aware of in the literature that traces a linguist's
analytical history in relation to a set of morphemes is found in Lowe (1990). Lowe
(1990, 544ff) states his problem as follows: "In Nambiquara, four morphemes, namely
–jau3-, jut3-, -kxai3-, -kxe3- figure centrally in expressing the interclausal relationship of
cause/reason. Each one of these morphemes occurs in two-clause constructions which
can be faithfully translated as expressing a cause or reason relationship. There is the
added complication, however that each one of these morphemes is used in other
contexts where a causal relationship does NOT [emphasis Lowe's, DLE] seem to be
present." Lowe proceeds to discuss the difficulties of analysing these morphemes,
showing clearly that looking at their uses exclusively in isolated sentences will lead
largely to confusion rather than conclusion. He then shows how an examination of their
functions in a wider discursive context leads to a fairly straightfoward analysis. Before
reviewing his findings, very useful from a methodological standpoint, I quote his
foundational assumption: "The analysis presented in this paper must start with the
assumption that each of the morphemes above has a unitary meaning, even though the
same morpheme may appear to have different meanings in different contexts. Such an
Linguistic Fieldwork Daniel L. Everett
page 157
assumption is justified provided it leads to consistent and insightful descriptions of the
function of each morpheme. The danger of assigning disparate glosses for the same
surface morpheme in different contexts is that each one of these disparate glosses will
be assigned from an English speaker's point of view; until one struggles to find what, if
anything, they have in common, the foreign viewpoint will be all that one has.'
[emphasis mine, DLE].
Lowe goes on to demonstrate, convincingly, that one cannot understand either
the notion of causality in Nambiquara, nor the ways that it is marked linguistically,
without understanding both the cultural meaning of causality in Nambiquara and the
various discursive uses of the different morphemes. In the course of his study, he also
strongly supports his assumption of 'unitary meaning' as a vital methodological tool.
Another, similar, example comes from my own study of the intersentential
connective, hoagá in Pirahã. Consider the following examples and the contexts in
which I initially recorded them.

(6.145) Hi toio/aagá, hoagá hi opaohoaibaáí.

he old is hoagá he works a lot.
'He is old .?.. he works a lot.'

(6.146) a. Hi gogiísoi /igí -ai?

he what-big:thing with is
b. Ti hoagá pi-obaibaai.

(6.147) Ti gi hoagá poogaihiai bagaboi.

'I you hoagá give banana.'

6.8.7. Bi-directional translate texts with language teachers

Another potentially useful method of verifying morpheme meanings is to translate
with native speakers, who are also fluent in the language of the translation, the words
containing the morpheme into the trade, national, or other relevant surrounding
language. Then with other speakers, you should translate back into the target language.
If you and they have fully understand the morpheme, the translations should produce
similar results. If you are wrong, however, you should get unexpected variations.

6.9. PERSPECTIVES ON MORPHOLOGICAL ALTERNATIONS

HOMONYMY, POLYSEMY, AND ALLMORPHY

The fieldworker will find it useful, but not always essential to distinguish
polysemy (a single word with multiple meanings) from homonymy (multiple words
with a single phonological form). For example, consider the different uses of the clitic
se in Portuguese. Are these multiple morphemes with one form or multiple meanings of
one morpheme? (Each 'se' is proceeded by a label for that particular function.)

Reflexive
(6.148) Sérgio se matou.
'Sergio killed himself.'

Argumental Impersonal
(6.149) Se recebe socos facilmente.
Linguistic Fieldwork Daniel L. Everett
page 158
'One receives punches easily.'
Linguistic Fieldwork Daniel L. Everett
page 159
Passive
(6.150) Maçãs se vendem por aqui.
'Apples are sold around here.'

Ergative
(6.151) A janela quebrou se (*a proposito).
'The window broke.' (*on purpose)

Inherent
(6.152) Tres meninos se desmaiaram. (Brazilian Portuguese)

Each of these uses of se from Portuguese (and the phenomenon repeats itself
throughout Romance) is different in function. So is there one se or five? Just about
every hypothesis imaginable has been proposed in the literature at one time or another
(see Everett (1996, ---) for a summary).
The answer to this particular problem is not relevant to our present concerns
(though I believe that this is only a case of apparent polysemy and that all these
functions are related in a simple way). However, the general problem is relevant. What
evidence could a fieldworker bring to bear on such a question if encountered in the
field (and the odds are high that something like this will confront most fieldworkers).
Historical evidence might help, but it is problematic for several reasons. First, it is often
unavailable to the fieldworker. Second, even if it could be proven that se entered Latin
as a single morpheme, that would not entail that it has been preserved as such in any
modern Romance language. Things change! We saw a similar problem with the Salish
facts in () – () above.
The ultimate answer to this question of homonymy vs. polysemy is found in the
linguist's analysis and argumentation. The most satisfying, clearest, and simplest
analysis will likely be the one linguists adopt as 'the answer'.

6.10. CLEANING UP YOUR MESSES –FOLLOW-UP VISITS

Field researchers must always allow time following the evaluation of data they
have collected to find 'holes' in the patterns of their data, for follow-up visits. Follow-
up field trips should be scheduled into a project in response to the possibility of
discovery, incomplete corpora, change in focus, etc. This is a way to build in error-
correction into your project, using these follow up visits to fill in missing bits of the
data and to 'clean up your messes', that is, verify conflicting transcriptions, recordings,
textual organizations or understandings, etc.
Linguistic Fieldwork Daniel L. Everett
page 160
CHAPTER 7. SYNTAX AND SEMANTIC FIELD RESEARCH
7.1. Introduction
Syntactic fieldwork, like all the other structural domains we have considered until
now is very complex and, again, exceeds the capacities of a single chapter. In this
chapter I follow the approach adopted in the previous chapter on morphology and
present types of phenomena that should be accounted for by any fieldworker. I present
these mainly in terms of communicative functions and the formal realizations of these
functions in the syntax.

7.1.1. What is syntax

Let's begin with a definition of syntax. I understand syntax roughly as the set of
restrictions on how words are put together into larger units, the structure of those units,
and the way that these units present and structure information.
What should the linguist give priority to among the potential areas of syntactic
investigation that emerge from this definition? Whatever they discuss about the syntax
and other aspects of the language under study, there will be large, disparate audiences
ready to pore over what has been said, looking for data related to themes they are
interested in, which may not overlap with the fieldworker's original interests in the
description. Of course, no linguist can satisfy everyone. But there is no getting away
from a certain responsibility that field workers incur to make their results useful and
clear to as wide an audience as possible. Therefore, the fieldworker, to repeat a
recurrent theme of this book, must endeavor to read widely and inform themselves with
a systematic reading program.41
In the first part of this chapter, I focus primarily on questions of structure, though
I attempt to give the reader a taste for all the other approaches just mentioned. This
reflects my own training, personal history, and continuing interests, just as any
scientist's decisions are affected by their background. In the next section of the chapter
I discuss the syntactic hierarchy and various assumptions that are widespread among
linguists, some of which need to be rethought, at least by the fieldworker.
The final section of this chapter addresses the role and method of argumentation
in linguistic description. All too often in my experience field linguists fail to recognize
that even descriptive accounts of linguistic items must argue for all their conclusions.
Therefore, I consider the components of good descriptive argumentation.
We begin with a discussion of syntactic field methodology, move to semantic
field methodology, and move, finally, to a discussion of argumentation. Before
beginning syntax proper, let us consider some additional relevant considerations for
syntactic field research.

7.1.2. The fieldworker is not a blank slate

Occasionally one hears fieldworkers advocate – or be advised to produce – so-
called theory-neutral descriptions, in order to serve as wide an audience as possible
(and so as not to be overly biased at the outset of the fieldwork). Insofar as this means
avoiding theory-specific jargon when writing for a general audience, it is reasonable
advice. But it is otherwise naive. Like newborns, fieldworkers are not blank slates.
Moreover, they should not be. The purpose of training is not to make us 'open-minded'

41
I suggest to my graduate students the program I maintained for a good part of my
career, namely, reading 50 pages per day, minimum, seven days a week, ranging across
topics e.g. phonology, syntax, semantics, descriptive linguistics, etc.
Linguistic Fieldwork Daniel L. Everett
page 161
but to make us keenly observant, able to recognize from among subtle alternatives the
best-motivated analysis. To maintain the 'observer's edge', we need to be slightly
biased.
The fieldworker cannot avoid learning and interacting with at least some
linguistic theory. This is because all of us accumulate a perspective on how language
works during our linguistic training and experience, either from our teachers, our
reading, or our natural predispositions coupled with our experiences analysing
linguistic data. And theories should not be avoided. They provide a framework of
expectations (a source of the 'abductive surprise' discussed in ___ below) and a source
of ideas (the 'abductive 'B's/hypotheses' also discussed in ___). And theories provide
strategies and boundaries for arguing for conclusions. They are essential training for
thinking more precisely. The choice of a specific theory is less important than the
recognition that some theory must be selected. However, other things being equal, it is
better to choose as your theoretical base a theory that has a track-record of
crosslinguistic usefulness (which is why I suggest that fieldworkers familiarize
themselves with ROLE AND REFERENCE GRAMMAR (VanValin and LaPolla (1997) in
addition to any other theory they might choose to work with).
Theories can be misused of course. They are misused when a linguist fails to
consider analyses or overlooks facts that have no obvious 'niche' in his or her own
theory. They are also abused when the linguist simply translates their analysis into
theoretical terminology, i.e. when the theory is only used 'for show' and is never
causally implicated in the analysis proper, an all too common occurrence. Syntactic
fieldwork always confronts the fieldworker with structures unlike any the have
confronted previously. Often a structure that is 'weird' and apparently irrelevant for a
particular theory's objectives will be ignored. But the same structure could turn out to
be very important for linguistics more generally. This is why a fieldworker must not
only familiarize him or herself with a particular theory, but also read widely, beyond
the boundaries of that theory.
Ultimately, the fieldworker should strive to simply give an honest day's work for
an honest day's pay. That means that they do their best to present the language (the
aspect(s) of it they are concerned with at any rate) clearly and fully (never intentionally
sweeping facts under the rug which appear to bear on analysis at the center of the
fieldworker's description). In return they receive career-related emoluments. To do this
work, they certain basic concepts about syntactic field research, to which we now turn.

7.1.3. Crucial syntactic concepts for fieldwork

Probably ever syntactic theory recognizes that the syntax of most, if not all,
languages will be composed of constituents of different sizes (though see ___ below on
recursion for a discussion of ways syntacticians differ). Almost all linguists agree on
the following, in one way or another:

(7.1) Syntactic hierarchy

a. words
b. phrases
(c. clauses)
d. sentences
(e. paragraphs)
(f. discourses)
(g. conversations)
Linguistic Fieldwork Daniel L. Everett
page 162

Some theories, not all, recognize clauses, paragraphs, and conversations (as
indicated by the parentheses). All linguists recognize that discourses and conversations
exist, but not all theories have a place for them (most lack any well-grounded treatment
of conversations). But the items in () not in parentheses are universally recognized. So
at a minimum the syntactic description of any language should include a discussion of
each of these constituent types, how they are put together, how they relate to one
another, how they are like one another, and how they differ from one another. In
addition, some linguists believe that there is a correct order of analysis, such that one
should begin with the largest constituent, e.g. the conversation (but this is theory-
dependent), and then work one's way down the hierarchy, one layer or level at a time.
Each constituent is built out of or 'manifested by' constituents at the next level
down, though the precise way in which one constituent is built out of another varies.
The standard article on the building or analysis of syntactic constituents is Wells (), still
worth reading today. The methodological implications of this hierarchy are discussed in
__ below.
Note however, that there are huge differences in the cohesiveness of syntactic
units as we proceed 'up' the hierarchy. This is important because it ultimately entails
different methodologies and forms of analysis and argumentation for different levels. It
is also important theoretically, because it could indicate that levels of this hierarchy
have different theoretical status (consider, for example, Chomskyan theory's refusal to
consider anything above the sentence as syntax proper. This is not an entirely
unmotivated position to take, as Everett () argues in depth. But, methodologically at
least, discourses and sentences must be studied in relation to one another.) A
conversation can vary tremendously in the units that compose each of its exchanges and
these could in principle be rearranged in a number of ways without loss of
intelligibility. At the other extreme, the word level, constituents are much more rigidly
ordered (but see Bickel, et. al. (2005) for a potential counterexample). Moreover,
speakers' varying ability in constructing these units becomes at once more obvious and
more crucial as the units get larger. For example, it makes little sense to ask 'Who
forms words best around here?' But it does make sense, and the fieldworker should ask,
'Who tells stories best around here?' before deciding on a language teacher for (certain
kinds of) discourse studies and text collection.
Therefore, each level of the syntactic hierarchy is associated with a specific
methodology. I want to begin the discussion of syntax proper by a review of
fundamental notions of syntax shared by most linguists.

7.2. Dominance and precedence (or Configurationality and linearity)

Two fundamental ways in which discourses and conversations differ from
constituents lower on the hierarchy are the relevant restrictions of configurationality
and linearity (or dominance and precedence) at each level. Consider the following:

(7.2) Words
a. running
b. * ingrun

Word-internal constituents, in general, perhaps universally, cannot be moved. So

placing –ing at the beginning of the word, as in (b) produces severe ungrammaticality.
Moreover words allow a range of configurational relationships, ranging from binary
Linguistic Fieldwork Daniel L. Everett
page 163
branching (as in the English examples above) to the periphrastic constructions
discussed in chapter ___. Phrases, on the other hand, have slightly less rigid linearity
constraints:

(7.3) Phrases
a. John's book.
b. *book John's
c. book of John
d. *of John book
e. book of John's

Phrase-internal order is also fairly rigid (and (e) shows that of and –s are not
simply 'allomorphs'. Yet, as (a), (c), and (e) show, it does allow some flexibility.
Clauses and sentences show even lesser constraints on precedence and, in some
theories, greater freedom in configurational relationships. So, for example, in some
languages prepositional phrases are only binary branching, with only one level of
branching, though clauses in the same language may have various levels of branching,
such as when prepositions allow no modifiers, only a head and a complement, whereas
clauses are complex. (Yarawara, Dixon (2005) is one such example.)

(7.4) Clauses and sentences42

a. John ran to the store.

b. ?Ran John to the store.
c. To the store ran John.
d. To the store John ran.
e. *To the ran John store.
f. *To the ran store John.
g. Yesterday, John ran to the store.
h. John ran to the store yesterday.
i. John ran, yesterday, to the store. etc.

And discourse orderings are freer yet:

(7.5) Discourse
a. John came in. Peter asked him if he wanted a drink. John sat down later.
b. %Peter asked him if he wanted a drink. John sat down later. John came in.
c. %John sat down later. John came in. Peter asked him if he wanted a drink.
d. John was writing. Peter was writing. Susy was writing.
e. Peter was writing. John was writing. Susy was writing.
f. Susy was writing. Peter was writing. John was writing.

42
Morphological constraints can supplant syntactic restrictions. So in Latin (and other
languages with rich morphology) word order is relatively free, because morphology is
so consistent.
Linguistic Fieldwork Daniel L. Everett
page 164
Discourse orderings are interestingly and significantly different from clause
orderings because they are constrained by very different principles. For example, (b)
and (c) are anomalous (%) because in the way we normally understand the discourse,
Peter cannot ask John anything until he comes in. And John can't sit down until he
comes in. But now imagine that John is just outside the window near a lawn chair. Peter
asks him, through the window if he wants a drink. John sits down after that, thinking he
is going to be served outside. He finds out he needs to come in for his drink, so he does.
This little scenario renders (b) just fine. And a bit of rethinking makes (c) similarly
fine. So the % only indicates that the order is strange in some contexts, but not in all.
And yet the ungrammatical examples of words, phrases, and clauses seems
unsalvagable by context. They are always bad. Conversational linearity is constrained
by yet different principles:

(7.6) Conversation
a. A Hi. Wanna beer?
B Yeah, sure.
b. % A Yeah sure.
B Wanna beer.
c. A So, what were you up to yesterday?
B Oh, about 5'8". Ha ha. Oh, not much really.
d. % A Oh, about 5'8". Ha Ha. Oh, not much really.
B So, what were you up to yesterday?

Generally, conversations are structured in pairs of utterances, at a minimum,

exchanges between different conversational participants. There are ways to think of the
%'d examples such that they become more or less acceptable, however, as with the
discourse examples. Although discourse does not share the 'interlocutor exchange'
structure with conversation, it does share with conversation the fact that context and
imagination can render otherwise (context-free) anomalous utterances or exchanges
sensible and acceptable.
Therefore, at each level of the syntactic hierarchy, the linearity principles differ.
Also, the overall structure, the configurationality of the constituents at different levels
of the hierarchy is achieved by different principles (one structures a novel differently
from a noun phrase, for example).
No one theory of syntax offers a comprehensive treatment of the syntactic
hierarchy (indeed, as indicated by the parentheses in (), not all theories even recognize
all the constituents posited). Tagmemics (Pike & Pike (1976)) attempted to do so, but
the results were hard to follow at best and unconvincing and confusing to most
linguists. Therefore, the fieldworker will have to supplement his or her chosen theory
of expertise with additional, eclectic readings on the other constituents, rather than
finding all the answers in any one study, book, or theory.
We now want to consider another notion that is basic for many linguists.

7.3. Basic word order

Linguists often talk about the 'basic word order' of a language. My guess is that
most linguists that use this phrase know that it is fundamentally imprecise. Each word
of the phrase is problematic. So, for example, 'basic' has no scientific significance, i.e.
it is not a technical term of linguistics. The term 'word' is also imprecise because what
most linguists mean here is 'immediate constituents of the sentence or clause', rather
Linguistic Fieldwork Daniel L. Everett
page 165
than words, which are usually themselves not immediate constituents of the sentence or
clause. And, finally, 'order' is not very useful either since to many (but by no means all)
linguists what is crucial is the hierarchical structure of phrases, not their linearity.
Given this imprecision, it is not surprising that what linguists actually mean by
'basic word order' includes several concepts, not all clearly related. At least the
following are included in one way or another:

(7.7) a. Information-theoretic default constituent ordering.

b. Discourse-functional constituent ordering.
c. Initial/underlying constituent configuration – the hypothesized configuration
that serves as the basis for non-default orderings.

Consider ( a). In most or all languages there is a default information-theoretic

ordering of constituents (see, for example, Van Valin (2005) and Lambrecht (1994))
and __ below. English usually reserves the subject position for old information and the
direct object position for new information. In fact, many languages (Van Valin and
LaPolla (1997--)) place new information immediately to the left or right of the verb and
the old or topical information farther out or on the other side of the verb. To get a feel
for how information structure works, consider the familiar examples in (), from Van
Valin (), in ___ below.
So for a given language, it is important that the fieldworker understand (by
examining constituent orderings and intonation), among other things, how topic and
focus are distinguished (see ___ below). To some people, this will be the core meaning
of 'basic word order'. But there is complementary concept that any fieldworker could
profitably consider, namely, discourse function of different constituent orders, not so
directly tied to information structure.
Discourse functions of sentences involve things such as the use of passive clauses
to mark backgrounded information or information with some other special purpose in
structuring the discourse, or the use of one constituent ordering to open a discourse but
a different ordering to mark denouement. Or subject-less or object-less sentences to
mark the discourse body but use of full NP subjects or objects to indicate the beginning
of end of a discourse. If such discourse notions are found to be useful in a given
language, they may very well differ practically from the other concepts in (). If such
different discourse functions exist, they may be discovered in one of several ways.
First, there is no available excellent software for discourse filing (see the websites
mentioned in ___). Second, there are some still useful manual methods. One method I
used to employ was to color code sentence types throughout texts (e.g. red for active,
blue for passive, etc., depending on the types of sentences you are interested in
tracking) and then compare discourse genres and discourse functions or positions (e.g.
beginning, climax, denouement, end) to see if there were patterns to the colors. This is
a psychedelic experience for those of my generation while also being a useful field
methodology.
Let's proceed in our discussion of major syntax notions to the concept of
dislocation.

7.4. Dislocation

Most formal theories, to one degree or another, concern themselves with the
phenomenon of 'dislocation', i.e. the situation where related meanings are expressed by
Linguistic Fieldwork Daniel L. Everett
page 166
different word orders or where some constituents are not where they might be in a
simple declarative. These theories often take one of the orders as underlying or default
or basic and derive the other orders from this. So consider examples like those in ()-():

(7.8) a. John saw James.

b. James was seen by John.

(7.9) a. The mail is here.

b. Is the mail here?

(7.10) a. John likes Mary.

b. Who does John like?
c. John likes who(m)?

(7.11) a. John thinks you like Mary.

b. Who does John think you like?
c. Who does John think likes Mary?

(7.12) a. John thinks that you like Mary.

b. Who does John think that you like?
c. *Who does John think that likes Mary?

In most theories, the (7.8a)-(7.12a) examples represent the basic forms and the
other forms the derived forms. The question words in () –() are said to be 'dislocated',
i.e. not in their 'expected' positions (where they would be in a simple declarative).
Harris (1947) proposed that such sentence alternations should be understood in terms of
specific discourse functions for each alternate word order (which he analysed as
'transformations', but without a sense that one was derived from the other). Formal
theories have more elaborate means of accounting for dislocation. But whether one
works in a formal theory or not, the fieldworker must come to grips with Harris's
observations, because they are vital to understanding the grammar of the language in
question. It is perhaps always the case that different word orders have different
discourse functions. And describing those functions is part of an adequate descriptive
grammar or understanding of a language's syntax. Whatever theoretical account of such
facts the fieldworker wishes to suggest is welcome, though many (myself included)
would see this as secondary to finding out the communicative functions of the
alternations or so-called dislocations.43
Also, it is important to note that dislocation, while a prominent part of the syntax
of many languages, is insignificant in many others (except in very abstract, theory-
internal ways that are of little use to the average fieldworker, at least in my experience).
Therefore, if the fieldworker goes to the field primarily trained in a theory with a strong
emphasis on dislocation, to the exclusion of, say, information structure, then, however

43
Urban () contains a number of suggestive analyses illustrating potential cultural
implications of discourse structures, e.g. the use of the passive vs. the active and the
types of heroes that emerge, say, in cultures with high percentages of passive sentences
for major characters, vs. those in which active sentences express the actions of main
characters. .
Linguistic Fieldwork Daniel L. Everett
page 167
interesting and important that theory may otherwise be, they are likely to find
themselves relatively 'unarmed' in the face of facts which are complicated, but where
dislocation is not part of the complexity.
Before proceeding to a consideration of discourse analysis and syntactic
elicitation methodology, I want to consider another notion that is widely assumed by
many syntacticians to be universal, but which also seems to have exceptions. Such
exceptions, if valid, require that the concept not be used as an obligatory mold for
structures in the grammar.

7.5. Tree-structure
7.5.1. Endocentricity
Another widespread assumption about syntactic structure, built into the
foundational assumptions of many formal theories, is that all syntactic structures are
endocentric. As Hudson (2002, 5) says: "Almost all modern theories of syntax accept
endocentricity: every phrase has a single head which determines the characteristics of
the entire phrase. " In this section I want to give reasons why the fieldworker should not
assume that all structures are endocentric. So consider the following sentences from
Wari':

(7.13) [S [Nuc(leus) [S [Nuc Ma' co mao] na -ini

that:prox:hearer m/frp/p go(sg) 3s:rp/p -3n
Guajarái]]
Guajará (Brazilian city)

naj -namk 'oro narimak' taramaxiconj].

3s:rp/p -3pf collective woman chief
'"Who went to Guajará?" (said) the chief to the women.'

(7.14) [S [Nuc [S Cao' xi' carawa]] nana hwijima'].

eat 1pincl:rf animal 3p:rp/p children
'The children will eat food.' (lit: '"We will eat food," the children (say).')

As Everett (2007) argues, these sentences have a predicate that is not a verb, the
embedded quotative clause, and therefore they are exocentric constructions, but fully
productive in Wari' grammar. This means that any attempt to analyze Wari' clauses
based strictly on endocentricity will fail.

7.5.2. Simple syntax vs. complex syntax

In an extremely interesting and important new approach to syntactic theory,
Culicover & Jackendoff (2005:5) state their hypothesis as follows:

"Simpler Syntax Hypothesis (SSH)

The most explanatory syntactic theory is one that imputes the minimum structure
necessary to mediate between phonology and meaning."

Culicover and Jackendoff go on to demonstrate throughout their book why the

structure hypothesized for a given construction should be guided by the SSH rather
than the standard 'theorems' of phrase structure in syntactocentric theories, e.g.
Minimalism (Chomsky 1995). Most linguists share the view that figuring out the
Linguistic Fieldwork Daniel L. Everett
page 168
hierarchical constituent structure of a given language is a very important and essential
step to understanding the language. Naturally, for such an important area of linguistics,
there are various theories on the constraints on possible tree structures, including the
SSH.
Since Kayne ()'s work on restricting tree structures, many generative linguists
have become persuaded that trees are binary branching, i.e. that (), but not (), is a
possible structure of a human language. The SSH, on the other hand, would allow
either, depending on the facts of the language under investigation.

(7.15) /\
/\
/\
etc.

(7.16) /|\
/\/\
/\
etc.

It may in fact be possible to fit any language into either of these structures, more
or less convincingly. In fact, it might even be possible (see section ___) to avoid tree
structures altogether and use only violable (e.g. Optimality Theoretic, Prince and
Smolensky (1993)) constraints on linear precedence in conjunction with semantic-
based adjacency to account for the syntax of some languages. We might call the binary
branching hypothesis 'complex syntax', the non-binary branching hypothesis in () the
'simple syntax' hypothesis, and the absence of tree structure in a language (whether the
entire language or only some structures) the 'no syntax' hypothesis. From my own
experience, each one of these views has advantages and disadvantages for the
fieldworker (though the inductive simple syntax and no syntax possibilities are spiritual
kin, complex syntax has nothing in common with them in its strong deductive
approach, forcing structures without considering alternatives).
Following the complex syntax hypothesis may be a requirement of the theory
you have adopted. But it is no less a procrustean bed for that. If a field linguist refuses
to consider non-binary structures, they may indeed produce a description fully
compatible with their chosen theory, yet utterly lacking in insight into various aspects
of the language and of little long-lasting value (remember the fate of the famous
Hidatsa grammar, discussed in ___).
The simpler syntax model may or may not be restrictive enough. My general
rule of thumb is to propose the most restrictive trees possible (binary) initially, but
always look for (and accept) evidence that this analysis is wrong, either relaxing it to (
b) or ( c) (on the latter also see ___). For example, consider the following discussion of
English syntax from Culicover and Jackendoff, based on English examples like those in
(7.17) and (7.18):

(7.17) I showed Maryi herselfi.

(7.18) *I showed herselfi Maryi.

Following the standard assumptions of complex syntax, i.e. imposing a deductive

structure on the data that requires c-command to capture binding facts, without concern
Linguistic Fieldwork Daniel L. Everett
page 169
for the thesis of SSH, Larson (1988) argues that such data require binary trees plus a
special 'verb shell', with subsequent movement from V1 to V2:

(7.19) VP

V2 VP

NP v'

Mary V1 PP

showed herself

However, Culicover and Jackendoff (2004, 52ff) argue that simple precedence
could account for such examples by requiring antecedents in the VP to precede their
anaphors (this grossly oversimplifies their account, so the reader is advised to consult
Culicover and Jackendoff (2004) for full details of their model).
Let's move now to briefly consider what a language without hierarchical syntax
might be like, drawing from some suggestions based on my own work on Pirahã.

7.5.3. No syntax
Almost every language manifests the property of recursion, what some linguists
consider to be the foundation of syntax (see Hauser, Chomsky, and Fitch (2004)).
Simplifying slightly, recursion has two forms:

Embedding:
(7.20) A → AB

System recursion:
(7.21) a. A → BC
b. B → DE
c. C → AF

Linguists vary in the importance they attach to recursion. So, for example,
Hauser, Fitch, and Chomsky (200??) argue that recursion may be the only (or core,
depending on interpretation) component of Universal Grammar. Others recognize it,
but do not attribute such genetic or species-defining importance to it.
Everett (2005) argues that there is no evidence for recursion in Pirahã. And there
is some reason to believe (Hale (), Dixon (), Austin (p.c.), Nordlinger (p.c.) that
recursion may be largely lacking from some Australian languages as well. In any case,
the fieldworker must exercise care and caution in attributing the property of recursion
(or indeed any property) to syntactic structures. This is because recursion is not the
only way of combining syntactic units of one level of the hierarchy into another.
Another way to put units together is parataxis (Bloomfield ()), where two more units
are set side by side as it were, with no further structure, in particular without one
structure being contained inside another of the same level. So, for example, () is the
result of recursion, () is parataxis, and () is arguably parataxis as well:

(7.22) [NP John's [NP father's [N hat]]] is nice.

Linguistic Fieldwork Daniel L. Everett
page 170

(7.22) is recursion, because one NP is found inside of another.

(7.23) The lion, the witch, the wardrobe, what have you, are all fictional objects in this
story.

The first part of (7.23) is potentially a case of parataxis. The phrases (the lion, the
witch, the wardrobe) are linked paratactically, rather than recursively (at least by my
analysis). As another example of parataxis, consider Tagalog.
Some linguists, e.g. Schachter and Otanes (1972 - PAGE) and others have
pointed out that some Tagalog sentence structures are like equations, where the two
sides are brought together in a larger unit (the sentence) via parataxis:

(7.24) Predicate/Comment Topic

(7.25) Titser ang babae.

teacher nom woman
‘The woman (is a) teacher.’

(7.26) Maganda ang babae.

stative-beauty nom woman
‘The woman (is) beautiful.’

(7.27) Umalis ang babae

leave-actor focus, completive nom woman
‘The woman left.’

Therefore, we can conclude that, at least for some constructions in some

languages, recursion is irrelevant. Further examples can be found in Pirahã, according
to Everett (2005). Consider the following:
First Pirahã lacks coordination, disjunction, and embedding. It uses non-
recursive 'circumlocutions' to express each one of these structural types.

COORDINATION:
(7.28) Ti ?isigihii kohoaipi. ?itii?isi pi-ai.
I eat meat. Fish does also/now.

(7.29) Ko?oi (hi) kohoaipi. ?aibigai (hi) pi-ai.

Ko?oi he eats. ?aibigai does also/now.

DISJUNCTION:
(7.30) Ko?oi (hi) kohoaipi. Hai, ?aibigai (hi) kohoaipi.
Ko?oi eats. Hmm. ?aibigai eats.

(7.31) Hi ?isigihii kohoaipi. Hai hi ?itii?isi kohoaipi. Ti kos-aaga.

He eats meat. Hmm. He eats fish. I am ignorant (do not know).

Notice that in the following quotative, it is the verb of saying that bears what
Everett (1986) analyzes as the nominalizer, rather than the otherwise 'embedded verb'.
Linguistic Fieldwork Daniel L. Everett
page 171
And yet in the subsequent apparently embedded clause 'arrow-making', the 'embedded'
verb has the nominalizer. Everett and Van Valin (in progress) analyze this –sai as a
marker of secondary or topical information, rather than as a nominalizer. The same
analysis would hold true for (). The suffix –sai is in fact a verbal form and the
language has a nominal corresponding to –sai, namely –si (found on the
autodenomination of the language of the Pirahãs, ?apaitíisi. The nominal suffix marks
nominal secondary or topical information (see Everett (1986)) for details. Many other
examples are discussed by Everett (in progress) on Pirahã and language evolution.

EMBEDDING:
(7.32) ti gái -sai kó'oí hi kaháp -ií
1 say -nom. name he leave -intention
'I said that kó'oí intends to leave.' (literally 'My saying kó'oí intend-leaves')

(7.33) hi ob -áa'áí kahaí kai -sai

he see -attractive arrow make -nom.
'He knows how to make arrows well.' (lit: 'He sees attractively arrow-
making')

One way to account for the apparent lack of recursion or configurationality of

Pirahã is to adopt a no syntax hypothesis and analyze the Pirahã facts in terms of lexical
semantics and linear precedence, as in (7.44):

(7.45) Linear Precedence + Lexical Semantics (> = 'immediately precedes'):

Agreement > NUCLEUS
NUCLEUS > Modifier
Topic > Comment
Argument > Agreement

These rules capture most of the facts of Pirahã 'syntax' without recourse to
recursion. And I would be surprised if Pirahã were the only language for which the no
syntax hypothesis is the simplest account of the data.This brings us to a more recent
development of syntactic theory, another place where hierarchical, recursive structures
have little role, namely, syntactic 'constructions'.44

7.6. Constructions
Words (and for some linguists morphemes as well) are considered by most
linguists to exemplify the Saussurian sign, an arbitrary association of sound and
meaning. Syntax, as the general reasoning (used to at least) goes is not arbitrary but
strictly the output of regular constraints or rules. Goldberg (2006, 3) summarizes an
alternative position,

44
We might well ask, of course, why such considerations are relevant for the
fieldworker, as opposed to the syntactic theoretician. The reason it is crucial for
fieldworkers to be knowledgeable and argue well is that if they simply apply the
complex syntax model or some other theory blindly, however 'basic' or 'uncontroversial'
they take it to be, their work can fail to teach all of us what it should have taught us
about the language in question. The fieldworker should take no structures for granted.
Linguistic Fieldwork Daniel L. Everett
page 172

"Many linguists with varying backgrounds have converged on several key

insights that have given rise to a family of ... constructionist approaches. ...
constructionist approaches emphasize the role of grammatical constructions:
conventionalized pairings of form and function."

In other words, there are many linguists who are persuaded that some or most of
syntax is convention and not merely the output of formal rules or constraints. (But cf.
Borer () for an alternative view of constructions.)
Some of the evidence for constructions does indeed seem very strong. So, for
example, consider the 'Xer the Yer' or co-variational construction of English:

(7.46) a. The more you think about it, the less you understand.
b. The hurrieder I go, the behinder I get.
c. The bigger they are, the harder they fall.

This co-variational construction, according to Goldberg (2006, 6) is "...

interpreted as involving an independent variable (identified by the first phrase) and a
dependent variable (identified by the second phrase)." These properties of the
construction do not follow merely by reading the semantics directly off of tree
structures, as some theories might maintain. Rather, they are 'added' to the sentence by
convention, the latter following from the matrix culture (see the papers in Enfield
(2004)). In the view of many linguists, all languages have constructions for which part
of the meaning is compositional and structure-based but where other parts of the
meaning are added by convention. If these linguists are correct, then a fieldworker
armed only with tree structures and form-based constraints will fail to account for vital,
construction-based components of the grammar.
So how would one go about identifying and describing constructions? I suggest
the conceptually simple, though methodologically rigorous, procedure in ():

(7.47) Methodology for identification of constructions

a. Understand the meanings of words and morphemes in the language
(morphology fieldwork, see chapter __).
b. Understand the constraints on syntactic object formation in the language
(chapter __).
c. Understand the syntax-semantics mapping in the language (section __
below).
d. See if any meaning is leftover that is constant for each instance of the
hypothesized construction.

