Formulaic Language and Second Language

Speech Fluency
Also available from Continuum

Perspectives on Formulaic Language

David Wood

Idioms and Collocations

Christiane Fellbaum
Formulaic Language and
Second Language
Speech Fluency
Background, Evidence and
Classroom Applications

David Wood
Continuum International Publishing Group
The Tower Building 80 Maiden Lane
11 York Road Suite 704
London SE1 7NX New York, NY 10038

www.continuumbooks.com

© David Wood 2010

All rights reserved. No part of this publication may be reproduced or

transmitted in any form or by any means, electronic or mechanical, including
photocopying, recording, or any information storage or retrieval system,
without prior permission in writing from the publishers.

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.

ISBN: 978–1-4411–5819-2 (hardcover)

Library of Congress Cataloging-in-Publication Data

A catalog for this book is available from the Library of Congress.

Typeset by Newgen Imaging Systems Pvt Ltd, Chennai, India

Printed and bound in Great Britain by CPI Antony Rowe, Chippenham, Wiltshire
 Contents

Introduction 1

Part I: Background
Chapter 1: Fluency 9
Chapter 2: Formulaic Sequences 38
Chapter 3: Cognitive Processing 60
Chapter 4: Social and Cultural Factors 72

Part II: Evidence

Chapter 5: Design of the Study 87
Chapter 6: Quantitative Results 115
Chapter 7: Qualitative Results 147
Chapter 8: Conclusions of the Research Study 171

Part III: Applications

Chapter 9: Fluency and Formulaic Language in
the Classroom 183

References 224
Index 235
This page intentionally left blank
 Introduction

Second language (L2) speech fluency is a language performance phe-

nomenon which, while integral to effective communication and ability
to thrive in an L2 milieu, is not particularly well understood by the lan-
guage teaching profession. Unfortunately, many L2 learners grapple
with the effects of inadequate fluency long after completing basic L2
study. As well, L2 teachers and assessors tend to bypass efforts to facili-
tate development of fluency and focus instead on language accuracy and
a hope that input and practice will help learners to speak “more
smoothly.” This may be due to the fact that fluency is a challenging con-
struct whose psycholinguistic foundations and place in the language
curriculum have not been investigated or discussed fully. The present
volume is an investigation of the nature of fluency development in an
effort to further our understanding of this important element of L2
performance, so instrumental for effective communication, yet so mar-
ginalized in the language curriculum. More specifically, the background
literature reviewed here, along with the empirical study presented, clar-
ify the role of formulaic language and psycholinguistic processes in the
development of speech fluency in a second language. Implications for
classroom teaching practice are presented, as are specific activities and
sequences of tasks.
To date, research on fluency has largely focused on temporal variables
of speech, namely, speech rate, repairs, amount and frequency of hesita-
tion, location of pauses, and length of runs of fluent speech between
pauses (Raupach, 1980; Möhle, 1984; Lennon, 1990a, 1990b; Riggenbach,
1991; Freed, 1995; Towell, Hawkins & Bazergui, 1996; Hansen, Gardner
& Pollard, 1998). Some recent research indicates that a possible key to
speech fluency lies in the mastery of a repertoire of formulaic speech
units, multiword strings or frames which are retrieved from long-term
memory as if they were single words (Nattinger & DeCarrico, 1992;
2 Formulaic Language and L2 Speech Fluency

Towell, Hawkins & Bazergui, 1996; Chambers, 1998; Wray, 2002). These
multiword units include, among other categories, 2-word collocations
such as good time, or first step, phrasal verbs such as run into, or come across,
idioms, routine expressions with social pragmatic functions such as have a
good day or how are you, whole clauses, discourse markers such as on the
other hand or in summary, and frames with fillable lexical slots such as a
(year/day/week . . .) ago or a (one/two/three . . .) step process.
In cognitive theory, a number of key concepts help to explain how
formulaic sequences can be learned and processed in the mind so as to
facilitate fluent speech production. The following paragraphs outline
these concepts, and for a more complete discussion see Chapters Two
and Three.
A central distinction in cognitive theory is made between two types of
knowledge, declarative knowledge and procedural knowledge. Declarative
knowledge has been defined as knowledge of content and information,
or as knowledge of things or as what is known. On the other hand, pro-
cedural knowledge is defined as knowledge of how to do or perform
things, and it is the basis of skilled behavior. Connected to these two
distinct types of knowledge is the concept of automatization or procedural-
ization, a process by which declarative knowledge may become procedural
knowledge. In this process, declarative knowledge is transformed into
procedural knowledge so as to enhance and speed up skilled performance
and permit declarative knowledge of information and content to be used
rapidly and efficiently in performance of skilled behavior (Anderson,
1983, 1993; Levelt, 1989). Given that formulaic sequences are multi-
word units which appear to be dealt with cognitively as single words,
knowledge of formulaic sequences may be automatized as single words
are. This could allow expression to occur fluently under the constraints
of time which real-life speech entails.
Research into how human memory operates also has yielded informa-
tion which can help to inform research into mental processing of language.
The concepts of long-term memory, short-term memory, and working memory
are key in this. In dealing with language processing, long-term memory
is a storehouse of knowledge of all kinds about language, including
lexis, syntactic and morphological rules, semantic information, and so
on. For utterances to be produced, however, items must be extracted
from long-term memory to express the concepts required. According to
Levelt’s (1989) lexically driven model of speech production, lexical
 Introduction 3

items must be retrieved relevant to the intended expression, and the

semantic, morphological, and syntactic rules applied to them, after
which phonological rules need to be employed to create the utterance.
This seemingly laborious process requires the use of short-term mem-
ory. However, short-term memory capacity in humans is restricted to
holding at most seven or eight discrete items at a time, which would
seem to make the construction of complex or lengthy sentences and
utterances time consuming and almost impossible (Anderson, 1983).
Formulaic sequences retrieved from long-term memory as wholes can
bypass the storage limit of short-term memory, allowing many words to
be held in short-term memory rather than merely seven or eight single
words at a time. Working memory (Baddeley, 1986; Gathercole & Baddeley,
1993), is proposed as the site of rehearsal, practice, and assembly of
language sequences. Working memory may be key to automatization of
language, specifically formulaic sequences, according to Anderson
(1993). Anderson’s model of automatization involves three types of
memory, declarative, procedural, and working memory. According to
Anderson (1993), automatization may develop over time as chunks of
language are separately taken from long-term memory and assembled
in working memory. Robinson (1995) outlines a similar process involving
novel language to which learners are exposed, by which language input is
converted into intake, or becomes part of language competence, as
chunks of input are rehearsed in working memory. With repeated prac-
tice of this, the chunks may become part of the production rule and
bypass declarative memory, or they may be retrieved directly from
declarative memory without need for working memory.
The distinction between controlled and automatic processing is another
important part of discussions about language processing. Controlled
processing is a conscious, effortful process of selecting and organizing
or manipulating language knowledge, using consciously learned rules
and parameters. Word-by-word construction of a novel utterance under
controlled processing would require following the steps outlined above
in extracting lexical items from long-term memory and applying rules
to them using short-term or working memory to create utterances.
Conversely, automatic processing is a result of automatization of such
knowledge, as well as automatic access to certain elements under given
conditions. It is a fast, virtually effortless process which occurs below the
level of the speaker’s conscious awareness (McLaughlin, Rossman, &
4 Formulaic Language and L2 Speech Fluency

McLeod, 1983). Later models of controlled and automatic processing

tend to see the two as more of a continuum than as two discrete systems,
and that language production processes may be strongly, partially, or
occasionally automatic or controlled, depending on aspects of the pro-
duction situation (DeKeyser, 2001). On the other hand, memory-based
models of retrieval and production such as Logan’s (1988) instance
theory hold that single-step memory retrieval is key. In such models,
what appears to be automatization of a previously controlled process is
in fact the replacement of a controlled process by holistic retrieval
from long-term memory. As instances of processing and retrieval
accumulate in memory over time, memory itself takes over from the
processing task and a conscious, laborious, and time-consuming task can
be accomplished in a single step of remembering. In explaining how
speech fluency occurs, it appears that some sort of automatic processing
which involves the use of formulaic sequences may be essential to fluent
speech production.
Cognitive theory can help explain how proceduralization or automati-
zation happens. In information processing theory, the power of language
input to drive the acquisition process is often emphasized. In connec-
tionist models of human cognition, frequency of experience with stimuli
such as language sequences increases the strength of mental connec-
tions between relevant features of the input and the categories to which
they belong (Ellis, 2002; Rumelhart, Hinton, & McClelland, 1996). Thus,
as one becomes more fluent in a language with increased exposure and
experience with language input, one develops an implicit statistical
awareness of the behavior of elements of language such as lexical items,
formulaic sequences, phonology, morphology, syntax, as well as discourse
aspects. According to Ellis (2002: 144), frequency of experience with
language input is the key determinant of acquisition because structural
regularities such as lexical choices, including formulaic sequences, and
rules of language emerge from a learner’s long-term analysis of the
distributional characteristics of language input.
Also important in discussions of the power of input frequency are
consciousness and awareness (Schmidt, 1990, 1993). Schmidt proposes
that noticing or awareness of aspects of input is key to acquisition, and
that noticing formal or lexical features of input causes input to be
converted to intake. It is possible that noticing or awareness of aspects
of input such as formulaic sequences, like single words, is the first
 Introduction 5

step in a process of acquisition based on frequency and strengthening

connections between elements of language and the categories to which
they belong. In short, it may be that language acquisition and automatiza-
tion are largely driven by frequency of input and noticing and awareness
of how words and patterns and sequences are used.
In fluent speech production, lexical items, including multiword for-
mulaic sequences, are likely automatized, meaning they are retrieved
from long-term memory so as to be used with less need for conscious
effort or control. Short-term or working memory use is reduced or the
lexical units processed therein are larger and multiword. Research on
formulaic language units, which are increasingly recognized as consti-
tuting a large part, perhaps most of everyday spontaneous speech, has
rich potential for helping to explain how spontaneous speech can occur
under the heavy processing and time constraints of real-life discourse
(Weinert, 1995; Miller & Weinert, 1998; Skehan 1998; Wray & Perkins,
2000). It appears that a great part of everyday spontaneous speech is
formulaic (Pawley & Syder, 1983).
A gap exists within the current body of knowledge about fluency between
the work focusing on temporal correlates of fluency and that which
describes formulaic sequences and infers their role. Part II of this volume
reports on an empirical investigation into how formula automatization
might contribute to fluency.
The study investigates fluency development and use of formulaic
language by linking developments in the temporal aspects of learner
speech with learners’ use of formulas. Speech samples from English L2
learners were collected over a 24-week period and were analyzed to
determine their fluency growth as measured by changes in key temporal
variables such as speech rate, hesitation phenomena, and mean length
of runs. The speech samples were elicited by means of short silent
animated films which the learners viewed, then recounted in English.
Their use of automatized formulas was studied, as it related to changes
in temporal variables. Statistical analysis for each temporal variable was
conducted to determine significant development of learners’ fluency.
Fluency development was examined in relation to increased use of
formulaic language. Comparison of the speech samples for the same
film prompt at two different points in time yielded comparable exem-
plars of speech situations in which the use of formulas could be related
to increased fluency.
6 Formulaic Language and L2 Speech Fluency

Part III of the book presents principles of a possible pedagogy

of second language fluency, based on the background and evidence
covered in Parts I and II. It also introduces types of activities and
a course or workshop sequence of tasks drawing on knowledge of
formulaic sequences and cognitive processing. Some pedagogical work
has concentrated on formulaic sequences and overall proficiency or
English for specific or academic purposes (Willis, 1990; Lewis, 1997;
Mudraya, 2005). However, there are very few such plans which concen-
trate on the facilitation of second language fluency, with the exception
of Boers et al., (2006), who found a fluency benefit from encouraging
the noticing of formulaic sequences.
 Part I

Background
This page intentionally left blank
 Chapter 1

Fluency

An overview and synthesis of research on fluency in second language

speech is first presented in order to define the construct and determine
what elements of language proficiency are most relevant to fluent speech
production. This research falls into four large categories: temporal
variables of speech; the nature and functions of formulaic language;
automatization and mental processing in language production; social
and cultural aspects of fluency.

Defining Fluency

Fluency is a term frequently used to describe oral language performance,

yet it is often only vaguely defined, if at all, and may be used as a substi-
tute for a cluster of aspects of proficiency. Despite this lack of precision
in use, the construct of fluency has been defined and studied in the
language learning field, and has been a subject of theoretical and
empirical inquiry. The body of literature concerning fluency has
identified key temporal variables of speech which can be linked to psy-
cholinguistic aspects of performance and production. It has become
increasingly clear over the past 20 years that spoken fluency in a second
language is largely a function of a speaker’s pauses and hesitations both
in temporal terms and in terms of their appropriate links with discourse
pragmatics and structure.
In general parlance, fluency is often used as a synonym for effective
spoken use of a language. It is frequently used to mean “native-like,”
having a high overall degree of proficiency, or having a “good com-
mand” of a language. In the language teaching profession, fluency is
generally more tightly defined. We tend to use the word to mean a
naturalness of flow of speech, or speed of oral performance. Explorations
10 Formulaic Language and L2 Speech Fluency

of fluency have ranged from studies of speech and thinking processes

(Chafe, 1980), to studies of production and temporal aspects of speech
(Dechert, 1980; Möhle, 1984; Raupach, 1980, 1984). As well, in the second
language literature, fluency and accuracy are often distinguished
(Schmidt, 1992), and fluent speech is sometimes seen as not necessarily
being representative of good linguistic competence (Sajavaara, 1987).
Indeed, the development of the construct of communicative competence,
integral to the evolution of teaching methodology through, in part, the
communicative approach, has done much to further awareness of
aspects of language performance beyond the merely grammatical.
Bachman (1990), in a highly developed model of communicative
competence and language proficiency drawing on earlier work by oth-
ers, notably Canale and Swain (1980), gives equal weight to pragmatic/
sociolinguistic and strategic aspects of competence as to grammatical
and textual aspects, expanding the notion of communicative competence
to include nonlinguistic realms of knowledge and skill which contribute
to spoken language performance. The concept of fluency has likewise
developed to include elements of performance beyond just accuracy of
syntax, lexis, and phonology. Based on a body of work in recent decades,
discourse coherence and cohesion, conversational pragmatics, sensitivity
to register and idiom, and communication strategies are all considered
integral parts of overall proficiency and, by association, fluency. It
appears that fluency is influenced by a broad range of competencies.
Just as classroom teachers need a clear concept of fluency, so do
those involved in the assessment of spoken language. Research on
proficiency testing has grappled with identifying fluency, and the
growth of performance-based testing over the years has necessitated a
clearer definition of oral fluency. If oral performance is to be assessed,
fluency is a key area of ability which needs to be pinned down to specific
proficiency criteria and levels. The Government of Canada Second
Language Oral Interview assessment (1983), using criteria largely
based on Foreign Service Institute descriptions, lumps fluency together
with accuracy in the descriptors of proficiency levels, together with
pronunciation, grammar, and vocabulary. The descriptors dealing with
proficiency levels referred to speed, hesitation, and “groping for words”
(p. 25). Fluency came to be defined more fully over time. Fulcher (1996)
refers to recent fluency scales which attempt to identify six levels, dis-
criminating among number and frequency of pauses in speech, and,
 Fluency 11

more importantly, positioning and perceived purpose of pauses (Fulcher,

1996: 50, 51). As shall be seen later, this refined scale of fluency levels is
evidence of how far we have come in our understanding of what fluency
is and how temporal variables in second language speech are seen to
link with underlying psycholinguistic mechanisms and elements of
competence.
Defining fluency as distinct from other aspects of oral proficiency repeat-
edly leads us back to temporal variables in speech, such as speed, pauses,
hesitations, and fillers. It is speed and pauses which receive the greatest
attention in the general literature on fluency. Lennon (1990b: 391)
states that fluency is distinct from other aspects of proficiency precisely
because it is a purely performance-based variable:

Fluency differs from the other elements of oral proficiency in one

important respect: whereas such elements as idiomaticness, appropri-
ateness, lexical range, and syntactic complexity can all be assigned to
linguistic knowledge, fluency is purely a performance phenomenon,
there is presumably no fluency “store.” Rather, fluency is an impression
on the listener’s part that the psycholinguistic processes of speech
planning and speech production are functioning easily and efficiently.
Dysfluency markers, as it were, make the listener aware of the produc-
tion process under strain. (Lennon, 1990b: 391)

While Lennon’s characterization of fluency as a strictly performance-

based phenomenon does not account for apparent disfluency, which
is often due to the pressures of complex thinking and decision-making
about what to say in certain situations, or variable individual fluency due
to task pressures or different knowledge domains or genres, it does shed
light on important aspects of fluency as it has been studied in empirical
research. This purely performance-based view of fluency allows mea-
surement through marking of temporal variables, but it is difficult to
analyze or explain since it has no direct link to a body of knowledge or
set of rules in language competence.
In fact, fluency has been defined as “automatic procedural skill”
(Schmidt, 1992: 358–359), meaning that it is speech produced at a speed
and ease which requires little or no effort or attention. It is characterized
by the fact that “the psycholinguistic processes of speech planning and
production are functioning easily and efficiently” (Lennon, 1990b: 391).
12 Formulaic Language and L2 Speech Fluency

Schmidt (1992: 358–359) goes on to state that fluency depends on proce-

dural knowledge, or knowledge of how to do something, not declarative
knowledge, or knowledge about something. Schmidt makes a further
distinction between procedural knowledge and procedural skill, plac-
ing fluent speech in the latter category, since it clearly has more to do
with the actual performance of something in real time than the knowl-
edge of how to do it (359). Rehbein (1987) would seem to agree in part
with Schmidt, asserting that “fluency of speech means that uttering and
planning/controlling are to be executed at least partly simultaneously”
(Rehbein, 1987: 99). A similar concept is treated by Levelt (1989), whose
model of speech production requires that various processes involved in
speech work more or less in parallel with each other. But before examin-
ing the psycholinguistic definitions of fluency and speech production, it
is important to review the empirical work on fluency and look at the
variables studied, particularly those which appear to correlate the most
strongly with fluency.

Temporal Variables

Empirical research focusing on fluency has generally involved the elici-

tation of a speech corpus and analysis of its temporal and qualitative
aspects. Some studies have attempted to link clusters of performance
variables with rater assessments of fluency (Lennon, 1990; Riggenbach,
1991; Freed, 1995); others have compared first and second language
speech performance (Deschamps, 1980; Raupach, 1980), or conducted
longitudinal examinations of the development of aspects of second
language spoken fluency (Dechert, 1980; Towell, 1987; Lennon, 1990a;
Hansen, Gardner, & Pollard, 1998). Across all of the studies of spoken
fluency and its development, there has been a remarkable degree of
agreement on the types of temporal variables to be tracked.
Over the several decades during which oral fluency has been studied
empirically, there has been a growth and development in the corpus
size of data analyzed, and more varied sources of oral performance
under different discourse conditions, but the variables have remained
largely consistent. The earliest such studies involved having one or two
learners retell stories, and the results were analyzed for a limited set of
performance variables, such as speech rate and the number and dura-
tion of pauses. Prime foci in these studies were the comparison of first
 Fluency 13

and second language speech (Deschamps, 1980; Raupach, 1980), and

the comparison of speech performance in a second language before
and after a stay abroad (Dechert, 1980). Later research continued the
story retell procedure and often used limited numbers of subjects, but
included interviews as well. Elicited speech came to be analyzed in
greater depth in these studies; while speech rate and pause features
continued to be the temporal variables most quantified, qualitative inter-
pretation of the data focused on possible speech planning strategies
(Raupach, 1984) and the use of formulaic speech and its relation to
underlying planning procedures (Dechert, 1984; Lennon, 1984a). In
other studies from the mid-to-late 1980s, researchers analyzed speech
using the same types of temporal variables, but interpreted the results
with reference to transfer of production strategies from the first lan-
guage (Möhle, 1984; Faerch & Kasper, 1989; Möhle & Raupach, 1989).
In the 1990s, the temporal variables which were used to analyze the
corpus of elicited speech were retained, but there was a greater variety in
the purposes and construction of the studies. During this time, cognitive
theory began to influence the studies. For example, introspective studies
of immersion language learners attempted to determine the extent of
improvement in automatization of lexical items including formulaic
sequences (Lennon, 1989), and elicitation tasks included story retell and
informal conversation to determine the effect of task on performance
(Lennon, 1990b). Some research used native-speaker judges to rate
the fluency level of second language speakers, then compared the rat-
ings to microanalyses of the temporal variables in the speech samples
(Riggenbach, 1991; Freed, 1995).
Some research has been conducted to investigate specifically whether
or how fluency gain occurs in study abroad contexts, notably influential
studies by Freed (1995), Riggenbach (1991), and Segalowitz and Freed
(2004). Key studies examining the fluency gain of learners in study
abroad contexts has established that time spent studying an L2 in a tar-
get language milieu has a positive effect, although most of the studies
involve the learning of languages other than English. Riggenbach’s 1991
longitudinal multivariate study of fluency development in Chinese EFL
(English as a Foreign Language) students found that key temporal
aspects of fluency correlated with perceived fluency as rated by
native-speaker judges. Freed’s 1995 study of the fluency effect of a
term abroad for American students of French found that the major
14 Formulaic Language and L2 Speech Fluency

temporal aspects of fluency all showed stronger improvement when the

term-abroad group was compared to a control group who remained in
the United States.
Segalowitz and Freed (2004) compared at home and study abroad
groups in Spanish as a second language and found that the study abroad
students made greater gains in the temporal aspects of speech fluency
but that amount of language contact, initial proficiency, and cognitive
abilities played vital roles as well. Evidence also exists that study abroad
experience in Spanish as a second language may facilitate greater gains
in the ability to tell extended narratives and produce semantically dense
language (Collentine, 2004). It is possible that increased use of formulaic
sequences may have played a part in the improved narrative productions
and semantic density of the study abroad group as well.
Other research compared the speech performance on a range of
tasks of learners, native speakers, and attriters of a language (Hansen,
Gardner, & Pollard, 1998), in an effort to determine how fluency
develops or can diminish over time under various circumstances.
Examining the variables used in the key studies will help to clarify
what constitutes an empirical correlate of fluency, and which of these
variables seem most relevant to the study of fluency development.
Möhle (1984), in a study of comparison of the spoken texts of
German students of French and French students of German, analyzed
the texts using four temporal variables which are now common in the
literature:

1) Speech and articulation rate, i.e. the average number of syllables spoken
per minute during recording time and the average number of sylla-
bles articulated per second during net speaking time, excluding
silent pauses.
2) length and position of silent pauses
3) length and quality of speech units, i.e. the string of syllables or words
between two silent pauses
4) number, type, and position of hesitation phenomena in the text, such as filled
pauses (euh, hmm, and other articulations used to bridge the gaps
between pauses), drawls, repetitions, and self-corrections.
(Möhle, 1984: 27)

In a longitudinal study of the fluency development of a British

student of French, Towell (1987) used a set of four temporal variables
 Fluency 15

related to those of Möhle:

1) Speaking rate: normally expressed in syllables per minute and calcu-

lated by dividing the total number of syllables produced by the total
time taken to produce the utterance (including pause time) and
multiplying the result by sixty. It allows comparisons of how much
time it takes to formulate and produce the speech.
2) Articulation rate: normally expressed in syllables per second and calcu-
lated by dividing the total number of syllables produced by the
amount of time taken to produce them (not including pause time).
This measure gives an indication of the speed at which speech is
being produced.
3) Pause/time ratio: gives the percentage of time spent speaking as a pro-
portion of the whole.
4) Lengths of runs between unfilled pauses: this is the mean (or median) num-
ber of syllables between unfilled pauses of no less than 0.2 seconds.
(Towell, 1987: 123, 124)

In his study of German students of English in the United Kingdom,

Lennon (1990b) took a qualitative approach to the study of fluency,
using ten native speakers to rate the fluency of the subjects and analyz-
ing the texts for markers of fluency. At the beginning of a 6-month stay,
then again at the end, subjects were asked to retell a picture story. Judges
made global decisions on their fluency according to 12 criteria:

1) words per minute unpruned

2) words per minute pruned.
“Pruned” in this study refers to the deletion of words which were
repeated, were asides, addressed to the interviewer, or self-corrections.
3) repetitions per t-unit
4) self-corrections per t-unit
5) filled pauses per t-unit
T-units were defined as “one main clause and all its attendant subordi-
nate clauses and non-clausal units.”
6) % of repeated and self-corrected words
7) total unfilled pause time
8) total filled pause time as % of total
16 Formulaic Language and L2 Speech Fluency

Filled pauses were non-words, three vocalizations represented as “er,”

“erm,” and “mm.”
9) length of runs between pauses
10) % of t-units followed by a pause
11) % of total pause time at t-unit boundaries
12) mean pause time at t-unit boundaries.
(Lennon, 1990b: 406)

A large multivariate study of fluency development was conducted by

Riggenbach (1991) with six Chinese students of English. In this case, a
taped dialogue was analyzed for five fluency-related variables, and judged
by native speakers for global fluency. Riggenbach’s selection of fluency vari-
ables differed somewhat from those in the previously described studies:

1. hesitation phenomena
– hesitation
– unfilled pause
– filled pause
2. repair phenomena
– retraced restart (including part of the original utterance)
– unretraced restart (original utterance is rejected)
3. rate and amount of speech
– rate (syllables uttered per minute)
– amount (total number of syllables produced)
– percent of speech native speaker/nonnative speaker
– percent of turns native speaker/nonnative speaker
4. interactive phenomena (by nonnative speaker)
– backchannels (encouragement)
– echo (repetition)
– questions
– repair initiation
– laughter particles
5. interactive features
– latch (speaking immediately upon the end of previous turn)
– overlap
 Fluency 17

– gap
– collaborative completion (one speaker tries to complete another’s
sentence).