If a particular phrase, clause, or sentence is a construction in the technical sense,

its meaning will not be exhausted by the preliminary steps in (a-c) above, but it will
instead be circumscribed by them (i.e. it will not be completely arbitrary, e.g. an idiom
like 'kick the bucket', but neither will its meaning be a simple sum of the meanings of
its consituents). In other words, the fieldworker is satisfied that they understand (a-c) in
the language under study, the part of the word or phrase that is left over, and that this
leftover portion follows regularly from the construction in its various manifestations.

7.7. Discourse-down analysis

Linguistic Fieldwork Daniel L. Everett
page 173
In this section we consider why it is important to collect the vast majority of
one's data in the form of natural texts. It is the best method for collecting acceptable,
natural examples. We discuss the basic text genres that should be represented in any
data sample – especially what I consider to be the four basic genres: expository (e.g.
teaching someone why certain things in the culture work the way they do), narrative
(e.g. a historical text), procedural (e.g. how to make a particular object or perform a
particular ritual), and hortatory discourses (e.g. exhortations on how to solve a
community problem, sermons, etc.).

Not all cultures have stories about topics we might expect them to have
stories about. For example, when I first told people that the Pirahãs had no
creation myths or discourses about their history, anthropologists were
skeptical.45 One anthropologist actually went to the Pirahã village with the
express purpose of collecting creation myths. His method was to ask in
Portuguese and hope for the best, since he immediately recognized the Pirahãs'
lack of knowledge of more than very rudimentary Portuguese. "How was your
world created?" he asked. He recorded his questions and the answers for me, in
order for me to help him translate the material he was collecting. After asking
the question in Portuguese, he waited for the Pirahãs to translate the question
into Pirahã. "The world is created," replies one of the assembled men in his own
language. "Tell me how your god made all this?" the anthropologist presses on.
"All things are made," comes the answer. The interview lurches on for a few
more minutes, until suddenly, the question and answer session is overtaken by a
deluge of excited banter as the assembled Pirahã vie to be heard.
"I've cracked it," said the anthropologist to me as he handed me his tape
recording a few weeks later. "Here is the Pirahã creation myth." I must admit to
being a bit dubious. In the past three decades, I have spent a total of seven years
living with the Pirahã in the Amazon rainforest and am one of just three
outsiders, along with my ex-wife and a missionary who spent time with them in
the 1960s and 1970s, who is fluent in their language. I have long maintained
that they are among the few people on Earth who have not devised a story to
explain their existence. Others, including this particular anthropologist, find the
idea difficult to accept.
So I listened to the tape. After the short, stilted exchange, some bright
spark points out that this guy asking them odd questions doesn't know their
language, so he will need to get help from me to translate the tapes. "Hello,
Dan!" comes a chorus of Pirahã voices. "How are you?" "When will we see
you?" "When you come, bring us some matches." "And bananas." "And
whisky." And so on. Nice try, but no creation myth here.

7.7.1. Discourse and its relation to syntax

Understanding what the principal formal genres or 'distinguishing parameters' of
discourse are in a given language, how they are structured, and how sentences are
distributed within them, what kinds of information are talked about, what kinds of
information are avoided, and what kinds are implicit, are each fundamentally important

45
This box on creation myths is take largely from the report on my work by Kate
Douglass in New Scientist, March 18, 2006, a story I told her during her research.
Linguistic Fieldwork Daniel L. Everett
page 174
to the fieldworker's quest to understand the language they are describing. As Chafe
(1994), among many others, demonstrates, there is much to learn about culture,
cognition, and grammar in different languages via the study of discourse. With regard
to content and cognition, as James (1890, 1:243) [The principles of psychology, 2
volumes, New York: Henry Holt. Reprinted 1950 by Dover Publications, New York.
]puts it (also cited by Chafe (1994, 3)):

"As we take, in fact a general view of the wonderful stream of our

consciousness, what strikes us first is this different pace of its parts. Like a
bird's life, it seems to be made of an alternation of flights and perchings. The
rhythm of language expresses this, where every thought os expressed in a
sentence, and every sentence closed by a period."

On the other hand, as the generative tradition has shown over the past several
decades, a large portion of syntactic structure does not simply 'emerge' as a by-product
of discourse study. Among other things, discourses do not provide systematic
ungrammatical examples – and yet these are vital for adequate analyses. The
fieldworker should therefore conduct careful investigations of discourse structures,
realizing, however, that these investigations will not be sufficient, however necessary,
for a full analysis of the syntax of the language.
However, there are significant methodological advantages in beginning one's
syntactic investigations by working 'downward' from the discourse. We consider a few
of these in the remainder of this section.
Before you have learned to speak the language or studied its structure extensively,
there is a plethora of opportunities to be misled as to the appropriateness,
grammaticality, or translation of different sentences. There is ultimately no way to
avoid this (another reason to remember Postal's Maxims). On the other hand, beginning
syntactic study with the discourse ensures that all data are embedded in an appropriate
context, are (likely) felicitous, and are grammatical (or at least acceptable). Thus, in
this sense, the data are more reliable at the initial stages of analysis than the isolated,
de-contextualized test sentences that might otherwise be used in elicitation.
Also, by studying sentences in their discourse context, the way the language
presents and structures information can be studied more easily, naturally, and
effectively in the initial effort to understand how it works. Eventually, other methods
will be necessary, but it is difficult to imagine a more effective way of beginning
studies of information structure than via the discourse.
Further, intonation can be more studied with more confidence at the outset (see
Chapter ___ below for many more detailed suggestions on the study of intonation)
because it is naturally contextualized when looked at initially in discourse. Initial
hypotheses on intonation can at least assume that the intonation recorded in discourse is
appropriate and natural and that it fits the particular information structuring of that part
of the discourse.
Another advantage of beginning syntactic analysis with the study of discourse is
that the entire array of sentential features, e.g. grammatical voices, constituent
orderings, moods, evidentials, pronoun omission, etc. are observed in natural
distributions. This is invaluable to undertaking their study. Finding them in appropriate
linguistic environments is almost inconceivable without initial study of natural
discourses. Of course, the initial analysis will not exhaust their function and structures,
but it will provide a better beginning point.
Linguistic Fieldwork Daniel L. Everett
page 175
Functions of different sentence types will also be more readily visible when
looked at in natural discourse initially. For example, are certain sentence structural
types, certain phrase types, or words more common at the beginning, middle, or end of
a particular discourse genre? Even such simple questions could provide important clues
and insights into the final analysis of specific structures and constructions. But this
simple source of insights depends crucially on studying discourse at the outset of
syntactic research. And, as we see in ____, Lowe () argues that understanding discourse
is vital to understanding morphology and lexical meanings in many languages.

7.7.2. Discourse genres and analytic methodology

This section makes no attempt to offer anything approaching a full introduction to
discourse analysis. The main reason for this lack of ambition is that discourse analysis
is an interdisciplinary set of research programs and not a simple step-by-step set of
tasks. This interdisciplinarity is natural since discourse is the nexus of culture,
philosophy, language, literature, cognition, social relationships and conventions, and
grammar, among other things. Arguably, our discourses are essential to identify and
distinguish languages, cultures, and psyches of individuals and groups. For that very
reason, discourse analysis is a vital part of field research. That is, if I am right to argue
that an understanding of a given grammar should go hand in hand with an
understanding of the containing culture, then discourse study becomes more important
yet, as the crucial link between the two.

7.7.2.1. Discourse genres

Discourse genres or types or varieties, etc. are hard to summarize. What might be
to one fieldworker two tokens of the same genre could be to another fieldworker
examples of very different types of discourse. Distinguishing discourse genres is
largely a matter of what your background, expectations, and objectives are.
Nevertheless, there are some distinctions that one should keep in mind, as a set of
suggestions.
One type of genre that might be encountered in many societies is what we can
label 'stylized or belletristic texts', that is texts that form part of the acknowledged
cultural traditions and oral literature of a particular culture. These texts, regardless of
their content, are very important because they represent careful reflection on both form
and content and can reveal a tremendous amount about the culture and about the
linguistics of discourse. However, at the outset of the field research they can be
somewhat problematic. The reasons include the facts that they often include stylized or
archaic language that is not idiomatic (which will be important to the eventual analysis
but confusing at first) and they can be very long. They will be much more involved,
harder to figure out in some ways. Moreover, since transcription time is roughly 5-6
times greater than actual recorded material and in fact much more at the beginning, this
type of text can compromise your freedom to look at other things in the grammar. And
if texts are not processed right away, at least not while in the field, they are simply
worthless. Therefore, though no grammar would be complete without a careful
discussion and analysis of such material, I recommend that one try to begin field
analysis of discourse with simpler texts, e.g. simple reports.
The field linguist should also be sensitive to the possibility that speakers of the
language may have their own recognized genres which may or may not correspond to
the field linguist's own categories. Treat these differently if you do encounter such
Linguistic Fieldwork Daniel L. Everett
page 176
cases. Almost certainly there is something in content or form that will distinguish them
from one another and it will be worth your time to attempt to figure this out.
For a linguist, the most important classification of discourses will be in terms of
their structural distinctions, things such as predominant tenses, aspects, construction
types, overall organizational principles, etc. For example, let us consider four types I
find useful to distinguish in my own fieldwork, namely, narrative, procedural,
expository, and hortatory. To these we can also add religious, brief reports and informal
discourse, and, last but by no means least, conversations.
Narratives are the easiest to study. They tend to use mainly indicative or a related
mood, e.g. realis, and an array of past tenses. It is interesting to observe how the
narratives are organized. For example, are events and participants introduced or
presented in chronological order or in some other way, e.g. cultural relevance, logic,
etc.? How are main participants distinguished from secondary participants? What kinds
of sentences mark the main event line, as you understand the text? Do different kinds of
sentences mark events that do not advance the main events of the discourse? How about
the way that new participants are introduced? Are all characters of the discourse
introduced in the same way? Or does it turn out that characters central to the telling are
introduced differently (e.g. more sentences about them, different aspects or moods,
more relative clauses, etc.) than peripheral characters? Are mythological characters
treated formally different from 'real' characters? And so on. The possibilities are almost
literally endless. Part of what distinguishes a mediocre field research from an insightful
fieldworker is the effort invested to interrogate structures from different perspectives,
formal, functional, cultural, societal, historical, etc.
Procedural discourses, things like how to make a bow and arrow, how to bake a
cake, how to catch a fish, etc. involve step-by-step procedures that can illustrate moods
and tenses (e.g. imperative and sequential, respectively) that are not found as readily in
narratives. Since the subject of most procedurals is the addressee ('You do this, then
you do that', etc.) then there is more likely to be subject omission in procedural
discourses, due to the facility of subject identification.
Expository discourses explain things about the world. They may not turn out to be
a lot different from simple narratives. But it is useful to begin by assuming that simple
story-telling could be different from explanations and to examine these types of texts
carefully for distinguishing formal and cultural characteristics.
And finally, hortatory texts are a superb source of cultural and linguistic insights
(if the language in fact formally distinguishes them, though it likely will). These will be
a rich source of deontic modals, different types and scopes of negation, cultural values,
aspects, subject forms, etc.
With this barest list of suggestions on how to approach the general issue of
discourse, let us consider in more detail an accompanying methodology.

7.7.2.2. Discourse analytical methodology

First, it is vital to have video or audio recording equipment present at different
events associated with different discourses or the whenever there is a likelihood of
recording specific texts, e.g. when someone returns from hunting, preparation for ritual,
etc. Second, you must be creative in getting what you are after across clearly to your
language teacher. Often in my experience, language teachers have felt that it is
somehow wrong to give complex answers and prefer to give short, one sentence or one
word answers to questions. Unless one can effectively communicate to the language
teacher what one is after, the field session can be very harrowing and unrewarding. And
Linguistic Fieldwork Daniel L. Everett
page 177
it is often just hard to get speakers to give a discourse in such cases. One suggestion is
to build up texts line by line, by going back and forth between speakers to get more and
more information on a particular topic and then ask someone to recap, rewarding the
person who tells the longest story about it.

Taperecorders have a use beyond recording of sounds. They are essential in discourse
analysis as well. For analyzing discourse, I recommend that you use at least two
recorders, or a recorder and a computer, etc. That is, that you have one device dedicated
to playback and another device dedicated to recording. Simple analog recorders with
'cue/review' buttons are the most convenient, but if you have a computer for playback
this can be almost as easy to use. First, record a text with a native speaker. Second,
transcribe the text and translate it with a native speaker. Third, go through the text with
another native speaker (there are some ethical considerations here: make sure that the
first speaker has given you permission to play the tape to others in the village, that the
tape contains no sensitive or embarrassing material, and that the genre is one
appropriate for a general audience, or carefully select language teachers that can and
are willing to help you with sensitive material). This is done by playing back a portion
of the tape and asking the new language teacher to repeat what was said, slowly.
Record what the new language teacher says, including their comments about the text.
This slower repetition will almost certainly introduce changes of pronunciation and
lexical choice, but that is good. You want paraphrases and alternative pronunciations.
The commentary provided will provide cultural and linguistic insights that are difficult
to obtain in any other way. For a particularly interesting or challenging text, I would
ask up to three additional speakers for comments, corrections, and translations.

After collecting texts, however obtained, they need to be processed. One set of
suggestions that I have found useful is: (i) record the text digitally, with the speaker
using a headset (this is discussed in the chapter on phonetics); (ii) transcribe it the first
time alone, with no speaker present; (iii) check your transcription with the speaker who
gave the text originally and write their corrections and comments in a different colour
of ink. This will require also the use of two recorders. One to play back the original
recording and one to record the speaker's comments. (Alternatively, one can record
directly onto the computer. But I recommend separate digital or analog recorders and
then transferring the data to the computer so as to have automatic backups of the data,
among other reasons.); (iv) check the second transcription with another speaker and use
two recorders, again, one to play the original text and one to record the second
speaker's comments, corrections, etc.; (v) do a four-line transcription of the entire
corrected text (all of this in the field!), with the following lines: Line 1 =
morphophonemic transcription; Line 2 = phonemic/underlying segments transcription –
these will differ occasionally; Line 3 = gloss; Line 4 = free translation. These are
illustrated in the following example from Wari':

(7.48) Mon womu cara ne

ma'-on womi-u
that:prox:hearer-3sm cotton-1s that:rec. rec:past
'Where are my clothes that were just here a minute ago?' (lit: '... my recently
absent clothes?')
 Linguistic Fieldwork Daniel L. Everett
page 178
Also, above the top, or morphophonemic, line one could also add an intonation
line, e.g. in ToBI format (see chapter ___ for much more detail). This entails a
significant amount for each text transcribed, but I will discuss why this is worth it and
suggest that not every single text needs this level of detail. (The reader should also refer
back to section 4.7.1. for more on metadata.)
Last but not least, all field linguists need to attempt to record conversation. The
box here gives some initial bits of a conversation recorded in July 2004, among the
Banawa people of Amazonas, Brazil.

Consider the following examples of a conversation transcribed and analyzed by Ms.

Julia Reinbold, a research associate on a grant of mine.

Speaker A: Batao
Speaker B: Bido

1. A: ta bom. e Bidoo. ebihiyarabone faiyama maniha / da Danieo badi

e Bidoo. ebi-hiyara-bone faiyama maniha / da Danieo badi
Bidu ?-converse-INT. this day ? Daniel big
‘Bidu, let us us talk today.’

2. B: m ee

3. A: e bihiyarabonee e faro ee at i y a / ebihiyarabone Danieo /

e bi-hiyara-bone e faro ee ati-ya ebi-hiyara-bone Danieo
converse-INT ? our ? voice-with ? converse-INT Daniel
‘Let us talk in our language.’

4. B: m

5. A: Danieo kamei hidabani fame / eremee e atini me n o f a fa i e

Danieo kamei hi -dabani -fame ere-mee

Daniel came 3p all together 1PL OBJ-3PL

atini me nofa fa i

language PL like this much

‘Daniel and the others came all together. They like our language much’
Linguistic Fieldwork Daniel L. Everett
page 179
This type of study is extremely important at various levels of study of the
language and culture, revealing not only formal components of conversational
organization in the language, but natural intonation, intersentenial anaphora, and so on.

By carefully looking at syntax, intonation, and discourse from natural data, we

have an extremely useful tool and data source for subsequent analysis.
Of course, getting two people to sit in front of a researcher and converse naturally
is not easy. The Banawa were particularly good at this, however, and although the
conversation began slowly, within a few seconds, both interlocutors were conversing
naturally, looking at each other and not at the researchers, the recorders, or the video
camera. Further, both speakers, Sabatao and Bidu, wore headset microphones and each
recorded onto a separate channel of our stereo digital recorder. If you are unlucky and
cannot get people to converse naturally in front of you (the Pirahã won't for me, even
after all these years), then there are other ways to approach the problem. You can get
the people's permission to leave a voice-activated recorder in their house. You can try
to record phone conversations (our first Suya text was one side of a phone
conversation, recorded more or less by mistake as the language teacher's cellphone rang
during a language lab session). These all have obvious disadvantages, but still they can
produce useful data for discourse and conversational analysis. Other possibilities are to
record radio broadcasts, video discussions of community beliefs, or rituals, etc. When
asking questions you hope will produce discursive answers, be sure that the questions
are framed broadly enough so that a free-ranging answer is more appropriate than a
simple yes or no. Watch how they ask questions on your favorite talk show (modulo the
rehearsal that goes into these overproduced spectacles). How does the talk show host
get guests to open up and talk freely (at least between commercials)? How does a
reporter interviewing a nervous subject? Take notes on technique and apply what you
learn to your discourse-collection techniques. It is particularly important that
researchers avoid framing questions so as to impose their own categories, values,
orderings, abstractions, and length expectations, among others, on the language teacher
(i.e. insofar as possible. It is never possible to completely eliminate our own culture,
nor do we want to because our grammar-writing goals are a cultural value of our own.
The very doing of field research and research more generally are themselves cultural
goals, so we are obviously as bound by our culture during the data-collecting process as
the language teacher is, if not more so. )
Finally, let me conclude with a brief mention of the importance of implicit
information – what is not said (e.g. sample Pirahã text) in studying natural language
discourse. This kind of information can be important in understanding how the culture
works. The reason for that is that what is not said often reflects values and knowledge
that all hearers and speakers of the culture hold in common. Consider for example, the
following hypothetical exchange:

(7.49) Speaker A: Do you want drinks at the reception?

Speaker B: If you want to clean up and referee.

This is a simple example. Members of American (and other Western) cultures

will know that this conversation is about alcoholic beverages and that people who drink
such beverages can get messy and rowdy. That is what the references to 'clean up' and
'referee' refer to, this implicit cultural information which is simply not present in any
theory of the literal semantics of this exchange. This seems obvious to members of the
Linguistic Fieldwork Daniel L. Everett
page 180
culture, but it is not obvious to outsiders and this kind of thing will be puzzling for the
field researcher until they have mastered the culture of the language under study (which
they may never do of course).

7.7.3. Syntactic methods

When a discourse begins, we can be sure that its beginning marks one and
probably several layers of morphosyntacic boundaries (e.g. morpheme, word, phrase,
clause, sentence, discourse). Likewise, the end of a discourse will be the right boundary
of at least one and likely several morphosyntactic constituents (and layers of
constituency).
Also, we can identify morphosyntactic boundaries via discourses by identifying
points at which speakers backtrack to reinterpret the sentence (e.g. for hard to follow or
garden-path sentences, etc.), to correct errors, and so on, even without sophisticated
experimental equipment) .
Spontaneous discourses also are marked by hesitations of different kinds by the
speaker.These hesitations will often come at morphosyntactic or prosodic boundaries
and so may be used as a source of hypotheses for investigating these boundaries.
Each discourse genre can also be expected a priori to be a likely source of certain
kinds of grammatical features. So, for example, in a narrative discourse, most sentences
are likely to be in past tense and indicative mood in most languages. In hortatory
discourses, a large number of sentences are likely to manifest imperative mood and
conditional or other special tenses. In procedural discourses we will get sequential
relationships between sentences, present tenses, indicative and imperative mood, and so
on. In other words, regardless of the theoretical status or relationship between
discourses and sentences (and see Everett () for a review of many ways in which they
might be profoundly different), the study of discourses can provide extremely important
methodological support to understanding sentences. Let's consider some
methodological rules of thumb that I have found useful over the years.
As I have said, textual information alone is inadequate to underwrite a grammar
of a given language. This is because individual sentences need to be studied and
permuted for tests of grammaticality. Nevertheless, sentences are usefully extracted
from discourses for further testing by the following procedures.
First, isolate clauses in the text, using phonology and phonetics to probe for
relevant constituent boundaries, as well as any knowledge available of the
morphosyntax. Once each sentence of a particular discourse has been identified, then
these can be used to generate paradigms, where each position and hypothesized
constituent of the sentence can be checked by substitution and distribution tests. So, for
example, assume that you have extracted the sentence, John will probably eat beans on
Wednesday. Then this sentence can be extracted and tested as per (), listing English
grammaticality judgments (my own) by way of illustration:

(7.50) a. John will probably eat beans on Wednesday.

b. ?John will on Wednesday eat beans probably.
c. *Beans will probably eat John on Wednesday.
d. The boy will probably eat beans on Wednesday.
e. The girl will probably eat beans on Wed.
f. Bob will probably eat beans tomorrow.
g. Bob will probably smoke beans tomorrow.
h. etc.
Linguistic Fieldwork Daniel L. Everett
page 181

As you work through every sentence in every text in this way, adding to this
process data observed in perambulatory elicitation, you will quickly come to have a
very solid foundation for analysis and 'attacking' the entire grammar of the language.
Filing your data is vital for your own subsequent analysis as well as for linguists
and others of the future for whom your data will be important.
Some linguists, usually those of the above 40 category, will prefer to file their
data on paper. There are in fact advantages to hard-copy filing. For example the linguist
works 'manually' with their data in this type of filing and for some people, this can be
an aid to remembering the data. Certainly your field transcriptions, hand-written notes,
etc. will be filed in hard copy.
By and large, however, I recommend the use of the very useful software programs
available for language data storage. A number of such programs are available through
SIL International, for example (http://www.sil.org/computing/catalog/index.asp). Other
software producers include the Nijmegen Max Planck Institute for Psycholinguistics
(http://www.mpi.nl/DOBES/INFOpages/applicants/dobes-techframe-main.html). The
general rule of thumb for filing of any kind is to provide ready and clear access to all
categories of grammar and semantics that the fieldworker has determined to be of
relevance and significance for understanding the language in question. Thus, filing will
not be the same from one language to another, but will require the judgment of the field
linguist. Categories such as discourse type, clause valency, subordinate vs. matrix
clauses, aspect, tense, mood, case, and so on are all obvious initial divisions in the
filing system adopted for any language. The software collected will also provide
numerous useful ideas for filing. On the other hand, since judgment is important in
establishing, maintaing, subdividing, and applying categories appropriate for each
language, the linguist will want the software to exhibit flexibility. Some programs can
turn out to be fairly 'wooden' in practice. Therefore, before investing time and money in
data filing software, you should inquire from the software producer and several users of
that software something of its problems, advantages, and overall rating or utility. On
the other hand, the more flexible the software, the less user-friendly and more
demanding of the fieldworker's computer knowledge is likely to be.

7.8. Semantic fieldwork

7.8.1. Introduction
Semantics fieldwork is difficult under the best of circumstances. Part of the
reason for this, in fact a large part of the reason for this, is that semantics, the study of
meaning, must involve knowledge of culture, since in a sense, culture is the source of a
great deal of meaning in natural languages. Another part of the reason is that for the
large part of how meaning is related to syntax in the language in question, as well as
various formal aspects of meaning (as discussed below) are extremely difficult to get at
in the best of circumstances. Here methodology is crucial. In this chapter, I provide a
brief overview of some of more salient issues in semantic fieldwork, drawing both from
my own experience and from the very important article on semantic methodology by
Matthewson (2005).
Semantic fieldwork can involve at least the following topics:
Ethnoclassification and Semantic fields – These related areas have to do with the
way that cultures organize the world around them into meaningful categories. The
world we live in can appear to be self-classifying to the naive observer. There are
things like water, forest, fire, sky, ground, animals, people and minerals in the world.
Linguistic Fieldwork Daniel L. Everett
page 182
Why would the linguist's task be any more complicated than working through 'Adam's
task' in reverse? (Where by 'Adam's task' I mean just the biblical reference to Adam
giving names to the things around him as though his pre-existing language matched
one-to-one the world around him, without the necessity for an intervening culture to
give meaning to that world.) That is, why doesn't the linguist simply collect the names
for things in the a priori universe? Let's consider this problem in more detail, via the
real world entities that we refer to in English as 'sky' and 'ground'. Isn't it perfectly
obvious that these are distinct entities in the world?
In my first day of research among the Pirahã, I learned that they used a single
word to refer to the entities labeled by 'sky' and 'ground' in English, their word bigí. But
it took many more years of research to find out why this was the case. Initially I
thought that perhaps I might have gotten the tones wrong or that there were actually
two forms that shared a single truncated form. But it wasn't too difficult to determine
that in fact it is the same word. So could the two words be homonyms, like 'too' and
'two' in English or 'janus' words (words that are their own antonyms) e.g. English fast
'to move quickly' and fast 'immobile' ('to stand fast'). These hypotheses seemed self-
serving and unrevealing. There was clearly something going on here that I was missing.
Ultimately, it emerged that the Pirahãs view the universe differently than we do. To the
Pirahãs, the universe is structured in layers and the name for the boundary between any
two such layers is bigí. So above the 'sky' is one layer of the universe. Below that is
another (our biosphere). Below that is another, the boundary corresponding in a very
broad way to what English calls 'ground'. Thus there is no simple mapping between
English words for nature and the Pirahã bigí. Pirahãs and English speakers structure
these parts of their universe differently.
The field linguist, if they are to give a comprehensive or accurate account of the
semantics of the target language must provide at least an initial list of categories and
subcategories of the natural and mythical worlds found in the language under study.
But this will clearly involve a study of the culture of the native speakers – at least a
synchronic study and often diachronic studies as well. Further, for detailed
ethnosemantic studies it is crucial to enlist the support and advice of ethnobiologists,
either in the field or in your prefield preparation. A great number of important pieces of
cultural knowledge, linguistic structures, and meaning relationships depend on careful
ethnosemantics.46
Semantics proper
Semantics-syntax interface: How does the syntactic form of a sentence constrain
its meaning? Is semantics simply a summation of the meaning of the words of a
sentence plus the contribution of a specific word order or hierarchical structure? We
have already seen in ___ that grammatical constructions are important because, among
other reasons, important aspects of their meaning are not derived by strict composition
in the sense just given, but also involve what we might refer to as 'annotations', specific
bits of noncompositional meaning associated with a particular construction which must
be memorized as part of one's knowledge of the language in question. This then raises
the question of what other types of meanings and meaning-form relationships might be
introduced in the syntax. To get at the meanings that are linked to different syntactic

46
A very good study of this kind in Brazil is Jensen's () study of bird classification
among the Wayampi of Northern Brazil. A huge portion of Brent Berlin's research
program has been dedicated to such classification, with important insights for all of us
interested in the language-culture-real world connection (see Berlin (), (), ()).
Linguistic Fieldwork Daniel L. Everett
page 183
forms and to tease apart different kinds of semantic meaning requires, to put it mildly,
careful and painstaking work. In this subsection, we look at some suggestions on how
to proceed in this area, borrowing heavily from Matthewson (2004).
Word and sentence meaning: Our basic question in this aspect of meaning is this:
what does one know when one knows the meaning of a word or sentence? There are
two broad components of the answer to this question of importance to most linguists.
First, to understand the meaning of a word or a sentence is to know when it is
appropriate to utter it. Saying 'John is a bastard' when meaning to say that 'John's
parents were not married at the time of his conception' might be literally true in some
archaic form of bastard. But this would not be a felicitous use of the term in general
since to most speakers 'bastard' is principally used to comment negatively on the
character of the individual in question, rather than the marital status of his parents. The
second basic component of our knowledge of the meanings of words and sentences are
what the world would have to be like for the utterance of the word or sentence to be
true. The first component of meaning mentioned here is referred to as FELICITY
CONDITIONS. The second component is known as TRUTH CONDITIONS. On the other
hand, as important as these two notions are in developing a theory of meaning or
describing meanings in a particular language, there are other fundamental concepts that
must be explored: ENTAILMENT; IMPLICATURE; AMBIGUITY; and VAGUENESS. For
sentences there is an additional meaning component, discussed in ___, viz., information
structure. Because we discuss information structure separately, I want to focus here
exclusively on these other concepts just mentioned, all of which are causally implicated
in an understanding of sentence or word meaning.
First let us consider the truth conditions of a sentence or word. Arguably the most
important component of meaning is found via the truth conditions. To see this, consider
the following examples:

(7.51) He ate all the fish.

(7.52) He ate none of the fish.

How must the world be like to native speakers of English to judge the truth value
of each of these examples? Let's first consider how we might go about determining this
in a monolingual setting. One idea would be to put some food in front of a helper (not
the language teacher), having previously instructed them to eat it all as soon as you give
the signal. Then you could repeat the experiment, having them eat most but not all of
the fish. In the simplest case, the language teacher will accept your description of 'He
ate all the fish' in the first instance and reject it in the second, where part of the fish was
not eaten. A very similar set of scenarios could be created to get at the truth conditions
of none. On the other hand, it is clearly vital that much thought needs to go into testing
for truth conditions. The linguist tests for truth conditions using complete, grammatical
sentences, to check meanings that are already largely understood. Checking for truth
conditions is not a tool for initial field research.
Using a metalanguage, the procedure is easier in some respects. For example,
consider how we might probe the truth conditions of 'all' using a metalanguage. We will
assume that the same third party eats the fish again. You ask, 'Can I say 'he ate all the
fish'? (in the metalanguage) The speaker will presumably then say 'Yes, you can say
that he ate all of the fish'. (Again, make sure that all exchanges between linguist and
native speaker are complete sentences. It is easy to misinterpret results if partial
utterances are given. And partial utterances cannot be appropriately labeled as either
Linguistic Fieldwork Daniel L. Everett
page 184
grammatical or ungrammatical.) Further probing will reveal that when the speaker eats
only part of the fish, even the major part, that the native speaker will say that 'No, then
you cannot say he ate all of the fish, because there is some left.'
To get at felicity conditions, consider the following example from Matthewson
(2004, 401):

"(7.53) Situation: There are two cats in the room, and they are both asleep.
The cats are awake. FALSE
The cat is asleep. INFELICITOUS"

How does one distinguish reliably between the falsity of a sentence vs. the
infelicity of a sentence? This involves discussion with native speakers, attempts to
paraphrase, or, as in ()-(), changing grammatical marking of some sort or another.

7.8.2. Diagnostics
There are a number of useful diagnostics for the study of meaning in well-known
languages. One such test is called the 'wait a minute' test, first introduced by Kai von
Fintel.
This test (Kai von Fintel, cited in Matthewson (2004, 34ff)) is useful for
distinguishing infelicity from falsity, a crucial distinction for the semantic fieldworker.
But like so many other diagnostics, it often does not travel well from one language to
another. Consider Matthewson's (2004, 34ff) discussion of this in the box.

"The issue of distinguishing falsity from infelicity is particularly tricky with

respect to St’át’imcets and Straits at least, and possibly other Salish languages, because
one easy language-internal clue to infelicity due to presupposition failure simply does
not work in these languages. The test is the so-called wait-a-minute test, invented by
Kai von Fintel (personal communication). Since presuppositions are propositions that
discourse participants are assumed to know already at the time of utterance, a
presupposition which is not known can be challenged with “wait a minute!” (or another
similar expression). On the other hand, an assertion which is news to the hearer cannot
be challenged by “wait a minute!” English examples are given
in (52) and (53).

(52) Presupposition of stop:

A: Felicia has stopped smoking
B1: Wait a minute! I didn’t even know she smoked!
(presupposition unknown)
B2: #Wait a minute! I didn’t even know she stopped!
(assertion unknown)

(53) Existence presupposition of the:

A: Barnaby won the semantics prize.
B1: Wait a minute! What semantics prize?
(presupposition unknown)
B2: #Wait a minute! I didn’t know he won it!
(assertion unknown)
Linguistic Fieldwork Daniel L. Everett
page 185
Unfortunately for researchers working on Salish, for some reason this test does
not work. Even items that are difficult to conceive of as lacking presuppositions (such
as ‘stop’) do not give rise to a “surprise” response with any consultant tested so far.
Even when consultants offer or accept overt denials or questionings of the failed
presupposition, as in (54), this does not distinguish presuppositions from assertions (as
the English “wait a minute” test crucially does). Exactly parallel denials and questions
are judged appropriate for assertions which are not already known, as shown in (55).

(54) A: plan cukw kw-a-s mán:x- ́m s-Bob

already finish det-impf-nom smoke-mid nom-Bob
‘Bob stopped smoking’.
B1: ?az L:u? kw- ́n-s-wá zwát- ́n kw-s
neg just det-1sg.poss-nom-impf know-dir det-nom
tu? mán:x- ́m s-Bob
past smoke-mid nom-Bob
‘I didn’t know Bob smoked’. (presupposition unknown)
B2: wa? ha tu? mán:x- ́m kw-s Bob?
impf ynq past smoke-mid det-nom Bob
‘Did Bob smoke?’ (presupposition unknown)

(55) A: plan cukw kw-a-s mán:x- ́m s-Bob

already finish det-impf-nom smoke-mid nom-Bob
‘Bob stopped smoking’.
B1: ?ay L:u? kw- ́n-s-wá zwát- ́n kw-s
neg just det-1sg.poss-nom.impf know-dir det-nom
cukw-s
finish-3sg.poss
‘I didn’t know he stopped’. (assertion unknown)
B2: wa? ha tu? cukw
impf ynq past finish
‘Did he already stop?’ (assertion unknown)

Why the “wait a minute” test should be inapplicable in some languages is at

present not known. For fieldworkers investigating a language that does allow “wait a
minute” responses, the test is a very useful tool for establishing infelicity."

Again, there is no magic set of crosslinguistic tests that will enable you to make
all the distinctions you want to make clearly. But there are numerous suggestions that,
when coupled with your own creativity and thought can provide useful inways into the
semantics of the language in question.
Let's move to a final area of study, one of the links between semantics and syntax
in fact, information structure.

7.9. Information structure

Information structure is to me the heart of syntactic analysis. It is the way that
language formally distinguishes between relationships among types of information. To
use common, though not completely accurate terminology, it is where we distinguish
new, contrastive, topical, and related types of propositional relations and content in the
language.
Linguistic Fieldwork Daniel L. Everett
page 186
In general, as students of language have recognized for centuries, the sentence
contains a basic pragmatic division roughly understood as the contrast between old vs.
new information. In general prosody (see ___) is used to provide clues as to the
location of focus in a sentence. Thus, since the predicate generally carries new
information it will likewise tend to bear the primary sentence stress. This can be seen in
the examples below, from Lambrecht (1994, __), in which predicate forms are
illustrated from different languages:

(7.54) Predicate focus

Q: What happened to your car?
A: a. My car/It broke DOWN. English
b. (La mia macchina) si è ROTTA.Italian
c. Auto se POKVARIO/POKVARIO se. Croatian
d. (Ma voiture) elle est en PANNE.French
e. (Kuruma wa) KOSYOO-si-ta. Japanese

(7.55) Sentence: My car broke DOWN.

Presupposition: ‘Speaker’s car is available as a topic for comment x’
Assertion: ‘x = broke down’
Focus: ‘broke down’
Focus domain: verb plus remaining post-verbal core constituents

The ingredients of information in Lambrecht's theory are the presupposition, i.e.

what the speaker assumes that the hearer knows; the assertion, the telling of what is
new; focus, the term of the assertion that is what the assertion is about; and the focus
domain, that structural part of the sentence that contains the focus. In Role and
Reference Grammar, this focus domain is known as the actual focus domain and a
further distinction is made of the potential focus domain, i.e. where the language allows
actual focus domains to occur.
To get a feel for focus domains in English and other languages, for example,
consider the following, extracted from Van Valin's work on these matters in Role and
Reference Grammar (and sent to me for use here by Van Valin in email, March 2006):
Linguistic Fieldwork Daniel L. Everett
page 187
SENTENCE
CLAUSE
CORE

NP NUC NP PP
PRED

Actual Focus V
Domain Chris presented a child with some flowers.
IU IU IU IU Basic Information Units

Potential Focus
Domain SPEECH ACT

Figure : Predicate-focus construction

SENTENCE
SENTENCE
CLAUSE
CLAUSE
CORE CORE

NP NUC NP NP NUC NP PP
PRED PRED
V V
Ogni ragazza ha baciato un ragazzo
CHRIS gave them to her.
IU IU IU IU IU IU
IU

SPEECH ACT SPEECH ACT

Figure : Predicate focus in Italian Figure : (Marked) Narrow focus in
English

Now let us consider other types of focus. If the entire proposition is the assertion,
then the focus is sentential. For example:

(7.56) Sentence focus

Q: What happened?
A: a. My CAR broke down. English
b. Mi si è rotta la MACCHINA. Italian
c. Pokvario mi se AUTO. Croatian
d. J’ai ma VOITURE qui est en PANNE. French
Linguistic Fieldwork Daniel L. Everett
page 188
e. KURUMA ga KOSYOO-si-ta. Japanese

(7.57) Sentence: My CAR broke down.

Presupposition: none
Assertion: ‘Speaker’s car broke down’
Focus: ‘Speaker’s car broke down’
Focus domain: Clause

Focus can also take as its scope only a single phrasal constituent:

(7.58) Narrow focus (what Lambrecht calls ‘argument’ focus)

Q: I heard your motorcycle broke down.
A: a. My CAR broke down. English
b. Si è rotta la mia MACCHINA. Italian (Lit: ‘broke down my car’/
it’s
È la mia MACCHINA che si è rotta. my car which broke down’)
c. AUTO mi se pokvario. Croatian
Pokvario mi se AUTO.
d. C’est ma VOITURE qui est en panne. French (‘it is my car which broke
down’)
e. KURUMA GA kosyoo-si-ta. Japanese
Linguistic Fieldwork Daniel L. Everett
page 189
(7.59) Sentence: My CAR broke down.
Presupposition: ‘speaker’s x broke down’
Assertion: x = ‘car’
Focus: ‘car’

English is relatively remarkable in the degree of flexibility it allows for focus

domains:

(7.60) a. Leslie sent the book to DANA yesterday.

b. Leslie sent the book to Dana YESTERDAY.
c. Leslie sent THE BOOK to Dana yesterday.
d. Leslie SENT the book to Dana yesterday.
e. LESLIE sent the book to Dana yesterday

As suggested by this example, languages differ (VanValin and LaPolla (1997, ---
)) according to the relative rigidity of their focus structures. Some languages allow
focus only in narrowly circumscribed syntactic positions, e.g. French and Italian
(immediately postverbal position is reserved for focus), whereas other languages allow
more flexibility in their potential focus domains, as illustrated in Table __ below:

Rigid Focus Flexible Focus

Structure Structure
Rigid Syntax French, Toba English, Toura
Batak
Flexible Syntax Sesotho, Russian, Croatian
Italian

Table : Typology of the interplay of focus structure and syntax (Van Valin 1999)

Moreover, as Van Valin (1999) has also show, there is an interesting interplay of syntax
and morphology with focus structure, as illustrated in (). As () shows, the more marked
a particular morphosyntactic object is in a particular information theoretic role, the
more phonology it has.

Markedness of Occurrence as Focus

Zero Clitic/Bound Pronoun Pronoun Definite NP Indefinite NP

Pronoun [-Stress] [+Stress]
Markedness of Occurrence as Topic
Figure 1: Coding of referents in terms of possible functions

7.10. Argumentation
All syntactic analysis is observation, guessing, belief-formation, and
argumentation. Argumentation is understood here as providing warrants for one's own
Linguistic Fieldwork Daniel L. Everett
page 190
beliefs while removing warrants from plausible opposing beliefs.47 Unfortunately, too
many grammars include only [observation + beliefs] or [observation + beliefs +
warrants for those beliefs]. And yet, leaving out the final stage, demonstrating a lack of
warrant for plausible alternative beliefs, renders a reference grammar or article much
less useful to many readers. It is easy to see why. The audience of a reference grammar
or professional article will be linguists and specialists from related disciplines. Good
readers must be persuaded, not merely informed. They will be constantly thinking of
alternative accounts of the facts. And those accounts will have warrants that are only
ostensible by your account. You need to say why, in anticipation of these being raised.
Let me give a very simple example. Let's say that you analyze the following
sequences as trochaic stress, oriented left-to-right within the word, i.e. 'stress every
even-numbered syllable from left-to-right in the word' (where boldface indicates
stress).

(7.62) a. ba ba ba ba ba
1 2 3 4 5
b. ba ba ba
1 2 3
c. ba ba ba ba ba ba ba ba ba
1 2 3 4 5 6 7 8 9

Your analysis works, quite obviously. But are there other analyses (structural
beliefs) you might form about these sequences? One possibility is that the system is
instead an iambic system oriented right-to-left, i.e. exactly the mirror-image of the
proposed trochaic analysis. Just as obviously, this new proposal works as well for the
facts as the trochaic analysis (stress every even-numbered syllable left-to-right).
Therefore, parallel to your belief about how stress works in the language, there is a
belief that it works in just the opposite fashion. The small amount of data in () warrants
both sets of beliefs. So how to improve the warrant for your belief while removing the
warrant for the opposing belief? Well, expand the empirical set under consideration.
What does each analysis predict for words with even numbers of syllables (those above
have only odd numbers of syllables)? Consider then the further hypothetical examples
in ():

(7.63) a. ba ba ba ba
b. ba ba ba ba ba ba
c. ba ba

If, ceteris paribus, words with even numbers of syllables are indeed stressed as
shown in (), then the alternative iambic hypothesis fails, its belief is no longer
warranted.
Now we need to ask how one comes to have ideas on how to select between
alternative hypotheses. Basically, the ability to argue effectively derives from knowing
your own analysis well.

47
A 'warrant' can be understood informally here as the linguistically ostensible reason
for your belief.
Linguistic Fieldwork Daniel L. Everett
page 191
Knowing my own analysis well enough to argue for its superiority to plausible
alternatives was hard for me in the early days of my career. And the effect was that all
of my initial submissions to journals were rejected. It was embarrassing to have
referees point out alternative analyses that worked as well or better than my own. I
realized that they were able to propose these counteranalyses because they understood
my own claims better than I did myself. So I eventually learned to work more slowly
and to think and reflect about my own analysis so that I would know its implications at
least as well as the referees. This enabled me to discuss the predictions of my analysis
and why these were superior to the hypothetical alternative analyses.

To know one's analysis requires that one know well the linguistic subfield in
which one is working, including the particular theory or framework one has selected to
analyze one's data and present one's research report. This means that the fieldworker
must read deeply and widely in linguistic theory, familiarizing themselves with the
predictions, constraints, and understanding of the relevant aspects of grammar urged by
a particular theory.
A good linguistic argument is based first of all on a good hypothesis and solid
knowledge of that hypothesis's predictions and implications. Again, good hypotheses
come from knowledge of linguistic theory, knowledge of the language, knowledge of
other, especially related, languages, as well as intelligence, aptitude, and luck.
One can approach hypothesis-formation in the field inductively, deductively, or
abductively, as in ()-():

(7.64) Inductive methodology

a. A fact is found. Analyze this fact. What is your understanding of this fact in
your theoretical framework and in your larger understanding of the language (etc.)?
b. Your understanding leads you to believe that future facts encountered in the
language will be constrained in specific ways.
c. You systematize facts according to what you think accounts for them best.

(7.65) Deductive methodology (best exemplar is the syllogism)

a. A theory is known. The theory constrains the notion of 'possible fact', along a
certain range.
b. Your understanding of the theory leads you to believe that facts encountered
will have a certain shape and to so classify the facts you find.
c. You systematize facts according to your understanding of the theory.

In fieldwork, as in all scientific research, both inductive and deductive

hypothesis-formation should be used. Using these as a beginning, it is inevitable that
some facts subsequently encountered will not fit the understanding mentioned in () and
(). When new facts do not fit our expectations, we are surprised. A great deal of the
success of field research is the result of being surprised, i.e. coming up with problems
to resolve. But, again, coming up with these problems to resolve comes almost
exclusively from the reasoning types in () and () in conjunction with the careful study
and knowledge on which they are based.
So when one comes up with a surprising fact, how is it to be dealt with, that is,
how is it to be integrated in one's emerging understanding of the language? The
American philosopher, Charles Sanders Peirce (1839-1914) suggested a method of
reasoning which has become extremely influential in accounts of how scientists reason
Linguistic Fieldwork Daniel L. Everett
page 192
about the world. Peirce () labled this new type of reasoning abduction or retroduction. I
summarize this in ():

(7.66) Abduction
a. You find a surprising fact, A, where by 'surprising' we mean that it fails to
conform to or contradicts your predictions.
b. But if another belief, B, a specific one you have imagined, were held, the A
would no longer be surprising.
c. The ability of B to remove the surprise of A is evidence, C, for B.

Let's consider an example. Consider that you have discovered several clauses
where the subject NP precedes the verb which in turn is followed immediately by the
object NP. So based on both your knowledge of theory (deduction) and your experience
with these examples (induction) you propose ():

(7.67) Information-theoretic based constituent order hypothesis: old information

precedes new information (where subjects carry old information and objects
new information).

Next, assume that you find a sentence in which the subject follows the object, e.g.
the simple form VOS. You are convinced that in this and other VOS examples, the
subject still realizes old or topical information and the object realizes newer or focused
information. Therefore, insofar as you believe the hypothesis in (), you are surprised.
However, you then have another idea, the hypothesis in ():

(7.68) Information-theoretic based constituent order hypothesis: new or focused

information immediately follows the verb.

Hypothesis () works for both sets of examples. If it were true. then VOS examples
would no longer be surprising (because () says nothing about the positioning of the
subject NP, only the object NP. Therefore, C, you have reason to believe (), the
abductive B.
In a sense, abduction is simply a formalization of guessing. But it is useful
because it gives us a clear, explicit visual representation of the nature of our task in
hypothesis formation. Peirce (1877) is a principal source on the concept of abduction.
Every fieldworker would profit by familiarizing themselves with reasoning models and
the notion of warranted belief.

Wood, Linda and Rolf O. Kroger. 2000. Doing discourse analysis: methods for
studying action in talk and text, London, Sage Publications.
Linguistic Fieldwork Daniel L. Everett
page 193
CHAPTER 8. MONOLINGUAL FIELDWORK
8.1. General considerations
Working monolingually can at times be like working on the hieroglyphics of lost
languages – except without a Rosetta Stone.
Some researchers (Berg (), Matthewson ()) refer to the importance of a
'metalanguage' in doing linguistics research. That is, a language which is not the target
language, but which both the linguist and the language teacher speak. If you have a
metalanguage to draw on, many tasks will be easier and perhaps even more effective,
e.g. getting translations from the language of study to the metalanguage, examining
'truth conditions' of sentences more easily, etc. Let us refer to situations in which there
is a metalanguage available as 'bilingual fieldwork'. If there is no metalanguage
available, we can refer to the situation as 'monolingual fieldwork'. In this chapter, I
want to provide some suggestions for working monolingually, as well as some reasons
that it might actually be superior to bilingual field research for a number of purposes.
Just to remind the reader of my philosophy of field methodology, however, I do not
believe that there is a 'best' way to do fieldwork. Many ideas and possibilities should be
explored, as many methodologies as you can think of or read about should be tested. In
fieldwork, the more ideas the better.
I first learned of monolingual fieldwork in 1976, when I took my first course in
linguistics from Kenneth Pike, one of the preeminent linguists of the first half of the
20th century. In one of my first linguistic classes, a speaker of a language that Pike had
never studied was brought in to the classroom. Pike switched to Mixtec, which the
speaker of this other language did not know. Pike then proceeded to speak to the
language teacher in Mixtec, showing her a number of natural objects, e.g. fruit, sticks,
stones, leaves, etc. and doing things with them, such as dropping them, throwing them,
breaking them, using them to hit people, etc. Within 30 minutes he had filled all the
available blackboard and overhead projector space with data. He then stopped and
thanked the teacher. He proceeded to tell us about the phonemic structure of the
language, the grammar (basic word structure and sentence structure), and even was able
to classify the language. For a new linguistics student, or even for an old hand, it was
most impressive.
To much applause, the speaker left, quite impressed herself.
Since then I myself have done such 'monolingual demonstrations' many times, at
the University of Pittsburgh, Presidency College in Madras, India, the University of
Campinas, Brazil, the University of Manchester, United Kingdom, and other places.
The two that most stand out in my mind, however, are demonstrations that I did at the
annual meeting of the Linguistic Society of America, in Manhatten, and a special forum
to honor the memory of Kenneth Pike at the University of Michigan. Both of the latter
lasted two hours and involved two speakers of the same language in each, man and
woman in each, on Kisi and Nepali, respectively, languages I knew absolutely nothing
about beforehand. At the LSA meeting, over 110 professional linguists were present,
many of whom had more field experience than I did (though not necessarily more
monolingual field experience). In fact, there were specialists on the Kisi language in the
crowd. In Michigan, there were specialists on Nepali in the audience and Pike's family
was present. There is a considerable amount of pressure on you each time you do such
a demonstration and many people hope that you will really get something wrong, or at
least the possibility that you might adds to the entertainment value and the suspense.
But these 'shows', though I consider them legitimate and very important teaching tools,
using them now in most of my introductory linguistics classes, do not begin to bring the
Linguistic Fieldwork Daniel L. Everett
page 194
linguist under the same amount of pressure to perform as real fieldwork in a community
of speakers where they have research objectives crucial to a particular state of her/his
career and crucial to the language community (perhaps) or to the relevant funding
agency, etc.
Before my demonstration at the LSA meeting, I flew to Dallas, to meet with Ken
Pike, not long before he died, as it turned out, and to ask him why he thought
monolingual fieldwork was so important. The main reason that stood out to me from
our conversations on the subject was that monolingual fieldwork and monolingual
demonstrations teach us about language as a holistic experience. They involve making
sense out of interdependent facts about communication, e.g. gesture, intonation, body
orientation, facial expression, accent, etc. in ways that simple elicitation, discourse
studies, or investigating natural corpora, all of these mediated through a metalanguage,
simply could never do. Pike's view of the importance of monolingual fieldwork thus
follows from his own theory of language as part of human behavior (Pike (1967)).
By being forced to figure out how language, grammar, the body, and the social
environment are all integrated in communication, the grammar in a sense (regardless of
theoretical perspective) becomes more 'concrete' or more intuitive and more easily
learnable. There are other advantages to monolingual fieldwork that I mention directly,
but this is very important reason – the 'phenomenology of language'.
There are various reasons why one might work monolingually. Here are some of
the more important ones.
First, it could be that you must work monolingually. This was my case with the
Pirahãs and it has been so for other field researchers. It is very rare today to find an
entire people group that speaks no outside language, except for a few phrases and
lexemes. But the Pirahã, for important cultural reasons (Everett (2005)) have chosen to
not learn Portuguese or any other language and have insisted on remaining
monolingual. Therefore, there was no choice for me. Either I conducted monolingual
research or I would have been forced to move on to another group. I believe that the
field of linguistics has been enriched by what it has learned about the Pirahã language
over the years, however, so I am very glad that I made the decision to tough it out and
work monolingually with the Pirahãs.
Another reason for working monolingually is that you may want to learn the
language better than you might under normal fieldwork conditions, even though you
might plan to eventually move to a metalanguage. In this sense, the monolingual
method of fieldwork is just a starting point for you.
A reason that some people work monolingually is because they believe that it is
superior all around to using a metalanguage, as difficult as it may be for some field
researchers to believe. That is, working through texts, semantics, pragmatics, and so on,
in only the language of study, is seen by some to give ultimately a better grasp of the
language, culture, grammar, and people than the 'cheat' of working through a
metalanguage. I do not believe this. But it can be a very well-reasoned position and it is
not one to set aside lightly. Usually, it is the view of people (such as missionaries) who
have much longer-term goals than the average linguist. But if you want to work
intensively to understand one language and people, and are willing to commit many
years to the effort, a case can be made that working monolingually is better.
If Pike was right in his own writings and lectures on the subject, a researcher
might adopt the monolingual method in order to develop a deeper intuitive grasp of the
language. How does monolingual research deepen intuitions about a language? Well, it
does this by, as I stated at the outset, summarizing Pike, forcing the researcher to
Linguistic Fieldwork Daniel L. Everett
page 195
approach the language, the grammar, and the people holistically, learning all
simultaneously. It does this because when there is no metalanguage, the researcher
must pay attention to every gesture, every expression, every outsider remark, every
response, all nuances of the utterance as communicative and cultural event (see again
Everett (2005)) in order to begin to make sense of what is said and begin to make an
inroad into the understanding of this new language and culture. To put this in terms that
Pike invented, but that have been quite influential, especially in the anthropological
literature, the fieldworker is trying to move from an etic perspective to an emic
perspective, i.e. from seeing only the surface, physical character of events and states, to
understanding the meaning of what is heard and seen, as someone on the inside of the
system (though, realistically, of course, one never is going to become an insider).
Yet another reason for working monolingually is to demonstrate greater respect
for the people whose language/grammar you aim to study. This respect emerges as the
people see that you are avoiding what may be to them 'languages of domination'. That
is, the national language of the country in which they are found, usually the language of
the 'conquerors', perhaps still (as with many groups I have worked with) having very
negative connotations. Morever, by forcing yourself to learn the people's language and
use it as the (reflexive) medium for studying itself, the field researcher demonstrates
very clearly and publicly their willingness to subordinate themself to these people. One
way, of course, that this is shown, is that the linguist willingly subjects her/himself to
becoming a laughing stock, at least temporarily during language-learning, as a by-
product of a genuine effort to learn, and thus, attribute value the people.
An additional reason for working monolingually that I will give here is one that
I have argued for elsewhere (Everett (2005)) but which is admittedly very
controversial, namely, that it is not always possible to translate between languages, i.e.
that not all languages have the same expressive power. In other words, a metalanguage
may simply fail you in two ways: (i) by not having the wherewithal to talk about
concepts in the target language and (ii) by misleading the linguist into putting concepts
that are in fact not understood in terms of the concepts expressed in another language
and culture, which are roughly, but too roughly, equivalent or in fact very different.
Of course, there are reasons why a field researcher might legitimately choose
not to work monolingually. These include at least the following. First, time is always
limited and it may be that in a particular research project there is simply not time to
work monolingually, that the researcher simply must use a metalanguage to get at the
data they need in the amount of time available. This is fine. But then, of course, all the
other advantages of monolingual field research discussed above will be forfeited. And
the fieldworker will have to avoid monolingual communities. There is nothing new at
all, though, in the idea that researchers cannot do everything but can only do what they
have time to do and nothing more. Another reason to avoid working monolingually is
to reduce the risks that the field research will fail to turn up anything useful. This is a
legitimate concern. If your objective is detailed semantic analysis then, as Matthewson
(2004) makes clear, the absence of a metalanguage could adversely affect your research
or, at least, require much more time for the same level of analysis. The benefits of
monolingual research in this case could be 'outranked' by the disadvantages of working
without a metalanguage towards certin goals, especially semantic ones. There is a risk,
then, that working monolingually with these goals could result in spending a lot of time
and coming away with nothing to show for it theoretically. I am skeptical that this
would happen, but it is certainly a reasonable concern for someone contemplating a
fieldwork methodology.
Linguistic Fieldwork Daniel L. Everett
page 196
A final reason I will mention here that might lead someone to avoid monolingual
field research is when the field situation could require moral, ethical, or political
understanding from the outset of the field research. If you are working in an area where
people are particularly suspicious of people from your home country, for example, it
will likely be to your advantage to use a trade language, the national language, or some
other language shared by you and the local community in order to more effectively
explain your purposes and to understand the relationship that the community expects
with you and the constraints that it expects you to operate under.
Choosing whether to work monolingually, therefore, is a complex decision that
will have numerous implications for the fieldworker's research success, trust from the
community, overall effectiveness, and, also important, enjoyment of her/his field
situation. There is no really 'correct' choice to make. Each person has to make the
choice that think best. The important thing is that people do field research.
Now I would like to include three anecdotes, with important lessons to teach the
potential fieldworker, from linguists who have worked monolingually. The first
anecdote is from my own experience.48
In early December 1977, when I was but a lad of 26, I flew in a Cessna 206
aircraft (which I fondly refer to as the 'pukemobile', given my problems with motion
sickness) for one hour and forty-five minutes Northeast of Porto Velho, Brazil, to make
my initial contact with the Pirahãs. Before I got on the missionary plane that was to fly
me to the village, I was full of thoughts about whether I had indeed chosen a good way
to spend this time of my life. Was I really up to the task? I had heard about the Pirahãs.
In fact, SIL had asked me to work with the Pirahãs because their language had stymied
two previous SIL teams and I had done very well in my linguistics training. As the
plane took off and we were flying low over the Amazon jungle (so much of it has been
cut down since that flight), my concerns intensified, aggravated by my growing motion
sickness. What would the people think of me? What would I think of them? Would the
monolingual methodology Pike urged me actually work in real life, apart from the
'shows' I had seen in classrooms? (At this time I had not done any such 'shows'
myself.)
Within the first thirty minutes I was feeling queasy. By the end of the first hour, I
was trying to find a happy place and not look out the window. Then suddenly, the pilot
offered me a tuna sandwich with onions. I had to talk and turn off my olfactory system
simultaneously. Half an hour later we were circling the Pirahã village, where the
previous SIL team had, at great expense and effort, built an airstrip (at the cost of about
1,500 trees or so). Just as I was beginning to hope that we would crash, anything to
relieve my misery, we touched down and came to a quick stop at the end of the airstrip.
We turned around and taxied back the 450 meters or so to the village path were a large
number of Pirahãs were waiting, yelling, and gesticulating. I was happy not to have
thrown up on the plane, though the temptation to do so was still present. Out of the high
altitude and cool air above the Amazon, my sickness was compounded by a
temperature of approximately 90 degees (Fahrenheit) and humidy of about 99%. The
plane stopped and the door opened. I got out. I didn't yet speak either Pirahã or
Portuguese, so I was literally speechless. As the pilot and two friends of his who had
flown along started to look around, I heard the pilot say that this was the weirdest place

48
I also used these anecdotes in Everett (), from which this chapter borrows heavily and
freely.
Linguistic Fieldwork Daniel L. Everett
page 197
he ever flew into. Walking along the path to the village, overgrown with grass up to my
knees and with river water up to my mid-calf, I smelled hair being singed. My head was
throbbing, I was perspiring heavily, and I was thinking that in a few minutes the plane
was going to leave me there. (One of the passengers who had visited the Pirahãs several
times, but spoke none of the language, had offered to spend the next ten days – the
length of this first, exploratory stay – with me. So I would not be completely alone.)
The Pirahãs were clearly asking me things, but I was paying little attention, just
thinking that this language sounded like a greater challenge than my brains were up to.
The fellow singeing the hair gestured for me to come over and have a look. He squatted
by a fire, in the sun, with no shade, and had a large rodent (a paca, I later learned) that
he had just thrown whole onto the fire. Blood was coming out of its mouth, dripping
from its protruding tongue, and the smoke from its hair was, let us say, pungent. I just
managed to control my gag reflex. But now I was beginning to recover. I remembered
that I had a notebook and a pen. I pulled out my notebook and pointed to the animal. He
looked at it and said something. I wrote it down and said it back. He smiled and
everyone else seemed pleased. So then I tried to refer to the whole process, the smoke,
the fire, the animal, in an effort to get 'singeing the hair'. He said something back. This
time I didn't bother to try to write it. It exceeded my short-term memory's capacity for
strange syllables. I stopped there in the hot sun and picked up a stick. I got the word for
stick, repeated by six or so Pirahã onlookers. Then I let the stick drop to the ground and
got that phrase. And so on. Within an hour after beginning my own private monolingual
demonstration, I was pretty sure that the language had only three or four vowels and a
small number of consonants, some very strange sounds among the latter, and that it had
two or three tones. I had also learned about twenty words. Over the next ten days, I
learned a number of expressions, none of them particularly useful for normal
conversation, but was coming to think that this job might actually be doable. I had
promised myself not to read anything about the Pirahã language until I tried to figure
some things out for myself first (this was not a good move and I strongly recommend
against this – you should read everything you can on the language before ever going to
the community). When I emerged from these ten days and started to read what Arlo
Heinrichs and Steven Sheldon had written on the Pirahã language, I was pleased to
learn that my ideas formed in those ten days were not so bizarre. This was going to take
a long time, but I could do it.

When I first went to the village of the Banawa people (Arawan), my expectations for
the first day were based, naturally enough, on my previous experience with the Pirahãs.
But when I got off the plane, I knew things were very different. The jungle, the heat,
humidity, and sounds were roughly the same. Yet men came up to me an addressed me
in very good Portuguese. By the next morning, I was working with Sabatao Banawa,
perhaps the best language teacher I have ever had. Though he had almost no formal
schooling, he not only gave me very natural texts, but as we went through them to
translate them together, his comments went something like this: 'This word means that
the words here were spoken long ago, by a woman. This part of the word means that
the one who is speaking is not sure. This part means that the pig was on a log, just
above the ground.' And so forth. In other words, Sabatao was able to give me nearly
morpheme-by-morpheme glosses. In my first three weeks among the Banawas, I felt I
had learned more about their language than in my first six months with the Pirahãs. So
even within a similar geographical area, field conditions can be radically different.
Linguistic Fieldwork Daniel L. Everett
page 198
Let's turn to another part of the world now, for a final anecdote. This one comes
from Loving (1975, 268) about her experience at the beginning of her career, working
monolingually with the Awas of New Guinea.

"We were especially on the lookout to learn to say 'What is this?' After two
weeks we were tired of pointing and we wondered if the Awas were not equally
tired of seeing us point. Evidently they were not, for they continued to be
gracious enough to give us new words as we continued to point. One day, we
were cooking some greens around an open fire. I pointed to the food, directing
my 'question' to an elderly man standing around looking into the pot. He turned
to the man next to him and said 'anepomo'. I repeated this thinking this was the
name of the greens. He and several others smiled and then leaning towards me,
he said 'tura'..."

What Loving had learned here was not what she had asked, but something much
better – the precious phrase, 'What is this'. It is difficult to overemphasize how
important this phrase is in learning a language. A companion phrase 'What is it/she/he
doing?' is also extremely important to learn and Loving's anecdote provides a useful
clue as to how both phrases could be learned in a monolingual field situation.
OK. We have seen some stories of how other people initiated their monolingual
field research. Now it is time to move on to consider the methodology in more detail.

8.2. The gavagai residue

It is crucial from the outset here to remember that Postal's Maxims outrank any
method. Moreover, it is important to recognize that all suggested methods for scientific
research are heuristic procedures, not algorithms. Unlike baking a cake, following a set
of methodological suggestions in science will not always give the same results. There
are too many variables. The best that one can hope for is that the methods will be very
useful. But the art of fieldwork (see Walcott ()) results from the fact that it is the
individual fieldworker's resourcefulness, resilience, and creativity that will also be
more responsible for success than any set of automated procedures. With these caveats,
I turn to outline a method that has worked for me in my own monolingual field
research.
Since this method involves beginning with simple words and working your way
'up' the grammar, progress assumes that you are actually understanding what they
people are telling you. That is, it assumes there is a reasonable semantic match between
the linguist's translation and the teacher's intended meanings. First, begin with
individual words. Speakers will, in every language and culture I have ever heard about,
almost immediately recognize when you indicate an object that you want the name of
the object and they will give it to you. On the other hand, monolingual field research is
always haunted by the spectre of Quine's (1960) 'indeterminacy of translation' problem,
brought out most clearly in his famous gavagai example. The fact that this issue does
not seem to be a tremendous problem in actual field experience is confidence-inspiring,
yet it should still be sobering to fieldworkers, because it does arise. Moreover, the
problem is such that it can arise without anyone noticing at first.
Summarizing loosely, here is the famous gavagai problem and how it emerges
from Quine's philosophy. This problem is extremely relevant to fieldwork, because
speakers may not always give the answers expected. How does the field linguist
recover from bad information? Is it even possible to do so?
Linguistic Fieldwork Daniel L. Everett
page 199
In Word and Object (1960) and Ontological Relativism (1977), Quine considers,
for its philosophical lessons, a hypothetical problem for field linguists attempting to
translate a hitherto unstudied language that the linguist does not speak. Here is how the
problem goes:
One asks the question of a native speaker, by gesture or by metalanguage, 'What
is that?', pointing, say, at a rabbit or some other object, and the native speaker utters
gavagai. How can the linguist match the utterance of gavagai to a specific meaning?
Does it mean 'rabbit'? Does it mean 'rabbit fur'? Does it mean 'token of a platonic rabbit
type'? And so forth. Even if the native speaker were to mean 'rabbit', the linguist might
never encounter a crucial counterexample to doubt their translation (unless they look
hard, which they only would do if they had reason to doubt their initial translation
which, as we have just seen, they would not have). How could you tell if you have the
right meaning? Quine answers that in principle you could not know and therefore there
is an 'indeterminacy of translation' problem. To some degree, this problem is a real and
sobering one for the fieldworker. And it is never completely solvable. However, there is
a bright side to the issue.
First, the gavagai problem never happens in exactly this way in fieldwork. Native
speakers, however naive linguistically, will almost always give you the name of an
object as a whole, e.g. 'rabbit', 'tree', etc. and will not intend by their answers parts or
salient features of objects.49 Of course, Quine would counter that you cannot ascertain
the exact meaning of an elicited word, in principle. But the fieldworker's reply is that in
the course of grammar writing and analysis of the language wrong interpretations 'come
out in the wash', especially following the methodology of this book. That is, the issue
might be important for philosophers, but it has perhaps no significant impact on
fieldwork.
On the other hand, in a monolingual field situation one can never be sure, initially
at least, that one has gotten the answer that one thinks they have gotten. Some 'false'
answers can be easily discovered, e.g. when one points at a tree and the speaker,
focusing on what is pointing, rather than what is being pointed at, says 'finger', instead
of 'tree'. Other meanings are not so easy to determine in any case, e.g. verbal mood,
aspect, etc.
Overall, of course, it is almost certain that no matter how long a fieldworker
conducts research on a given language, there will be many words where the meaning
the linguist has recorded is not quite right. And the same goes for many grammatical
constructions. Some of these will, as I said, be resolved in the course of the grammar
writing. Other errors will be recognized and resolved as the linguist learns to speak the
language better. On the other hand, some mistranslations will never be recognized.
There will always be a 'gavagai residue' in fieldwork between any two languages.50
It is simply vital for the linguist working monolingually (and, to a lesser degree,
linguists working with a metalanguage) to always remember Quine's lesson, namely,

49
What does this tell us about human nature? Well, as Searle (to appear) notes, Homo
sapiens share a strong, evolutionary attention to objects. I doubt that anyone could
function well thinking of 'rabbit essences' when they saw a rabbit in most
circumstances, instead of just 'rabbit', since only the latter is edible.
50
QUOTE Grace on indeterminacy of translation.
Linguistic Fieldwork Daniel L. Everett
page 200
that you might not be getting what you think you are getting. For this reason, Postal's
Maxims and careful reasoning, checking, and re-checking become ever more important.
Let's consider some suggestions for monolingual fieldwork that might enable you
to avoid or at least not be completely misled by the gavagai problem. The procedure
below is very simply. Clearly the resourcefulness of the linguist is the main ingredient.
I usually begin my monolingual elicitation, whether in a monolingual community
or during a demonstration, with objects from nature. So pick up a leaf, stick, rock, or
some such to begin with. Get what you think its name is by pointing at it and saying
what it is in your language. Don't just grunt. Use your language freely. This is natural.
Then repeat back what you were given for correction. Now say the word again as you
let the rock, whatever, drop to the ground. Write down (don't use a taperecorder yet.
Just you and your ears and paper) what you get. Now say it back again. Now pick up
another object of the same type and roughly same size and color. Show two fingers
while letting both drop. Imitate. Write this down. Did the form of the verb change? Can
you recognize any differences in the form of the nouns or noun phrase? You likely now
have a form for plural or dual. Now pick up two other objects (same type, size, color)
and repeat the process from the beginning. Now do it all over again with three, then
four objects. You should be getting numerals and grammatical number, articles, etc.
Always build up slowly, so that you can feel in control of what you are getting. Look at
how much you are learning! And this is all with just a few natural objects and a single
verb.
Now work with colors, sizes, and conjunctions (which you can get by mixing the
object types, e.g. 'a rock and a stick fell to the ground'). After exploring these aspects
of noun phrases for a bit, you can try some transitive verbs. Begin perhaps by having
your language teacher hit you. You can do this by taking his hand (if appropriate) and
hitting it lightly on your shoulder. Then you pretend to hit him. Work with this for a
while until you feel fairly secure that you are getting a transitive construction. Now
take a biggish stick. Hit yourself with this. Now pretend to hit your language teacher
with it. Now have him hit you with it. Now have him it someone else. For every single
action, get a description by your language teacher. Repeat after him and make sure that
you watch reactions. It is very easy to confuse 'I' and 'you' in these circumstances. But
as you repeat as you perform the action, if you describe what you are doing with the
wrong pronoun, you will almost certainly be corrected. But if you don't repeat and
make sure you are following, you could easily confuse the pronouns and confuse
yourself for a while.
You are prepared, if you haven't already gotten this, to get the paradigm for
pronouns (at least those corresponding to interlocutors in your environment). You can
get at this or add to your knowledge by now switching to intransitive verbs, e.g. 'jump',
'stand up', 'walk away', 'crouch down', 'sit down', and so on.
These ideas should be enough for you to start your monolingual fieldwork well!
And it is so much fun! You are learning things no one maybe ever has. And it is all
your brain and creativity and the friendship you are building with your language
teachers that is responsible for this! Organize your future sessions. Each day you will
be building, making huge steps in your understanding. Work simultaneously on
vocabulary memorization and greater fluency. Your comprehension, as is usual, will
exceed your ability to produce. But it will all come, slowly but eventually and surely.
Just takes a lot of hard work. But no secret to that.
From these data, that you should be collecting for two-three hours per day, you
will begin your phonetic, phonological, morphological, semantic, and syntactic
Linguistic Fieldwork Daniel L. Everett
page 201
analyses. After perhaps 12 weeks of hard work of this type, and accompanying
language-learning, you should be in a position to begin collection of simple texts,
following the methodology of chapter ____.
Finally, let's review the implications of the fact that working in a monolingual
situation there is no 'metalanguage' (Matthewson (2004)) to use. So how does one come
to understand verbal or other meanings? One thing is certain – you cannot tease apart or
even discover the full range of verb meanings (including affixal meanings), e.g. tense,
aspect, valence, aktionsart, argument structure, case, etc. simply going through texts. In
a monolingual situation you cannot get translations for texts, for one thing. The best
you can hope for are paraphrases. But since analysis of verbs requires subtle and
accurate distinctions of verb meanings, you must come up with a method to help you do
this.
Here are some suggestions:
First, act out scenarios. Let's say you're following a guide, e.g. the Lingua
Descriptive Questionnaire, and you want to try to distinguish directional actions or to
see whether such distinctions are made on the verb in the language. Try this. Assemble
a few objects from the local environment, e.g. sticks, rocks, leaves, bones, tools,
necklaces, etc., things that are all in regular use and seen as normal objects that one
would handle within the community. Next, begin work initially with individual
teachers, moving later to work with multiple teachers (see ___). It is important to use
culturally relevant objects and tasks at the beginning of the research to build on
people's confidence in their knowledge of their culture (there is so much opportunity
for uncertainty on all sides in any case).
So take, say, a necklace, some beads, and the string (or whatever) used to make
necklaces. Place the necklace on yourself. Get the description of what you did. Record
and film all of this. Place the necklace on your language teacher, on a child, on a man,
on a woman, etc. That is, as you have time and opportunity in this session, try to act out
paradigms. (Always be careful to obey cultural constraints.)
Place the necklace on the ground. Place it on a table. Hang it from the roof. Drop
it. Throw it. Toss it up and catch it. Take the beads and place one on the string. String
several in succession. Let one bead fall off the string to the ground. Let another fall off
onto a table or chair. Let all of them fall of on the table, chair, ground, etc.
simultaneously. Get various (3-6) speakers to describe each one of your attempts.
Filming and recording all of this will give you good data, help you find
directionals, numbers, positions, aspect, valency, transitivity, etc. And this is just a
trivial example. Tremendous amounts of high-quality data can be collected in this way.
But you need a plan. There are various questionnaires and aids to fieldworkers (see the
LDQ in appendix ___ or visit the website for the Max Planck Institute for Evolutionary
Anthropolgy: http://lingweb.eva.mpg.de/fieldtools/tools.htm).
You can also use films, of the type prepared by the Max Planck Institute for
Psycholinguistics in Nijmegan. Using films, rather than acting things out, has both
advantages and severe disadvantages. The advantages include the fact that film can do
things you cannot (like make a bowl of beans suddenly appear on a table where there
was nothing before).
There are, however, a number of disadvantages to using films instead of acting
out scenes for elicitation. These are largely cultural. First, some language teachers may
have a difficult time following two-dimensional electronic images. Second, they may
not recognize the gender of the person on the film (many peoples consider long hair a
sign of feminity or flat-chestedness a sign of masculinity, to cite two problems I have
Linguistic Fieldwork Daniel L. Everett
page 202
encountered) and this is surprisingly distracting. Finally, many of the activities make
little sense to some language teachers watching the films. Therefore, though I certainly
recommend that films be used in elicitation, as well as other media, where applicable,
they must be used very judiciously.