In a similar study, Freed (1995) examined the speech of 30 native

English-speaking students of French, and had native speakers globally
rate their fluency. Half her subjects spent one semester abroad in
France, while the other half continued French study at home in the
United States. In the analysis of fluency-related features she focused on
seven points:

1. amount of speech
2. rate of speech
3. unfilled pauses
4. filled pauses
5. length of fluent runs between pauses
6. repairs
7. clusters of disfluencies.

In looking at empirical correlates of fluency across recent studies, it is

obvious that researchers have had a high degree of agreement on the
relevant temporal variables. Amount, rate of speech, repairs, pauses,
and the length of runs between pauses are common elements of analysis
in such studies. These are discussed in turn below.

Amount of Speech

Data elicited in empirically based fluency research relating to increases

in the amount of speech produced has generally not been shown to be
related to other measures of fluency. Only a few studies have included
measures of amount of speech produced as part of a list of features to
quantify.
Riggenbach (1991) included a raw measurement of the amount of
speech produced by six Chinese university students of English in the
United States. The subjects participated in a taped dialogue with a native
speaker, and a 5-minute segment of the corpus was analyzed for fluency
markers: hesitation phenomena, repair phenomena, rate and amount
18 Formulaic Language and L2 Speech Fluency

of speech, interactive phenomena and interactive features. The samples

were also evaluated for global fluency by a panel of native-speaker
judges. All semantic units were counted, including filled pauses and
partial words. Quantitative analysis showed virtually no significant dif-
ference between amount of speech produced between subjects rated as
highly fluent and those rated as having low fluency, despite the fact that
other fluency markers did show a difference.
Freed (1995) also looked at this variable with her 30 American students
of French. Half the group spent a term in France, while the other half
stayed home. The speech of the two groups was compared on an oral
interview based on the American Council for the Teaching of Foreign
Languages/Inter-language Round Table (ACTFL/ILR) Oral Proficiency
Interview; a microanalysis of fluency was conducted and a panel of
native-speaker judges evaluated their fluency. One of Freed’s measures
of fluency was amount of speech, calculated as frequencies of non-
repeated words or semantic units. Amount of speech turned out not
to show any statistical significance in discriminating among levels of
fluency, although other temporal variables did show statistically signifi-
cant differences.

Rate of Speech

Many more studies have focused on the speed or rate of speech as a

feature of fluency. Measured as syllables uttered per minute or second,
speech rates tend to increase over time along with certain other
measures of learner fluency or to correlate with judges’ perceptions
of fluency. It is commonly believed that perceived fluency has to do
with increased speed of speech, and the empirical studies generally tend
to show that this is true. Measures of speech rate and articulation are
relatively easy to link to fluency.
Möhle’s 1984 study comparing the fluency of French and German
language students was influential in its selection of temporal variables
to be studied, among which speech and articulation rate featured large.
Möhle calculated speech rate as the average number of syllables uttered
per minute, and articulation rate as number of syllables produced per
minute excluding silent pauses. She found that both speed indicators
showed increases for German students of French, but that for French
speakers of German as a second language, speech complexity, rather
 Fluency 19

than speed, increased. Clearly, then, features such as speech rate tell
only a part of the fluency story.
Towell (1987), in his longitudinal, 4-year study of fluency development
of a student of French, found a significant improvement in both speak-
ing rate and articulation rate over time. In this study, speaking rate was
calculated as syllables per minute, indicating how long it took to formu-
late and produce speech. Articulation rate was calculated as syllables
per minute excluding silent pauses, and was intended as a measure of
the speed of actual production. Over the 4 years, the subject increased
her speaking rate by 65%, and her articulation rate by 20%.
Lennon (1990a, 1990b), in his longitudinal study of fluency develop-
ment in four German students of English, measured words per minute
produced in tasks involving text retells and conversation. The data showed
that speed of delivery was rather stagnant over the 23 weeks of the study.
Interestingly, however, Lennon found that the speech of the subjects
increased in complexity, and he attributes lack of speed increase to the
greater processing burden of producing longer clauses and t-units.
Riggenbach’s 1991 multivariate study of fluency in Chinese ESL
(English as a Second Language) students found that words per minute
(semantic units) correlated with perceived fluency as rated by native
speaker judges. Of the six subjects, words per minute ranged from a
low of 102 per minute for the lowest rated, to 253 per minute for the
highest rated.
Freed’s 1995 study of the fluency effect of a term abroad for American
students of French found that rate of speech was the only fluency
measure to show a significant difference when the term-abroad group
was compared to a control group who remained in the United States.
Calculating speed as the number of nonrepeated words or semantic
units per minute of speech, Freed found that “not only was this the most
salient of all the factors of fluency analyzed, it was also the most striking
in listening to the differences in the pre and post samples of the Abroad
students” (Freed, 1995: 137). The average number of words per minute
for the “at home” group in her study was 83.7, while that for the “abroad”
group was 115.2.
Using measures of speaking and articulation rate similar to those of
Möhle (1984) as described above, Towell, Hawkins, and Bazergui (1996)
analyzed speech samples of 12 French students of English before and
after a period of residence in Britain. The group as a whole increased in
20 Formulaic Language and L2 Speech Fluency

average speech rate from 136.61 syllables per minute to 156.88 syllables
per minute. Articulation rate also increased, on average, from 3.85 syllables
per second to 4.17 per second. Clearly, speech and articulation rates
were important markers of fluency development in this study.
In an ambitious study measuring fluency acquisition and attrition in
speakers and learners of Japanese, Hansen, Gardner, and Pollard (1998)
tracked words per minute as a measure of speech rate related to fluency.
It was discovered that the mean number of words per minute increased
over time for learners of Japanese. The reverse time effect was just
as significant for attriters of Japanese, who had left Japan and were
living in English in the United States without contact with Japanese.
In story-retelling tasks, the attriters produced fewer words per minute
over time.
From this survey of empirical research on speech rate as a marker
of fluency, it appears that it is a fairly sound indicator. In most of the
studies, speech and articulation rates increased with overall fluency
or correlated well with evaluations of fluency, time spent learning
the language, or composite measures of overall fluency. In Lennon’s
longitudinal studies (1990) speech rate failed to show significant
improvement, but complexity of speech did improve. Similarly, Möhle
(1984) found fluency improvements linked to speed increases in some
groups of learners, but that complexity of language was a stronger
fluency indicator for others. This is an important point to which we
will return later in the discussion. It seems that speed gives us little infor-
mation about the workings of fluency unless it is viewed in interaction
with certain other variables.

Repair Phenomena

In the fluency literature, many studies have tracked how learners self-
correct and repeat or restart utterances. There is a focus on such repair
phenomena in the empirical data from research on fluency, and the
results are mixed.
Möhle (1984) paid some attention to repair phenomena in her influen-
tial study of German and French learners, noting briefly in her results
that as other measures of fluency showed improvement, there were also
“fewer linguistic signs which repeat or correct words or syllables already
articulated” (Möhle, 1984: 44).
 Fluency 21

Lennon’s 1990 study of the speech performance of four Germans

studying English in Britain looked at repetitions per t-unit, self-corrections
per t-unit, and percent of repeated and self-corrected words over 23 weeks.
Lennon found that the learners generally had fewer repetitions per
t-unit as time passed, but that the number of self-corrections stayed
fairly constant as did the overall percent of repeated and self-corrected
words. Lennon concludes that this self-correction provides some useful
information about fluency:

. . . certain sorts of self-correction, particularly those involving refor-

mulation of discourse, do not make for perceived disfluency at all, and
are a feature of certain sorts of fluent native-speaker performance.
Indeed, that 3 of the 4 subjects had increased self-corrections per t-unit
at week 23 may even suggest that part of fluency development in the
advanced learner may involve increased ability to reformulate, moni-
tor, and self-correct production on-line. (Lennon, 1990b: 412, 413)

Other studies have drawn less positive conclusions about repairs and
fluency indicators.
Riggenbach (1991) included an analysis of repair phenomena in
her study of fluency development in Chinese learners of English.
Riggenbach’s measures of repair included counting retraced restarts
and unretraced restarts:

retraced restart—reformulations in which part of the original utterance is

repeated
– repetition—exact adjacent repeats of sounds, syllables, words or
phrases
– insertion—a retraced restart in which new unretraced lexical items
are added
unretraced restart—reformulations in which the original utterance is
rejected (= “false start”).
(Riggenbach, 1991: 427)

The results showed a complicated pattern of repairs when comparing

learners rated as highly fluent to those rated as less fluent. The subjects
at the low end of the fluency scale in Riggenbach’s group seemed to be
rated less fluent for various reasons. One learner repaired frequently,
22 Formulaic Language and L2 Speech Fluency

but overmonitored and was rated as having low fluency due to other
temporal features. Another low fluency subject rarely repaired and
produced flowing speech, but had so many grammatical and syntactic
inaccuracies, she was rated low by judges. From Riggenbach’s research,
we are left wondering how repair phenomena can link to fluency.
Freed (1995) also studied repair phenomena in her research on
30 American French students. In this research, four repair features
were tracked:

repetitions of exact words, syllables, or phrases . . .

reformulations/false starts—repairs which suggest a decision to rephrase
because the speaker perceives that the best form for the intended
meaning was not originally selected . . .
corrections/grammatical repair—specific correction of a structural feature
...
partial repeat: realized as a search for the appropriate form but distinct
from a part of a false start . . .
(Freed, 1995: 130)

The results indicated that the presence of repairs in non-native speech

was not particularly indicative of a lack of fluency. Generally, the group
rated by judges as highly fluent, the “Abroad” group, displayed more
repair markers in their speech than did the “at home” group. Reformu-
lations and repetitions were the two repair markers the fluent group
used more (Freed, 1995: 140). Freed states that this may be a sign of
greater linguistic sophistication:

There is a tendency for students who have been abroad, especially

those whose speech is more advanced, to attempt linguistic expres-
sions which they sometimes find don’t work: they reformulate their
speech producing more false starts than is evidenced in the speech of
those who have never been abroad. (Freed, 1995: 142)

It is possible that the enriched input to which such learners have been
exposed, in addition to the varied interaction with native speakers, has
given them a broader repertoire of language which will manifest itself at
varying degrees of accuracy in speech. So far, it seems that repair phe-
nomena in second language speech provide mixed and inconclusive
 Fluency 23

information about competence and performance. It appears that while

repair phenomena may have something to tell us in qualitative terms
about how fluency develops or occurs, repairs are only weakly linked in
the literature with overall development of fluency.

Pause Phenomena

The most complex and one of the most informative elements of fluency
studied so far in empirical research involves pause phenomena. There
are two aspects of pauses which have been studied, namely, frequency
and location. It can be seen from a survey of the research that certain
elements of pauses, particularly where they occur, can provide us with a
great deal of information about the nature of fluency.

Pause times and frequencies

Comparisons of the pause times and frequencies in first as opposed to
second language speech have yielded some relevant results. Möhle
(1984) looked at the length and number of silent and filled pauses in a
study of French and German second language learners in cartoon
descriptions and interviews. She found differences between the first and
second language performance of the subjects in the number of pauses,
but no great difference as regards length of pauses. Lennon (1984) had
12 German students of English retell a story from listening. Their
performance was compared to a native-speaker model, and it was
found that there was a higher ratio of pause time to speech time in the
performance of the second language speakers, due to more frequent
and longer pauses.
Towell (1987) tracked pauses in the speech performance of a British
learner of French over a 4-year period. The only measure of pausing in
the study was a calculation of total pause time as a ratio of total speaking
time. The ratio improved 37% over the course of the study, meaning
that total pause time reduced in relation to total speaking time.
Lennon (1990a) studied the pause time of the four German English
students in his longitudinal research. He found that total unfilled
pause time as a percentage of total speech decreased by an average
of 25% in three of four subjects. Total filled pause time showed no
significant change.
24 Formulaic Language and L2 Speech Fluency

Riggenbach (1991) used four categories of pauses in her study of flu-

ency development in four Chinese learners of English:

Micropause—a silence of .2 seconds or less

hesitation—a silence of .5 seconds or greater
unfilled pause—a silence of .5 seconds or greater
filled pause—voiced “fillers” which do not normally contribute additional
lexical information
a) nonlexical . . . fillers that are not recognized as words and that
contain little or no semantic information
b) sound stretches—vowel elongations of .3 seconds or greater . . .
c) lexical . . . fillers that are recognized as words but in context con-
tribute little or no semantic information.
(Riggenbach, 1991: 426)

The results indicated that unfilled pause frequency was an important

discriminator between subjects rated as highly fluent and those rated as
less fluent. In general, pause frequencies in total were higher for the
less fluent subjects.
Freed (1995) also tracked unfilled and filled pauses in the speech of
her American learners of French, comparing learners who spent a term
abroad with those who stayed in the United States. She measured
unfilled pauses not occurring at clause junctures, and filled pauses
including “drawls” and sound stretches. She found some differences
between the fluent group and the less fluent group, in that the fluent
learners had generally shorter and fewer silent pauses and filled pauses,
although the total percent of speech time taken up by pauses was similar
for both groups.
Hansen, Gardner, and Pollard (1998), in their study of acquisition
and attrition of Japanese among English-speaking learners and those
who left Japan to live in English in the United States, found that the
length of pauses was longer for students in early years of acquisition
than in later years. It was also found that attriters paused longer in their
Japanese speech the longer they lived in English outside of Japan.
The findings of previous research about the importance of pause times
and frequencies tells us a great deal about speech fluency, particularly that
related to the value of unfilled pauses. It appears that analyzing filled
pauses yields mixed and inconclusive results.
 Fluency 25

So what does the information about pause times and frequencies tell
us about fluency? It appears that length and frequency of pauses, be
they filled or unfilled, is of some significance to fluency. These results
do not, however, inform us about how fluency works or how it relates to
psycholinguistic mechanisms of production. The empirical research on
the positioning of pauses in speech is more important in this regard. To
appreciate the significance of the placement of pauses, it is useful to
combine a survey of empirical evidence with some explanations of why
pauses occur where they do in fluent and nonfluent speech.

Pause location
Dechert (1980), in a study analyzing the speech performance of a
German student of English who retold a story in English before and
after a stay in the United States, found some importance in pause
location. He noted that the second speech sample showed that pauses
tended to be located at breaks corresponding to what are termed
“episodic units,” or before and after segments of a story which have
specific narrative functions such as establishing setting, location, reac-
tion, attempts, and so on. The more fluent second recording displayed
more pauses at these junctures and fewer within the episodic units.
Dechert notes that the subject was able to use the structure of the
narrative to provide himself with natural breaks in which to search for
words, phrases, and so on (Dechert, 1980: 274).
Lennon (1984), in a comparison of second language learners’
retelling of a story after listening to a native-speaker model, found
significant differences in pause distribution between first and second
language narration. In the model narration, 100% of the pauses
occurred at clause breaks or after nonintegral components of the clause,
with no pauses within clauses. The second language narrators, however,
showed different patterns, pausing frequently within clauses. Lennon
concludes that they are “planning within clauses as well as in supra-
clausal units” (Lennon, 1984: 61). It seems, then, that locating pauses
within clauses and not at clause junctures is a discriminator between
fluent and nonfluent speech.
A similar finding was reported by Deschamps (1980), in a compari-
son of students’ performance in their first language, French, and in
English, their second language. It was found that the second language
26 Formulaic Language and L2 Speech Fluency

speech showed more pauses within sentences, and even within verbal
phrases:

A comparison between French L1 and English L2 shows an important

decrease in the number of pauses at “end of S” for English L2, and a
corresponding increase in “inside S” . . . hesitation phenomena marked
by an increase in the number of pauses at non-grammatical junctures.
(Deschamps, 1980: 261)

Deschamps finds not only an increase in pauses within sentences, but a

decrease in pauses at the end of sentences.
Riggenbach (1991) noted that the pause locations in the speech of her
six Chinese students of English gave some useful information about
fluency. She remarks that the location of pauses had a significant effect
on the perceived fluency of some subjects:

A number of the short pauses produced by high-group subjects . . .

were considered native-like, because they occurred at junctures and
often in isolation . . . for members of the low group, most of the
regular unfilled pauses . . . occur in chunks of disfluencies . . . it is
possible, therefore, that a high number of disfluency chunks in a
NNS’s (non-native speaker) speech may indicate low fluency . . .
whereas a low number of disfluency chunks may indicate high
fluency or proficiency. (Riggenbach, 1991: 431–432)

The location of pauses at clause junctures, and the avoidance of clus-

tered pauses (disfluency chunks) seems to play a role in perceived
fluency.
Freed (1995), in her study of American students of French, found a
similar force at work in the pause distribution of highly fluent speakers
as opposed to those rated less fluent. The presence of clusters of pauses
in the speech of less fluent subjects is significant:

. . . the presence of pauses is not exclusively associated with a lack of

fluency in a second language . . . the function of pauses in the context
of discourse is as important as their frequency . . . a single pause might
be accepted as normal or fluent . . . but clusters of such disfluencies
create other impressions. (Freed, 1995: 138–139)
 Fluency 27

The location and the clustering of pauses, then, are much stronger indi-
cators of relative fluency than the number or the duration of pauses.
One can conclude from the above that the location of pauses in
speech is an important indicator of fluency. The clustering of pauses is
a correlate of reduced fluency, and the syntactic location of pauses appears
salient as well. Highly fluent second language speakers and native speakers
tend to pause at sentence and clause junctures, or between nonintegral
components of clauses and clauses themselves. Pausing at other points
within sentences and clauses gives the impression of disfluency.
What does this information tell us about speech production and the
speaker’s ability to juggle the cognitive processing loads which charac-
terize fluent and native-like speech performance? It has been posited
that there is a pattern of pausing in first language speech performance
which is a natural consequence of the weight of psycholinguistic
processing needed to produce speech. Chafe (1980) states that first
language speech occurs in “spurts” of 2 seconds, containing an average
of five words. Pauses occur at these junctures, usually after a single
clause, also marked by the intonation contour. Pawley and Syder (1983)
state that the norm in native-speaker production is to pause or slow
down near clause boundaries generally after four to ten consecutive
words, and only extremely rarely in mid-clause. In conversational speech
in English, an average 270 to 300 syllables per minute are produced, and
over 50% of fluent units are complete and grammatical clauses. It is
uncommon to pause more than 0.5 seconds in mid-clause, generally for
emphasis or to breathe. Pauses of less than 2 seconds are the norm for
pauses at clause boundaries. The second language performances in the
empirical studies reviewed indicate pause patterns which deviate from
these native-speaker norms.
Chafe (1980) conducted an influential study of the pause structures
of native speakers retelling the story of a brief film. The speakers’ per-
formance exhibited common features having to do with attention focus
and pausing. Chafe noted that sentence-final intonation indicated the
shift between what he terms “focus clusters”:

The majority of foci of consciousness are expressed in linguistic

phrases or clauses which end in rising pitch contour . . . but approxi-
mately one-third of those in our data end instead with the kind of
falling pitch contour interpreted as “sentence-final intonation.” This
28 Formulaic Language and L2 Speech Fluency

conspicuous sentence-final contour, as well as various other criteria

including the syntactic properties associated with a “grammatical sen-
tence,” define what I will call a “focus cluster.” In other words, foci of
consciousness appear to cluster together to form longer units which
are verbalized as sentences. There are various kinds of coherence that
bind the foci within such clusters together. (Chafe, 1980: 173)

Chafe goes on to note that the majority of pauses occur between focus
clusters and not within them, and that the clusters themselves focus into
“episodes,” which act like paragraphs in oral production (Chafe, 1980:
176). Therefore, pauses serve a blend of rhetorical and syntactic func-
tions in speech, as human consciousness and awareness activate small
chunks of information and formulate speech to encode them. Second
language speech is characterized by disfluent pause distributions likely
because of the difficulty of the encoding.
Hieke (1981) sees hesitation and pauses in speech as a means of
content and quality control. He states that they “serve as devices by the
speaker to produce more error-free, high-quality speech” (Hieke, 1981:
150). He classifies pauses as serving either a stalling function, as in silent
or filled pauses, drawls, and so on, or a repair function, as in false starts.
Presumably, second language speech is characterized by more of both
sets of hesitation phenomena, occurring more frequently and within
clauses, sentences, and focus clusters.

Length of Fluent Runs

The final, and most important variable of speech associated with fluency
is the size and quality of the runs of speech which occur between pauses.
Together with the distribution of pauses, this feature not only serves as
a discriminator of fluent and disfluent speech, it also provides us with a
key to the means by which fluency development can be facilitated
through instruction.
One of the earliest studies of temporal variables in second language
speech is that of Raupach (1980), which includes useful data on the
significance of the length of runs between pauses. When Raupach had
French and German students tell a story in their first and second lan-
guages, the second language speech exhibited shorter runs between
pauses, as expected.
 Fluency 29

Möhle (1984) was another of the early fluency researchers to track

the length of runs in an empirical study. Both the French and the
German speakers in her study produced shorter runs between pauses in
second language speech than in first language speech.
Towell’s (1985) study of a British learner of French, involving analysis
of speech samples over a 4-year period, focused on length of runs
between pauses. For the subject of this research, the mean length of
runs increased a remarkable 95% over the first 3 years.
Lennon (1990b) noted that, in his study of the second language flu-
ency development of four German students of English, their mean
length of runs between pauses increased markedly in three cases. Over
23 weeks, three subjects increased the mean length of runs by 20 to
26%. One subject exhibited a decline in the mean length of runs, per-
haps due to methodological procedures.
Freed (1995), in her large study of fluency development in American
students of French, also included length of runs in her analysis. She
found no significant correlation between this feature and overall
fluency ratings, but the raw data of her results indicate a trend in the
direction of longer runs. Due to issues of complexity of language, two of
the more fluent subjects in the study exhibited a mean length of runs
equal to those of the less fluent subjects, a phenomenon which skewed
the data somewhat (Freed, 1995: 139,140).
Towell et al. (1996) focused quite rigorously on mean length of runs
in the speech of English learners of French. Mean length of runs did, in
fact, increase over time in the productions analyzed in the study,
although there was variation among the performances of the subjects.
The general consensus in these studies seems to be that mean length
of runs between pauses is a significant indicator of fluency in a second
language. Why is this so? The answer likely has to do with the need to
balance skills, attention, and planning during speech, and the fact that
advanced, fluent speakers and native speakers have a greater repertoire
of automatized chunks of language to use to buy time in order to
formulate the next sequence or phrase. In fact, an increasingly skillful
blend of automatized chunks of formulaic strings and frameworks of
speech, together with newly assembled strings of words, is thought by
some to be what enables speakers to produce the longer runs between
pauses which distinguish fluency.
As Chafe (1980) notes, fluent speech occurs in spurts, punctuated by
pauses at meaning and syntactic junctures. The ability to perform in this
30 Formulaic Language and L2 Speech Fluency

way necessitates a facility in handling plans which often could compete

for attention and “jam the system.” When this jamming happens, the
result is disfluent speech, characterized by slow speed, pauses at
mid-clause, sentence, or phrase, and brief, incomplete or simplified
language runs between pauses. Rehbein (1987: 104) notes that “one
may propose that fluency in a second language requires the capability
of handling routinized complex speaking plans.” Routinized speaking
plans are those plans which have become more or less automatized, that
is, stored in long-term memory in such a way as to be easily pulled from
a repertoire and encoded into speech. Simultaneous to the encoding
and production of the automatized strings, the speaker must generate
new words and constructions to encode the new or novel elements of
the message.
This complex process needs to happen in real time, under all the
temporal, perceptual, and contextual constraints which genuine oral
communication entails. The planning which occurs in speech genera-
tion needs to be conducted on several levels. It was Lennon (1984: 67)
who, based on observation of learners completing a story retell task,
noted a possible distinction between planning at the level of syntax, and
planning at the lexical level:
The indicators are that planning is conducted on at least two levels:

1. topic and overall syntax structure are planned in advance chunks,

ideally identical with the clause/statement breakdown of the passage.
In practice, however, the subjects are forced to break these units down
still further . . .
2. Planning at the level of lexical selection would appear to be on more
of an ad-hoc basis. The self-corrections at this level would indicate the
late stage at which this planning takes place.

The advance frames appear, then, to be generated early as frameworks

within which concepts can be mapped, with the attendant lexical items,
both single and multiword, fitted in closer to the moment of articulation.