Say it back to several speakers.

Many times I have received crucial corrections on my third or fourth repetition to
different native speakers. Inexperienced linguistic teachers often find it hard to know
what the linguist is after (so does the linguist at times!) and find it even harder at times
to correct the linguist (who is, after all, paying their salary in many cases and should
have the 'right' to say whatever they want, or at least this is the impression I get
sometimes). So checking the data with speakers other than the one who provided the
original data is crucial to processing the data. Get these other speakers to paraphrase
what the previous speaker said, as well as saying it back identically. This way, you
have pronunciations of the same data by different speakers, as well as various
alternative syntactic expressions of the same or similar (usually, but not always)
content. If you can work via a metalanguage, ask each speaker for a translation. But in
all cases, make sure you get paraphrases of difficult to understand data. Almost always,
translations and paraphrases will vary from the original's meaning in subtle, yet
significant and revealing ways.
If you overhear something and speak no language in common with the speech
community, describe the context in a few lines next to the transcription (e.g. time of
day, who was speaking, what they were doing, what you think they were saying, etc.).
Then test your understading by trying to use the expression, as you understand it, when
and where you think it would be appropriate to do so. Almost certainly you will
discover that your initial guess was wrong (and if it wasn't, congratulations, but you
might not know one way or the other at this stage anyway) and people will laugh at
you. (I am laughing at you now just thinking about the mistakes you are going to make.
Ha Ha.)

Generating Paradigms
A word about paradigms: lists or theoretical constructs? Wari' (Everett 2005)

The next thing you must do is to take the expressions you want to study and
generate paradigms from them. Let's take our squeat example from the outset of the
morphology chapter. What does it mean? Well, to get at this we write its context as
uttered by someone, say, as they rise to leave, about 12ish, and as they walk towards
the cafeteria. I ask someone else to paraphrase it and they say 'he's hungry and he thinks
you are too'. Then I ask someone to say the phrase slowly and they something like
Let's go eat. I ask someone else to repeat this and they also say Let's go eat or,
perhaps, Let us go eat. Now I can try to build some paradigms from this expression. To
do this, I divide the utterance into positions or 'slots' and try to put other words I have
learned in each of these slots. An example of what someone studying English for the
first might do is given in ():

(8.1) a. Let us go eat.

b. Try us go eat.
c. Make us go eat.
d. Help us go eat.
Linguistic Fieldwork Daniel L. Everett
page 203
e. See us go eat.
f. Believe us go eat.
g. Run us go eat.

(8.2) a. Let me go eat.

b. Let you go eat.
c. Let him go eat.
d. Let them go eat. etc.

(8.3) a. Let us do eat.

b. Let us try eat.
c. Let us make eat.
d. Let us can eat.
e. Let us want eat.
f. Let us be eat. etc.

(8.4) a. Let us go run.

b. Let us go fish.
c. Let us go please.
d. Let us go work. etc.

Next, I work through all of these, asking whether each is

acceptable/pretty/crooked/etc. and how they compare to one another, i.e. which would
the native speaker think that they are most likely to use. Never be satisfied with an
answer that simply says a form is 'good'. Get the native speaker to say it themselves. If
they will not, then the form is almost certainly not 'good'. If they do, do they say it with
natural speed and intonation and a solid sense that they understand it?
We can now ask what the linguist is supposed to learn from these paradigms?
First, they learn that the verbs that can go in the first slot are limited. Try and believe,
for example, are not allowed. Why not? What makes these verbs different from let,
make, and the others? Let's say you hypothesize that try and believe are excluded
because they are interpreted as second-person imperatives by native speakers. How
could you test this hypothesis? That is, what data would you need to refute it, do you
have such data, and is it refuted?
The point, of course, is not whether this initial hypothesis is right or wrong. All
your initial hypotheses will likely be off to one degree or another. The point is that by
developing these paradigms, you generate grammaticality differences and different
hypotheses about the grammatical structure of the language, a way of implementing
Postal's Maxims.
The development of paradigms must be restricted so that the initial data used to
generate them come from natural texts, not elicitation. Using elicited data, which itself
can be artificial, can propagate artificiality throughout the examples.
Next, test each paradigm with several speakers. If you see little or no
disagreement in judgements between speakers, then you can move much more quickly
to the next paradigm. If you do encounter disagreements (say, for example, that some
speakers allow try and believe in the sample paradigms above) then ask them about
this, howbeit indirectly, never directly. For example, what does the controversial
example mean to the speakers who claim that it is acceptable to them? Can they be
Linguistic Fieldwork Daniel L. Everett
page 204
gotten to say it with intelligibility? What does the same example mean to those who
reject it?
Now let's assume that you have worked your way through a paradigm and that
you now believe that ou have a reasonable understanding of what a morpheme means.
What next? One example of what to do next comes from the beautiful study by Ivan
Lowe (), in which he shows how to 'track' the morpheme through texts, as reported
above.

Can the morpheme be isolated? In fairly mechanical terms, does the same
sequence of segments occur in different positions in the utterance? Can it occur with
other segments, e.g. in answer to a question? Does this sequence correlate with a
constant meaning in the utterances in which it occurs? Consider in this regard another
example, from English:

(8.5) They were running but stopped suddenly.

(8.6) When I am done eating, I'll talk to you.
(8.7) I would rather be playing my guitar.
(8.8) Q: Are you done?
A: *No, I am ing.

Does it appear in some parts of texts more than other parts (e.g. introduction vs.
conclusion; denoument vs. setting vs. build up, main theme 'line' vs subsidiary
information lines, etc.)?
Deliberately show people things from your baggage that you do not think that
they will have had previous experience with. If they allow it, you should have a video
camera, preferably, or an audio recorder (keep it running as you unpack, settle, etc, and
keep it pointed at the people speaking). Listen as you unpack, try to imitate, go back
over your tapes later and try to figure out what was being said. Test your hypotheses by
trying out phrases, based on your understanding of what they mean, with the people.
Do you see or hear question-like behavior? Look for things like hand-gestures,
eyebrow-raising, intonational changes, and question-like actions that might provide
clues. Do any of these seem to be focused on potentially novel items among your
possessions? In particular, you after things like 'What is that?', 'What is he doing?', etc.
These are vitally useful phrases for your research, for getting along with the people, for
negotiating your way through the community and the language. These are not usually
easily to get by direct elicitation, yet they are uttered spontaneously in exactly the kind
of situation your initial arrival in the community will create. Pay attention.
Linguistically, such phrases are vital even if the linguist otherwise plans to work
bilingually (i.e. using a trade language for lab work).
Linguistic Fieldwork Daniel L. Everett
page 205
CHAPTER 9. INTEGRATING ETHNOGRAPHY AND FIELD LINGUISTICS
9.1. The neglect of ethnography
In this chapter I discuss ways in which culture and language can interact that go
beyond standard sociolinguistics and in which the ethnography of communication can
affect the formal grammar of the language, as per the Boas quote in __ above. The
point is that the linguistic field researcher must be an observer of culture, not just
language, and she must keep a careful record of the connection between cultural
contexts and uses of different constructions. Let' get a feel for what is at stake here by
considering a classic example from Sapir:
Sapir (1921,172) writes of the need to understand the 'genius' of each language.
By this Sapir refers to that which makes each language unique, the essential core of a
language, that part less subject to historical change (a sort of Heraclitus-inspired
question of what changes vs. what remains). Judging by his intellectual output, Sapir
was always concerned with what Everett (2004) refers to as 'coherent fieldwork', i.e.
fieldwork that integrates specific phenomena in the larger cultural context and the
background of the researcher. Sapir's concern was with difference, the relative value of
a given language, as opposed to seeing it merely as an exemplar of one variant of an
absolute Universal Grammar. One good example of what I mean is found in a study he
undertook of Nuu-Chah-Nulth (then known as Nootka, Wakashan, Canada) consonant
alternations. In this language, as Sapir (1915, 181) observes, there are extremely
interesting consonantal alternations that cannot be explained grammar-internally.:

"It is possible and often customary in Nootka to imply in speech some physical
characteristic of the person addressed or spoken of, partly by means of suffixed
elements, partly by means of 'consonantal play'. Consonantal play consists
either in altering certain consonants of a word, in this case sibilants, to other
consonants that are phonetically related to them, or in inserting meaningless
consonants or consonant clusters in the body of the word. The physical classes
indicated by these methods are children, unusually fat or heavy people,
unusually short adults, those suffering from some defect of the eye, hunchbacks,
those that are lame, left-handed persons, and circumcised males."

Sapir exemplifies this 'consonantal play', concluding that to understand the

grammar of a language, we must therefore understand the culture in which that
grammar is found. Sapir's study of Nuu-Chah-Nulth is well-known, of course. This is
the kind of study that illustrates the solid connection between culture and language,
though it is only a single connection between one rule of phonology and culture.
Perhaps it has failed to exert modern influence because it is considered to be a marginal
example.
But the main reason that Sapir's and others' studies of of these connections have
failed to influence theoretical grammar as they might have because in the mid-50s, as
discussed in chapter 1, linguistics underwent a conceptual change, from a discipline
concerned with documenting, describing, and explaining observed language behaviors
in terms of linguistic and social structures and dynamics (e.g. variation and change), to
one in which the primary focus was on inferred mental structures and abstract
principles supposedly common to all human languages. This is a priori neither good nor
bad, simply a change of interest. The change, as stated in ___ above, has produced
some of the most exciting scientific results in the history of the study of language.
Linguistic Fieldwork Daniel L. Everett
page 206
However, a posteriori, there have also been nontrivial implications to this change
in focus. One is a reduced emphasis on fieldwork as the natural activity of linguists.
But another is the near complete lack of attention given to culture and its relation to
language, a full rejection in practice of the views of Boas and Sapir cited above. And
with this change has come a loss of any truly evolutionary perspective on language, in
particular, a loss of the idea that individual languages might follow different
evolutionary paths, adapting to different cultural-social niches.

9.2. Ethnography, Cognition, and Grammar

The purpose of this chapter is, therefore, to equip the fieldworker for beginning
field research on the culture-language nexus. As a conceptual point of departure, let's
consider Table 1 below, which lays out some of the potential connections between
language, cognition, and culture, all of which are ultimately necessary to an
understanding of language (which one could, but need not, equate with Chomsky's () E-
language) and grammar (Chomsky's () I-language):

Table 1
CONSTRAINT RELATIONSHIP SAMPLE RESEARCH PROGRAMME
1. cognition → grammar Chomsky's Universal Grammar
2. grammar → cognition Linguistic Relativity (Whorf, Lucy, etc.)
3. cognition → culture Berlin & Kay (1969) on colour terms
4. grammar → culture Urban's (1991) work on discourse-centred
culture
5. culture → cognition Anthropological research on semantic
fields
6. culture → grammar ETHNOGRAMMAR; individual forms
structured by culture (e.g. evidentials or
Sapir 1915)

This table is clearly inadequate because, among other things, it is two-

dimensional. A three-dimensional display would indicate connections between all three
domains (cognition, culture, and grammar) simultaneously. Nevertheless, it provides a
useful basis for discussion, so let's consider each of the rows in Table 1 in turn.
We can situate a good deal of modern linguistics and anthropology relative to this
table. For example, simplifying tremendously, the shift in linguistic research focus that
Chomsky was rightfully credited with moving the focus from Cell 6 to Cell 1. And the
questions raised by the relationship in Row 1 are indeed fascinating. How does
cognition constrain language (whereby 'cognition' in Chomskyan theory has
traditionally focused on the properties required of the initial state of the part of the
mind responsible for language-learning). This is a vital question, obviously.
Row 2 highlights research which examines how linguistic structures, lexical
structures, constructions, etc. can affect thinking. The idea is old (much older than
Whorf, for example) and in some senses obviously true. There are many tasks we
engage in that are affected by language, as numerous researchers have shown.
Nevertheless, I think that the usefulness of linguistic relativity has been overstated in
much current research. Research on this relationship is exemplified today in the work
by Lucy (), Levinson (), and their colleagues, among others.
Linguistic Fieldwork Daniel L. Everett
page 207
Row 3 again interprets 'cognition' as static, the built-in mental parameters for
distinguishing colors which affect how cultures categorize their world. So the way that
cultures classify and identify the colors of the world around them is seriously
constrained by color-perception biology, i.e. static cognition (what is known or given in
the brain and that underlies behavior).
Row 4 indicates the research on the effects of language on cultural values. So, for
example, Urban (1991, --) has argued that a high percentage of passives in the
discourse of a language, themselves resulting from independent structural constraints,
can affect the cultural concept of 'hero', producing heroes that are better known for
what happens to them (they are the Undergoer subjects of passive clauses) rather than
for what they do (if they were instead Actors of active clauses). This is fascinating
research that linguists by and large show no interest in – a result, in my opinion, of the
shift in interest in most linguistic training and research since the mid-50s.
Row 5 indexes research on how cultural values can affect how members of the
culture think about the world. An example of this kind of research can be found in
Seeger () and related research where, say, types of body adornment are at once a
reflection of cultural concepts (communication in Seeger's () ground-breaking study of
Suya/Kisedje)) and how members of society come to value or use the relevant body
parts (ears, lips, tongue, etc) in communication. Another example is Bradd Shore's
excellent 1996 study, Culture in mind: cognition, culture, and the problem of meaning.
Row 6 indicates the kind of research that I have been doing for the past few years,
namely, the effect that the culture can exercise on grammar. My interest is in whether
culture can exercise an architectonic effect on the grammar as a whole (and I conclude
that it can, Everett (2005)). A different, but very important, approach to this problem is
found in Enfield (2003), Sapir (1915), and other studies where convincing evidence is
adduced that culture is responsible for particular structures, constraints, or rules in a
grammar.
Rows 3, 4, and 6 are areas of potentially rewarding research by the linguistic
fieldworker with a background in ethnology. But these of course do not exhaust the
possible domains of research in linguistics, they simply list some salient
interdisciplinary connections.

9.3. Content of an ethnography of communication

Perhaps the single most important broad topic of relevance for linguistic-cultural
connections is the 'ethnography of communication'. References in this regard include
Hymes (1974), Saville-Troike (2003), Gumperz (1986), and various others. Specific
research topics in the ethnography of communication are many, but some of the more
prominent ones are:

(9.1) Topics in the ethnography of communication (see Saville-Troike 1982, 51ff)

a. Communication and social structure
b. Language and culture
c. Rituals and communication
d. Phatic language and cultural relationships
e. Varieties of language
Registers
Social and regional dialects
Different languages
Channels of communication
Linguistic Fieldwork Daniel L. Everett
page 208

Let's consider these briefly, as examples of some of the kinds of connections

between language and culture have been shown to be productively studied by linguists
interested in the causal-connection between the two. First, under communication and
social structure we may consider such topics as the nature and types of discourses
associated with different social roles (e.g. those in political power, those without power,
those with specialized roles, e.g. shaman, physician, pastor, official spokesperson, etc.).
A simpler way of putting this issue is 'Who says what and when do they say it?' This
can impinge on studies of pragmatics, semantics, and even morphology and syntax (use
of different moods, evidentiality markers, etc. according to type of discourse and
discourse utterer), but clearly to study these manifestations of language requires
cultural observation and understanding. Ultimately, the ethnography of communication,
like all studies involving a member of one group studying members of another group,
involves attempting to move from an Etic perspective to an Emic perspective (Pike
(1967, ); see also Headland ()). That is, if we understand a particular ethnography of
communication we shift from the perspective of the outsider to an understanding of the
perspective of those within the system, group, culture, etc. Item (--b), language and
culture, is a more specialized type of study, one that I have come to refer to as
Ethnogrammar, and I offer a case study of such a study from Everett (2005) in ___
below. Other areas of () are less difficult to see the point of, so let us close this
introductory discussion by considering what it means to study 'varieties of language'.
Recall the accounts in chapter ___ above about my monolingual demonstrations with
speakers of Kisi and Nepali. Both revealed the need to understand varieties of
language. In the Kisi demonstration I conducted at the annual meeting of the Linguistic
Society of America in (), I noticed that most of the word endings from the female
language teacher differed from those of the male language teacher. During the
demonstration I noticed this, but of course I had no way of knowing why the forms of
her words were different. It might have been a women's dialect; it might have been
social register (a woman addressing an unknown and older male – me), it might have
been something else. It turned out to be something else – Kisi was not her native
language! She simply didn't know the endings. That can be a trivial observation, but it
requires understanding of who is who in the community and under what circumstances
people learn or use different languages. In the Nepali case, the differences in the
woman's responses to my questions, as opposed to the male Nepali language teacher,
were in fact a matter of register. I was older and a stranger, so she used the formal
register with me. To understand different choices of languages and registers (and by
'understand' I mean more than simply observing that they exist as I have done here)
requires observation of the culture. The importance of understanding different varieties
of language in analyzing the grammar of a given language can perhaps best be seen via
a case study of discourse channels in Pirahã, to which I now turn.
As pointed out in Everett (1982) Pirahã phonology cannot be fully described or
understood without a knowledge of how it interacts with culture. There are other
examples from Pirahã phonology. Let me present two of the strongest, in ascending
order of importance for coherence.
Imagine that a language could have various systems/modalities of sound structure,
beyond its phonetics and phonology. And then consider the possibility that one
modality can affect another, but not necessarily via constraint-rankings or rules, the
standard devices of phonological theory proper. If so, then to understand the sound
system of language, L, at any level (e.g. 'what happens' or 'what native speakers know
Linguistic Fieldwork Daniel L. Everett
page 209
when they know the sound system of their language') we must look carefully at the
modalities of expression made available via an ethnography of communication and not
merely at a supposed universal formal apparatus. Corollaries of this scenario might
include, e.g. the appearance of new roles for old constraints (e.g. mode-faithfulness of
segments being highly ranked to mark syllable types; syllables are maintained, a form
of prosodic faithfulness, in order to parse the larger speech stream, not merely to
enhance the perception of segments; and thus arguments for syllables may go beyond
phonotactics and segmental enhancement and the syllable may have roles not
envisioned by the so-called 'phonological hierarchy'). If this were true, the coherent
fieldwork would evolve from a curiousity or desideratum to an imperative. Is there such
a case? Indeed. Consider the following facts about Pirahã phonology, beginning with its
phonemes.

Table One
Pirahã Phonemes
Consonants () = missing from women's speech
p t k ?
b g
(s) h
Vowels
i
o
a

The first thing to notice about Table One is that the segmental inventory is one
of the smallest in the world. The next is to recall that it includes allophonic sounds
found in no other language, subject to cultural constraints. The third is that the /s/ is in
()s because it is not found in women's speech, but only in men's.
Though this is one of the simplest segmental phonemic inventories in the world
(the women's inventory does seem to be the simplest known), we should juxtapose
alongside this simplicity, the complexity of Pirahã's prosodies. Pirahã's stress rule is a
good place to begin, since it is well-known.
This rule, from Everett & Everett (1984), is considered one of the more complex
and unusual stress rules in the literature, mainly for its phonological consequences
(rather than, say, any difficulty in stating or recognizing it):

(9.2) Pirahã stress rule: stress the rightmost token of the heaviest syllable type in the
last three syllables of the word.

The phonetic basis of 'heaviness' in (1) is just this: Voiceless consonants are
always longer than voiced consonants and there are five syllable weights based partially
on this contrast:

(9.3) Pirahã 's five syllable weights: CVV>GVV>VV>CV>GV

Pirahã is a tonal language, as well. But stress, tone, and syllable weight vary
independently in the language. To see this, I will just review one simple set of
examples, in (3), from Keren Everett (1998). In the examples in (3), tone is independent
Linguistic Fieldwork Daniel L. Everett
page 210
of stress. ´ = high tone; no mark over vowel = low tone. The stressed syllable is marked
by !. There are no secondary stresses (7=glottal stop).

(9.4) a. !tígí 'small parrot'

b. !pìgì 'swift'
c. !sàbí 'mean, wild'
d. !7ábì 'to stay'
e. tíí!híí 'bamboo'
f. 7ì!tì 'forehead'
g. tì!7í 'honey bee'
h. tí!hì 'tobacco'

Thus alongside Pirahã's extremely simple segmental phonology, it manifests an

extremely rich set of prosodies. This leads us to ask a reasonable question, namely,
does the language exploit this differential complexity in any way? Indeed, as Everett
(1985) describes it, Pirahã communication makes crucial use of the CHANNELS in (4),
below, where Hymes (1974) defines a channel as 'sociolinguistically constrained
physical medium used to carry the message from the source to the receiver'. The four
principal modalities or channels in Pirahã after 'normal' speech are:

(9.5) CHANNEL FUNCTIONS

a. HUM SPEECH Disguise
Privacy
Intimacy
Talk when mouth is full
Child language acquisition relation

b. YELL SPEECH Long distance

Rainy days
Most frequent use – between huts &
across river

c. MUSICAL SPEECH ('big jaw') New information

Spiritual communication
Dancing, flirtation
Women produce this in language teacher
sessions more naturally than men.
Women's musical speech shows
much greater separation of high
and low tones, greater volume.

d. WHISTLE SPEECH (sour or 'pucker' mouth' Hunting

– same root as 'to kiss' or shape of mouth Men-only (as in ALL whistle
after eating lemon) speeches!)
One unusual melody used for
aggressive play

Example (9.5) is illustrates how prosodic information in Pirahã is exploited to

create these channels. The inventory in Table One above, also partially shows how little
Linguistic Fieldwork Daniel L. Everett
page 211
the segments contribute to the total set of phonological information in a given Pirahã
word. In (5) we see that the phrase in (5a) has the quasi-musical representation in (5b),
the basis for the channels just summarized.

(9.6) a. kái/ihí/ao -/aagá gáihí

paca poss/exist-be there
'There is a paca there.'

. ●. ● ●. ●.
b. | ● ●. |   | | ● |
| | |
^ ^ ^
All channels must include the information in (5b), though only the consonant
and vowel channel needs to include the information in (5a). The notes represent
syllables, with 'ties' indicating unbroken falls/rises in whistle speech.
In the musical form in (5b) there is a falling tone, followed by a short low, with
a preceding break in the whistle (where the glottal stop would have been in kai?ihi),
followed by another short break (where the h would be) and a short high tone, and so
on. Thus, the syllable boundaries are clearly present in whistle (humming, and yelling)
channels, even though the segments themselves are missing. The syllable in this case
indicates length, offers an abstract context for tone placement, and the overall word is
stressed according to syllable weight (see Everett (1988) for details). The syllable in
these cases is vital to communication in differing channels, primarily in parsing the
input.
But does the discovery of channels like this imply any causal interaction
between culture and grammar? Or are these channels outside the grammar proper?
Notice that these channels rely crucially on the syllable weights and stress rule in (1)
and (2) above. So, if nothing else, they help account for what is otherwise an
anomalous level of complexity in the stress rule. Yet the facts cut deeper than this.
Consider the following example of what Everett (1985) calls the 'sloppy phoneme
effect' :

(9.7) tí píai ~ kí píai ~ kí kíai ~ pí píai ~ /í píai ~ /í /íai ~ tí píai, etc. (*tí tíai, * gí gíai,
*bí bíai) 'me too'
(9.8) /apapaí ~ kapapaí ~ papapaí ~ /a/a/aí ~kakakaí ~(*tapapaí, * tatataí, *
bababaí, * gagagaí) 'head'
(9.9) /ísiihoái ~ kísiihoái ~ písiihoái ~píhiihoái ~kíhiihoái ~ (alternations with /t/s or
involving different values for [continuant] or [voicing] are unattested) 'liquid
fuel'

Pirahã allows a tremendous amount of variation among consonants, though not

for the features [continuant] or [voice]. This can be accounted for, but only if we refer
to Pirahã’s channels. The ungrammatical examples in (6)-(8) show that the features
[continuant] and [voice] are linked in the sense that they may never vary in the effect.
Only place features may vary. With no reference to channels this is without
explanation. But in light of the channels this follows because [continuant] and [voice]
are necessary for stress placement (Everett (1988)) which in turn must be preserved in
every discourse channel, or the constraint in (9) is violated:
Linguistic Fieldwork Daniel L. Everett
page 212

(9.10) Constraint on functional load and necessary contrast (Everett (1985)):

a. Greater Dependence on the Channel → Greater Contrast Required
b. Lesser Dependence on the Channel → Less Constrast Required

Notice that I am not claiming that the absence of variation for different values of
[continuant] is predicted by 'channels' alone. This case in fact demands that we further
investigate the connection between [continuant] [voice]. There is no claim that
ethnography replaces phonology! But I am claiming that without the study of channels
and their role in Pirahã culture, not even an understanding of Pirahã’s segmental
phonology is possible.
The lesson for the field researcher and theoretical linguist to be drawn from these
examples is just this: first, language and culture should be studied together; second, as a
modality-dependent channel, phonology may be subject to constraints that are (i)
language specific and (ii) grounded not only in the physical properties of the
instantiating modality (the phonetics) but also or alternatively on the culture-specific
channels of discourse employed. This is a very important result because it shows that
the 'interface conditions' of the HUMAN COMPUTATIONAL SYSTEM, in Chomsky's (1995)
terms, may range beyond PF and LF, if we define an interface system as a system
setting bounds on interpretability for HCL. Such examples also show how coherent
fieldwork can be useful for theory. Thus not only the fieldworker, but also the
phonologist must engage the language as forming a coherent whole with culture. And
this in turn entails more culturally informed fieldwork.
Before turning to another case study from Pirahã, on Ethnogrammar, it would be
worthwhile to conclude this particular section with a consideration of some
methodological suggestions for studying ethnography of communication.

9.4. Methodology for ethnogrammatical studies

Some suggestions in Saville-Troike (1982, 108ff) include the following:

(9.11) Methodological suggestions for the ethnography of communication:

a. identify recurrent events
b. analyze these events, examing their function, form, and relationships
between different constituents.
c. examine the relationship between these events to other speech events and
to the society and culture in which they occur.

For example, one might study the use of whistle speech on the Canary Islands.
One variety, Silbo Gomero is used in and around La Gomera. In relation to (), each use
of whistle speech is thus an event. Some questions that might be asked about these
events are: When is it used? Who uses it? What are the constraints on its intelligibility?
(e.g. Can two people understand Silbo under any circumstances or does a topic of
conversation need to be established first to provide context?) How many other channels
of discourse are there among speakers who use Silbo? Are there contents or types of
discourse in which the people prefer to use Silbo? Are the contents or types of
discourse in which the people prefer not to use Silbo? What are the phonetic details of
Silbo and how is it possible (since the language it is based on is not tonal, does it use
inherent segmental frequencies as a basis, intonation, etc.)? How does it relate to the
consonant and vowel channel (i.e. normal speech)? etc.
Linguistic Fieldwork Daniel L. Everett
page 213
We now move to consider a different type of connection between language and
culture, namely, Ethnogrammar.

Ethnogrammar
Enfield (2002,3) makes the important observation that "Grammar is thick with
cultural meaning". Enfield defines ethnosyntax (part of ethnogrammar, in my
conception of the latter as including ethnosyntax, ethnophonology, ethnomorphology,
and ethnosemantics, among other possibilities) as "... the study of connections between
the cultural knowledge, attitudes, and practices of speakers and the morphosyntactic
resources they employ in speech..." And "This field of research asks not just how
culture and grammar may be connected, but also how they may be interconstitutive,
through overlap and interplay between people's cultural practices and preoccupations
and the grammatical structures they habitually employ." To better appreciate the nature
of what is meant by 'Ethnogrammar' let's consider another case study from Pirahã, one
that has received a good deal of attention from linguists, anthropologists, and the
popular media, the connection between 'immediacy of experience' and the group's
cultural and grammatical structures. Following this, I conclude the chapter with
methodological suggestions on the investigation of Ethnogrammar.
Everett (2005) notices the following facts about Pirahã:

(9.12) Pirahã facts in need of explanation by anyone's account:

a. Pirahã lacks number, numerals, or a concept of counting.
b. Pirahã lacks color terms.
c. Pirahã lacks syntactic recursion.
d. Pirahã has the simplest pronoun inventory known and evidence suggests that
Pirahã's entire pronominal inventory may have been borrowed (see Appendix
Two).
e. Pirahã has no perfect tense.
f. Pirahã has perhaps the simplest kinship system ever documented.
g. Pirahã has no creation myths – its texts are almost always descriptions of
immediate experience or interpretations of experience; it has some stories about
the past, but only of one or two generations back.
h. The Pirahã in general have no individual or collective memory of more than
two generations past.
i. Pirahã people do not draw, except for extremely crude stick figures
representing the spirit world that they (claim to) have directly experienced.
j. Pirahã has no terms for quantification, e.g. 'all', 'each', 'every', 'most', 'some',
etc.
k. The Pirahã are monolingual after over 300 years of regular contact with
Brazilians.

Some examples of this follow:

LACK OF NUMBER, NUMERALS, AND COUNTING

(9.16) hiaitíihí hi kaoáíbogi bai -aagá
Pirahã people he evil spirit fear -be
'The Pirahã are afraid of evil spirits.' OR 'A Pirahã is afraid of an evil spirit.'
OR 'The Pirahã are afraid of an evil spirit.' OR 'A Pirahã is afraid of evil
spirits.'
Linguistic Fieldwork Daniel L. Everett
page 214

(9.17) kó'oí, kóhoibiíhai, hi píai, 'aáibígaí, hi

name name he also, name 3

píai, hi koabáipí
also, he die
'Kó'oí, Kóhoibiíhai, and 'aáibígaí died.'

(9.18) báigipóhoaá 'i 'óooí kobai -baaí

name:f she tarantula watch -intensely
'Báigipóhoaá watched the tarantula(s) closely.' (this can refer to one woman
named 'Báigipóhoaá or several)

LACK OF NUMERALS
There are three words in Pirahã that are easy to confuse with numerals, because
they can be translated as numerals in some of their uses.51 These are listed in (9)-(11):

(9.18) a. hói 'small size or amount'

b. hoí 'somewhat larger size or amount'
c. bá a gi so 'lit: cause to come together (loosely
'many')
touch-causative associa nominalizer
-tive
Some examples which show how Pirahã expresses what in other cultures would
be numerical concepts:

(9.20) a. tí 'ítíi'isi hói hii 'aba'áígio 'oogabagaí

I fish small pred. only want
'I only want {one/a couple/a small} fish.' (NB: This could not be used
to express a desire for one fish that was very large, except as a joke.)
b. tiobáhai hói hii 'small child/child is small/one child'

(9.21) a. tí 'ítíi'isi hoí hii 'oogabagaí

I fish larger pred. want
'I want {a few/larger/several} fish.'
b. tí 'ítíi'isi báagiso 'oogabagaí
I fish many/group want
'I want {a group of/many} fish.'
c. tí 'ítíi'isi 'ogií 'oogabagaí
I fish big want
'I want {a big/big pile of /many} fish.'

51
The 'translation fallacy' is well-known, but field linguists in particular must be ever-
vigilant not to be confused by it. Bruner, Brockmeier, and Harré (2001, 39) describe it
as the supposition that there is only one human reality to which all 'narratives 'must in
effect conform – be they fiction or linguistic theories, say. Throughout this paper, I will
urge the reader to be on guard against this – the mistake of concluding that language x
shares a category with language y if the categories overlap in reference.
Linguistic Fieldwork Daniel L. Everett
page 215
There are likewise no ordinal numbers in Pirahã , e.g 'first', 'second', etc. Some of
the functions of ordinals are expressed via body parts, in a way familiar to many
languages:

(9.22) ti 'apaí káobíi 'ahaigí hi tíohió'ío/gaaba káobíi

1 head fall same generation he towards me/there stay fall
'I was born first then my sibling was born.' (lit: 'I head fall sibling to
me/there at fall.')

The two expressions in (14), tíohió'ío/gaaba, are interchangeable in most

contexts. They refer to both intermediate points in a succession of participants, events,
etc. or to the final position. But we need to be clear on one thing, namely, that the word
'head' does not really mean 'first', not if we assume that 'first' derives its meaning
partially in opposition to 'second', 'third', etc. but overlaps with 'first' in referring to
something 'at the beginning of a spatial or temporal sequence'.52
The Pirahã language has no words for individual fingers, e.g. 'ring finger', 'index
finger', 'thumb', etc. They occasionally refer to their fingers collectively as 'hand sticks',
but only when asked by an insistent linguist. By the same reasoning, there is no word
for 'last'. Moreover, they do not point with individual fingers. If they use any part of
their arms for pointing, they tend to extend a flat hand, turned sideways, or an open
palm facing up or down. More often, they point, as is common around the world, with
their lower lip or jaw, or a motion of the head. When discussing a large
quantity/number of objects, they do not make tallying motions on individual
appendages, etc. If they use gestures, they hold the flat hand out, palm down, varying
the distance between hand and ground to indicate the size of the 'pile' or amount under
discussion. However, a seated Pirahã man or woman (though women rarely do this)
occasionally will extend both feet and hands, with toes and fingers also extended to
indicate a large number of individual items (they would not do this in my experience
for a non-individuated quantity, such as manioc flour, but rather for bags of manioc
flour, etc.). Other than these gestures, there is no other use of body parts, objects, or
anything to indicate a concept of 'tallying'.
Very suprisingly (see Davidson () and Wierzbicka () for the predictions that such
quantifiers will be found in all languages), there are no quantifier terms like 'all', 'each',
'every', 'most', 'few' in Pirahã. There are also no 'WH-quantifiers' per se.53 To appreciate
this, let us consider the examples in (15)-(18), to see the closest expressions Pirahã can
muster to these quantifiers:

ALL
(9.23) hiaitíihí hi 'ogi -'áaga -ó pi -ó
Pirahã people he big -be (permanence) -direction water
-ó kaobíi

52
Part of the conclusion of this paper, agreeing with Gordon (2003), is that much of
Pirahã is largely incommensurate with English and so translation is simply a poor
approximation of Pirahã intentions and meaning, but we do as well as we can do.
53
One reviewer suggests that these Pirahã words are quantifier words, but have
different truth conditions from their English counterparts. But having different truth
conditions just means have different meanings in this context so if it could be shown, as
I do here, that they have different truth conditions then they are different words. Period.
Linguistic Fieldwork Daniel L. Everett
page 216
-direction entered
'All the people went to swim/went swimming/are swimming/bathing, etc.'

MOST
(9.24) ti 'ogi -'áaga -ó 'ítii'isi 'ogi -ó
I big -be(perm) -direction fish big -direction

'i kohoai -baaí,

she eat -inten.

koga hói hi hi -i kohoi -hiaba

nevertheless small amount intens. intens. -be eat -not
'We ate most of the fish.' (lit: 'My bigness ate (at) a bigness of fish, nevertheless
there was a smallness we did not eat.'

Example (9.25) is the closest I have ever been able to get to a sentence that would
substitute for a quantifier like 'each', e.g. 'each man went to the field'.

EACH
(9.25) 'igihí hi 'ogiáagaó 'oga hápií; 'aikáibaísi, 'ahoáápati
man he bigness field went name, name
pío,
also,
tíigi hi pío, 'ogiáagaó
name he also bigness
'The men all went to the field, 'aikáibaísi, 'ahoáápati, tíigi all went.'

FEW
(9.26) gáta -hai hói hi -i
can -foreign object small intens. -be
'aba -'á -ígi -o 'ao -aagá
remain -temp -associative location possession -be
(temporary)

('aba'áígio can often be translated as 'only', though I give its full morphological
breakdown here to show that it is not really equivalent in meaning to 'only'. Nor
does it share the full range of meanings of 'only')

'agaoa ko -ó
canoe gut -direction
'There were (a) few cans in the foreigner's canoe.' (lit: smallness of cans
remaining associated was in the gut of the canoe')

There are two words, usually occurring in reference to an amount eaten or

desired, which by their closest translation equivalents, 'whole' báaiso and 'part' gíiái
might seem to be quantifiers:

(9.27) a. tíobáhai hi bá -a -i -so

child he touch -causative -connective -nominalizer
Linguistic Fieldwork Daniel L. Everett
page 217
'whole'

kohoai -sóog -ab -agaí

eat -desiderative -stay -thus
'The child wanted/s to eat the whole thing.' (lit: 'Child muchness/fullness eat
is desiring.')

b. tíobáhai hi gíi -ái kohoai -sóog

child he that -there eat -desiderative
'part' (in the appropriate context)

-ab -agaí
-stay -thus
'The child wanted/s to eat a piece of the thing.' (lit: 'Child that there eat is
desiring.')

In (19) báaiso and gíiái are used as nouns. But they can also appear as
postnominal modifiers:

(9.28) a. tíobáhai hi poogaíhiaí báaiso kohoai

child he banana whole eat

-sóog -ab -agaí

- desiderative -stay -thus
'The child wanted/s to eat the whole banana.' (lit: 'Child banana
muchness/fullness eat is desiring.')

b. tíobáhai hi poogaíhiaí gíiái kohoai -sóog

child he banana piece eat -desiderative

-ab -agaí
-stay -thus
'The child wanted/s to eat part of the banana.' (lit: 'Child banana piece eat is
desiring.')

Aside from their literal meanings, there are important reasons for not
interpreting these two words as quantifiers. First, their Truth Conditions (see chapter
___ above) are not equivalent to those of real quantifiers. For example, consider the
contrast in (21) vs. (22):
Context: Someone has just killed an anaconda. Upon seeing it, (21a) below is
uttered. Someone takes a piece of it. After the purchase of the remainder, the content of
(21a) is reaffirmed as (21b):

(9.29) a. 'áoói hi paóhoa'aí 'isoí báaiso

foreigner he anaconda skin 'whole'

'oaboi -haí
buy -relative certainty
'The foreigner will likely buy the entire anaconda skin.'
Linguistic Fieldwork Daniel L. Everett
page 218

b. 'aió hi báaiso 'oaob -áhá; hi 'ogió

affirmative he whole buy -complete certainty 3 bigness

'oaob -áhá
buy complete certainty
'Yes, he bought the whole thing.'

Now, compare this with the English equivalent, where the same context is
assumed:

(9.30) a. STATEMENT: He will likely buy the whole anaconda skin.

b. OCCURRENCE: Piece is removed (in full view of interlocutors).
c. STATEMENT: %He bought the whole anaconda skin.

It simply would be dishonest and a violation of the meaning of 'whole' to utter it

in (22b). But this is not the case in Pirahã, (21b).

Next, there is no truly quantificational-abstraction usage of báaiso 'whole':

(9.31) *Ti 'ísi báaiso 'ogabagai, gíiái 'ogi -hiaba.

1 animal 'whole' want, piece want -negative
'I prefer whole animals to portions of animals.' (lit: 'I desire (a) whole
animal(s), not piece(s).')

Sentences like (9.32) cannot be uttered acceptably in the absence of a particular

pair of animals or instructions about a specific animal to a specific hunter. That is,
when such sentences are used, they are describing specific experiences, not
generalizing across experiences.
It is of course more difficult to say that something does not exist than to show that
it does exist, since in the former instance a skeptic can always reply that you have not
looked hard enough. Nevertheless facts like those discussed in this discussion, in the
context of my nearly three decades of regular research on Pirahã, lead me to the
conclusion that there is no strong evidence for the existence of quantifiers in Pirahã.
ABSENCE OF RECURSION54
With regard to recursion, as mentioned in ___ above, there are two varieties,
where I claim that Pirahã lacks either type. I repeat these here:
(9.32) A → AB
(9.33) a. A → BC
b. B → DE
c. C → AF
The lack of rules like (1) explains why Pirahã lacks the following:
54
Everett (2005) was not always clear on which of these was in focus. Due to space
limitations, I only gave evidence against (1), even though (2) is the principal claim. In
work in progress, I tighten up the arguments (thanks to suggestions from David
Pesetsky and David Adger) so as to rule out system recursion as well, thus
strengthening the case against the view of 'creativity in human language' advocated in
Hauser, Chomsky, and Fitch (2002).
Linguistic Fieldwork Daniel L. Everett
page 219

(9.34) Coordination:

a. [NP John and Bill] came to town yesterday.

b. [S [S I saw [NP Mary, Sue, and Willy] [PP in town and at the mall]
and [S I saw some other people too]].

(9.35) Disjunction:

a. Either [NP Bob or Bill] will come.

b. I had [NP chicken or pork], some white meat.

(9.36) Embedding:

a. I doubt [S whether they will come].

b. John says [S that Bill thinks [S that Mary will agree [S that Sue
should come too]]].

(9.37) Basic phrase structures:

a. I came [PP with the Pope].

b. [NP The big man] got sick.

c. The [AP very big] man got sick.

(9.38) Semantic scope requiring embedding:

a. He didn't say it was raining.

b. He will not say if he is coming.

Everett (2005) argues that these facts follow from the constraint on 'immediacy of
experience' in ():

(9.39) Immediacy of experience expression in Pirahã: Declarative Pirahã utterances

contain only assertions related directly to the moment of speech, either experienced (i.e.
seen, overheard, deduced, etc. – as per the range of Pirahã evidentials, as in Everett
(1986, 289)) by the speaker or as witnessed by someone alive during the lifetime of the
speaker).55

55
The original formulation was: "Grammar and other ways of living are restricted to
concrete, immediate experience (where an experience is immediate in Pirahã if it has
been seen or recounted as seen by a person alive at the time of telling), and immediacy
of experience is reflected in immediacy of information encoding—one event per
utterance." David Adger (personal communication) rightly points out that I will need to
Linguistic Fieldwork Daniel L. Everett
page 220

To use an old term of anthropology in a new framework, the constraint in ()

is in effect a cultural taboo. If this proposed taboo were the correct way of looking at
what is going on, it would automatically and exactly capture the facts that were listed in
Everett (2005), i.e. the lack of:

(9.40) Embedding: since embedded sentences are not assertions (Cristofaro (2003)),
they cannot be used. To avoid these, the grammar of Pirahã will not have rules
of the type in (2) above. This will explain all the anti-recursion effects in the
paper.
(9.45) Number & numerals: These are skills that have both immediate application and
wider application, ranging beyond immediate experience. Since the latter uses
would violate the cultural principle in (3), however, these are not available in
the grammar (interestingly, counting and numerals involve recursion, which
could be taken as additional evidence that Pirahã lacks recursion).
(9.46) Relative tenses: These involve assertions defined in terms other than the
moment of speech. So when I say in Pirahã 'When you arrive, I will go', as I
show in Everett (1993), both 'arriving' and 'going' are defined relative to the
moment of speech, (however, one could argue that relative tenses involve
recursion and so are for this reason unavailable). More complex tenses would
violate (2).
(9.47) Kinship terms: All kinship terms are related directly to the one speaking (the
controller of the 'moment of speech', i.e. ego) and none are defined in terms of
other relations (i.e. no kinship terms involve recursion, e.g. grandfather,
grandson, etc.).
(9.48) Color terms and quantifiers: Color terms and quantifiers can identify
immediate experiences, as can numbers, but, like numbers, are avoided by the
grammar because they also entail a significant component of ranging beyond
immediate experience.
(9.49) Myths and fiction: These violate the evidentiality constraint in (3).

One reader of Everett (2005) wondered why, if the above is correct, Pirahã has
nouns, since there could be, for example, abstract nouns, i.e. ranging beyond immediate
experience. This objection doesn't follow, though I see the point. First, Pirahã does lack
abstract nouns. Second, it cannot do away with the (semantic) category of nouns,
because all languages must have terms that represent entities and terms that provide
information about them. Pirahã is a language, so cannot get by without noun-like
elements.

9.6. Linking Ethnography and Grammar

Whether or not Everett () is right about capturing all these facts under the single
generalization in (), the Pirahã facts nevertheless provide strong evidence for the
importance of investigating the culture-language nexus as part of fieldwork. Let us
therefore consider some methodological suggestions for investigating this connection.
First, I suggest a few questions that might lead to a methodology.

(9.50) Pre-methodological ethnogrammatical questions:

discuss how this applies to modality.

Linguistic Fieldwork Daniel L. Everett
page 221
a. Are there irregularities that have no obvious structural explanation?
b. Are there examples 'free variation', i.e. where there are choices between two
structures which are not constrained by the structures or the grammar, in so far
as can be determined?
c. Are there unusual facts about the cultural events, values, or explanations that
involve principles or phenomena that at any level look similar to principles
operative in the grammar?

As to the methodology that follows from such questions, Enfield (2002, 14ff)
offers some cogent and very important considerations and suggestions for the study of
ethnogrammar. First, he recommends that the fieldworker "Examine specific
morphosyntactic structures and/or resources and make explicit hypotheses as to their
meaning." Second, following development of this and related methodological
considerations, he raises the crucial issue of 'linkage', namely, how can we establish a
causal connection between facts of culture and facts of grammar? I turn to this directly.
Before doing this, however, I want to point out what seems to be the biggest lacuna in
the study of ethnogrammar, whethere in the studies in Enfield (2002) or elsewhere.
This is the effect of values, especially cultural taboos like () above, in restricting both
culture and grammar. That is, previous studies, like those in Enfield (2002), while
reasonably focusing on meaning, which is after all a principal contribution of culture
(i.e. guiding its members in finding meaning in the world), fail to consider cultural
prohibitions or injunctions, however deeply or shallowly embedded in the community
system of values. The Pirahã example of this section is evidence that such values
should also be considered in ethnogrammatical studies. However, before we can draw
any conclusions at all about ethnogrammar in a given language, we need to consider the
vital issue that Enfield refers to as 'linkage', i.e. the establishment of a causal
connection between culture and language. That is, how can we convince someone or, at
least, effectively argue that property p of culture C causally determines feature f of
grammar G? According to Clark & Malt (1984), cited by Enfield (2002, 18ff) there are
four prerequisites to establishing linkage between culture and language:

(9.51) Culture – grammar linkage prerequisites

a. Empirical grounding
b. Structure independence
c. Theoretical coherence

And also (9.52):

(9.52) Avoid circularity

A circular argument in ethnogrammatical studies would be to claim that a

particular linguistic feature is simultaneously determined by an aspect of culture and
evidence for that aspect of culture (so, for example, 'The language has evidentials
because the culture values empirically-based reasoning.' And then 'We know that the
culture values empirically-valued reasoning because it has evidentials'). The way to
avoid this is to first establish, using nonlinguistic evidence, particular values or
meanings in a certain culture. Next, using noncultural evidence, establish the meaning
and structure of the relevant linguistic examples. Finally, show how linking the two
Linguistic Fieldwork Daniel L. Everett
page 222
provides a conceptually and empirically (in terms of predictions where possible, or
explaining independent domains such as historical change) superior account of the facts
that leaving them unconnected.
Ethnogrammatical studies thus range from showing that, say, a language has
honorifics because of a severe social structure, or a particular set of kinship terms
because of its restrictions on marriage, to (what most researchers have overlooked), the
kinds of global, architectonic constraints on grammar from, e.g. taboos like () above.
Linguistic Fieldwork Daniel L. Everett
page 223
Chapter 10. The written record: grammars, dictionaries, and text collections
10.1. Grammars
10.1.1. Why write a grammar?
I once heard a colleague remark that no one should write an article about a
language until they have written a grammar of the language. I think he was serious. It is
good advice, even though it is exceedingly stringent. For one thing, writing a grammar
deepens and broadens one's knowledge of a language. Also, writing a grammar helps a
linguist get the 'big picture', seeing how things fit together – kind of like helping to
understand each tree better by understanding its place in the forest. This is what Pike
(19--) would have called the 'field' view of language.
To my way of thinking, writing an article about a single construction or rule
without a good grasp of the grammar as a whole is like writing about the function and
form of a puzzle piece without knowing what the whole puzzle looks like. Of course,
many languages already have grammars. So one could read a grammar instead of
writing one. Certainly, noone should write about a language without at least reading
entire grammars of that language when available. I would, on the other hand, tend to
avoid working on languages where a grammar has been written, simply because so
many other languages need to have descriptive grammars written.
In any case, reading a grammar is a poor substitute for writing one, if you really
want to figure out how a language works. Writing a grammar teaches you that every
part of a grammar is a series of decisions on how best to bring order from the chaos of
data that the linguist has collected. There is more or less security in the usefulness of
these decisions in different parts of the grammar. Often only the grammar-writer knows
this, even when they are not at all trying to sweep recalcitrant facts under the rug. I can
think of no more challenging, rewarding, important or urgent task for individual
linguists than writing a grammar of a language based on their own field research.
Ultimately, a grammar is the linguist's theory of how a specific language works,
atomistically and holistically, i.e. what are its 'bits' and how do these 'bits' fit together.
A grammar is the result of careful methodology, lots of hard thinking, innumerable
conversations with native speakers, artistic flair, boldness (to propose connections or to
say that 'x' does not exist in language 'y', etc.), lots of luck, and huge amounts of
reading, planning, testing, and interpreting. A grammar may even identify a 'theme' of a
particular language – recurring properties, semantic or formal or cultural or societal –
that are found in multiple constructions, text types, social patterns, etc. throughout the
language. A grammar should ideally include an ethnography of communication, for
reasons presented in chapter __ above.56
So how does one do fieldwork with writing a grammar in mind? First, have a
basic outline of a grammar in mind from the start of your fieldwork. One such outline is
the Lingua Descriptive Questionnaire by Norval Smith and Bernard Comrie (this can
be found on the Department of Linguistics website for the Max Planck Institute for
Evolutionary Anthropology: http://lingweb.eva.mpg.de/fieldtools/linguaQ.html).
Another is my own phonology questionnaire (see appendix ___, also on the MPI site
above) or the phonology outlines of one of the volumes in the Oxford Phonology of the
World's Languages series.
Another way to do fieldwork guided by questions relevant to writing a grammar
is to invest time learning the craft of grammar writing by reading grammars that are

56
Give sources on writing ethnographies of communication.
Linguistic Fieldwork Daniel L. Everett
page 224
considered exemplary by other linguists, either for a given region of the world or in
general. Linguists should read grammars for fun and for professional development. If
you do not like to read grammars, you may be in the wrong profession.
However, ultimately, the questions you ask will reflect your own interests and
background. What attracts your attention most in this grammar or language-culture
pair? What kinds of things do you believe are in most need of being said about this
language? What have you learned from native speakers about their view of their own
language? What would they liked highlighted? (Alternatively, would they want
anything omitted from public discussion, even grammatical forms, certain kinds of
texts or semantic domains, etc.?)
First I begin with a discussion of the different kinds of grammars, including the
following three main types: Reference, Descriptive, and Pedagogical Grammars, as
well as the usefulness of 'grammar fragments'. I will emphasize that quality is more
important than quantity, but that a certain coverage is required to get a 'feel' for the
'genius' of the language and knowing how and where to 'hang' each piece in the
grammar. I discuss grammar-writing as a literature genre in relation to the remarks on
its fallibility in Chapter One and the implications of this for different methods and
attitudes towards grammar-writing. I also discuss ways of testing the grammar as a
whole with native speakers, e.g. reading it to groups of them as it is written and after
the entire grammar is in draft. I next turn to consider the task of dictionary-making,
comparing and contrasting different kinds of dictionaries. I then discuss the importance
and methodology of compiling representative collections of texts, framing the
discussion in terms of at least the following parameters:

10.1.2. Grammars are the nucleus of documentation and description

The goals of a project of documentation and description are to provide a record of
the sounds and meanings of a language and how these are associated with one another
and with the culture in which they are embedded. Sound files of high quality, texts, and
a useful and fairly extensive dictionary form the core of such documentation. But the
grammar provides not only the description (the analytical account of how the pieces of
the structure of the language fit together) but also is the tie that binds together the
various components of the documentary evidence of the language's sounds, texts, and
words. A complete documentary and descriptive project for a language will include the
following components:

(10.1) a. Descriptive grammar

b. Sound files (all prosodies and segments exemplified)
c. Dictionary
d. Texts
e. Ethnography of communication
f. Pedagogical grammar portions

So for most fieldworkers, even if their own objectives are not a full
documentation and description of a language, contributing to the construction of a
grammar of the language under study is at least an important goal to which their
research should contribute. In other words, if the linguist is doing research on the
phonetics of a language's segments, on information questions and dislocation, or on
voice alternations, etc., the research should be written up in such a way that it can be
Linguistic Fieldwork Daniel L. Everett
page 225
incorporated into a grammar of the language eventually. This is done by providing
careful glosses, high-quality sound files, and relative jargon-free descriptions.

10.1.3. Types of grammars

10.1.3.1. Pedagogical vs. Reference Grammars
There are at least three types of grammars – pedagogical, descriptive or reference,
and theoretical. A comparison between pedagogical and reference grammars is given in
Table 1 below (see also the SIL website
(http://www.sil.org/linguistics/GlossaryOfLinguisticTerms/WhatIsAReferenceGramma
r.htm)
Table One: Reference vs. Pedagogical Grammars

Type of Audience Purpose Organization Style of

Grammar Presentation

Pedagogical Native speakers To provide In terms of short Textbook like,

looking for a lessons in how lessons on for learners, to
tool to preserve to speak and specific topics inform and to
and teach their
language write the of develop skills.
language pronunciation,
word usage, etc.

Reference Linguists who To persuade and Things are Clear for a wide
want specific to inform. To presented in the readership, but
kinds of provide a way the linguist technical when
considers most
technical cohesive picture effective to necessary and
information and of the language understand the with sufficient
who will not that allows for language, but argumentation
believe both will usually be in to convince
everything you comparisons the order well-informed
say. with other phonology- readers.
phonetics-
languages and morphology-
understanding syntax-
the 'genius' of semantics-
the language at discourse
hand.

A pedagogical grammar is an educational tool and as such requires different skills

and training from a reference grammar. Also, since the pedagogical grammar is largely,
if not exclusively developed for the needs and uses of the language community, it must
be built up slowly, based on experience in using it as a teaching tool, discussing it with
community teachers, getting advice from educational experts, and so on. Moreover, a
pedagogical grammar should be based carefully and solidly on a comprehensive
Linguistic Fieldwork Daniel L. Everett
page 226
reference grammar. It should not be developed in isolation from the larger effort of
describing, arguing, and reasoning about the structures of the language.57
Pedagogical grammar development can require skills in motivation as well. For
example, if the community leadership wants the linguist to lead an effort for language
revitalization, where a pedagogical grammar is to play a major role, the linguist will
have to engage the active participation of native speakers and those trying to learn the
traditional language of their people, to help shape and improve the pedagogical and
analytical aspects of this specialized grammar.
A reference grammar, on the other hand, is designed to provide analyses,
descriptions, and data in a cohesive, relatively standardized form to a general audience
of professional linguists and other linguistically sophisticated readers. These can follow
rigid guidelines (e.g. the Croom Helm descriptive series, see Everett & Kern (1997)) or
allow greater freedom for the linguist to describe the grammar as they think best (e.g.
the new Cambridge series of descriptive grammars (see Dixon (2004)). The former
model, when part of a series, has the advantage of producing a range of grammars to
serve as a comparative reference set, an encyclopedia of language structures
appropriate for rapid comparisons. This is also its disadvantage, however, because its
detailed, meticulous, and highly constraining outline can be too confusing or force the
linguist to distribute across various subpoints of a detailed outline a discussion that
would have been best presented as a whole, a single portion of the grammar. Reference
grammars have the general advantage of being directed towards all linguists and thus
do not generally require any special jargon or theoretical background to read.

10.1.3.2. Theoretical grammars

The final type of grammar I want to discuss in this section is the theoretical
grammar. The first in-depth application of generative grammar was G.H. Matthews's
grammar, Hidatsa Syntax (Matthews 1964). Hale (1967, 341) concludes his review of
this book by claiming that this grammar is "... an outstanding landmark in American
Indian studies as well as a highly significant contribution to general linguistic inquiry.
This work is very much a credit to modern linguistics." Chomsky further situates the
theoretical grammar in modern theory:

"LSLT [Logical Structure of Linguistic Theory, Chomsky (1975), DLE] and other
detailed work of the 1950s (particularly G.H. Matthews, Hidatsa Syntax) at once
revealed a tension between descriptive and explanatory adequacy. As soon as serious
descriptive work was undertaken, it was discovered that available accounts of
language, however extensive, barely scratched the surface; even the most
comprehensive grammar provided little more than hints that sufficed for the intelligent
reader; the language faculty was tacitly presupposed (without awareness, of course).
The same is true of the most comprehensive dictionary. To attain descriptive adequacy,
it seemed necessary to construct extremely intricate and complex grammars, radically
different for different languages. On the other hand, to approach explanatory adequacy

57
In my Arts and Humanities Research Council project for the documentation and
description of the Kisedje language, one of my research associates was an expert in
indigenous education with more than fourteen years of experience in education among
indigenous groups of the Xingu Park, where the Kisedje live.
Linguistic Fieldwork Daniel L. Everett
page 227
it was necessary to assume that the states attained are determined to an overwhelming
extent by the initial state, which is language-invariant. Thus languages must all be cast
to the same mold, differing only superficially. The major research project was aimed at
overcoming this tension by showing that the apparent complexity and variety of
language was only superficial, the result of minor changes in a fixed and invariant
system." Chomsky (1994)

Moreover, as Hale's review shows in considerable detail, Matthews's book

provides detailed argumentation and explicit structures for most major aspects of
Hidatsa syntax. In addition, for researchers working within Generative Grammar in the
early to mid 1960s, Hidatsa Syntax provided a model for how to write a grammar.
Judged when it appeared, therefore, this volume was a pioneering, extremely important
study, almost universally accepted by theoreticians as a significant improvement over
previous grammars.
The problem is that today, forty-one years on from the publication of Hidatsa
Syntax, this grammar is almost universally lamented as being of very little use to
anyone wanting to learn about either Hidatsa or Generative Grammar (the latter
because the field has changed so dramatically in the intervening years). So this once
proud grammar is today relatively useless, exercising almost no influence at all,
whereas the so-called 'taxonomic' (i.e. reference or descriptive) grammars it was once
so favorably compared to continue to be read and serve to enlighten new generations of
linguists about the languages they describe. And in my opinion and experience the
same fate has befallen every theoretical grammar written since.
This is not to say that theoretical grammars should be avoided. But we need to
draw at least two lessons from the fate of Hidatsa Syntax and its ilk. First, if one does
write a theoretical grammar, the theoretical discussion contributes to very short-term
(howbeit potentially important) goals, a sort of rock in the stream that can divert the
flow around it, though it itself is long forgotten downstream. Another lesson is that it is
useful to reconsider the organization of a theoretical grammar. My PhD on Pirahã, for
example (Everett (1990 [1983]), was divided into two major sections, a reference
grammar (also published in English as Everett (1986) and a theoretical discussion. The
theoretical discussion is no longer particularly relevant, but the reference grammar
portion can still be used by linguists regardless of theoretical orientation. To sum up,
the ideal grammar should be useful for present and future generations; it should be
clear; it should anticipate reader questions and objections and deal with these by
detailed argumentation (except in the case of a pedagogical grammar).
Let us conclude by considering a couple of ways to enhance the usefulness
of the grammar. First, the grammar must be tested. This can be done by re-checking the
entire grammar, or at least the parts that are controversial, theoretically significant, or
for which the linguist feels less secure, with native speakers in groups and with
individuals. Second, be sure that the grammar includes argumentation for every major
analytical assertion. If the linguist says that 'Stress goes on every other syllable from
right to left within the word', then they should explicitly alternative analyses (e.g. stress
first picks out the rightmost syllable and then goes from left to right), by considering
hypothetical or ungrammatical forms, by spectrographic analysis, etc. (see ___ above).
Argumentation is at once a display of erudition and reasoning and tremendously
enhances the reader's opinion of the grammar-writer, the usefulness of the grammar,
and the likely reliability of the grammar as a whole. In a reference grammar,
argumentation will be less theory-internal than in a theoretical article, but it is
Linguistic Fieldwork Daniel L. Everett
page 228
nonetheless crucial. Finally, tell the reader something about the empirical basis for the
grammar. What size is the corpus? What is the corpus like? How long was the linguist
in the field? What was the role of the community in the analysis (including how
language teachers were involved)? How many hours a day did the linguist work and
how did they work (e.g. 'I was in the community for six months and I worked three
hours per day with language teachers and five hours a day filing, analyzing, and
retesting examples by walking around the village trying out my knowledge). And so on.
I turn now to consider the corpus upon which the grammar is based.

10.1.4. The corpus for the grammar

To write a grammar of any type requires a significant corpus of data. What is the
nature and size of an 'adequate' corpus for a grammar? The answer is simple: an
adequate corpus must be varied, natural, and big enough. Let's consider each of these
terms.
First the corpus should be varied. That is, it should contain the greatest possible
number of distinct form, meaning, and construction types. It should vary for ages of
language teachers, emotional states, content, social class of teacher, genders, speeds of
utterance, topics of discourses, genres of discourses, styles of discourses, and channels
of discourse (see ___ for the concept of 'channel').
The corpus should be natural, providing data that native speakers utter (in the
appropriate context) or naturally reject (That is, a good corpus contains (tested)
ungrammatical examples as well). This latter point is very important theoretically and
methodologically. Theoretically, the presence of ungrammatical utterances recognizes
that there are many facts about speaker's knowledge of their language that cannot be
learned simply by observing what they say or do not say. Methodologically, however,
this raises the issue of how to collect grammaticality judgments. So before considering
the final question, i.e. when is the corpus 'big enough', we need to take up more
carefully the implications of ungrammatical examples in field research.
Ungrammaticality is a difficult notion, as Schutze () and Cowart () demonstrate at
length. The reason that ungrammaticality is problematic is that utterances can be
rejected by native speakers for a variety of reasons. Let me give an example from my
fieldwork on the Wari' language. As Barbara Kern and I were working on the grammar
of Wari' (Everett & Kern (1997)), a particular construction (see Everett (2005c)) type
struck me as extremely weird. I could not believe that the facts were exactly as Barbara
had described them to me. Even though Kern's knowledge of the language seemed
impeccable and even though I had never had reason to doubt her before, I just found
these facts hard to take in. When I raised this concern, Kern's response was typical of
her – "Then go to the village and check them out with the Wari' for yourself." Since we
were working only a day's travel from the village, I took her suggestion and departed
the next morning. I met thirty or so Wari' speakers on the banks of the Mamoré river on
the Brazil-Bolivia border and we sat down to consider the examples.
As I read a couple of examples out loud, the people seemed mystified. "That
makes no sense at all", said the man on my right. "Who told you we say that?" asked
the man on my left. So just as I thought, these example are ungrammatical! Then I
answered the man on my left, "Barbara told me you say these things." Surprised looks
all around. Someone then asked, "Barbara said that? Hmm. Then we must say that.
Let's think." So people began discussing the example and then they began to smile.
"Oh, yeah, we do say that. If we are talking about something and want to say something
about it like this, then we say it just like you said there. Just like Barbara told you." The
Linguistic Fieldwork Daniel L. Everett
page 229
utterances were now perfectly fine, we discussed them, they corrected my
pronunciation (segmental and prosodic) and I was satisfied that they were indeed
grammatical utterances. But this was clearly a nonstandard way of getting at their
grammaticality. In the normal course of events in, say, English, linguists simply ask
individual language teachers if this or that utterance is grammatical. Yet the Wari'
example shows that speaker's initial reactions can be quite misleading.
Why and when is an utterance accepted or rejected? The Chomskyan distinctions
between competence (grammatical factors) and performance (cognitive factors, among
others), as well as the Saussurean distinction of parole vs. language (social factors),
when applied to such judgments predict that speakers can accept or reject utterances for
a variety of reasons. For example, the sentences in () and () are grammatical but
unacceptable, both famous examples in the literature:

(10.2) Buffalo buffaloes Buffalo buffaloes buffalo buffalo Buffalo buffaloes.

Meaning: Buffaloes from Buffalo, NY who hoodwink buffaloes from Buffalo,
NY hoodwink buffaloes from Buffalo, NY – i.e. this is a 'center-embedded' tautology.

(10.3) Oysters oysters eat eat oysters.

Meaning: Oysters that oysters eat (also) eat Oysters.