Clause Chaining

The structures of discourse as to the way clauses are chained is another

aspect of fluent speech which has been examined by Pawley and Syder
 Fluency 31

(1983). Apparently, everyday speech of a fluent nature is characterized by

strings of more or less independent clauses, with little grammatical inte-
gration with earlier or later clauses. Subordination, for example, is mini-
mized in spontaneous speech (Pawley & Syder, 1983: 202–204). Pawley
and Syder analyzed two stretches of native-speaker production to see how
fluency relates to clause-chaining. One speaker, George Davies, produced
speech in which fluent units were separate clauses:

/we had a /fan tastic time - --

[slows] (1.1)
/there /were/ all kinds of re/lations /there/
[accel] [slows ]
/I dun/no where they/all come /from/
[accel] [slows ]
I didn’t know /’alf o’ them - -
[accel] (0.9)
and’ ah - the kids/sat on the floor - -
(0.2) (1.5)
and ol’/ Uncle Bert/he/ah
o’/course /he was the life and soul of the party
[accel] [slows ]
/Uncle /Bert ‘ad a /black bottle - - -
[accel] [slows ] (1.5)
an ah - ‘e’d t/tell a/few stories
(0.2) [accel] [slows]
an ‘e’d/take a /sip out of the /black bottle
[accel] [slows ]
n’ the / more sips he /took /outa / that bottle –
[accel] (1.0)
the worse the /stories got - - -
(1.6)
(Pawley & Syder, 1983: 203)

Another speaker, Q, produced comparatively non-fluent speech, in a

PhD dissertation oral defence:

and it / seems to be –
[accel]
32 Formulaic Language and L2 Speech Fluency

if a /word is/fairly - - /high on the frequency /list/ -

[slow] [accel]
I /haven’t /made /any count -
[accel]
but - /just – im/pression istically – um
[slow]
um – the /chances are -
that you get a - com /pound -
[slow]
or - a /nother – phono /logically deviant - - form -
[slow]
with ah / which is al/ready in other /words
[accel] [slow]
/which is /fairly frequent - ly the /same - /phono /logical
[accel] [slows]
shape –
(Pawley & Syder, 1983: 201)

It appears from the pattern of his disfluencies that, unlike Davies, Q is

planning only a few words at a time. It is a conceptually novel discourse
for him, and he has to think through the content while under stress. Q
uses a clause-integrating strategy, which means each new clause is depen-
dent on the structure of a previous one, for example, his struggle (false
start or reformulation) with the final clause in the sample, beginning
with “with,” repaired to begin with “which.” He has little choice, given
the genre, register, and relative lack of interactivity in the context of
his speech production. Davies, on the other hand, is speaking more
spontaneously, and exhibits clause-chaining of independent clauses
linked by “and,” in most cases. Pawley and Syder state that this style is
more effective in narrative than Q’s integrated style:

With the chaining style, a speaker can maintain grammatical and

semantic continuity because his clauses can be planned more or less
independently, and each major semantic unit, being only a single
clause, can be encoded and uttered without internal breaks . . . we
may speak, then, of “a one clause at a time facility” as an essential
constituent of communicative competence in English: the speaker
must be able to regularly encode whole clauses in their full lexical
 Fluency 33

detail, in a single encoding operation and so avoid the need for mid-
clause hesitations. (Pawley & Syder, 1983: 203, 204)

Given this norm of clause-chaining in native-like conversational/narra-

tive speech in English, a speaker should be able to encode whole clauses
in order to avoid hesitations in mid-clause. The hallmarks of fluent
speech include pauses at clause junctures and a certain length of speech
runs between such pauses. The way to accomplish this seems to be the
recall of most clauses as more or less intact, or automatically chained.
Pawley and Syder assert that only a minority of spoken clauses are novel
(Pawley & Syder, 1983: 205), and that memorized chunks form a high
proportion of the speech of everyday conversation. As expected, the
memorized sequences need little encoding. Since speech is therefore
not produced word-for-word, the speaker can focus on rhythm, variety,
combining memorized chunks, or producing creative connections of
lexical strings and single words.

Transfer and Strategies in Fluency Development

Evidence of transfer of aspects of fluency can bear out this distinction

between the two types of processing and knowledge. Raupach (1980)
found that, when French and German subjects were compared when
speaking in their first language and in a second language, they tended
to transfer a preference for silent or filled pauses from the first language
into the second. Möhle and Raupach (1989) analyzed the first and
second language speech of a German student of French. Comparing
first and second language production as to length of unfilled pauses,
intonation contours, pronunciation, and hesitation phenomena, they
found evidence of transfer of these phenomena from the individual’s
first language production patterns. They concluded that second language
declarative knowledge, that is, knowledge of content or information, was
blocked by the predominance of first language procedural knowledge, or
awareness of how to perform or execute a skill:

The omnipresence of the first language procedural knowledge . . .

“disposes” the learner to select and organize second language mate-
rial in a quite specific way and, therefore, may have a considerable
impact on the nature of the “productions” . . . or processing units in
34 Formulaic Language and L2 Speech Fluency

the individual’s second language performance. In some instances, the

learner is able to process second language linguistic material directly
. . . just as he or she proceeds in the native language, that is, without
any need for retrieving items from the declaratively encoded knowl-
edge. The highest level of automaticity is attained with the use of pre-
fabricated patterns and routines; here the linguistic data are intimately
tied to the procedures . . . (Möhle & Raupach, 1989: 207–208)

Möhle and Raupach go on to note that the application of procedural-

ized productions from the first language contributes to disfluency, and
they also suggest that if declarative knowledge from the first language
is used repeatedly in second language speech it could become auto-
matized and lead to fossilization. Möhle and Raupach conclude that
transfer of procedural knowledge from the first language likely occurs
at several levels:

. . . one has to distinguish between different levels of transfer: the

learner might conceptualize and plan parts of his or her L2 perfor-
mance “automatically” in an L1 manner and thus transfer procedural
knowledge; in addition to this, the linguistic data used as slot fillers
within the chosen frame can likewise be processed with L1 knowledge
involved, be it in the form of another transfer of procedural knowl-
edge at a lower level or in the course of retrieval process working on
L1-affected declarative knowledge. (Möhle & Raupach, 1989: 213)

The authors give the example of a German learner of French using

German phrase constructions inaccurately transferred to French, such
as declining the indefinite pronoun on and translating idioms, as well as
correlations between temporal patterning in the first language and the
second language which lead to nonnative-like speech (210, 211). So,
the transfer of procedural knowledge could occur wholesale, in an auto-
matic way, or it could confuse the retrieval and use of second language
declarative knowledge, even in tandem with first language declarative
knowledge. Faerch and Kasper (1989), in an analysis of first and second
language performance of a Danish learner of English, found evidence
of three types of such transfer:

1) strategic transfer, in which learners use first language speech

routines when faced with a gap in their second language or
interlanguage knowledge
 Fluency 35

2) subsidiary transfer, in which performance constraints such as speed,

work against the learner’s ability to use second language knowledge
3) Automatic transfer, in which attention is diverted to some other
aspect of the production process, such as articulation or conceptual-
izing, and highly automatized first language plans take over
(Faerch & Kasper, 1989: 175–176)

The predominance of automatization and proceduralization of first

language speech appears to lead to transfer of some temporal aspects
of speech into second language production. The competing and
challenging constraints of second language speech are characterized
by less automatization and more need to resort to controlled process-
ing, the step-by-step construction of utterances based on assembly
from component lexical items into syntactic patterns. This makes the
transfer of automatized speech routines from the first language virtually
inevitable.
There is evidence as well that learners tend to use strategies such as
avoidance to facilitate the automatization process. Clahsen (1987), in
an analysis of the naturalistic acquisition of German, discovered that
apparent simplification of performance rules takes place at certain stages
in the acquisition process. Learners in the study tended to avoid use of
structures which would seem to involve complex learning tasks. The
example of subject-verb inversion in German is discussed:

Deletions not only accompany, but also prepare for the acquisition of
inversion. During a period of uncertainty a learner systematically
reduces the number of obligatory contexts for the application of
the rule . . . inversion can be seen as a complex learning task. As a
result, this operation cannot be automatized immediately . . . during
this transitional period the complexity of the task is reduced by dimin-
ishing the number of contexts. After having automatized inversion,
there is no need anymore for simplifying these structures, and the
deletions vanish. (Clahsen, 1987: 73)

This simplification and task-management strategy could help to account

for the short runs characteristic of disfluent speech. Learners may be
using avoidance of this type to reduce the complexity, and, by implica-
tion, the length, of runs spoken. This would free up attention and time
for managing the lexical planning which Lennon (1984) posits as the
36 Formulaic Language and L2 Speech Fluency

second and more immediate planning stage, after syntactic planning

has created a framework. It seems that the evidence of disfluency which
is observable in the pause locations and short runs between pauses in
second language speech may be evidence of some sophisticated under-
lying processes.
Other information about the development of fluency over time,
reflected in the correlates of disfluency most present at each stage of
development, accords with this interpretation. It seems that lower level
proficiency learners show a greater variety of largely idiosyncratic plan-
ning behavior, whereas more advanced learners show more regularity
and consistency in planning (Raupach, 1984; Lennon, 1990a, 1990b;
Freed, 1995). Raupach’s study of formulaic speech and its effect on flu-
ency (1984) identified a sort of simple taxonomy of disfluency markers
linked to speech planning activities:

At a certain level of second language competence, most planning

activities have to take place during unfilled pauses and in connection
with standard pause fillers . . . With the adoption of new forms of hesi-
tating (e.g., drawls) combined with a new way of segmenting speech
stretches . . . learners gradually abandon, in their second language
speech productions, the “temporal patterning” they use to follow in
their native language . . . part of the planning activities that previously
had been reserved for silent and filled pauses is now processed in con-
nection with other hesitation phenomena and at other places than
before.(Raupach, 1984: 135)

It was noted earlier that duration of silent pauses was a relatively signifi-
cant correlate of lower fluency levels. This is likely due to the need to
plan based more on declarative knowledge, rather than proceduralized
skill, which leads one to fall back on first language pause patterns. As
well, frequent hesitations, unfilled or filled with nonwords, give time for
processing to be completed. Later, the more advanced speaker is able to
use chunks of formulaic speech and lexical “hedges,” such as you know
and I mean as ways of buying time for processing. In their qualitative
analysis of the results of a study of the speech fluency development of
advanced learners of French, Towell, Hawkins, & Bazergui (1996) take
 Fluency 37

exactly this view, that first language transfer and then proceduralization
and automatization are key to second language production:

We have taken the view that second language production may be

based on multiple knowledge sources: these include setting L2 param-
eters on the basis of external evidence, transfer of L1 parameter
settings, transfer of L1 surface structures (transliteration), situation-
ally based lexical phrases, and consciously learned rules. These
knowledge sources, available to post-7-year-old learners, have to be
developed . . . in proceduralized and tuned productions . . . if speech
which even approximates to native fluency is to be produced. (Towell,
Hawkins, & Bazergui, 1996: 103)

Thus, use and transfer of various knowledge sources may be involved in

speech production by second language speakers. This knowledge must
also be proceduralized to help achieve fluency. There is evidence that
one of the knowledge sources listed by Towell, Hawkins, & Bazergui, a
type of formulaic language unit called lexical phrases, may constitute one
of the building blocks of fluency and nativeness for second language
learners. Equating fluency with proceduralization of knowledge is a
plausible way of conceptualizing the process of fluency skill develop-
ment, but the use of formulaic sequences may constitute much of the
raw material of the fluent speech itself.
 Chapter 2

Formulaic Sequences

There has been an increased interest recently in the nature and

functions of formulaic lexical units. A growing body of work suggests
that ready-made chunks or preferred sequences of words play a
significant part in language acquisition and production. Numerous
researchers have attempted to define and categorize this generally
overlooked aspect of language, and there is growing evidence that
formulaic language is basic to language learning, processing, and
production.
The following is an investigation of the phenomenon of formulaic
language, which defines the construct through a survey of literature
and examines the research evidence on the role of formulaic sequences
in acquisition and production.

Defining and Categorizing Formulaic Language

While researchers have studied the phenomenon of formulaic language

from different perspectives over the years, there is a certain amount of
agreement on basic definitions of what constitutes a formulaic sequence
and what characteristics such sequences share which makes them dis-
tinct. Definitions typically focus on cognitive storage and processing of
formulas. The consensus has seemed to be that formulaic language
sequences are multiword units of language which are stored in
long-term memory as if they were single lexical units. Wray and Perkins
(2000) define formulaic sequences as multiword units of language:

A sequence, continuous or discontinuous, of words or other meaning

elements, which is, or appears to be, prefabricated: that is, stored and
retrieved whole from memory at the time of use, rather than being
subject to generation or analysis by the language grammar. (2000: 1)
 Formulaic Sequences 39

This type of definition is common in the literature. Pawley and Syder

(1983) refer to formulas as “sentence stems” which are lexicalized, that
is, which are “regular form-meaning pairings” (p. 192). This notion of
lexicalization is echoed by Nattinger and DeCarrico (1992) in a highly
influential work which focuses on lexical phrases, a term for particular
formulaic language units which serve pragmatic functions. They also
indicate that lexical phrases function as a sort of bridge between lexis
and grammar:

. . . lexical phrases [are] form/function composites, lexico-grammatical

units that occupy a position somewhere between the traditional poles
of lexicon and syntax; they are similar to lexicon in being treated as
units, yet most of them consist of more than one word, and many
of them can, at the same time, be derived from the regular rules of
syntax, just like other sentences. Their use is governed by principles
of pragmatic competence, which also select and assign particular
functions to lexical phrase units. (p. 36)

The notion that storage of formulaic units is as single lexical or lexico-

grammatical units is key to a common definition. Hickey (1993) puts it
most succinctly by noting that definitions of formulaic language units
are often expressed in terms of processes, referring to multiword or
multiform strings produced and recalled as a chunk like a single lexical
item rather than being generated from individual items and rules.
Definitions of formulaic language have tended to evolve and shift
somewhat over time as researchers have taken new and varied approaches
to the task. Wray (2008) modified the standard definition as presented
above, positing a new term for a subcategory of formulaic sequences,
Morpheme Equivalent Unit, or MEU. She defines the MEU as “a word or
string, whether incomplete or including gaps for inserted variable items,
that is processed like a morpheme, that is, without recourse to any
form-meaning matching of any subparts it may have” (p. 12). While this
definition is more specific about the nature of presumably prefabricated
multiword strings of language, it is important to note that this definition
shares with most others in the literature a focus on the psycholinguistic
processing of multiword units, and the way in which these units may
be stored in the mental lexicon. However, there are some significant
differences in how researchers operationalize a definition of formulaic
sequences in research. Wray (2008: 94) points out that in addition to the
40 Formulaic Language and L2 Speech Fluency

definition used by Wray and Perkins (2000) as quoted above, and Wray’s
own 2008 definition of MEU, researchers may use frequency of reoccur-
rence of a word string in a corpus (e.g., Biber, Conrad, & Cortes, 2004),
or inclusion in a corpus-derived dictionary of idioms (e.g., Conklin &
Schmitt, 2008) as bases for analysis.
Two general approaches to the study and analysis of formulaic language
have emerged over time, the phraseological and the frequency-based
or distributional (see Granger & Paquot, 2008 for a review). The phraseo-
logical approach, with roots in Russian and European linguistics, focuses
on the structure and functions of multiword language units. Researchers
in this tradition have sought linguistic labels and criteria to describe
formulaic sequences, and classified them according to syntactic and
semantic criteria (e.g., Melĉuk, 1998; Cowie, 1998). More recent
approaches have been driven by corpus studies, using frequency of
a string in a given language corpus as a guiding principle. This has
uncovered a wide range of word combinations previously overlooked
by traditional phraseology, and has actually shown that the fixed items
such as idioms and proverbs are much less frequent. An example of the
type of previously overlooked formulaic items uncovered by corpus
studies is the lexical bundle (see Biber & Conrad, 1999; Biber, Conrad &
Cortes, 2004). Lexical bundles may consist of strings such as what I want
to say is, crossing phrase or clause boundaries and having observable
and classifiable functions in discourse. This made them virtually invisi-
ble to phraseologists in the past, who were concentrating on formulaic
sequences most often as unitary items with some internal variability.
It is interesting to posit the existence of multiword strings or chunks
of language which are stored in long-term memory as if they are single
units, but how does one recognize a formulaic language sequence in
productions? Some researchers have attempted to elaborate criteria for
the identification of formulaic language sequences. Coulmas (1979)
outlines conditions which need to be met if a sequence is to be consid-
ered formulaic. Two conditions, that the unit must be at least two
morphemes long and cohere phonologically, are identified as necessary
for formulaicity. Utterances which are formulaic, then, are polymorphe-
mic and produced without internal hesitation or pausing. Coulmas
also specifies that a formula may be more grammatically advanced than
surrounding language, exhibiting a level of syntactic and phonetic
complexity beyond the norm for the language produced by the learner.
 Formulaic Sequences 41

Other criteria laid out by Coulmas for formulaic sequences are that they
are typically shared within a community, situationally dependent, and
repeatedly used in the same form:

1. at least 2 morphemes long (i.e., two words)

2. coheres phonologically
3. individual elements are not used concurrently in the same form
separately or in other environments
4. grammatically advanced compared to other language
5. community-wide formula
6. idiosyncratic chunk
7. repeatedly used in the same form
8. situationally dependent
9. may be used inappropriately. (p. 32)

Similarly, Peters (1983), in an effort to elaborate criteria for identifying

formulas in learner first language, focuses on phonological coherence,
greater length and complexity than other output, nonproductive use of
rules underlying a sequence, situational dependence, and frequency
and invariance in form.
The most current set of criteria for identifying formulaic sequences is
that of Wray, 2008, which draws on the criteria noted above to a great
extent. Elaborating on an earlier list used in empirical research on child
L2 acquisition (Wray & Namba, 2003), she lists 11 diagnostic criteria
useful in making intuitive judgments, especially for spoken language
corpora:

1. There is something grammatically unusual about the word string

2. All or part of the string lacks semantic transparency
3. The string is associated with a specific situation or register
4. The string performs a function other than or in addition to the mean-
ing of its component words
5. This formulation is typical of this speaker in conveying this idea
6. This word string has an associated action, orthographic phenomena,
or phonological pattern, and/or the speaker/writer is repeating some-
thing just heard/read
7. This word string has been marked grammatically or lexically to give it
status as a unit
42 Formulaic Language and L2 Speech Fluency

8. It is highly likely that the speaker/writer has encountered this pre-

cise formulation in communication from other people
9. While this string appears novel, it is clearly derived from something
which is formulaic
10. This string is formulaic but unintentionally applied inappropriately
11. This string contains linguistic material which is too sophisticated or
not sophisticated enough to match the speaker’s general competence.
(pp. 119–121)

These criteria overlap with those of Coulmas (1979) and reflect

current thinking about what makes a formula stand out from other
language production. For the study described in Part II of this volume,
lists of criteria like this were valuable, and many items of Wray’s list
were used.
Clearly, formulaic sequences are characterized by certain key features.
They are multiword or polymorphemic units of language, stored in
memory as if they are single lexical units, and recalled and produced
as wholes. This production is marked by a degree of phonological
coherence, and the unit may outstrip other output in terms of length
and complexity. As well, formulaic units can be invariant in form and be
used for specific situational purposes.
Formulaic sequences have been classified in a variety of ways in the
research literature over the years. They have been labelled by as many
as 40 different terms (Wray & Perkins, 2000), and a number of distinct
categories of the sequences have been identified. Across these classifi-
cations, themes of degree of syntactic and structural flexibility, and
pragmatic function emerge.
A simple, early function-based classification of formulas is that of Yorio
(1980), who identifies four general categories of formulaic sequences.
The four types are situational formulas, which fit to certain conversational
parameters, for example, how are you, or excuse me, stylistic formulas, which
are specific to particular registers of language, for example, in conclusion,
or by way of conclusion, ceremonial formulas, which are ritualistic and
required by certain formal settings, for example, may I have your attention
please, or ladies and gentlemen, and gambits, which organize interactions or
activities, for example, what do you think or it’s your turn.
Other researchers later elaborated classifications of formulaic
language sequences. Cowie (1988) distinguishes between phrases
 Formulaic Sequences 43

which are pragmatically specialized and those which are grammatically

specialized:

Word combinations can be divided into two major groups, which dif-
fer according to the kinds of meaning which their members convey
and to the structural level at which they operate. The first category, of
which good morning and how are you are members, have evolved
meanings which are largely a reflection of the way they function in
discourse . . . In so far as those discourse meanings have stabilized, the
expressions are pragmatically specialized (Leech, 1983, p. 28) The
second category, of which kick one’s heels and pass the buck are
examples, have developed more or less unitary referential meanings
by virtue of their use as invariable units in grammatical constructions.
To the extent that their meaning and form have stabilized in this
way, the expressions have become semantically specialized, or idiomatic.
(pp. 132, 133)

Cowie goes on to refine the classification further, referring to formulaic

sequences as composites

Composites are word combinations, more or less invariable in form

and more or less unitary in meaning, which function as constituents
of sentences (as objects, complements, adjuncts, and so on) and con-
tribute to their referential, or propositional, meaning. They are lexi-
cal building-blocks comparable in their syntactic functions to nouns,
adjectives, adverbs, and verbs. (pp. 134, 135)

The idea that formulaic sequences such as composites or lexical phrases

have grammatical functions as units shows their utility and ubiquity in
language.
Sinclair and Renouf (1988) identify a particular aspect of language
based on corpus study, in which particular verbs in English possess a
quality they refer to as dilexicality. This leads to the possible creation of
large numbers of formulaic sequences of various function:

A major feature of the language . . . is the phenomenon known as

“dilexicality,” the tendency of certain commoner transitive verbs
to carry particular nouns or adjectives which can in most cases
44 Formulaic Language and L2 Speech Fluency

themselves be transitive verbs. In general, the more frequent a word

is, the less independent meaning it has, because it is likely to be
acting in conjunction with other words, making useful structures,
or contributing to familiar idiomatic phrases. (p. 153)

They give the examples of the verbs give and have, which tend to often
collocate with nouns in regular patterns, as in give advice, give a look,
give information, or have a good look, have a deep longing, have a heart to
heart talk, have a strange feeling. Another example is that the primary
function of make is to carry nouns such as decision, discovery, arrangement.
Sinclair and Renouf go on to note that other types of regular word
combinations are characteristic of everyday language, for example, the
regular collocations of happy marriage, accidental death. As well, function
words often have grammatical restrictions. For instance, each occurs
with units of time, of leftward with kind, part, sort. Furthermore, combi-
nations of grammatical words produce discontinuous frameworks,
such as a ____ of, attracting particular lexical insertions such as lot, kind,
number, and so on.
Nattinger and DeCarrico (1992) provide a complex and broad
classification of lexical phrases, a pragmatically specialized subset of
formulaic sequences. They outline two large categories of the phrases:
strings of specific lexical items and generalized frames. The former are
generally unitary lexical strings and may be canonical in the grammar
or not, while the latter consist of category symbols and specific lexical
items. Four criteria help in classifying the phrases: length and gram-
matical status; canonical or noncanonical shape; variability or fixedness;
whether it is a continuous, unbroken string of words or discontinuous,
allowing lexical insertions (pp. 37, 38). They also identify four large
categories of lexical phrases which display aspects of the four criteria:
polywords, which operate as single words, allowing no variability or
lexical insertions, and including two-word collocations (e.g., “for the
most part,” “so far so good”); institutionalized expressions, which are
sentence-length, invariable, and mostly continuous (e.g., “a watched
pot never boils,” “nice meeting you,” “long time no see”); phrasal con-
straints, which allow variations of lexical and phrase categories, and are
mostly continuous (e.g., “a ___ ago,” “the ___er the ___er”); sentence
builders, which allow construction of full sentences, with fillable slots
(e.g., “I think that X,” “not only X but Y”) (pp. 38–45). Nattinger and
DeCarrico see a great deal of variety and diversity in formulaic sequences,
 Formulaic Sequences 45

and their comprehensive taxonomy covers a large proportion of the

types of utterances which are produced in a language.
A more recent descriptive scheme for formulaic sequences is that of
Wray and Perkins (2000), in which they focus on semantic and syntactic
irregularities of the sequences. A vital aspect of formulaic sequences,
according to Wray and Perkins, is their semantic irregularity. They
are not composed semantically, but are holistic items, like idioms and
metaphors. Another key element of formulaic sequences is their
syntactic irregularity, which is manifest in two qualities: a restriction on
manipulation, for example, one cannot pluralize beat around the bush
or passivize face the music or say you slept a wink or feeding you up; the
fact that, in formulaic language, normal restrictions are flouted, such
as the sequences which contain an intransitive verb + direct object, for
example, go the whole hog or other gross violations of syntactic laws like by
and large.
Wray (2002) elaborates a set of criteria for determining if strings are
prefabricated. Structure or form of the sequences is one such criterion,
and she notes that such strings often begin with conjunctions, articles,
pronouns, prepositions, or discourse markers (p. 31). Compositionality
or internal structure of strings is another such criterion, in which “the
string is no longer obliged to be grammatically regular or sematically
logical” (p. 33). Fixedness, or the tendency for prefabricated sequences
to be of invariable form is another such criterion, although Wray does
allow that a large subset of formulaic sequences often have fillable slots
(p. 34). She also identifies intonation contour and speed of articulation
of sequences as being markers of prefabrication, and fluency criteria
such as lack of internal pausing during utterance as indicative as well
(p. 35). Of most relevance to the identification of formulaic sequences
in spoken language is Wray’s statement that “it may simply be that
identification cannot be based on a single criterion, but rather needs to
draw on a suite of features” (p. 43).
Wray (2008) later emphasizes that the processing of formulaic
sequences as wholes likely stems from the ways acquisition processes
operate with respect to input. She notes that a great deal of language
input in L1 acquisition is left unanalyzed unless necessary, a phenome-
non she terms Needs Only Analysis, or NOA (p. 17). If there is a strong
form-meaning link with a particular string, for example, How do you do
as a standard greeting among previously unacquainted adults, with no
variation, then the string will remain unanalyzed. Over the course of L1
46 Formulaic Language and L2 Speech Fluency

acquisition acquirers may note some variation in such strings, such as

lexical insertion, (e.g., Have you seen my boots/shirt/watch?, or I’d like a
Coke/cheeseburger/3-month plan) but analysis will likely stop at the recogni-
tion of the existence of a fillable slot and the possible word types which
may fill the slot. For adult L2 learners, this process may be much less
frequent or slower, since the tendency of adults and language programs
is to analyze L2 input for patterns, not to mention the fact that L2
learners receive greatly reduced input compared to children in L1.
Some researchers have linked formulaic sequences to the lexico-
grammar (Tucker, 2005) and to systemic models of functional grammar
(Butler, 2003). These researchers have noted that formulaic sequences
have a place in models of language which prioritize the lexico-grammar
and levels of structure related to speech act realizations. They encour-
age acknowledgment of the role of formulaic sequences in integrating
extraclausal or partially clausal expressions into functional grammars of
discourse.
Classifications and descriptions of formulaic sequences tend to con-
centrate on functional-pragmatic aspects of their use, or on syntactic
and linguistic aspects of their structure. The research underlying all of
these classifications and descriptions demonstrates that formulaic
sequences are common, widespread, and necessary parts of human
communication.