Both examples are hard to process but are grammatical, as linguists have known
for decades. The traditional distinction, based on the competence-performance
dichotomy, is to say that both are grammatical but unacceptable. Sometimes one hears
too that an utterance can be acceptable but ungrammatical, as in the fairly trivial
example I have constructed in ():

(10.4) Q. Who would do such a childish thing? (Wife to husband)

A. Me do it. (Husband replies sarcastically to wife.)

The answer in () is not grammatical, but it is an acceptable utterance in that

circumstance for the purpose of expressing sarcasm. Although we can thus make the
distinction between 'acceptable' and 'grammatical' conceptually clear, separating them
practically in fieldwork is nontrivial (see also Schutze ()). Moreover, the difficulty of
distinguishing these vital concepts, in a Whorfian twist, becomes even more intense in
the lack of any terms for these concepts. That is, there are relatively few languages in
the world that actually have metalinguistic terms like this such that the linguist can
actually ask "Is this grammatical?" or "Is this acceptable?" And circumlocutions such
as "Is this 'pretty'?", "Is this OK?", "Can I say this?", "Is this good English?", etc. are all
extremely unreliable.
Consider "Can I say this?" (or "Can you/one/a Pirahã/a speaker of your language
say this?"). In my experience language teachers often say "Yes, you can." without the
slightest regard for the acceptability or grammaticality of the utterance. This is partially
because the linguist is paying to talk so, to some language teachers, the linguist can say
anything they damn well please. None of these expressions in fact can be relied upon to
give clear evidence of either the acceptability of an utterance or, a fortiori, its
grammaticality. There are simply too many variables for the fieldworker to know why a
particular answer has been given – does the speaker's judgment reflect pragmatics,
semantics, syntax, phonology, lack of concentration, desire to please, or feelings of
different status? There is no foolproof way to know the answer to these questions.
Linguistic Fieldwork Daniel L. Everett
page 230
So what can the fieldworker do? Well, one thing is to get the speaker to confirm
his or her broad assessment (e.g. "Yeah, that is OK.") by uttering the sentence
themselves. Than the linguist can repeat it slowly and ask for yet another repetition. In
my experience, the native speaker will not repeatedly utter an ungrammatical sentence.
They will 'edit' it, i.e. change it slightly to make it grammatical. Or they will refuse to
say it. (If the utterance is grammatical there often will be no problem for the speaker to
repeat it.) The kind of correction made should identify what was wrong with the
example structurally. On the other hand, if the example is pragmatically bad (roughly,
if it is inappropriate in the environment of the lab session), it is quite likely that the
speaker will still not say it. In this case, the linguist cannot tease apart the
ungrammatical from the inappropriate.
One way around this latter problem is to work with multiple language teachers
simultaneously when checking for grammaticality. Recall from the earlier example
from Wari' that I only discovered that a particular kind of construction was possible by
working with a group of speakers who discussed the example among themselves.
Usually language teachers attempt to be helpful and when several are working together
they tend not to accept snap judgments by one of their group. They discuss and refine.
Therefore, checking grammaticality judgments with a group of speakers has benefits.
On the other hand, none of this guarantees the field linguist that they know exactly,
precisely why a given utterance is rejected. Moreover, even utterances that are
acceptable could be ungrammatical, as we have seen. So what is a poor linguist to do?
They must carry out grammaticality tests with multiple speakers (serial language
teachers and groups, see ___ above). Where the results of these various interviews
produce a unanimous judgment, one way or the other, the linguist can have moderate
confidence in the result. But it is still necessary to test and interview for truth
conditions (see ___ above), for naturalness (in speakers' opinions), look for related
examples in natural texts, etc. But in fact it is likely that the results of all these
interviews will not produce a unanimous judgment but, rather, a mixed judgment.
Speakers are quite likely to disagree, especially when they are not in the room together
as they offer their judgments. In this case, the linguist could go with his or her best
judgment (using their theoretical predictions for example), advising the reader of their
grammar or article that this is what they have done and why. (Always, always be honest
with your reader.) Or a more scientifically respectable approach can be adopted and the
linguist can employ a statistical approach to grammaticality, following suggestions in
Cowart (). After all, other social scientists deal with conflicting judgments and accounts
of behavior all the time. Their reports, therefore, present statistical analysis of these
different judgments, etc. and distinguish statistically significant vs. insignificant results.
But using statistics and standard social science methodology requires more and
different training of linguists in general and fieldworkers in particular.
The upshot is that determining grammaticality is never a straightforward matter.
Judgments, testing, statistical analysis, text tracking, eavesdropping, etc. are all crucial.
As a consumer of grammars, the linguist reader must always take all crucial
grammaticality judgments with a huge grain of salt, in the absence of the corroborating
studies just mentioned. All grammars ever done, past, present, or future, need retesting
and can never be understood as a 'God's eye view' of the language in question. Once
again, grammars are stories we tell about a language as we see it.
Let us turn now to the final question concerning the corpus for writing a grammar,
namely, 'When will I have enough data?' Samarin (1967, ---) gives a number of useful
Linguistic Fieldwork Daniel L. Everett
page 231
suggestions in this regard. I have borrowed some of Samarin's suggestions and added
some of my own to come up with the list in ():

(10.5) A complete corpus is obtained when:

a. All the closed classes of linguistic elements are fully accounted for
b. When there are no 'holes' yet in the data needed for analysis (partially,
therefore, the answer depends on theory)
c. When there are multiple tokens of all types.
d. When it is maximally useful for other disciplines, as well as linguistics
e. All new material collected only contains structures and meanings already
found in the corpus collected.

Let's consider each of these points in more detail. First, what does it mean to say
that all the closed classes are fully accounted for? This simply means that when you
have all the prepositions, all of the adjectives, all of the verbs, i.e. classes with a small
number of members that do not expand in membership. In some languages verbs will
be in the open classes of lexical items (e.g. English), while in others they will be among
the closed classes (e.g. Mosetén, Sakel ()). How do you next determine that there are no
'holes' in the data? Well, you have to have a view of how language works, partially
based on general principles shared by most linguists and partially based on the
particular theory that you are most influenced by. And you must be able to argue for
your conclusions. On the latter, see ___.
Table –
The phonetic segments of hypothetica

p t k
b d
m n N

There is a missing segment in Table -, i.e. a voiced velar, [g]. Is this an accidental
gap or an actual asymmetry in the segmental inventory? The linguist will need to look
for examples of [g]. At some point, they might conclude that the system is indeed
asymmetrical, certainly not all that uncommon. But until they can say with confidence
that this is the case, the corpus is incomplete.
It is also important to ensure that for every segment, prosodic pattern, syntactic
construction, suffix, etc. in the language, that the corpus includes multiple tokens of
each. And the linguist's analysis must be the guide as to when there are enough tokens
of each. One useful criterion in answering this question is 'Are all tokens I am now
recording simply repeating the patterns that I already have?' If so, then there are
probably sufficient tokens in the corpus. However, one cannot simply rely on texts to
magically produce all the tokens and their distributions that are necessary for a
complete corpus. The fieldworker must think, based on his or her analysis and ask
questions like the following: 'If my analysis is correct, then there will be forms of
interpretation/shape x but never forms of interpretation/shape y.' Then the linguist must
look for the missing forms, both those they predict to be missing (no matter how long
they search) and those which they predict to be found eventually, but which are
currently absent (i.e. accidentally) from the corpus. The linguist must be able to assure
the readers of the grammar that the corpus is complete by this metric.
Linguistic Fieldwork Daniel L. Everett
page 232
The corpus should also be maximally useful for other disciplines. The fieldworker
may be working on a rare language that few people are likely to have access to. In this
case especially, but in all cases ultimately, the linguist should collect texts and data
relevant to other disciplines insofar as they have time and knowledge to do so. Text
collections should include all important cultural values, to the degree that the
community is willing to allow access to these. Claims about numerals or counting
should be accompanied by experimental evidence corroborating the claims (even if this
means bringing in an expert consultant). And so on. Finally, once all new data appears
to contain no new structures, etc. then the linguist can consider that, with respect to his
or her current working hypotheses and purposes, the corpus is complete. But, as we
have been saying, the 'complete corpus' is a relative, never an absolute concept.
We move now to another core component of language documentation, the
dictionary.

10.2. Dictionaries
The traditional view of the dictionary in linguistic theory up until twenty-five
years ago, and still widespread, is that the dictionary is an asylum for the misbehaved,
i.e. where we put forms that are not derivable by regular rules of syntax or phonology.
People who work under this view may be tempted to produce trivial dictionaries that
are little more than lists of words, idioms, and morphemes. But this would be a mistake,
even for those with the 'asylum' (or 'jail' – see Williams and DiSciullo ()) view of the
dictionary, because it renders the dictionary less useful. A dictionary is formed by a
view of its potential users, not merely by a particular theoretical perspective.
In their volume, Making Dictionaries: Preserving Indigenous Languages of the
Americas, the editors address the purpose of making a dictionary:

"A reasonable person might ... ask, Why do it? One way to read the
contributions to this volume is as personal answers to this question. But a more
general response can be discerned in all the chapters and, indeed, in the work
of every lexicographer. There is something at once both marvelous and
practical about producing a guide to the mind, world, and behavior of a group
of people. The benefits that accrue from such a handbook – literacy,
preservation, history, discovery – only add to the excitement of seeing the
published dictionary standing upright on the shelf." (Frawley, Hill, and Munro
(2002, 2-3)).

The editors go on to suggest that the ten most important issues in compiling a
dictionary, with immediate application to languages of the Americas, but with a clear
pan-geographic relevance are as follows:
Linguistic Fieldwork Daniel L. Everett
page 233
(10.6) Ten crucial issues in dictionary compilation:
a. Entries
b. Theory
c. Literacy
d. Graphics
e. Role of the community
f. Types and numbers of dictionaries
g. Historical information
h. Technology
i. Presence or absence of dictionary-making tradition
j. Handling exceptions

Let's consider these in turn:

Entries: how is the headword (the objects of definitions, possibly with subentries)
determined? Should the headword be a 'basic form' (e.g. citation form)? How can the
fieldworker decide on a basic form?
According to the editors, ultimately the choice of headword for a dictionary will
result from a "... tradeoff between the pressures for maximal explicitness and the desire
to match the users' minds to facilitate their inferences as they fill in what must be left
implicit." And further, "In the end, entries are a wager that the tension between the way
the dictionary ought to look to the compiler and the way it feels to the user will not be
too great." (Frawley, Hill, and Munro (2002, 5)).

Theory: How much information in each entry should be there for theoretical vs.
applied reasons? How much should linguistic theory affect the overall form of the
dictionary? Each entry should contain as a minimum sufficient phonetic, phonological,
morphological, syntactic, semantic, and pragmatic information for the reader to know
how to use and pronounce the entry and where it fits in the grammar and culture of the
language. Some linguists build large amounts of additional theory into entries, while
others see the dictionary more as a service to the language community and prioritize its
utility rather than writing a lexicographic treatise on each entry. I favor the user-
friendly view, though it is conceivable that linguist-only dictionaries can be done, in
addition, if the linguist has inclination and time to do so.

Literacy: The dictionary may be the first or one of the first documents ever produced
in the language under study. In this case its impact on literacy and discussions of
representations of the language will be massive. But regardless of when the dictionary
appears in the literary history of the the people, it will be or can be an important part of
their self-identification and 'represents' their language to themselves and to people
outside the community, the latter especially if the dictionary is bilingual (see ___
below).
An issue that arises in this regard is the extent to which the national or other
major language(s) should influence the orthographic representation of the vernacular.
For example, in Romance languages, as in many other languages, vestiges of history
are included in the national orthography. So, consider the words in () from Brazilian
Portuguese:
Linguistic Fieldwork Daniel L. Everett
page 234
(10.7) a. casa 'house'
b. cicatriz 'scar'

As is well-known, in Romance languages, the old palatalization rule from Latin,

whereby the voiceless velar occlusive and the voiceless coronal affricate are in
complementary distribution such that the latter occurs before high front vowels and the
latter elsewhere, continues to be encoded in the orthography by using a single letter, 'c',
for both. (Also imported to some degree into the representation for loan words in
English, e.g. 'electric' vs. 'electricity'.) So in some cases 'c' is [ts], in others it is [k]. And
to complicate matters further, [k] is represented in still other cases, i.e. to represent the
voiceless velar before a high vowel, by [qu], as in ():

(10.8) a. quinze 'fifteen'

This representational inefficiency is OK for a language with a long history of

literacy and in which it is apparently desirable to continue to maintain some of the
history of the language (and avoid 'imperialist' symbols, e.g. 'k'!). But should it be
imported into a language that does not share its history, a minority indigenous
language, say? Should the language of field study be obliged to share the inefficiencies
of the national language? The questions become more radical and urgent when
considering whether to use national scripts, such as Chinese, Devanagari, etc. or
Western symbols. These are just some of the reasons why the dictionary is at once an
important milestone in the literacy of a community and why it must be subject to
community discussion and approval. The linguist cannot simply present a dictionary as
a 'gift' to the community, but must develop it in conjunction with them.

Graphics: Will your dictionary include illustrations, photos, different colors, etc?
These will add to its expense (in print media in particular), but funding for well-
planned dictionaries is readily available from a number of sources, so the expense
should not be allowed to become overly discouraging. Well-designed and utilized
graphics can enhance the dictionary's usefulness as a tool for literacy since beginning
readers can use different graphics to figure out on their own the meanings and form of
certain words. Of course, graphics can get out of hand and create complications out of
all proportion to their benefits. So once again the use of graphics has to be constrained
via discussions between the linguist and the speech community.
Expense of graphics can be tremendously reduced of course if the dictionary is
electronic, rather than (or complementary to) print media. See __ below for more on
this.

Role of Community: This is discussed throughout this section.

Number of dictionaries: By and large the average documentation project could be

rightly proud of producing a single high-quality dictionary. But other types of
dictionary are possible, e.g. dictionaries of place names, flora and fauna, thesauri,
cultural concepts and artifacts, etc. More specialized additional dictionaries will not be
something the average fieldworker expects to produce, though communities could
change the linguist's mind and plans by requesting these. In such a case, the linguist
needs to determine what the community wants in terms of specialized dictionaries and
figure out how to produce it.
Linguistic Fieldwork Daniel L. Everett
page 235

History: How many dictionaries and how much linguistic and, particularly,
lexicographic work has been done? To what extent should dictionaries track changes in
lexical meaning? How much, if any, space should be dedicated to etymologies in the
dictionary? The answers to such questions will depend partially on the linguist's
training and interests, partially on community wishes, and partially on the view of the
readership most likely to use the dictionary outside of the community. It is obvious that
the more historical information the better. This enriches the dictionary and its role in
the cultural heritage of the people. However, at the same time, it is nontrivially
complex to provide etymologies and such an endeavor can get out of hand rapidly in
the hands of someone with little training in etymology.

Technology: There are a number of advantages to constructing an electronic dictionary,

whether as a complement to a print dictionary (recommended) or as the sole dictionary
(not recommended). Just a few of the things that an electronic dictionary can do better
and more cheaply than a print dictionary are: better graphics, audio files (providing
pronunciations of each entry and many examples), easier dissemination, easier
connections and cross-references from entries to the text or portion of field data from
which they come, much greater and easier searchability, permanence (if done properly,
an electronic dictionary can last forever), accessibility (more people have access to the
dictionary), lower cost, easier to look for and predict trends in semantic change of
lexical entries, and much greater interactivity – changes, corrections, and updating the
dictionary become much easier when it is in electronic form. Anyone who has
consulted an on-line dictionary source, e.g. one finds on internet sites such as
Yourdictionary.com (http://www.yourdictionary.com/) knows that being able to listen
to the pronunciation of entries, being able to find entries almost instantly, etc. are
tremendous advantages of electronic media dictionaries.

Ultimately, however, the crucial thing to remember once again is akin to Postal's
Maxims – the linguist and the community must see dictionary making as a fluid
process, subject to negotiation, constrained only by imagination and budget.
We conclude the chapter now with a discussion of the nature and importance of
text collections.

10.3. Text Collections

What is the purpose of collecting texts? Well, for some linguists texts should be
the core component of the corpus upon which analysis and the grammar are based. To
some degree, I agree with this view, although as has been pointed out numerous times
in this and other chapters, texts are but one type of data that is essential for a good
corpus. They are important because they are revealing with respect to the interaction of
language and culture and they provide natural data, often providing examples and
structures that the linguist would never have discovered on his or her own. It ought not
to be difficult for any fieldworker to understand why text collections are important to
the enterprise of linguistics in general and fieldwork in particular. Accepting the
importance of texts, then we come to questions about the nature of textual
documentation. First, what kinds of texts should be collected? Second how many texts
are needed? As to the kinds of texts, my suggestions are in ():

(10.9) Kinds of texts needed:

Linguistic Fieldwork Daniel L. Everett
page 236
a. All genres (narrative, procedural, hortatory, expository, and any other type
that the fieldworker identifies as relevant for the language).
b. Texts that cover all significant cultural topics, e.g. life, death, harvest,
hunting, dealing with the outside world, creation, fiction, history, etc.

Different genres will reveal different kinds of linguistic information. Different

moods, aspects, tenses, participant coding and tracking, different kinds of logical
connectives, etc. all are associated to a greater or lesser degree with different text
genres. And anthropologists, ethnographers of communication, and linguists will all
benefit from a rich array of text topics, linked to the culture (which implies that the
linguist must understand the culture more than superficially, in order to know which
topics to ask for texts on).
Next how many texts does the fieldworker need to collect? This gets us back to
the issue of the 'complete corpus' and is answered by the considerations given in the
previous section. However, as a rule of thumb, I would suggest at least two hundred
pages of transcribed texts with morpheme by morpheme glosses, single spaced, no
greater than 11 point font.
Linguistic Fieldwork Daniel L. Everett
page 237

Chapter 11: Modeling and Managing Field Research

11.1. Individual Fieldwork
As we consider some of the issues in individual field work, it is useful to ask what
it means to work alone in the field or, indeed, even if it is possible to work alone. To
work completely alone is a fiction; it is not possible. If it were, this would imply, for
example, that there is no speech community you are working on – you are obviously
not working alone if you are learning from another person (just as there is no such thing
logically as a 'self-help' book. If the author of this book helps you, it is not 'self-help').58
As Mithun puts it, 'The product of fieldwork will ultimately be shaped not only by the
nature of the language, but also the methodologies chosen, by the roles assumed by the
speakers, and by the preparation and sensitivity of the linguist.' (Mithun, p34, in
Newman and Ratcliff). As experienced fieldworkers know, and as Mithun goes on to
point out in her useful discussion, the linguistic record is a product of the collaboration
between linguists and speakers. This means in practice and in theory that there is no
way to really work alone. But let's consider this a bit more, because there is some
intuitive appeal to the idea that some people are loners, while others work in groups.
If you go to the field with your partner or family, then you clearly are not going
alone. On the other hand, if you travel into the community without any companion from
your own culture, background, etc. can't you say that you are alone in the sense that
there is no one to help you with your work? No, once again, because the language
teachers are in fact your partners in the crucial senses (as Mithun aptly states it above).
And you have, if you follow this guide, books with you that will help you, so you are
still not working 'alone' in the strictest sense.
So let's try to be clear: field linguists do not work alone, not ever. In his excellent
book, The Art of Fieldwork, the anthropologist Harry Walcott (2001, 37) puts it this
way:

'A moment's reflection reveals how extensive are the strands that link, and
important ways bind, seemingly lone and independent researchers, in the field
or at their desks, to larger and more embracing social systems.'

As I write this book, for example, I have potential readers, students, and reviewers
looking over my shoulder. Still, though, if a fieldworker can be alone professionally, if
they are the only linguist around. In this latter, special sense, they are alone. Working
alone in this sense of 'individual' field research has both advantages and disadvantages.
These are not frequently considered, however, because the 'individual' model is the
default, usually unquestioned way to do fieldwork for most linguists. In any case, let's
consider the pros and cons of working alone in this sense, beginning with the
advantages.

Advantages of the 'Individual' model

Less Pressure and more flexibility
When you work apart from other linguists, you can experience less pressure,
depending on your individual psyche, by not having people 'looking over your

58
This is another bit filched from George Carlin's HBO special at the Beacon Theatre
in about 2000.
Linguistic Fieldwork Daniel L. Everett
page 238
shoulder'. When you work alone you are not immediately accountable to anyone else
for how you spend your time. Your language teacher will put time constraints on you to
some degree, but nothing like members of your own culture would if they were there to
work with you towards a common research objective. Nobody in the field, when you
are there without a linguistic research team, is going to watch you to see if what you are
doing daily conforms to official project objectives. This is simultaneously an advantage
and a disadvantage. It is a disadvantage because it removes the field researcher from
some potentially useful and immediate quality control. However, it is an advantage in
living in the field because it reduces strain, pressure, and potential conflict. If you get
up in the morning, you can in principle change your schedule. Let's say you were
planning to analyse the previous day's data and you have no lab sessions planned this
day. But now that you think about it, it would be nice to spend the day baking cookies,
going for a walk in the jungle, paddling a canoe, or reading the trashy crime novel you
brought with you. If there are other linguists working with you in the field, you will
probably feel constrained by them, to make the day useful. But psychological well-
being in the field is often based on feeling like you are in control of your time (another
illusion, but a helpful one). I do not like routine, to be frank. I do not like feeling that I
need to do something because of others' expectations. When I am the only linguist
around it is easier for me to give time to myself than when I have colleagues with me,
especially if I am the senior colleague and the one the others are looking to for
guidance and example.
Also, there is less temptation when you are alone to sit around and talk with
friends from your culture about topics unrelated to the research. This kind of
camaraderie can become a detriment to research by siphoning time and by making the
linguist less dependent on the culture and language under study (see chapter ___ below
for why this dependency is crucial).

More learning (since you have to do more on your own)

I hate mechanical things when they break down. And in the field they break down
a lot. For my field research among the Pirahãs, I needed a motorboat, provided for me
by the National Science Foundation. One morning, after a heavy rain, I woke up to the
smell of gasoline. I went to the river, over thirty meters deep in the peak of the rainy
season, which this was, and looked. There, about six meters down, held only by a very
taut nylon rope, was my motorboat, my only transportation to the road (the
Transamazon) where my car was stored, 100 miles upriver. One night of hard rain had
sunk it. With the help of many Pirahãs and a Brazilian missionary, I was able to raise
the boat, just until the edges were about one inch above water. Pirahã women came
running with gourds and bailed the water out. But even though the boat was up, I had
water mixed with my gasoline and the cylinders of my boat motor were full of water. I
had to fix the motor and get the water out of the gasoline and cylinders on my own, or I
would not get out of there. I did. And in the process I learned more about motors.
As a linguistic example of forced learning when working alone, consider my
attempt to discover comparative structures in Pirahã. First, I gathered a number of
sticks and leaves of varying sizes. Then working first with the sticks I got the words for
'long' and 'short' (all work with the Pirahãs is monolingual, as described in Everett
(2001) and chapter ___ below). Then I would hold up a short stick and a shorter stick
from my carefully size-arranged array in front of the language teacher. Then I would
hold up an even shorter stick and try to get 'shorter/more short', the comparative form.
What I got were answers like the following:
Linguistic Fieldwork Daniel L. Everett
page 239

(11.1) a. /ií tio/óíhii

woodshort
b. /ií pii/isái
woodlong
c. /ií tio/óíhii /iahá
woodshort go

I had to conclude, after hours and hours of this, that either (i) Pirahã had no
comparatives or that (ii) it used /iahá 'go' to indicate that one stick was particularly
short, though with no specialized morphosyntactic comparative marker, or (iii) I was
incompetent (and the latter feeling rarely goes away entirely). Before this experience
(and countless others like it), I thought every language had certain structures, e.g.
comparatives. Since then, I have read more, but as a graduate student my ignorance
was, perhaps, even greater than it is now. So I had to struggle with this and, in the
struggle, learn more than I might have otherwise have learned had others been with me.
A partner in the field might simply have told me that some lack comparative structures.
But finding this out on my own made the learning more effective.
It can also be an advantage to work alone because this simplifies interpersonal
and intercultural interactions between linguist and speakers. The more outsiders that are
present at one time in a community, the more complex the social dynamics. Working
alone avoids many of these complexities (see Chapter ___ for a discussion of some of
the ethical issues that can arise).

Few barriers between the researcher and the speech community

The more people present with the linguist who do not belong to the speech
community, whether government officials, fellow linguists or friends – i.e. anyone
except family (see ___ below on how children and family can be of tremendous help in
field research), the more insulated from the target language culture the fieldworker
becomes. And insulation slows, impedes, and halts learning. If there are multiple
linguists present in the community at a time, for example, the people can feel less
comfortable approaching them, especially if they stay together in the same house or
location, for example, reducing chance for observing and engaging in spontaneous
conversations, etc.59 But a 'lone' field researcher will have a more direct and constant
connection to the community and, an extremely important correlate, will need members
of the community socially more if they have no colleage or partner or family member
along.

Getting more credit (or more blame)

The final advantages of working alone that I will mention have to do with money
and fame. The more people working on a project, the more money needed and the less
control over the project's funds will the fieldworker exercise. And the wider the credit
will have to be shared for discoveries and the work of the project. So if you work alone,
you have the advantage of keeping the money and fame to yourself. OK. Those are
some of the advantages of working alone. Let's now consider a disadvantage of
59
This is only true, of course, if the entire team goes to the community at the same
time, which is not at all a necessary correlation of team research.
Linguistic Fieldwork Daniel L. Everett
page 240
working alone. We will then see how team work need not have the disadvantages
associated with it above and can have the advantage of multiple individual field
research.

A disadvantage
My favourite quote on fieldwork is the following from Margaret Mead, the
distinguished American anthropologist on her first field trip. It is found in a letter to her
thesis supervisor, Franz Boas:

'I have no idea whether I am doing the right thing or not, or how valuable my
results will be. It all weighs rather heavily on my mind.' (January 16, 1926)

This is honesty. And just about anyone who has done fieldwork alone has felt
somewhat like this at some point. But notice how this is expressed – in a letter to a
colleague thousands of miles away. Alone in Samoa, Mead had to grapple with the
responsibility of her (thesis) research without local academic interlocutors or any local
anthropologist to share her field responsibility with her. The advantage of this for the
researcher is that it makes them think harder and learn more than they might otherwise
have learned. But the disadvantage is obvious – working alone can at times nearly
overwhelm the field researcher with self-doubt and emotional pressure. I will take up
other disadvantages when we move to the section on team research.
Before closing this section, I want to urge prospective field researcher to read
others' accounts of their own field research on the individual model. Such exercises
remind us of our commonalities, that no one's experience, however good or bad, is truly
unique. In general ethnographies contain accounts of the ethnologist's own field
experience. The lessons our predecessors have learned and their very experiences can
teach important lessons before going to the field and save hours of wasted effort, as
well as providing an emotional vaccine against some of the hardships of different field
experiences. Here is a list of some of my favorites, though this is only the tip of the
iceberg: Dixon (); Chagnon (); Macaulay (); Conklin (); Malinowski (); Mead ();
Darnell (); and Darnell ().
These all tell stories of independent, resilient, and resourceful fieldworkers who
entered 'alien' environments alone and thrived as scientists and as human beings. They
discuss their research and, often, explicitly discuss their personal challenges as lone
field researchers. I recommend that all fieldworkers read at least the chapter by
Macaulay before going to the field, especially women.

11.2. Team fieldwork

Fieldwork has long been considered the domain of the individualist, the strong
personality who is self-motivating, able to withstand loneliness, well-rounded, etc. A
sort of Indiana Jones. That is, it is not normally thought of as an environment for
teamwork (at least not with members of one's own culture), even though, as I have
shown above, this is a fallacious concept. What I want to do in this section is to argue
that a research team has a better chance of success than a lone researcher.

Composition of the team

The team is composed of at least the speech community, the professional linguist
or graduate student who is responsible for the written form of the results to a host
institution or funding agency, the program director of any relevant funding agency, the
Linguistic Fieldwork Daniel L. Everett
page 241
government officials or local academic representatives who must monitor the research
project for local government agencies, academic officers at the linguist's home
institution who must authorize his or her absence and activies, as well as his or her host
insitution's ethics committee (and any other ethics committee, e.g. funding agency,
government of host country, etc.), consultants, PhD students, co-Principal Investigators,
secretarial staff, and postdoctoral research associates, depending on each individual
research situation. When the team contains more than one designated fieldworker,
management becomes even more crucial.

Advantages
The advantages of team research are best framed in terms of the tasks the linguist
takes on in the field. One of the most important tasks facing linguistics today (though it
has always been important in spite of recent more general awareness) is the
documentation and description of endangered languages. At one time 'documentation'
was seen as little more than collecting word lists. Early explorers, naturalists, and
missionaries, among others, often seem to have collected almost random lists of
words.60 But to document and describe a language in the modern sense draws on a
number of specialized skills that may go beyond those possessed by any one linguist.

More help, more specializations present

When you work in a team, there is more knowledge available to you. If your team
is balanced in terms of specialities and experiences, you could have solid knowledge of
most of the field of linguistics with just, say, three people, along with different kinds of
practical and information-processing knowledge. For example, if the team leader is a
phonologist, working alone they would struggle a bit more with syntax analysis,
perhaps, than if they had a syntactician on the team. Likewise for other subdisciplines.
Ever since tape recorders began to make their impact on field research, we have
come to expect more from documentation than mere word lists. Today, the
documentation of a language requires a wide variety of skills. Photography,
videography, sound recording and storage, web design, computer programming, and
other specialized knowledges may all be called upon in documenting a language. Even
if one person had all the requisite skills, to use them all in a particular project would
take so long that it is unlikely that a full (e.g. dictionary, text collection, and grammar –
see ___ below) project would be completed. But people do not commonly have all the
skills and training necessary to fully document and describe the parts of a language
considered crucial in a documentation project. And, I say this from raw experience,
people change as they age. I no longer feel inclined to learn agricultural vocabulary, for
example, by working all day chopping trees in the tropical sun. But I do know how to
collect and analyze data better than ever before in my career. And there could be
graduate students on my team who would jump at the chance to swing an axe among
flies and mosquitoes in 99% humidity, in exchange for an array of chopping/cutting
verb paradigms and plant (and insect) names.
These are advantages which are difficult to overcome for any one working as an
individual linguist in the field.

60
In spite of their obvious shortcomings, these early efforts have been much-
appreciated by posterity because they are often the only record left of lost languages. In
many cases these lists provide bases for historical reconstruction and understanding of
language change. One ought not to look a gift horse in the mouth after all.
Linguistic Fieldwork Daniel L. Everett
page 242

More work can be done (in theory)

Just as there are more specialities available within a team than to an individual
worker, a team can do more work. The task of, say, writing a grammar, can be
distributed and each member of the team can work on a section of the grammar. To be
sure, everyone should read everyone else's work on the grammar and comment on it,
but there is no way that a single individual can do the same amount of work in the same
amount of time as a well-organized, well-selected team. One member can take, say,
primary responsibility for a dictionary, another for a pedagogical grammar, another for
the reference grammar. There are many ways the task could be divided so as to increase
both its scope and quality.

Replicability
Another important advantage of team research is that it can help make results
more replicable for non-team members, as well as facilitating replication of results
prior to publication. This is so because working as a team enables all results,
methodologies, and analytical decisions to be subjected to discussion, criticism, testing,
refinement, and development by the group as a whole. If the research team is well-
managed, with each team member contributing a needed speciality, then each member
will teach the other members, challenge the other members, and strengthen the project
as a whole, beyond what any single field research could accomplish working alone.
The team concept is hardly new, of course. Linguistics, like other sciences, has
always benefited from cooperation and group-based research, whether in small local
groups such as particular labs, or in international theoretical ventures, e.g. chomskyan
linguistics and functionalism. But in field linguistics the common assumption has been
that research will be conducted by a solitary field researcher 'figuring out' how a
language works on their own. If I am correct, however, it is important to challenge this
assumption.
For the sake of completeness, however, let us briefly review the disadvantages of
team fieldwork and how to counter them.

Disadvantages
The advantages listed for the individual model earlier would be disadvantages for
the team model if everyone on the team went to the same community location at the
same time. For this reason, I recommend instead that members of the team with a
responsibility for team work only go to the field as individuals, trying to cover the
entire year in the village, at least in the first year of a multiyear project. (Another reason
for this is that many aspects of culture and language are seasonal and only by being
there for the entire year can the team hope to observe and learn these seasonal aspects.)
The principal disadvantage of the team model is that it entails group dynamics.
People might not get along, they might be different in practice than they are on their
CV, they might turn out to do terribly in the field situation, they might decide that they
do not like you, they might complain about their salaries, they might not want to follow
the team objectives after all, etc. There is no way around the issue of 'group dynamics',
which is why I have added the following section on management. Management is vital
in any project, team or individual, but it becomes especially critical in a team situation.
The principal ways of overcoming the disadvantages of team research are selecting the
right team members and managing the team well. Whether you are a team
manager/Principal Investigator on a grant or a team member (postdoctoral research
Linguistic Fieldwork Daniel L. Everett
page 243
associate, PhD student, or language teacher) solid management principles can help you
contribute to building and understanding what makes a successful team for field
research.

11.3. MANAGEMENT
I want to review a few principle management principles in directing and working
together in teams, though many of them apply to individual workers as well. Not
everyone who is part of a team will perceive themselves as directly concerned with its
management, but all members of a team should be familiar with some of the issues
involved in getting the best out of team work.

No algorithm for success

First, it is important to realize that in team fieldwork, just as in individual
fieldwork, there is no algorithm for success. Good judgment is always more imporant
than lists of priorities and principles. What follows are points that have been important
to me. Since this is not a management manual, it is likely that I am omitting principles
that business administrators would consider vital. But this reflects my experience as a
fieldworker.

Mission statement
The team must have a clear statement of what their objectives are, some might
refer to this as a 'mission' statement or goal statement, or some such. There should be
no ambiguity as to what you as a team are trying to accomplish. If you have a funded
project, the written project can offer guidance, of course, but it is not sufficiently
succint to serve as a mission statement. It is very important that you be able to
summarize in a sentence or two what your entire project is about and what it hopes to
contribute to the world. The entire team, secretaries, PhD students, postdoctoral
research associates, and co-Principal Investigators should all know and agree to this.

Criteria for success

All members of the team also need to know and agree on what the criteria for
success are. How will the team know that it each individual and the team as a whole has
finished its job and finished it successfully? These criteria must be explicit, specific,
and, above all, measurable.'Finishing a dictionary', for example, is not a well-stated
criterion for success. 'Publishing a 3000 entry, tri-lingual dictionary, with publisher __
is', however. The criteria can and should be refined and reassessed as research goes
along, in discussion involving all team members.

Expectations
It is important that expectations for all team members be reasonable. If, say, the
team leader is an experienced researcher and regularly works at a particular rate of
accomplishment, clearly measurable in terms of project goals, it must be recognized
that other team members with less experience may not work as fast or as well, initially
at least. Expectations must therefore be negotiated at the outset between the team
leader and team members, on an individual basis, and then evaluated and reset as
necessary as time goes on. The team leader should be very hands on in the beginning of
a research project to ensure that all members are using their time well, working with a
clear sense of mission, a commitment to the to success of the project, and that they are
Linguistic Fieldwork Daniel L. Everett
page 244
able to meet the expectations that the team has of them. The project must be seen as
belonging to the entire team, not merely the Principal Investigator.