Formulaic Language and Child Language Acquisition

There is a certain amount of evidence of formulaic sequences being

used as a strategy by children in first and second language acquisition
(e.g., see reference to Wray 2008 above). It appears that initial first
and second language acquisition in children includes attending to
formulaic sequences in language input, adopting them for use, and
later segmenting and analyzing them. The analysis may take place later
partly as a result of neurological development and a resultant increase
in analytic cognitive skills.
Wong-Fillmore (1976) was one of the first to study the second
language acquisition of a child and find that one prominent process
involved formulaic chunk acquisition followed by a process of segmenta-
tion or syntactic and semantic analysis and breakdown which fostered
development of overall linguistic competence. Hakuta (1974) conducted
 Formulaic Sequences 47

a 60-week study of the second language acquisition of a Japanese child

and found evidence of initial acquisition of prefabricated chunks later
analyzed and used to facilitate overall language development. Hickey
(1993), in a longitudinal examination of the acquisition of Irish Gaelic
of a child also discovered a role for formulas in acquisition. Again, they
were later broken down and analyzed, providing grist for the linguistic
competence mill.
Peters (1983), in a seminal piece of work on child first language acqui-
sition, documents how this process of formulaic chunk acquisition and
later segmentation might work. She claims that there is evidence for eight
assertions about the process:

1. First acquisition units by children often consist of more than one

morpheme;
2. there is no difference between these units and minimal ones in terms
of storage;
3. all of the polymorphemic units can be segmented (broken down);
4. smaller units from segmentation are stored in the lexicon;
5. both the original unit and the segmented ones can be stored in the
lexicon;
6. segmentation produces structural information, starting with simplest
frames with slots, then generalized into patterns;
7. the lexicon grows through units perceived in conversation and their
segmentation, as well as fusion (storage of combinations);
8. fusion continues into adulthood.

According to Peters, early on the child develops strategies for extracting

meaningful chunks from the flow of conversation. She is able to remem-
ber them, compare them phonologically with others, and remember
them as new lexical units. They are stored as wholes in the lexicon as
individual words or as multiword units. Later in her cognitive develop-
ment, she is able to analyze the stored chunks and then recognize and
remember structural patterns and information about distribution classes
revealed by the analysis. She is then ready to develop an ability to utilize
lexical and syntactic information already acquired to analyze new chunks
in the linguistic environment.
These dynamics of acquisition of formulaic sequences and their use as
a basis for creative construction have been examined by Myles, Hooper,
48 Formulaic Language and L2 Speech Fluency

and Mitchell (1998) and Myles, Mitchell, and Hooper (1999) in child
learners of French as a second language in a classroom context. Myles
and her research associates found that the young learners in their
studies did in fact not only acquire and use formulaic sequences as
wholes, but that they also used segmentation of the formulas to enhance
their increasingly complex communication needs over the 2 years of the
research project. Initially, the learners were able to use unanalyzed
wholes to communicate simply, but they began to break the formulas
apart and use components in different ways as their routine classroom
communication needs developed beyond simple communication of
personal information into a need to discuss third person activities
and characteristics. When the third person communication needs
grew, the segmentation process began and then accelerated (Myles,
Hooper, & Mitchell, 1998: 359).
Other researchers have been able to determine that processes related
to pragmatic competence are at work when children acquire formulaic
sequences. Bahns, Burmeister, and Vogel (1986) investigated the second
language acquisition of a group of children and found evidence of a
formula segmentation process at work. They found two particular prag-
matic factors at work in the use of formulas by the children, namely,
situational frames requiring their use, and frequency of occurrence of the
formulas. The authors note that it was common to discover exceptionally
sophisticated language in stretches of child learner speech in research:

In their attempts to write grammars for different stages of development,

mainly in structural areas like negation or interrogation, child language
researchers were very often confronted with utterances of a rather com-
plex nature. The structure of these utterances was somehow “outside”
the rules written to account for the bulk of data representing syntactic
development for the stage in question. (pp. 696, 697)

In their study, Bahns et al. found a large range of formulas used by the
children, accounting for the complex utterances noted by earlier
researchers. The categories found included:

1. Expressive formulas—indicators of a sudden state of mind, e.g., shut

up, stupid idiot, thank you;
2. directive formulas—intended to change the hearer’s behavior, e.g.,
let’s go, knock it off, wait a minute;
 Formulaic Sequences 49

3. game or play formulas—tied to specific play activities, e.g., who’s up,

you’re out;
4. polyfunctional formulas—exceed a single semantic-pragmatic value,
e.g., what is it? I don’t know;
5. question formulas—elicit information, e.g., how come? What time is it?
6. phatic formulas—to establish, prolong, or discontinue interaction,
e.g., good bye, see you later, You wanna see X?

The study also found evidence of a progression of development of use

of the formulas, starting with use of the simpler expressive and game
formulas. This was followed by a broadening of the range of formulas as
pragmatic awareness and ability grew, and, eventually, full native-like
selection and use of formulas with more precise knowledge of when an
expression is pragmatically target-like.
It is interesting to note the double role of formulaic sequences as an
element of child language acquisition. They are acquired and retained
in and of themselves, linked to pragmatic competence and expanded as
this aspect of communicative ability and awareness develops. At the
same time, they are segmented and analyzed, broken down, and com-
bined as cognitive skills of analysis and synthesis grow. Both the original
formulas and the pieces and rules which come from analysis are retained.
The next question to attend to is whether or not adult language devel-
opment has been shown to take place following a similar process.

Formulaic Sequences in Adult Language Acquisition

A great deal of evidence has also been collected over the years of a role
for formulaic sequences in the process of adult language acquisition,
but the development processes uncovered by researchers in this area
are not exactly like those found in the child language acquisition
studies.
Yorio (1980) was an early investigator of adult language development
and formulaic sequences. Examining several longitudinal studies based
on instructed adult learners’ written work, he found that unlike
children, adult learners do not make extensive use of prefabricated
formulaic language, and when they do, they do not appear to use it to
further their language development. Instead, they appeared to use it
more as a production strategy, to economize effort and attention in
spontaneous communication.
50 Formulaic Language and L2 Speech Fluency

Schmidt (1983) conducted an in-depth case study of the English

language development of a Japanese adult in Hawaii and found a
definite role for formulaic sequences. In fact, the learner under study
used a great and ever-increasing number and range of formulaic
sequences as a communication strategy, while appearing fossilized and
grammatically inept in other aspects of language. Schmidt found that
his highly motivated, rapidly acculturating research subject was resistant
to error correction and was able to develop linguistically and adapt
socioculturally through use of formulaic sequences almost exclusively.
There was little or no evidence of the processes of segmentation and
analysis so marked in the child acquisition studies.
Ellis (1996), in an overview of sequencing in language acquisition,
finds a role for formulas in adult language acquisition. He asserts that
much of language acquisition is really acquisition of memorized
sequences, and that short-term repetition and rehearsal permit the
development of long-term sequence information for language. In turn,
this information allows chunking of working memory contents to these
established patterns. Long-term storage of frequent language sequences
allows them to more easily serve as labels for meaning reference, and they
can be accessed more automatically. The result is more fluent language
use, freeing attentional resources for dealing with conceptualizing and
meaning. Ellis asserts that multiword units in long-term storage serve as
a database for grammar acquisition.
Bolander (1989), in a study of acquisition of Swedish by adults, found
that formulaic sequences contributed to a greater facility and economy
in learning and use. The adults in this longitudinal study consistently
used prefabricated language units which contained target language struc-
tures well in advance of demonstrating that they had actually acquired the
structures themselves. Like the child subjects of Peters (1983) and Hickey
(1993), they produced formulaic sequences which contained language
which outstripped their normal abilities. As well, Bolander noted that
the learners appeared to sometimes use canonical formulas to help in
acquiring specific rules of Swedish syntax.
It appears that adults in naturalistic L2 learning environments, like
children, tend to acquire and use formulaic sequences. However, the
established cognitive and learning styles of adults, their diverse acquisi-
tion contexts, knowledge of L1, and other factors make for more variety
in the route of language acquisition generally, and with regard to use of
formulaic sequences specifically. Some adults may be more analytic and
 Formulaic Sequences 51

seek to infer rules from chunked units or from pieces of input, while
others, such as Schmidt’s (1983) subject, may rely heavily on acquired
formulas and not attempt to break them down or analyze them. Further-
more, degree of literacy and type and degree of instruction may play a
part. One important fact remains, however: formulaic sequences are
extremely important for language performance, and it is to this part of
the picture that we turn to next.

Pragmatic Aspects of Formulaic Language

Many researchers have noted the links between formulaic language use
and pragmatic competence. Nattinger and DeCarrico (1992) state that
a particular type of pragmatic competence is required in the use of the
subset of formulaic sequences they call lexical phrases:

(. . .) selection is accomplished by (. . .) competence (. . .) a particular

type of pragmatic competence, which takes specific strings generated
by the syntactic component and assigns them functional meanings, so
that these strings not only have syntactic shapes, but are capable as
well of performing pragmatic acts, as for example basic forms selected
for speech acts such as promising, complimenting, asserting and so
on. Pragmatic competence thus selects the form/function composites
required for particular circumstances. (p. 11)

Nattinger and DeCarrico are not alone in focusing on the pragmatic

value of formulas. The link between specific types of formulas and
pragmatic functions has been examined, and it appears that, while
we have come far in many ways from the evolutionary source of this
behavior proposed by Wray (1998) of protohumans uttering meaning-
specific sound strings while engaging in grooming behaviors together
(see below), we are still assigning specific formulaic sequences to quite
particular social needs.
Formulaic sequences are highly context dependent for their meaning
and selection. As Coulmas (1979: 241) puts it, “As they provide the
verbal means for certain types of conventional action, their meanings
are conditioned by the behavior patterns of which they are an integrated
part.” Specific cultural situations provide specific contexts for the use
of particular formulas, and only an understanding of the relevant
52 Formulaic Language and L2 Speech Fluency

dimensions of certain social situations, and their relative value, guaran-

tees an understanding of the meanings of the formulas which are highly
likely to occur in them. In fact, it appears that, typically, formulas
are used in situations with highly specific and regularly occurring and
continuing patterns of behavior and communication (Coulmas, 1979:
242, 243). They help one cope with the complexity of many social situa-
tions, help structure orderly and unambiguous communication, and
help with a sense of group identity.
Various studies have looked at the functions of particular sets of for-
mulas in communication. Bahns, Burmeister, and Vogel (1986), in a
study of the use of formulas in child language acquisition, found six
main pragmatic categories of formulas emerging, expressive, directive,
game or play, polyfunctional, question, and phatic formulas. Bygate
(1988), in a study of formulas in adult learner classroom interaction,
found a wide range of syntactic and pragmatic uses of formulas. These
are some of what he terms satellite units - conversational utterances which
lack a finite verb and are syntactically dependent. Further conditions
are that they are finite or nonfinite, uttered in a turn which does
not include a main finite clause to which they may be attached, and
the turn does not contain a related main finite clause for which the
unit is syntactically superfluous. A large range of types and functions
were exhibited in their speech. The types can be classified into six
main groups:

noun group—e.g. a little toy

prepositional phrase—e.g. at the door
adjective group—e.g. polite really polite
adverb group—e.g. more or less
verb group—e.g. compare, looking down
sub. Clause— e.g. because he is running with a handbag. (p. 68)

These formulaic units were used in a remarkable range of conversa-

tional functional contexts and for a wide variety of pragmatic purposes.
The purposes included repetition, questioning, agreeing, confirming,
clarification, and focusing attention, among others. Bygate notes as
well that oral productions such as these must be supported by the
surrounding spoken discourse which provides a meaning context, and
that speakers tend to use their interlocutor to adjust their choice of
expression. Obviously, conversation of this type is glued together and
 Formulaic Sequences 53

built of prefabricated particles of speech which depend heavily on

the conversational context and the ongoing co-construction of mean-
ing. After all, these are the features of spontaneous spoken interaction
as a whole.