Records
Teams must keep records of team administration as well as team intellectual
output. Therefore, a procedure of record-keeping and filing is essential to the well-
being of the team. Some types of records that need to be kept are given below.
DATA: How are you going to record the data that you collect? Are you going to
use proprietary software, e.g. Shoebox (see
http://www.ethnologue.com/tools_docs/shoebox.asp), or software developed to serve as
a long-term record, universally usable, e.g. Elan (http://www.mpi.nl/tools/elan.html)?
How will you store backups of data? On a server, on DVDs, or some other means?
These decisions will no doubt be required from the early days of writing your proposal
for funding, since most funding agencies these days are very careful about data-storage
and data-ownership. Most field researchers these days will be required to turn over
copies of all of their data in an XML-compatible format. (For further information, the
reader is urged to consult, as an entry point into the general area, the website of
EMELD (http://emeld.org/), a project developed in connection with the LinguistList
(http://linguistlist.org/)). There are many linguists working in various places around the
world to maintain long-term storage facilities and standards for the preservation of
linguistic data. Before going to the field, all researchers should be familiar with the
basic standards for data-storage, an issue discussed briefly in Chapter ___ below.
WORK MEETINGS: Team members should meet to discuss the project face-to-
face on a regular basis. meetings can be less frequent perhaps as the project goes on and
people know each other better, but they should not stop. Regular contact, discussion,
and assessment must be a vital part of the project from start to finish. Minutes must be
kept of all these meetings, in the language of the funding agency, so that if any
problems develope or if conflicts arise, it may be determined whether such issues were
foreseen, previously discussed, etc. Was any action taken to avert such consequences?
Why not? If your project has a secretary, they can be appointed to keep records of
regularly scheduled project meetings. Otherwise appoint someone at each meeting to
keep the minutes and to distribute them for approval and then filing within forty-eight
hours of each meeting.
CORRESPONDENCE: All project correspondence must also be stored in an easily
accessible manner. Correspondence with the speech community, between team
members, with the funding agency, with government officials, etc. must all be carefully
stored. If it is email, then it can be stored as electronic media using the filing system on
your mail program, though I also recommend that hard copies be kept in an old-
fashioned drawer filing cabinet.
FINANCES: Your project will also generate financial records which must be
stored. If you are ever audited (an unpleasant experience), you must have careful
records of project expenditures. In most cases, your home institution will keep records
of financial transactions. But the Principal Investigator, and team members to a less
degree, ultimately bears responsibility for all financial decisions and use of funds in the
project, so I suggest that the project office keep duplicate records of all transactions.
AGREEMENTS: Everyone on the project, as well as the relevant members of your
home institution's personnel department should receive copies of the agreed-upon
structure for project accountability and administration. This is important. Even the most
reliable team members can forget or selectively remember what the agreements are in
Linguistic Fieldwork Daniel L. Everett
page 245
this regard. It is vital that everyone be clear on how the project is administered and who
is accountable to whom. And this must all be written down and read and approved by
all. The originals of these agreements should be filed in the project office (which may
simply be the Principal Investigator's university office). Documents on accountability
and administration should also include a record of time commitments, job descriptions
for the overall project, job descriptions for each individual field trip, and any other
expectation of individual team members that could affect the overall results of the
project. It should also be made clear in these administrative documents what the
publication and presentation guidelines and constraints are for the project. For example,
how many papers should the project produce per year according to its funding
commitments? Who should write the bulk of each paper? Does everyone know that the
Principal Investigator (and this is very important!) has veto power and editorial control
over every project publication or presentation? These matters must be spelled out
clearly in advance and a copy of the agreements filed in the project office and with each
individual team member.
SUCCESS CULTURE: Aside from filing and documents, there are even more
important considerations for team management. First, a team is responsible for creating
a 'culture of success'. What does this mean? It means that each member on the team is
committed to the success of the project, that she knows what the criteria for success are,
that she feels that her contribution will be a success – because she feels well-trained,
well-equipped, well-guided, and well-supported for this success – and that the
contributions of the other team members will also be successful. People should enjoy
the challenges of the project and be excited about the likely scientific and career
benefits of successfully concluding the project.
PROJECT CALENDAR: All research projects must have a (somewhat but not
completely fluid) project calendar. A project calendar should lay out the responsibilities
of the entire team, the timing of its deliverables for the funding agency and the
language community and a detailed breakdown of each team member's responsibilities
in the field and out of the field. Ideally, each team member should provide the team
with a breakdown of how she plans to spend each week in the field, giving the amount
of time estimated for each task. This is done by breaking the principal task(s) assigned
for each field trip into subtasks. These can then be translated into daily, weekly, and
monthly 'work quotas', based on project duration. So, for example, if it is my
responsibility to analyse the segmental phonology, then how much time should I give to
that task (I suggest one month of regular fieldwork), what are the subtasks (e.g.
consonants, vowels, syllable structure), how much time should I give to each? And so
on. It is important, however, that you realize that you will be regularly interrupted in
the field, by visitors, by the community, by your health, by many unforseeable factors.
Use every minute. Field time is precious. Don't waste it. Therefore, I recommend that
you double your estimates for all project tasks and schedule your time accordingly. For
example, if you believe that you can collect five new words per day, learn them, and
analyze them, then you need to shoot for ten words per day or double the amount of
days that you have dedicated to this task, such will be the unforseeable variables that
affect your time.
TRUST & RESPECT: Nothing can kill a project, ruin its chances for success, and
embitter all participants more easily or more permanently than a lack of trust or respect
by team members for one another. The project participants must trust one another. The
PI must have the trust of all subordinate members of the project. And, perhaps even
more importantly, the PI must give all project members trust and respect. Do not try to
Linguistic Fieldwork Daniel L. Everett
page 246
micromanage their assigned responsibilities. If there is evidence that someone is not
fulfilling their responsibilities, this will emerge in the regularly scheduled reports. But
micromanagement can be/is a symptom of distrust and paternalism (or maternalism).
Everyone should be trusted as a highly-qualified professional. If appointments to the
project are made carefully, this will be true.
FINANCIAL MANAGEMENT: Research projects also require financial management.
If an organization gives you money to complete a research task then you must have a
process in place for managing the income. In general, your home institution will have
procedures already in place. If you follow these procedures, your home institution will
support you in your financial accounting to the funding organization. If you do not
follow these procedures, you will not receive the support you need. And you also open
yourself up to potentially very unpleasant consequences if and when you are audited.
Therefore, each team member must be fully aware of and careful to follow, the
financial regulations (for receipts, advances, equipment purchases, etc.) of the home
institution and the funding organization. Managinging money is not something that all
academics are good at or want to be bothered with. But these days getting external
funding is a strong expectation that universities and many research institutions place on
their staff and therefore all academics must attempt to get funding and, when
successful, must know the operative regulations for managing this money. One very
sensitive issue, for example, that arises is the use of per diem funds, i.e. funds that are
used to pay the living expenses of team members in the field and at conferences. Each
team member must, again, be very clear on how these funds are to be used. They must
also be very clear on how much is to be paid to language teachers, what amount of
money is available for gifts to the community as a whole, and they must be able to
resist pressure to increase such payments and gift amounts. The project cannot invent
money. It must live within its budget.
CONFLICT RESOLUTION: A project must be prepared for conflicts to develop.
Someone is going to get upset at someone else or disagree strongly with some aspect of
the project's management, etc. This is unavoidable. Therefore, a project must have
some way of resolving conflicts. In my own projects, I have taken advantage of
services already available from the University of Manchester, which allow members of
my own research team to discuss their role in the project regularly with other faculty in
the department or members of the personnel department. I also maintain regular and
informal contact with other team members, in the field when possible (certainly with
language teachers and the language community in the field), and also at the home
institution.
ROUTINE: In the field, it is also vital that you have some way to avoid boredom.
I offer a number of suggestions in Chapter __ below. But here, let me say that the most
important component of mental health, in my experience, in field research is the sense
that you are making solid, measurable progress towards your goals, and that you have a
workable routine established that you follow daily.
Among the Pirahãs, in my early years, I fell into a routine quickly. I would get up
about 545AM, just as it was starting to get light, get my flip-flops and two 25-liter
containers and walk down the path to the river. I would then make four trips, hauling up
roughly 200 liters of water to a large barrel mounted above our kitchen sink (which I
built) and then poor the water into the barrel, add chlorine, and then recover the barrel
(looking for frogs that often got into the barrel). Then I would help my wife get
breakfast for our three children and us, most often oatmeal, powdered milk, and coffee.
After that, usually about 815AM, I would work with a language teacher (arranged the
Linguistic Fieldwork Daniel L. Everett
page 247
day before usually, but sometimes just opportunistic, depending on who was in my
house) for two hours. After that, I would help with our children's correspondence
courses, usually teaching my oldest daughter language and history. Then I would visit
Pirahã in the village, doing more elicitation, but mainly just visiting and practicing the
language. I helped with lunch. We always took a brief nap after lunch, to help cope
with the jungle heat. After lunch, I worked most of the afternoon on processing and
clarifying (with an additional language teacher) data that I had collected in the morning.
In the evenings, we went to the river to bathe, after exercise (jogging when trails were
wide enough, skipping rope, lifting weights, etc.). Our home, completely open as it
was, with no doors, no full walls, etc. would fill up with Pirahãs at night. They often
slept in the house with us. After a time with the Pirahãs, we would go to the back part
of the house, the one part that had a door, walls, and screen, and spend time as a family,
usually with me reading to the children. The Pirahãs respected this space usually and
would not come back to that part of the house, though of course they could have done
and occasionally did. As you can see, my day was hardly a morning-to-night linguistic
marathon. This is partially because I had gone to the village with my entire family, but
also partly because I could not absorb more than that. I don't think it is reasonable to
plan long days and seven-day weeks in the village. You can 'burn out' that way. I
recommend a much easier pace, one that allows you to develop an enjoyment of village
living, at least in the first months of your research.
And time for recreation is crucial. So weekends, or at least one day per week,
should be reserved for doing nothing whatsoever on the project. Fieldworkers differ
about this. But I need at least one day a week in the field to myself, to read novels, to
play my guitar, etc.

Case study: the Suyá project

If field researchers were to attempt to develop more team-centered approaches,
what would be needed to achieve success with this type of model? Let me provide a
case study of team-based field research, my own projects for the study of the Suyá
language (the people prefer the autodenomination, Kisedje). Kisedje is a Ge language
spoken by approximately 378 people in the Xingu Park region of Brazil. Although their
culture has been studied extensively, especially by Seeger (), documentation and
description of their language is very sparse (see dos Santos () and Guedes ()). An
apparently related language, Tapaiuna, is spoken near the Kisedje villages. In 2004 I
received grants from the Arts and Humanities Research Council and the Economics and
Social Research Council, both of the UK, to prepare a reference grammar, dictionary,
pedagogical grammar, and theoretical & descriptive articles on the language, along
with a comparative study of Tapaiuna. For this research, I received two postdoctoral
research associates, a PhD student, and a secretary (also a PhD student in Linguistics,
but not doing research on Kisedje). This was a huge change for me. All of my previous
research followed the 'Indiana Jones' model.61
My first tasks were to hire the postdoctoral research associates. What qualities
should I look for in my partners? Analytical ability? Theoretical interest and insight?

61
The project also began with a Brazilian consultant who later dropped out, citing my
bad management, though I think that there were other motives. Still, if I had been a
better manager, perhaps this would not have happened.
Linguistic Fieldwork Daniel L. Everett
page 248
Field experience? Data-base experience? Knowledge of Brazil? Knowledge of
Portuguese? In the end I hired two Brazilian PhDs who turned out to be a marvelous
combination of all these various desiderata and more. But they did not always find me
all that easy to work with. Apparently, my management style can be at once dictatorial
and laisser-faire, which naturally led to misunderstandings and occasional hurt feelings.
That is, after explaining the project objectives and my view of what everyone should do
and could do on the research, I tended to leave people on their own, only getting back
to team members when they failed to meet my expectations, an experience we all found
unpleasant. But my team wanted regular meetings, clear feedback, assessment,
encouragement, and linguistic help. All reasonable items. And they pointed out my
problematic management style. Without management and leadership from me,
therefore, the project would slow down. I could no longer just do my own thing, as it
were, on analysis and elicitation, as per the Indiana Jones model, but needed to be
concerned about others. And I needed to give thought to project management
principles.
Linguistic Fieldwork Daniel L. Everett
page 249
CHAPTER 12. FIELDWORK ETHICS
12.1 INTRODUCTION
Establishing a 'best practice' for ethics crossculturally can be complex. On the one
hand, there are universals or near universals of conscience such that the fieldworker can
go a long ways towards establishing a sound ethical basis for his or her fieldwork by
following his or her conscience and 'doing what seems right'. But regardless of how
well-developed one's conscience is, this subjective rule of thumb is far from an
adequate basis for governing the ethical aspects of any given field project. Ethical
fieldwork is not simply avoiding gaffes or the giving or taking of offense, or failure to
commit criminal acts. Just as peace is more than the absence of war, so ethics involves
a positive, pro-active code of behavior and right-thinking intended to leave the field
situation and language community better off than when the linguist first 'found' them.
Moreover, no researcher can avoid explicit consideration of ethical issues because all
major universities and funding agencies in the Western world require that all research
projects associated with them receive pass rigorous ethical review that requires that the
reseacher(s) deal with the issues discussed in this chapter.
There are several sources available on ethics and fieldwork on the internet. 62 I
provide the URLs to several of them in the footnote above. This chapter is unable to
provide more than the barest of overviews of ethical considerations. I therefore strongly
recommend that the prospective fieldworker also consult the web sources in the
footnote and their references. Nevertheless, this chapter does attempt to cover the basic
issues of the major ethical considerations in fieldwork. We begin with a discussion of
issues affecting the language community and then move to issues concerning
government relations and then to the very sensitive issue of relating to missionaries, an
issue that few fieldworkers will be able to avoid.

12.2. THE LANGUAGE COMMUNITY

CONSENT
The first issue that arises in any fieldwork session is the consent of the language
community to allow and participate in the research and the subsequent dissemination of
the research results and data. The key issue here is that the community and each
individual participating in the research give informed consent to participate and that this
consent be registered in such a way that it is accessible and clear to all that the consent
was voluntary and informed. One way of doing this, when working with a literate
society, would be to spell out the details of the research in writing and ask the
community leadership and/or all language teachers to sign an agreement to work on this
project as it is presented in the document. The details of the research that should be
made clear include:

62

http://www.lsadc.org/
http://www.als.asn.au/
http://www.stanford.edu/dept/linguistics/fieldwork/info/ethics.html
http://www.eva.mpg.de/lingua/files/ethics.html
http://www.aaanet.org/committees/ethics/ch1.htm
http://www.chass.utoronto.ca/linguistics/LIN458H1F/04.458.ethics.pdf
http://www.geneva.quno.info/pdf/tkmono1.pdf (very useful paper, perhaps the best I
have seen on Traditional Knowledge and Property Rights).
Linguistic Fieldwork Daniel L. Everett
page 250
() Research details for discussion with language community (all of these will ideally be
made clear in writing with written consent from the community leadership):
a. The objectives of the research – what are your scientific reasons for being in
the community? How did you make this clear to whoever is funding you? Work hard to
ensure that the community really understands why you are there. Do not be satisfied
with giving them vague ideas about 'studying your language' or some such.
b. The methodology of the research – how do you plan to go about finding out
the things you said you wanted to know under (a) above? How do you expect the
members of the community to be involved? What will they do? What are the possible
risks to them in terms of physical safety, group or individual prestige or loss of face,
etc?
c. The funding of the research: who is funding it; what are the categories of
expenses; for how long is it funded; what percentage of the funding is for the
community? People are rightly suspicious about where the money is coming from. And
they may have heard all sorts of rumors about the linguist's nationality and its activities
in the world. This has the potential downside of confusing people.63 Why, for example,
might the linguist have a grant for $100,000.00 if their share is only $8,000.00? Why is
the linguist staying in a hotel, spending their money instead of staying with a
community member and paying the savings to the community? This is a very sensitive
issue. It is not always a good idea to tell everyone everything about the finances. On the
other hand, it is not a good idea to ignore the issue either.

An interesting example of the kinds of unexpected issues that can arise in this
regard comes from a recent research project of mine. As is common nowadays, the
funding agencies listed my projects along with the total amount of funding in each on
their public webpages. The language community found these pages and read about the
projects. During my next visit, the people were all concerned that I was profiting from
the research because the total funds listed by the funding agencies greatly exceeded the
amount to be received by the community. I had thought before this that it made little
sense to explain things like indirect costs, overheads, postdoc salaries, secretaries, and
university office supplies to the community. But suddenly, I needed to do this. So I did.
Along the way, I explained to the community that I received nothing in salary at all for
the research and that I would earn exactly the same if I never returned to their
community or indeed if I never did field research again. This was important news to
them, because they thought that I was able to make a living only because of their
consent to work with me and that therefore they should be seen as my employers. This
was a very new experience for me, because previous groups I had conducted research
on, in the Amazon, would never have looked at the internet or questioned me in this
way. This is why it is essential to consider each group anew and not to lecture the

63
Many PhD dissertations in the US, like much of the early work on Transformational
Grammar, was funded by branches of the US Military and the US Department of
Defense. This is common knowledge. Although there apparently were no strings
attached to these grants and though there was no connection I am aware of between the
linguistics research of that time (1960s by and large), still, people in countries
suspicious of the USA might want assurances that the funding behind the project have
no connection with the US military. This is an extreme example, but not implausible.
Linguistic Fieldwork Daniel L. Everett
page 251
community but to dialog with it, giving plenty of time and opportunity about issues of
finance, etc.

d. The outside participants in the research: The community will want to know
if there are any 'silent' partners in the research, i.e. postdoctoral research associates,
professorial colleagues, graduate students, etc. that are involved in the research, but
who never come to the community.
e. The potential for profit of the research: Some anthropological studies make
a profit. Anthony Seeger's book, Why Suyá Sing, (Seeger ()) has sold thousands of
copies and makes a reasonable profit, a large part of it (perhaps all, I am not sure) goes
to the Suyá community. As we discussed my linguistics research and the publication of
a grammar and dictionary on Suyá, I assured that any profits at all would go to the
community but that it was unlikely that this would amount to much (ergativity is a less
likely marketing tool than exotic music). (Barbara Kern and I gave all the profits of
Everett & Kern (1997) to the Wari' community. I think that this amounted to a few
thousand dollars. Not a lot, but at least something.)
f. Payment to the community and to individual language teachers: It is crucial
that everyone involved in the research understands how much they can be paid and
what the parameters are, if any, for salary increments, non-budgeted requests, etc.
Agreements should be reached and put in writing or on video about all finances. There
is a serious issue that occsionally arises when a community wants to renegotiate
financial terms as the project goes along, but that is an issue that each fieldworker will
have to confront on their own. The only solution to that is to reason with people and
stand firm within the parameters of the research budget.

On the DVD of the excellent movie 'The Mission', one of the special features –
which I have used in my classes on Amazonian languages for some time – includes the
documentation of a dispute over pay and working conditions between the production
company of the movie and the Waurana people who played the part of the Guarani in
the movie. The people refused to work at a crucial juncture for the movie, in the last
days of production, because they believed that they had been lied to. Apparently, what
had happened with regard to the payment was that the production company had thought
that it could pay the community a lump sum which could then be used or divided as the
community saw fit. However, the community had not understood things quite this way
and individual Waurana were expecting to be paid at the proportionate amount that had
been promised to the community. Moreover, the production company seems to have
thought that the Waurana were committed to the making of the film, as something
benefitting indigenous peoples, when in fact the Waurana were concerned about getting
back to their villages, to their fields, etc. all of which were much more important to
them that the goals of the movie. Eventually an understanding was reached. Part of the
problem is that no one in the production company spoke Waurana. Another problem
was, in my opinion as someone who simply watched the DVD, that the production
company, like many Westerners, assumed that the community had a socialist rather
than individualist view of economy and wealth. Whatever the reasons, it is an excellent
example of why it is vital to explain all financial arrangements with the people before
more serious problems develop.

g. Personal gain of the linguist: The linguist will get something out of the
research, e.g. tenure, a job, promotion, a raise, fame, a book, etc. It is important for the
Linguistic Fieldwork Daniel L. Everett
page 252
community to understand these advantages to the linguist as well as possible. But it is
equally important that no one have the idea that the linguist is going to get rich off the
research or that the linguist's higher standard of living is a result of his or her work with
this particular language community. As I said earlier, I have made the effort to explain
to groups that I work with that my salary, income, job security, etc. does not depend on
my fieldwork, at least not on my fieldwork in their specific community. It is important
to explain this at times so that the people do not get the mistaken impression, as I have
heard said of me, that 'You live on their backs', i.e. that the linguist is exploiting the
people since they live relatively well and can travel, whereas the community that is
perceived as paying his or her salary does not.
It is also important that the form of consent from the community to work on the
project with the linguist is a form that they have access to. And this can vary quite a bit
for field linguists from, say, field psychologists or anthropologists, since linguists often
work with languages where there have been few previous studies and where the people
are pre-literate. There is little point in getting a headman to 'make his mark' on a
contract that no one in the community can read. What I usually do in pre-literate
communities is to record their consent on video and then let them edit it and rerecord it
until they have explained the conditions as they see fit. I then leave a copy of this with
them, whether or not they have the means to play it. At least then they can find
someone with the means to play the tape or DVD and make sure that it says what they
think it says. And these recordings should include the linguist explaining the goals of
the project to the community prior to their giving of consent.
TRAINING AND EDUCATION: See section __ above. The project must include a
component of training for all community participants. This can, but need not be, spelled
out in the agreement with the community. It is important to emphasize again, however,
that no one 'size fits all' in this regard. Each different field situation will favor or require
different training and education relations between the linguist and the community.
EMPLOYMENT CONDITIONS: The community and all language teachers must
understand the conditions under which they are employed. When and for how long are
they to work? How much will they be paid? Will this pay exceed the national minimum
wage (I recommend that people be paid at least five times the minimum wage in the
developing world or an above average wage in the developed world)? Do they have to
leave their community? What rights for return do they have if they fail to honor the
agreement on time outside the village? Other questions will arise in different local
contexts, of course. Much care must be taken to avoid misunderstandings. And all
understandings and agreements should be recorded in a community-accessible fashion.
ADVOCACY: The researcher will often be associated with community aspirations,
at least in the minds of the community if not in their own mind. The community may
want the linguist to help them raise funds for community projects, speak to the
government about encroachments on their reservation, travel with them to negotiate
with people or groups that impinge on the community's well-being in some way, etc.
Pro-active application of ethics will motivate the fieldworker to serve as they are able
as an advocate for the group. This can be a very important service as the linguist, a
prestigious intellectual, for example, can offer that will be deeply valuable to and
appreciated by the community.
INTELLECTUAL PROPERTY RIGHTS & CO-AUTHORSHIP: One issue that is becoming
more and more important for all indigenous communities is ensuring that they control
the applications of their knowledge and community outputs. Who owns the linguistic
data in a text for example? Who has the right to prohibit or constrain the use of all data?
Linguistic Fieldwork Daniel L. Everett
page 253
The answer is that the community controls all data. The linguist must not use any texts,
examples, etc. without the permission of the community (though blanket permission
can be given at the outset of the research, it is still appropriate to seek community
permission for data in individual publications, at least data heretofore unpublished).
And the community has the right to demand the return of all tapes, videos,
transcriptions, and so on – all forms of their language gathered by the linguist – at any
time. And they must be apprised of this right and this understanding must enter the
written or videod record of agreement between the fieldworker and the community.
For further discussion and contemplation of this issue, I recommend a careful reading
of the material in the websites mentioned above, especially the discussion paper by
Correa (1999).
MEDICAL: This is discussed in __ above.
GOING NATIVE: Occasionally I meet linguists or anthropologists with the quaint
idea that they can become just like any other member of the community. So if they see
the people going nude to take their baths, for example, these people cheerfully skip to
the river with the rest of the community, in all their European nakedness. Most
communities are puzzled by such behavior. The fieldworker should show respect and
not give offense. And they should adapt to an appropriate degree to the culture and
behavior of the community, but they are usually expected to respect the simple fact that
they are not one of the community and will never be exactly like them. There can be
exceptions, of course, but it is important that the fieldworker enter the community
aware that 'going native' can be and usually is both silly and offensive.

The chief of the Kisedje people, Kuiussi, raised this issue with me at the beginning of
my project to study their language. He said "I don't want to see you or any of your
people naked in our village. We go naked, this is our custom. But it is not your custom
and it means something different to you than it does to us. And I don't want any of your
group have sex with any of our group because you are not part of our group and we are
trying to preserve our language, our culture, and our identity."

PERSONAL MORALITY: All the same ethical standards that would apply to the
fieldworker in their home country apply to them in the language community. Moreover,
to these can be added the standards of the language community and the laws and
expectations of the country in which the community is located. 'Sex, drugs, and rock n'
roll' are not what the fieldworker has gone to the community for. Most communities
have extremely conservative values concerning personal dignity and morality. Some
have more liberal standards. The general rule in this regard that I would like to suggest
is that the fieldworker figure out the moral standards of the community and follow them
closely, even if some members of the community appear not to do so.
INTERFERING: A final issue that I will raise here has to do with the fieldworker's
potential 'temptation' to interfere in moral actions of the community. For example in a
recent well-known case in Brazil, a missionary couple interfered with a community's
practice of infanticide, by taking the baby that was about to be killed out of the village
by emergency flight, causing some offense among some in the community (even
though in this particular case the parents of the baby wanted it taken out of the
community). This is roughly like an anthropologist studying US culture blocking access
to abortion clinics. It may be a deeply felt moral conviction on the part of the
fieldworker but it will always have extreme consequences.
Linguistic Fieldwork Daniel L. Everett
page 254
Other issues like infanticide abound, though different issues will affect different
people differently. For example, I have seen Amazonian Indians torture animals,
roasting them alive, plucking out their feathers and chasing them around, giving them
to little children to pull apart while they are still alive, etc. I have never interfered. I
have been bothered by this many times and deeply. Perhaps another person would have
interfered. I did interfere once when a group of men were about to rape a young girl in
the village. Should I have done that? I felt that I should have at the time. And there
were no bad consequences to anyone, though I had no way of knowing what the
outcome would be.
These are hard choices and I have no wisdom to offer. But if practices regularly
arise that so deeply offend the fieldworker that they feel compelled to intervene, then I
suggest that the fieldworker is in the wrong line of work and should consider leaving at
least this community and, likely, fieldwork altogether.

13.3. INTELLECTUAL OUTPUTS AND CO-AUTHORSHIP

The primary results of scientific research are publications, webpages, or other
forms of public dissemination of one's findings. The decision as to who should share
co-authorship (and thus blame or fame) for a particular article, book, report, etc. is a
very important one. Should all language teachers automatically be entitled to co-
authorship? Or if the linguist is helped in their living in the village, introduction to the
people, or data-gathering by a third-party is this third-party (e.g. local missionary,
government official, non-indigenous citizen of the country, etc.) entitled to co-
authorship? This is a thorny issue. But I suggest the following rules of thumb:

() Guidelines for co-authorship credit

a. Was the person essential to the research in whole or in part?
b. Did the person make a major intellectual contribution to the research?
c. Was the person actively involved in the writing of the piece?
d. Was the person responsible for obtaining funding or the PI of the grant?
e. Does the person want co-authorship?
f. Does the person believe that they have earned co-authorship?
g. Would it be intellectually dishonest to award or deny co-authorship?
h. Did the person collect the data for your study?

Let us say that your entry to a community, your having a place to live in the
community, and the people's willingness to help you learn their language are all the
results of someone else's efforts. This would be deserving of credit in the
acknowledgments to the research, but by themselves do not entitle the person to co-
authorship. However, there could be additional factors, e.g. if they spoke the language
and ran your experiments for you, served as third-party interpreters, or otherwise
became essential to your data-gathering. Further, if during the course of data-gathering
they offered suggestions that were crucial to the final shape of the research, then this
would also cause re-thinking of whether or not they deserved co-authorship. At this
point, the other factors would 'kick in'.
Did this person give you ideas, suggestions, and help intellectually without which
the research could not have been conceived or carried out or had anything like its final
form? Then co-authorship seems appropriate, subject to the other considerations in ().
If the person was actively involved in the writing up of the research (either by
actually contributing original sections or by consulting with the linguist on many major
Linguistic Fieldwork Daniel L. Everett
page 255
or subsidiary points), then, subject to the other constraints in (), this person should be
considered for co-authorship.
If the person assisting the linguist is the actual PI of the grant then, for some
people, this would automatically entitle this person to co-authorship. In some of the
natural sciences, this is common. However, although I believe that PIs rightfully
demand veto power over any research coming out of their funded projects, I do not
believe that PIs are automatically entitled to co-authorship if this would violate (g), e.g.
if they made no intellectual contribution to the particular research of the report.
It is also important for the fieldworker to know what a person (language teacher,
government official, etc.) wants with respect to co-authorship. Do they believe that they
have earned it and do they want it? Then, subject to (g), they should receive it. If it
would be intellectually dishonest in the eyes of the fieldworker to award co-authorship,
then perhaps it should not be offered. I suggest that in such a case the linguist contact
someone from their home institution or at the journal or other outlet they plan to submit
their research to for publication, to get advice. On the other hand, if the person feels
very strongly, there could be cultural value differences involved, and likely are, and
these are likely to have long-term effects on the researchers, the research, and feelings
of exploitation. For this reason, I believe that co-authorship should be offered when the
person feels they have earned it and they want it.
Occasionally it arises that one researcher collects data and does not analyse it.
Then another linguist may find the data (openly and honestly with no violations of any
ethical standards), analyse it, and publish the results. In such cases, does the original
gatherer of the data deserve co-authorship? Not if the data are published. However, if
the data are unpublished, then the linguist should contact them, ask them about the
analysis that the linguist is proposing, asking if the collector of the data knows of any
counter-examples or other problems in the linguist's analysis and then offer co-
authorship, if the original collector of the data feels that they have earned it and if they
want it.
It is always better to err, modulo intellectual honesty, on the side of giving too
much credit than not enough.

12.4. THE GOVERNMENT

LEGAL ENVIRONMENT: In a new country, the fieldworker must not forget that the
legal environment is not the same as in his or her home country. They must be sensitive
and aware of local laws and customs and avoid anything illegal insofar as possible.
This may require reading on the legal system, major laws, etc. and in discussing
potential issues with friends from the country in question. Permissions and visas are
discussed in __ above.
RELATIONSHIPS WITH NATIONAL SCHOLARS: The fieldworker should (and may be
obliged to) develop close relationships with national scholars. It is important that the
local scientific community know about the linguist, about the research being conducted,
about the community's response to this, etc. There may be legal requirements to this
effect in the country in question (as there are in South American countries, for
example), but beyond the legal requirements it is simple professional courtesy to
develop and cultivate relationships with linguists or related disciplines in the country of
research.

12.5. MISSIONARIES
Linguistic Fieldwork Daniel L. Everett
page 256
No matter where you decide to do field research, there is a good chance that in
the course of your time in the field, you are going to come into contact with
missionaries, especially Christian missionaries. The work of Christian missionaries is
very controversial and most field researchers will have to decide how they are going to
relate to them professionally and personally. The choice made in this regard could
affect your relationship to the language community and to host country intellectuals and
partners dramatically. Moreover, the linguist will also likely be forced to consider the
ethics of the missionary enterprise whether or not they have ever considered it before
and whether or not they want to consider it. So let us consider this here. To do this, for
the potential field researcher who knows little of Christian missions, I will need to
provide some background. This background will include generalizations based on my
many years of experience in a missionary organization. Some of this discussion touches
on uncomfortable issues of religious and political beliefs. But the discussion must cover
these issues, I believe, in order to fully understand why missionaries and missionary
work are controversial.
First, let me say that I began my career in Brazil as a missionary with what was
then known as the Summer Institute of Linguistics (SIL). My assignment with SIL
included the following responsibilities: (i) produce a solid descriptive study of the
grammar of Pirahã; (ii) produce at least two hundred pages of translated, glossed,
publicly accessible texts of the language; (iii) produce a dictionary of the language – as
large as I could, but certainly one of at least 2,000 or so entries; (iv) develop a literacy
programme in the Pirahã communities where every Pirahã man, woman, or child
wanting to learn to read and write their own language could do so, subject to the
permission and direction of the local community; (v) translate at least the New
Testament into the Pirahã language; (vi) share my Christian faith, on an individual basis
with Pirahãs, but without any proselytising, pressure, or church-related activities. The
aim of the latter was to see an 'indigenous church' (one initiated by, managed by and
formed doctrinally by the Pirahãs themselves). Had I been with another missionary
agency working with Amazonian Indians, my goals might have been to found a church
by preaching, baptizing, and never hesitating to let the people know when their culture
violated 'God's law'. If I had been a Catholic missionary, depending on which 'order' or
organization I belonged to, my goals might have been to baptize the people and see a
chapel established for semi-regular visits to hear confession, provide communion, etc.
Often missionaries from one mission are suspicious of missionaries from other
missions and across the Protestant vs. Catholic divide their is a yawning chasm of
centuries of mistrust. So if you befriend or associate with one missionary or group, you
may automatically, however unintentionally, alienate another group with consequences
potentially adverse for your research.
Most Christian missionaries are motivated primarily by verses like the
following:

Matthew 28: 19, 20

"Therefore go and make disciples of all nations, baptizing them in the name of
the Father and of the Son and of the Holy Spirit, and teaching them to obey
everything I have commanded you. And surely I am with you always, to the very
end of the age."

Acts 1: 8
Linguistic Fieldwork Daniel L. Everett
page 257
"But ye shall receive power, after that the Holy Ghost is come upon you: and ye
shall be witnesses unto me both in Jerusalem, and in all Judaea, and in
Samaria, and unto the uttermost part of the earth."

No matter how these beliefs emerge practically, they still are there and knowing
about them can help understand missionaries better.
Whatever anyone may eventually conclude about missionaries, they are doing
what they believe to be right. Many if not most of them are also motivated by love of
their god and love of the people with whom they work. Contrary to most stereotypes,
including the utterly silly portrayal of missionaries in films, e.g. At Play in the Fields of
the Lord, missionaries are generally kind, reasonable on many issues, attractive, and
loving, often with a deep satisfaction with Life and a sense of security that is many
ways enviable. It is important to say this, because most of us find it easy to befriend
and enjoy the company of such people, especially when otherwise isolated from our
native language, culture, and countrymen. Moreover, in many, many parts of the world,
the only medical, educational, economic, and political assistance available to the local
community is from missionaries. Many missionaries sacrifice personal comfort, the
company of their families and friends, their own career aspirations, their health, and
material well-being (contrary to many popular stories, missionaries I know who work
with Amazonian Indians are poor by US standards and live very simple lives
materially) for the sake of the people to whom they 'minister'. Most missionaries that I
know would die for the communities they believe they have been called to serve. This
latter may sound melodramatic, but the emotions on this issue can run high and it is
worth bringing out some of the depths of feelings that people have. Moreover, the
members of the language community are likely to support the continued presence of the
missionary because many of them will now be believers and because none of them will
want the missionary to leave if this removes all the medical and other material support
that the missionary has been providing. Of course, this latter source of support for the
missionary is part of the problem. One reason that missionaries do 'get converts' in
many tribal areas in my experience is because of this support. The missionary may use
this support as both carrot and stick to convince people of their 'need' for the church
and its doctrines (though certainly not all missionaries do this). The missionary is a
powerful person by local standards and it is almost never the case that members of the
local community are able to make fully objective decisions about what role to provide
to the missionary in their community in the absence of alternative forms of medical and
material support.
In this regard, many intellectuals have concluded that missionaries' 'retrograde'
thinking (to use a common expression) is harmful for the peoples where they work,
largely forced on them, overtly or covertly, and that the paternalism, condescension,
and backwardness of much missionary thinking, should be stopped from reaching the
language communities facing severe problems in the modern world, problems serious
enough to threaten their very survival. And this fairly natural view is likely to be one
that many readers of this book share.
The linguist must therefore be prepared for how they are to relate to
missionaries. Because of the close relationship that often obtains between the
missionary and the community, open hostility or rudeness is not a good choice. This is
divisive for the community in almost every case, leading to potential antagonism
between 'believers' vs. 'unbelievers' or, from another perspective, the 'rational' vs. the
'superstitious'. On the other hand, completely unguarded, unreserved friendship and
Linguistic Fieldwork Daniel L. Everett
page 258
support of the missionary can lead to the linguist being classified as a missionary, since
to many indigenous peoples with very limited experience with other nationalities, for
example in Brazil, 'American' can already be a synonym for 'believer'. The default
assumption is often that all those who share the missionary's nationality share his
beliefs. And this assumption can extend to local government officials as well.
How then does one sort this issue out? First, be careful of generalizations. By
and large, attempt to evaluate your potential relationship to people in the field not based
on their occupation, religious beliefs, nationality, or culture, but based on your
assessment of the person as a specific, sui generis whole. Second, maintain your
distance and build relations very carefully. I suggest an initial attitude that I often urge
on my junior academic colleagues when they first begin teaching, namely, 'maintain
polite aloofness'. The practical application of this 'aloofness' will need to be worked out
in individual cases. In some cases, for example, the linguist may be forced to choose
between paying for space on a missionary plane to the language area, though this is
likely to associate the linguist with the missionary, or travelling in by car or boat or on
foot. Travelling to some of the peoples I work with, for example, this can mean a plane
trip of one hour vs. a boat trip of one week. Alternatively, the linguist may have to
choose between a local 'air taxi' vs. a local missionary plane. The former will almost
certainly be more expensive and will often be less safe than the latter (missionary pilots
and planes tend to be the best trained and maintained, respectively). Whatever the
decision, the missionary presence presents the linguist with an ethical issue with stark
choices over several issues, some clear, some nebulous.
In my very close connection over a period of more than twenty-five years with
so-called missionary linguists, I can think of only a relative handful of missionaries
who can in any sense be called linguists by training, interest, or motivation. But the
'handful' is not insignificant and these tend to be professional linguists by anyone's
standards and quite unlike most missionary stereotypes I have encountered.
On the other hand, some missionaries do produce excellent linguistics. And no
linguist has any ethical justification for either refusing to cite that linguistic research or
to acknowledge its importance in their linguistics research (as I have known some more
radically anti-missionary linguists to do). Moreover, in the case of SIL, the organization
produces some of the best software for linguistics research available anywhere in the
world, including fonts, data-management programs, electronic libraries, phonetic
analysis, and so on. And some missionary-produced grammars and research is among
the best in the world as well. It is unethical for researchers to use such software and
intellectual output (including grammars and articles) from missionaries on the one
hand, while on the other accusing the missionaries of not being linguists in any sense,
as some are wont to do.
The world is complex and people are complex. Simplifications that involve
complete vilification (or beatification) of entire groups is not only unethical (since it is
the basis of prejudice), this kind of behavior leads to the worst (and most infantile)
kinds of intolerance. This is not to say that people should avoid negative conclusions of
this or that enterprise or organization and eschew any contact with it. Even so, it may
be hard to avoid, say, the work on morphology by Nida () or the work on phonetics by
Pike (), simply because they were doing missionary work at the time they wrote these
books, although, according to Anderson (), exactly this has been done to some degree,
by the way that modern linguistics has failed to acknowledge the debt it owes to these
two missionary linguists.
Linguistic Fieldwork Daniel L. Everett
page 259
Trying to apply an 'orthodoxy' test to writers before reading their writings on
any subjects is akin to book-burning and has a long and sad history in Western culture
that we have, hopefully, come to reject in modern society. That said, missionaries are at
once people and an issue. There will be times that opposing them or befriending them
could either be appropriate in particular circumstances.
Linguistic Fieldwork Daniel L. Everett
page 260
APPENDIX ONE: WRITING FUNDING PROPOSALS FOR FIELDWORK
Writing a funding proposal is hard work. Funding is never guaranteed and
always highly competitive. The work and long wait involved in a funding proposal will
more often than not end in a rejection letter from the funding agency. So why should
one subject oneself to this?
The reason is simple. You need money to do field research. And well-funded
projects can produce better results than poorly-funded projects. In terms of your career,
well-funded projects also bring money and student support to your home institution.
They also are necessary to underwrite team research and to enable you to train new
fieldworkers.

I have received as Principal Investigator (PI) or co-PI more than six million
dollars (US) in funding since 1984, with an overall 'hit rate' of 70%. That is, 70% of all
my grant proposals have been funded. I have received funding from the National
Science Foundation, the National Endowment for the Humanities, the Fundação de
Amparo a Pesquisa do Estado de São Paulo, the Arts and Humanities Research
Council, the Economics and Social Research Council, and the European Union. This
has not all been for field research, but it has given me a solid basis for the advice I want
to offer in this section.
At the same time, as I once heard Peter Ladefoged say, I have been turned down
for more money than most people ever ask for. The 30% of proposals that I have had
rejected, however, have, due to useful feedback from referees and program officers,
made me a better grant writer.

The first step in preparing a proposal is to identify an appropriate funding

agency. Agencies differ widely in the kinds of research that they fund, the kinds of
methodologies that they favor, the amount of additional personnel, indirect costs, etc.
they will fund. And they differ in their constraints on the form of proposals. So before
writing a proposal, familiarize yourself with the agencies most likely to be interested in
your proposal.
All funding proposals will be evaluated by at least the following criteria:
(i) Track-record of the PI – do they deliver what they promise? Are they
active, publishing researchers? Is their work respected as of high quality?
(ii) Evidence of preliminary preparation, e.g. contracts and permissions
obtained ahead of proposed project start date, etc.
(iv) Budget – is it reasonable, non-lavish, yet well thought out, covering all
likely expenses (from batteries to hotels), etc.?
(v) Criteria for success – how will the scientific community know when the
project is a success? How and when will project personnel, including the PI, know this?
How will the funding agency know this? What are the follow-up plans if things do not
go exactly according to plans (they will not)?
(vi) Deliverables – has the PI promised too little or too much from what they
expect to learn from this project (publications, websites, blogs, community
contributions, etc.)?
(vii) Intellectual quality of the proposal – the reviewer wants evidence that
the PI is fluent in his speciality and the matters to be investigated and that the PI has
outlined an interesting problem that the referee would like to know the answer too
(ideally) and has contextualized it appropriately within the field of study.
Linguistic Fieldwork Daniel L. Everett
page 261
With regard to your track record, if you are a new PhD with few or no
publications yet, reviewers will be willing to give you the benefit of the doubt,
depending on how long it has been since you received your PhD and how well-written
the current proposal is. The larger and more ambitious your project, the more reviewers
will require from previous publications of the PI. Your track record, as mentioned, will
also be extremely important. If the PI says 'We will publish the results of our findings
in Language, NLLT, and Linguistic Inquiry', the readers will ask if they have ever
before published in these journals.
It is also important that once the PI has identified a funding agency, that they
contact the relevant administrative personnel of the agency to discuss their potential
project with them to confirm that this agency is interested and to request advice on
what the agency looks for in a successul proposal. Occasionally, it is even worthwhile
to travel to the headquarters of the agency to meet the program officer in person or to
the relevant professional meetings, e.g. the Linguistic Society of America annual
meeting, where linguistic funding agencies and program directors often agree to meet
with potential PIs.
It is also useful to bear in mind that for many agencies, at least in the USA, the
submitter has the right to request that some individuals not be asked to review the
proposal or that some individuals be requested to review the proposal. If you know
someone hates your type of research (or, worse, hates you), then you should say simply
'I prefer that the following people not be asked to review my work.'
Linguistic Fieldwork Daniel L. Everett
page 262
Appendix 2: Ethics statements from the Linguistic society of America and the Max
Planck Institute for Evolutionary Anthropology

Linguistic Society of America: (http://www.lsadc.org/):

"Human Subjects in Linguistic Research

Studies of a human language often depend upon a continuing relation with

speakers of the language. Such a relation comes to be defined as much by the speakers
as by the linguist. Their patterns of life govern when work can be done. Their
expectations, and those of their community, shape what is to become the results of the
work. Understanding of the nature of linguistic inquiry grows in the course of the
relationship. Sometimes lifelong friendships are established.

Such work must be conducted with respect for those who participate, with
sensitivity as to their well being, and with concern for consequences of publication or
sharing of results.

Certain considerations may make the study of a language different from much
research in the sciences and social sciences. One asks many questions in discovering
the features of the language, of a kind the collaborator learns to expect and even
anticipate. They are seldom of a sort that can be disturbing or injurious. Moreover,
fruitful work may depend upon the linguist learning and observing the norms of
politeness and friendship expected by those with whom he or she is talking. Those who
participate in such a work often do so with pride in their command of their language
and may wish to be known for their contribution. Not to disclose their names would do
them a disservice. Native Americans sometimes justly criticize earlier work with their
language for not having adequately proclaimed the contributions of the Native
Americans themselves. Fairness to speakers of a language is very much a matter of
understanding their viewpoint, and what is appropriate in one situation may not be in
another.

Such considerations make it difficult to apply general rules in a mechanical

way."

Max Planck Institute (http://www.eva.mpg.de/lingua/files/ethics.html)

"The following guidelines are binding on all members of the department. They cover
relations between the fieldworker and native speakers of languages under investigation
and / or the communities in which those languages are spoken. They do not cover the
following aspects that might nonetheless be relevant to the work of the department but
which should be covered elsewhere:
a) Matters relating to scientific plagiarism and falsification. These are covered by the
Max Planck Society's publication Rules of Good Scientific Practice / Rules of
Procedure in Cases of Suspected Scientific Misconduct.
b) Matters relating to the collection of genetic samples. Where members of the
department are involved in the collection of genetic samples, the ethics policies of the
Department of Evolutionary Genetics must be followed. Researchers should note, with
respect to point 1 below, that additional official permission to collect genetic samples
will be needed in many parts of the world, in addition to permission to conduct
Linguistic Fieldwork Daniel L. Everett
page 263
linguistic work.
c) Matters relating to work with children. The policies set out below must be followed
as far as possible in relation to work with children, but other ethical issues relating
specifically to the status of children must be taken into account, including but not
restricted to the need to obtain the consent of their parents / guardian(s). Members of
the department are required to adhere to the highest ethical standards in their
research. In particular, they must show respect for the individuals, communities, and
cultures with which they work. The following are designed to implement this general
recommendation more specifically.
1. Members of the department must comply with all legal requirements for the conduct
of their research in the relevant area. This includes but is not restricted to obtaining
necessary visas, residence permits, and research permits, from both national and local
authorities, and complying with requirements on the collection and dissemination of
materials.
2. Members of the department must ensure that they have the informed consent of the
individual(s) and community(ies) concerned to carry out the research in question and
to disseminate the results of that research. In this connection, it is essential to note that
consent must be informed. In particular, explanation must be given of the uses to which
the material will or might be put and of the access that will or might be made available
to the material. Under appropriate circumstances, individuals or communities may
place restrictions on the use or accessibility of material, and such restrictions must be
adhered to. Researchers will often have to rely on their own judgements as to which
individuals or
communities must be asked for their consent, but such judgements should be formed in
a responsible and accountable manner. Researchers should note that informed consent
may need to be obtained not only from the source of material (e.g. the narrator of a
story) but also from others who are affected by that material (for instance, persons who
are mentioned in the story). Under no circumstances should individuals or communities
be subjected to coercion to give their consent; researchers will need to be sensitive to
local circumstances in this respect.
3. The agreement between the researcher and the individuals / communities involved in
the research must be documentable, i.e. if a question arises as to the validity of the
agreement the researcher must be able to produce evidence of that agreement. In some
cultures and circumstances a written agreement will be appropriate, whereas in other
cases some form other than a written agreement will be needed.
4. Especially given the increasing importance of intellectual and cultural property
rights, individuals or communities participating in research should be informed that the
institute and the researcher seek the right to store, use, and disseminate (with
restrictions where appropriate) the material in question, but do not assert ownership of
the intellectual or cultural materials entrusted to the institute or the researcher. When
stored and disseminated, such materials should always make due acknowledgement to
their authors and performers. Authors / performers should be named explicitly only
where their informed consent to this has been obtained; otherwise, an anonymous
acknowledgement is appropriate. It is appropriate for the researcher to pay the
individuals involved in research for their time and travel and other out-of-pocket
expenses. It is not appropriate to make payments that might be construed as payments
for the transfer of ownership.
5. Members of the department must, wherever possible, ensure that they contribute to
the communities in which they work. Exceptions to this policy can only be considered in
Linguistic Fieldwork Daniel L. Everett
page 264
truly unusual circumstances where implementation of the policy is impossible, and such
exceptions require detailed justification and the approval of the department director.
Contributions to the community would include but not be restricted to:
a) documentation of the language in a way that is accessible to the community, for
instance through the preparation of primers or printed or audio recorded collections of
traditional stories;
b) development of a writing system for the language;
c) training of members of the community in appropriate ways, for instance in text
transcription, linguistic analysis, literacy, audio and video recording.
Linguistic Fieldwork Daniel L. Everett
page 265
Appendix Three: Phonology Questionnaire
Caveat: This questionnaire is not intended to be even nearly exhaustive. It is simply a
list of some suggestions from my own fieldwork as to what I have found useful. It is
not really for beginners, however. I intend it to serve as a framework, a useful prod, for
writing detailed phonologies of languages, something missing from most grammars
(though the detailed phonologies of the OUP x series, edited by Jacques Durand is a
useful model). Like all questionnaires, as Nikolaus Himmelmann has pointed out to me,
this one has the disadvantage of asserting categories when in fact many of the
categories themselves need to be argued for and established independently. So it
certainly isn't meant to be followed slavishly. Hopefully, however, it will provide
useful suggestions for the 'phonologically challenged' fieldworker.

1. Segmental phonology
1.1. List the distinctive segments of the language. Give rules of allophonic distribution.
Summarize the (articulatory) phonetic realization of each segment.
1.2. What are the nonallophonic restrictions on the distribution of these segments? For
example, do any segments appear exclusively in loan words? Are any subject to
sociolinguistic or cultural restructions (e.g. "Do not use /x/ in the presence of
foreigners")? Are there differences in the segmental inventory according to gender (e.g.
men use /s/ where women use /h/ or variation in points of articulation between women
and men)?
1.3. Are some segments restricted as to which word class they may appear in (e.g. /b/
only in Nouns and Adjectives)?

2. Syllabic structure
2.1. What are the syllable types (e.g. CCV, CV, CVC, etc.)?
2.1.1. Describe any restrictions on syllable distribution. Are some syllables allowed
only in word/phrase-final position (or medial or initial)?
2.1.2. Discuss the evidence for these syllables.
2.1.2.1. Phonotactics:
Are consonant sequences allowed? Where? Do allowable consonant clusters
vary according to where they appear in the word (e.g. st only in word-initial position,
but ts in word-final position)? Are there any restrictions as to the type of
vowel/semivowel which may precede/follow consonant clusters?
Are there word-final consonants?
Are vowel sequences permitted? Where? Do allowable vowel clusters vary
according to where they appear in the word (e.g. ai only in word-initial position, but ia
elsewhere)? Are there any restrictions as to the type of consonant or semivowel which
may precede or follow specific vowel clusters?
How many vowels or consonants may appear in a single cluster, if clusters are
allowed? In adjacent vowels are there restrictions on vowel features (e.g. all the vowels
have the same value for height, roundedness, etc.)? Are some sequences banned (e.g.
aa)?
2.1.2.2. Phonetic evidence
Is there phonetic evidence in favor of syllables (e.g. chest pulses)?
2.1.2.3. Do native speakers segment words into syllables in slow speech?
2.1.2.4. Do phonological rules crucially refer to syllable structure, e.g. stress placement,
nasal spreading, tone distribution, etc., as in (i) and (ii):
Linguistic Fieldwork Daniel L. Everett
page 266
(i) Stress the rightmost (C)VC or (C)VV syllable in the word, otherwise stress
the penult?
(ii) Lower tautosyllabic, adjacent high tones to mid tones in (C)VV syllables.

2.2. Interpretation of glides

2.2.1. Do semi-vowels, such as [y] and [w], appear in both or either syllable-initial and
syllable-final positions?
2.2.2. If the language allows vowel sequences and semi-vowels, may the first vowel be
[i] or [u]?
2.2.3. In vowel or semi-vowel sequences, are any of the following orders prohibited?
Preferred? (X and Y are variables and thus may represent any segment type):
X iy Y
X yi Y
X uw Y
X wu Y
X yu Y
X uy Y
X wi Y
X iw Y
etc.

2.3. What are the allowable sequences of segments within the syllable, according to
their articulatory classification or generalizable acoustic properties? For example, are
there ordering restrictions such as the following (just as a few suggestions):
(i) The onset of a syllable may begin with any consonant, but the second member
of a complex onset must come from a more restricted class of segments (e.g. voiced
continuant)?
(ii) In a complex nucleus, the first vowel must be a high vowel.
(iii) In a complex coda, the order of consonants is more (or less) restricted than
in complex onset.
(iv) The order of consonants in the coda is the mirror-image of the order in the
onset.
(v) etc.
2.4. If the nucleus contains a diphthong, can it also contain another vowel?
2.5. If the language has CVC syllables, can V be a diphthong? If so, are there any
restrictions on the following C?

3. Tone
3.1. Does the language have contrastive pitches which distinguish lexical meanings of
words?
3.2. Do contrastive pitches have a fairly constant F0 or does their F0 rise or fall or
'undulate' significantly?
3.3. If F0 of pitches varies, yet is significant in distinguishing lexical items, does the
variation correlate with position in the word, preceding or following segments,
preceding or following pitches, or the word's position in the sentence or discourse?
3.5. Can consonants bear tone or only vowels? Which consonants? Under what
circumstances (e.g. 'w and y bear tone following a rule of asyllabification').
3.6. Does consonant voicing affect tone? How?
3.7. Does vowel quality affect tone? How?
Linguistic Fieldwork Daniel L. Everett
page 267
3.8. Can tone patterns of individual words vary arbitrarily or do there appear to be tonal
melodies assigned on words or classes of words (e.g. High Low Mid for one class of
nouns, HLH for another, LHL, etc.)? Do the tonal melodies change according to the
number of syllables?
3.9. What happens to a tone if its associated segment is deleted? For example, does the
tone delete or appear on another syllable?
3.10. If the language does not allow contour tones (those with an underivable but
constant change in pitch, e.g. rise and fall) on short vowels or sequences can they arise
from morphological or phonological processes? Consider the Pirahã example in (i):
(i) tíi /ísitoí /ogabagaí → tíi /ísito(ogabagaí
I egg want

In the case of (i), in normal speech the direct object and the form form a close
phonological unit, deleting a verb-initial glottal stop and the noun-final vowel.
Notice that the tone does not disappear, however. In the example in (i) no mark
over a vowel indicates low tone and the acute accent marks high tone. The
wedge over the first /o/ to the right of the arrow indicates that it bears both a low
tone and high tone simultaneously. This is in my analysis the result of the high
tone remaining even after its original vowel host, /i/, has been deleted. This is
otherwise prohibited in the language.
3.11. Can a tone ever shift to the right or the left in a word? Across words? Can one
tone ever replace another, e.g. in (i) immediately above?
3.12. Is there complementary distribution among the tones, e.g. H → M/___ L
3.13. Are the frequency distances between tones (especially in a language with three or
more tones) fairly constant or are some tones closer in frequency than others
(e.g. tone Mid and tone High being closer in average frequency than tone M and
tone L, in a three tone system)? Is frequency distance affected by how many
different tone levels are present in a given word or phrase?
3.14. Is tone affected by phrasal intonation? How?
3.15. Does the language have other channels of discourse that exploit linguistic tone,
e.g. whistle speech, drum communication, hum speech, etc. Please describe this in
detail, as well as the social/cultural restrictions on its use.

4. Intonation
4.1. What is the most common intonational pattern (e.g. rising, falling, fall-rise, rise-
fall, etc.) at the end of utterances?
4.2. How are different intonational patterns distinguished? By end points? By
beginning and end points? By relative height of the entire intonational phrase? By
beginning, middle, and end points?
4.3. What functions does intonation serve? For example, does it distinguish:
syntactic phrasal types (e.g. interrogative, declarative)?
illocutionary acts (e.g. indirect request vs. direct request)?
other?
4.4. Is intonation affected by tone, stress, syllable patterns, or other phonological
phenomena? How?
4.5. Does intonation affect tone, stress, syllable patterns, or other phonological
phenomena? How?
4.6. What is the largest grammatical unit for which you can identify a distinct
intonational pattern? Phrase? Sentence? Paragraph? Discourse?
Linguistic Fieldwork Daniel L. Everett
page 268
4.7. Does intonation serve to unite two or more phrases in parataxis?
4.8. Can intonation mark subordination/superordination relations between clauses?
4.9. Are there step accents in the language, i.e. where the highest pitch of one
intonational contour appears immediately prior to the stressed syllable, which itself
bears a relatively low pitch? Are other correlations between stress and intonation
placement observed? Describe these carefully, paying attention to the syntax,
semantics, and pragmatics of the utterances as you do so.
4.10. What are the quantitative variations allowed in basic intonational contours? That
is, can the same contour appear with more or less prominence by manifesting greater
pitch distances between its distinctive points? When? What is the F0 evidence like?
4.11. Is it more common for frequency to decline at the end of utterances? How many
syllables or words are in the domain of this declination? Is there an accompanying
rhythm (slow down, speed up, etc.)?
4.13. How can the different intonational contours be affected in their overall ranges of
pitch, amplitude, duration, etc. by different ways of speaking, e.g. 'speaking up',
whispering, etc.?
4.14. How is intonation manifested across different prosodic channels (e.g. whistle
speech)?

5. Stress
5.1. Are some syllables in the language more prominent (for example, by using more
acoustic energy, e.g. louder, higher pitch, longer, etc.) than others?
5.2. Do such syllables appear in every word?
5.3. Is this prominence predictable? How?
5.4. Are there different patterns of prominence on different classes of words, e.g. nouns
vs. verbs (if there are, describe them)? Or is it constant across all lexical categories?
5.5. Are there secondary (tertiary, quaternary, etc.) stresses?
For example: multiplication
2ary 1ary
5.5.1. Do nary stresses occur at regular intervals? How are these intervals determined
(e.g. every other syllable in the word from left-to-right, etc.)
5.5.2. Can primary or secondary (etc.) stresses ever appear on adjacent syllables in a
word or phrase?
5.5.3. Does the stress of one word/syllable ever seem to move away from the stress of
another word when it would otherwise be adjacent? Which of the otherwise adjacent
stresses shifts, the one on the left or the one on the right? (e.g. Thir'teen + 'women →
'Thirteen 'women)?

5.6. Are 'heavy' syllables more frequently stressed than nonheavy syllables (e.g. (C)VC,
(C)VV, vs. (C)V)? Under what circumstances, if any, can a lighter syllable bear
primary stress if primary stress is normally restricted to heavy syllables?
5.7. What are the acoustic correlates of stress (e.g. loudness, pitch, length)? Are the
correlates constants or variable across utterances or across speakers?
5.8. Do any (morpho)phonological processes interact with stress in a systematic way?
What is the nature of this interaction (e.g. segmental lenition, voicing, vowel harmony,
vowel reduction, etc.)?
5.9. If heavy syllables bear stress, what happens if the syllable-final consonant or vowel
of the stressed heavy syllable is deleted?
Linguistic Fieldwork Daniel L. Everett
page 269
5.10. If stress shifts for any reason, in which direction does it shift, leftwards or
rightwards? Is its 'final destination' predictable in such stress shifts? How?
5.11. Does stress behave identically in longer and shorter words or utterances?
5.12. Is there any evidence of native speaker sensitivity to stress, such as correcting you
misplacement of it, tapping on stressed syllables as the say the word (Ladefoged,
Ladefoged, and Everett ()), etc.?
5.13. How does (or does) stress interact with tone? Does stress shift also cause tone
shifts? Does stress placement perturb (raise, lower, metathesize, etc.) tones?

6. Morphophonology
6.1. Do affixes affect stress or tone patterns in words?
6.2. Do the affixes which do and not not affect stress (if there are such distinctions
among affixes) fall into natural semantic, phonological, or morphosyntactic classes
(e.g. syllable structure, inherent tone, prefixes vs. suffixes vs. infixes, derivational vs.
inflectional, etc.)?
6.3. Do segmental rules (e.g. devoicing, assimilation, vowel-harmony, deletion, etc.)
affect affix shapes? Which and how? Again, what are the differences between affixes
which are affected vs. those which are not?
6.5. Does the language have clitics? (Like affixes, these are phonologically dependent
on another word, never appearing alone. Unlike affixes, a single clitic can appear on a
wide variety of word types, e.g. N, V, A, P).
6.6. Do these clitics appear in various locations within the sentence or do they cluster in
a given position?
6.6.1. If clitics appear in different positions, does their placement depend on
phonological (e.g. stress) or syntactic (e.g. a clitic must appears on the word to the
immediate left of the word with which it forms a syntactic consituent. Consider
English, where ()s = phonological boundary and []s = syntactic boundary: ([I]['ll) (go)].
6.6.2. If clitics cluster in a given position, which clitics may cluster and where this
takes place in the phrase or sentence (e.g. 'all clitics expressing tense and mood appear
following the first constituent of the sentence).

6.7. Are some phonological processes peculiar to particular types of affixation (e.g.
prefixation, suffixation, infixation, simulfixation, circumfixation, etc.)?
6.8. Is there reduplication?
6.8.1. Is reduplication monosyllabic, disyllabic, or larger or smaller (e.g. a single
vowel, consonant, or mora)?
6.8.2. Do the consonant-vowel sequences in the reduplicated morphemes follow a
constant order and shape for all reduplicated affixes?
6.8.3. Are there subregularities of CV order (e.g. CVC) for one type of reduplication
and others (e.g. CVVC) for others?
6.8.4. Does reduplication interact with any other phonological processes, e.g. stress,
nasalization, vowel-harmony, etc.?

7. Other prosodies
7.1. Do any other phonological elements take a domain larger than individual
segments? Some possibilities are: aspiration, nasalization, labialization, voicing, vowel
features, and so on.
7.2. Do such elements take larger domains only under certain circumstances? That is,
can they 'spread' to surrounding phonological material?
Linguistic Fieldwork Daniel L. Everett
page 270
7.2.1. In what direction can they spread?
7.2.2. What can trigger this spreading?
7.2.3. What can block this spreading, e.g. 'nasalization spread is blocked when it
reaches a voiceless consonant).

Is there a minimal word size (e.g. no word in isolation can be less than two moras in
length)?
Linguistic Fieldwork Daniel L. Everett
page 271
REFERENCES
Bickerton, Derek. 1990. Language and species. Chicago: University of Chicago Press.
Chomsky, Noam. 1980. Rules and representations. New York: Columbia University
Press.
Chomsky, Noam. 1986. Knowledge of language: Its nature, origin, and use. New York:
Praeger.
Chomsky, Noam. 1989. “Some notes on economy of derivation and representation.”
(MIT Working Papers in Linguistics X, edited by I. Laka and A. Mahajan. 43–
75.) Cambridge, MA: MIT Department of Linguistics and Philosophy.
Darwin, Charles. 1859. The origin of the species. (Great Books of the Western World.)
New York: Britannica.
Donald, Merlin. 1991. Origins of the modern mind: Three stages in the evolution of
culture and cognition. Cambridge, MA: Harvard University Press.
Everett, Daniel L. 1985. “Syllable weight, sloppy phonemes, and channels in Pirahä
discourse.” (Berkeley Linguistics Society Proceedings, Volume 11.) Berkeley:
University of California, Department of Linguistics.
Everett, Daniel L. 1989. On floating feature nodes in Pirahä phonology. (Talk presented
at the annual colloquium of Generative Linguists of the Old World.) The
Netherlands: Utrecht.
Everett, Dan, and Lucy Seki. 1985. “Reduplication and CV skeleta in Kamaiurá.”
Linguistic Inquiry 16:326–330.
Fodor, Jerry. 1975. The language of thought. New York: Thomas Y. Crowell.
Fodor, Jerry. 1983. The modularity of mind. Cambridge, MA: MIT Press.
Fodor, Jerry. 1987. Psychosemantics: The problem of meaning in the philosophy of
mind. Cambridge, MA: MIT Press.
Gibson, Edward. 1991. A parallel model of human parsing. Unpublished dissertation.
Department of Philosophy, Carnegie Mellon University.
Jakobson, Roman. 1990. On language. Cambridge, MA: Harvard University Press.
Longacre, Robert. 1964. Grammar discovery procedures. The Hague: Mouton.
Marantz, Alec. 1982. “Re reduplication.” Linguistic Inquiry 13:435–482.
Rodrigues, Aryon. 1953. “Morfologia do verbo Tupí.” Letras I:121–152. Curitiba,
Paraná, Brazil.
Tarnus, Richard. 1991. The passion of the Western mind. New York: Harmony Books.

References (seriously incomplete)

Beckman, Mary E., 1986, Stress and non-stress accent, Dordrecht: Foris
Chafe, Wallace L., 1994, Discourse, Consciousness, and Time, Chicago: The
University of Chicago Press
Croft, William, 1995, “Intonation units and grammatical structure”, Linguistics 33:839-
882
Dixon, R.M.W., 1972, The Dyirbal Language of North Queensland, Cambridge:
Cambridge University Press
Donohue, Mark, 1997, “Tone system in New Guinea”, Linguistic Typology 1:347-386
Hayes, Bruce, 1995, Metrical stress theory, Chicago: The University of Chicago Press
Heath, Jeffrey, 1980, Nunggubuyu myths and ethnographic texts, Canberra: AIAS
Laver, John. 1980. The phonetic description of voice quality. CUP
Linguistic Fieldwork Daniel L. Everett
page 272
Pawley, Andrew and Frances Syder, 2000. “The one-clause-at-a-time hypothesis”. In
Heidi Riggenbach (ed.). Perspectives on Fluency, pp. 163-198. Ann Arbor: U.
Michigan Press.
Selkirk, Elisabeth O., 1984, Phonology and Syntax: The relation between sound and
structure. Cambridge, Mass.: The MIT Press
Steedman, Mark, 2000, The syntactic process, Cambridge, MA: The MIT Press.
Trubetzkoy, Nikolaj S., 1939, Grundzüge der Phonologie, Prague (Travaux du Cercle
Linguistique de Prague 7)
Yip, Moira, 2002, Tone, Cambridge: Cambridge University Press

Gussenhoven 2004
Güldemann, 2005 – Shona CHECK

Payne, Doris (ed.). 1990. Amazonian linguistics: studies in lowland South American
languages. University of Texas Press.

Rappaport, Gilbert C. 2002. Numeral phrases in Russian: A Minimalist Approach,

Journal of Slavic Linguistics 10: 329-342.

Zwicky and Anderson [from the second edition of the Oxford International
Encyclopedia of Linguistics, to appear]

Japanese honorifics as emotive definite descriptions

Christopher Potts and Shigeto Kawahara
University of Massachusetts, Amherst
Salt 14. 2004
SALT XIV
Held at Northwestern University May 14-16, 2004
Edited by Robert B. Young 322 pages, Cornell Linguistics Circle (CLC)

The Independence of Syntax and Phonology in Cliticization

Judith L. Klavans
Language, Vol. 61, No. 1. (Mar., 1985), pp. 95-120.

On Linguistics and Politics

Noam Chomsky interviewed by Günther Grewendorf
Protosociology, Vol. 6, 1994, pp. 293-303
Dixon, R.M.W. Searching for Aboriginal Languages : Memoirs of a Field Worker
University Of Chicago Press; Reprint edition (July 15, 1989).

Chagnon, Napoleon A. 1996. The Yanomamo (Case Studies in Cultural

Anthropology), Wadsworth Publishing; 5 edition
Conklin, Beth 2001. Consuming Grief Compassionate Cannibalism in an
Amazonian Society University of Texas.
Margaret Mead. Blackberry Winter: My Earlier Years, Kodansha America;
Reprint edition (April, 1995)
Linguistic Fieldwork Daniel L. Everett
page 273
Darnell, Regna. 1990. Edward Sapir: Linguist, Anthropologist, Humanist, Univ of
California Pr (February, 1990)

Darnell, Regna. And Along Came Boas: Continuity and Revolution in Americanist
Anthropology (Studies in the History of Liguistic Sciences, 86) John Benjamins Pub
Co (January, 1998)

References

Dixon, R.M.W. 1972. The Dyirbal language of North Queensland. Cambridge:

Cambridge University Press.
Durie, Mark. 1985. A grammar of Acehnese. Dordrecht: Foris.
-----. 1987. ‘Grammatical relations in Acehnese’. Studies in Language 11.365-399.
-----. 1988. ‘The so-called passive of Acehnese’. Language 64.104-113.
Lambrecht, Knud. 1986. Topic, focus and the grammar of spoken French. Unpublished
Ph.D. dissertation, UC Berkeley.
-----. 1994. Information structure and sentence form. Cambridge: Cambridge
University Press.
Rögnvaldsson, Eiríkur. 1982. We need (some kind of a) rule of conjunction reduction.
Linguistic Inquiry 13.3:557-61.
Van Valin, Robert D., Jr. 1999. A typology of the interaction of focus structure and
syntax. In E. Raxilina & J. Testelec, eds., Typology and linguistic theory: from
description to explanation, 511-24. Languages of Russian Culture: Moscow.
[available on RRG web site]
Zaenen, Annie, Joan Maling and Hölskuldur Thráinsson. 1985. ‘Case and
Grammatical Functions: The Icelandic Passive’. Natural Language and
Linguistic Theory 3.441-483.

RRG web site: linguistics.buffalo.edu/research/rrg.html

Blair, Frank. 1991. Survey on a Shoestring: A Manual for Small-Scale Language

Survey, Dallas, SIL.

Wackernagel, Jacob (1892). 'Über ein Gesetz der indogermanischen Wortstellung',

Indogermanische Forschungen, 1: 333-436. Reprinted in vol. 1: 1-104 of id. (1953,
1979). Kleine Schriften, eds. Kurt Latte (vols. 1-2) and Bernhard Forssman (vol. 3).
Göttingen.
Linguistic Fieldwork Daniel L. Everett
page 274
Notes