Protolanguage and Formulas

Wray (1998) elaborates a theory of the evolutionary origins of formulaic

sequences in human language and their connection to aspects of
acquisition and production. In producing a model of the evolution
of language from primate protolanguage, she highlights the value of
formulaic sequences in the development of pragmatic knowledge and
the ability to engage in real-time performance.
Wray points out that many high level primate species produce a range
of holistic utterances which serve to help with social behaviors such as
grooming, as well as inciting desired behavior in others. She posits that
our human ancestors, at an early stage of evolution, developed more
complex systems of such vocalizations. Then, as the human brain evolved
and cognitive skills developed, analytic, generative language became
available, and segmentation of holistic utterances began, but was stymied
by the limitations of short-term memory. In her view, formulaic sequences
were born of a protolanguage from the earliest stages of human devel-
opment, and were retained as a feature of language production because
analytic language abilities proved to be of limited use in real-time
production due to short-term memory boundaries. Similarly, holistic
utterances are still used to maximize limited memory processing space
during speech production.
According to Wray, protolanguage originally had no grammar, but
rather, large numbers of specific utterances were used for specific
pragmatic purposes. Chimpanzees, for example, use different vocaliza-
tions to refer to things, states, or events, with no internal morphological
structure, producing strings and chunks of sounds functioning as free-
standing wholes to achieve a limited number of specific communication
goals. Presumably the first humans communicated in the same way, with
language perhaps evolving during grooming to talk about other mem-
bers of the group, to manipulate or control the behavior of others, to
move communication and awareness beyond mere personal experience
and to get a secondhand perspective on things.
54 Formulaic Language and L2 Speech Fluency

The transition from single-meaning to structured phonetic sequences

was likely very gradual, developing along with the physical evolution of
the speech tract. As well, each individual would have to memorize all
of language piece by piece, in the absence of a grammar or formal
structure of any sort. However, along with cognitive development over
thousands of years, it is probable that rudimentary grammars emerged
from the strings of sounds associated with specific meanings. The emer-
gence of grammar would likely have happened through a gradual
linking of various areas of the brain as awareness of various aspects
of experience developed, such as thematic analysis, conceptualization
of actor, recipient, action. These awarenesses would then carry over
into language so as to express and describe the more complex aspects
of perceived reality. The originators of this rustic grammar probably
had a brain like ours today.
This proposed development in many ways parallels the development
of language ability in children as documented by Wong-Fillmore,
Hickey, Peters, and others. A process of segmentation likely occurred
in the evolution of human language as formulaic sound sequences
were broken into pieces using pragmatics and context, inserting
morpheme boundaries. Holistic sequences may have been refined
into components with regular patterns and functions to accommo-
date grammar, which led to construction by rule. Perhaps the use of
fixed formulaic frames with fillable slots was the first step in this move
toward grammar.

Formulaic Language and Performance

As noted earlier, the importance of formulaic sequences in language

performance, particularly speech performance, has been documented
quite extensively. Studies which have investigated the nature of fluency
in speech, discussed below, have revealed a strong facilitative role of
formulaic sequences in the production of fluent, running speech under
the time and attention constraints of real life communication.
Early research in the area of fluency, in the 1980s, produced mention
of notions of elements of speech fluency which may be interpreted to
imply that formulaic language was of importance. Raupach (1984), in a
study of an adult learner of French, found evidence of formulaic con-
structions contributing to fluency, particularly modifiers and rhetorical
 Formulaic Sequences 55

organizers. He noted that two types of such formulae had primarily

been identified in the literature:

“routines” . . . are utterances which seem to be memorized wholes,

such as what’s that? How are you? and which allow little or no internal
structural variation. In studies on first and second language acquisi-
tion, a clear distinction is often established between routines and
patterns. Utterances in which segments of sentences operate in con-
junction with a moveable component, such as where’s (plus slot
for different noun phrases), are called prefabricated patterns . . . or
sometimes “frames.” (p. 115)

This distinction between two types of such formulae is a useful one, in

that it allows us to see that the formulaic phrases can have at least two
distinct uses. For one, they can, in the case of prefabricated routines,
express more or less complete functions and operate as complete
clauses, allowing the speaker time and attention to planning the
subsequent utterance. The second way is that they can create a struc-
ture within which to insert novel items relevant to the content and
context of speech, buying some syntactic and lexical retrieval and
encoding time.
Formulae of these types are likely generated at many points in the
planning and execution of speech. Dechert (1984a) observed that the
most fluent German students of English, required to retell a narrative in
their second language in his study, appeared to establish “islands of
reliability” of ideas and language, around which they pieced together
a spoken narrative. Sajavaara (1987), in a reflection on a wide range of
factors affecting second language speech, observed that retrieval of a
concept or a single lexical item could trigger the release of other lexical
items and phrases for the learner:

A “word” activates, for example, certain frequent and prefabricated

phrases, word combinations, grammatical constraints, selectional
restrictions, semantic concepts and fields. (Sajavaara, 1987: 54)

A store of many aspects of conceptual items and links, lexical items,

phrases, and patterns of language and ideas may be activated by stimuli
in the input or the context. Automatized, proceduralized strings of
56 Formulaic Language and L2 Speech Fluency

language could then be generated appropriate to the ideas linked to

the stimuli, while more specific items and constructions can be placed
with or within the formulae. In this way, fluent speech may be
generated.
It seems likely that fluency is greatly enhanced by the control of large
numbers of formulaic sequences and sentence stems, at least in English,
a relatively uninflected language, although fluency can also be achieved
by using automatized individual words in familiar situations. Pawley
and Syder (1983) refer to the need for mastery of a body of lexicalized
sentence stems to achieve fluency:

A lexicalized sentence stem is a unit of clause length or longer whose

grammatical form and lexical content is wholly or largely fixed; its
fixed elements form a standard label for a culturally recognized
concept, a term in the language. (Pawley & Syder, 1983: 191)

Thus, a string or frame is needed for expression which links to the

concept or part of the concept to be expressed. These prefabricated
pieces are often strung together in a way appropriate to the communi-
cative situation. This allows the speaker’s energy or attention linked
with single lexical units in the speech run to be freed up to plan larger
stretches of speech. Many of the most familiar concepts and speech acts
can be expressed formulaically, and, if a speaker can pull these formulas
readily from memory, that is, if they are proceduralized or automatized,
fluency is enhanced. This reduces the amount of planning, processing,
and encoding needed within clauses. It gives the speaker time to pay
attention to the multitude of other tasks necessary while speaking, such
as generating specific lexical items, planning the next unit of discourse,
syntactic processing of novel pieces, and so on.
The hallmarks of fluent speech usually include pauses at clause junc-
tures and a certain minimal length of speech runs between such pauses.
The way to accomplish this seems to be the recall of many clauses as
more or less intact, or automatically chained. Pawley and Syder assert
that only a minority of spoken clauses are novel (Pawley & Syder, 1983:
205), and that memorized chunks form a high proportion of the speech
of everyday conversation. As expected, the memorized sequences need
little encoding. Since speech is therefore not produced word-for-word,
the speaker can focus on rhythm, variety, combining memorized chunks,
 Formulaic Sequences 57

or producing creative connections of lexical strings and concepts. The

degree of novelty in expression is often due to the proportion of use
of 2-word lexical units or collocations and the use of entire clauses
which are formulaic.
Peters (1983) sees formulas as being primarily a shortcut in
communication:

For mature speakers of a language . . . formulaic speech may serve as

a shortcutting device: It saves processing time and effort, allowing the
speaker to focus attention elsewhere, for instance on the social
(opposed to the linguistic) aspects of a situation . . . (p. 3)

Peters also states that certain expressions or variations on them are so

useful it is convenient to be able to retrieve them in as prefabricated
a form as possible. (p. 85)
It appears that, at least in English, the retrieval and combining of
lexical chunks and frames is the foundation of fluency. Fluency in a
highly inflected language such as Finnish might involve less reliance on
clause-chaining (Sajavaara, 1987). In any case, any stretch of discourse
can only be planned a certain number of words ahead of articulation,
so, for the second language speaker there is a high degree of risk of
breakdown inside clauses, a typical feature of disfluent speech. Having
a broad and highly automatized store of memorized clauses and clause
stems or frames gives the second language speaker a chance at approach-
ing native-like fluency. Given that, as found by Pawley and Syder, normal
conversational English does not entail a high degree of syntactic inter-
action between clauses, planning is simplified. If clauses can be retrieved
and produced automatically, without need for individual planning, then
the speaker generally should be able to be a clause or two ahead of
himself as he articulates. This can buy a considerable amount of time
and freedom of effort to attend to issues of vocabulary, articulation,
suprasegmental aspects of pronunciation, and so on.
Clearly, there are some genres of speech in which the cognitive load
and need for creative construction are so high as to mitigate against
such heavy use of automatized chunks. Situations involving sustained
spontaneous monologic speech or which require processing of novel
or complex content would normally require a greater proportion of
controlled processing and complex formulation of utterances, as was
58 Formulaic Language and L2 Speech Fluency

the case with the speaker Q in Pawley and Syder’s examples cited earlier
(1983: 202–204).
Wray (2002: 97) summarizes the value of formulaic sequences in
speech production by listing functions of formulaic language in speech.
She sees formulas as aiding in the manipulation of information content
in speech, that is, as a means of controlling the nature and flow of infor-
mation. They also serve as tools for allowing time for mental processing
of creatively constructed strings or in conceptualizing ideas to be uttered
later. She uses the term “textual bulk” to describe the quality of this
function of formulas, in that they allow for a continuing flow of speech
to occur while the conscious mind is focused elsewhere in the commu-
nication process. In addition, Wray notes that formulaic sequences have
the function of shortening the processing route of speech by bypassing
the need for assembly of components or use of short-term memory.
A final function of formulaic sequences is that they help in signaling the
organization of spoken discourse.
Boers et al. (2006: 246, 247), in a study of learner fluency develop-
ment, note three categories of benefits L2 speakers may gain from a
command of formulaic sequences. First, many sequences may assist
learners to sound native-like because of their idiomatic nature—their
meanings are not predictable by rules of grammar or the sum of the
meaning of their lexical components. Secondly, retrieving formulaic
sequences directly from memory can improve spoken fluency by reduc-
ing hesitations and increasing the length of runs. Thirdly, the sequences
which are recalled and stored correctly may “(. . .) help speakers reach
a degree of linguistic accuracy, because these prefabricated chunks
constitute ‘zones of safety’ and appropriate use of them may thus con-
fine the risk of ‘erring’ to the spaces in between the formulaic sequences
in one’s discourse.” (p. 247)
As speech fluency research seems to have also discovered, the value of
the role of automatized formulas is hard to overstate. It is important
to note that formulas are not just common clichés or very frequent col-
locations, but that they function in all registers and in highly specialized
content domains. They may have quite low general frequency of occur-
rence but be quite frequent within certain cultural groups or in specific
genres. For example, among vocabulary specialists or applied linguists
the collocations lexical phrase or formulaic sequence are frequent and
accepted, whereas these particular multiword units are virtually unknown
 Formulaic Sequences 59

outside of these academic contexts. In any event, the wholesale produc-

tion of chunks retrieved automatically from long-term memory appears
important in everyday spontaneous communication. In discussions of
the role of formulaic sequences in production, the need for economy in
mental processing is seen as primary. The consensus among those who
have studied formulas in language production seems to be that their
prime value is in lightening the attentional and processing burdens of
utterance construction and allowing for fast and fluid communication.
Phonological reduction may also occur, involving phonological fusion,
reduction of syllables, deletion of schwa, all of which are common
in production of most high frequency phrases in English, but much
rarer in lower frequency or novel utterances, according to Bybee (2002).
Phonological reduction can be seen as a sign that “much of the produc-
tion of fluent speech proceeds by selecting prefabricated sequences
of words” (Bybee, 2002: 217). See Lin (2010) for an up-to-date review
of the research related to phonological characteristics of formulaic
sequences.
 Chapter 3

Cognitive Processing

Thus far, a picture of the role of formulaic sequences in language

acquisition and production has emerged. Formulaic language is basic to
communication, likely having its origins in an early protolanguage of
our primate ancestors. In addition, the limitations of working memory,
combined with the time and attention constraints imposed by real-life
spontaneous communication, have ensured that formulaic language
continue to be important. Both children and adults appear to make use
of formulas in language acquisition. Children use them as a database
for grammar acquisition, and there is evidence that some adults may do
so as well.

Mental Processing and Fluency

Key to the entire process of spontaneous speech production is the work-

ings of mental processes and skills. In cognitive theory in general there
is agreement on a number of concepts which can help in understanding
how mental processing of language contributes to speech fluency.
One key element of cognitive theory which relates to mental process-
ing and fluency is the distinction between two types of knowledge,
declarative knowledge and procedural knowledge. Declarative knowledge
relates to consciously known content and information, while procedural
knowledge relates to knowledge of how to do things, and is related to
performance of skilled behavior. A related concept is automatization
or proceduralization, by which declarative knowledge is converted into
procedural knowledge through repeated recall and use. In this process,
knowledge which has been stored in the mind declaratively is trans-
formed into procedural knowledge so as to speed up skilled performance
and allow a more efficient use and implementation of content and
 Cognitive Processing 61

information in real-time performance (Anderson, 1983; Levelt, 1989).

See below for a discussion of theoretical perspectives on how automati-
zation may take place with formulaic language to facilitate fluent speech
production.
The concepts of long term and short-term memory also help to provide a
means of understanding how language may be processed mentally.
Long-term memory is a repository of all kinds of knowledge, including
knowledge about language such as syntactic and morphological rules,
semantic information, and lexis. For this information to be used in
speech performance, however, it must first be assembled in short-term
memory. Lexical items need to be selected to express concepts, and syn-
tactic, morphological, and phonological rules applied to construct utter-
ances. However, short-term memory capacity in humans is limited to
roughly seven or eight discrete items at a time, making construction of long
and complex sentences and utterances laborious and time-consuming
(Anderson, 1983). Baddeley (1988) posits the existence of working mem-
ory, a different conceptualization of short-term memory, and the site
where declarative knowledge taken from long-term memory is assembled
or rehearsed. In particular, he proposes a phonological loop of working
memory where necessary lexical elements encountered in language
input or from long-term memory are rehearsed. The piece or string of
knowledge or language is silently repeated and rehearsed in the phono-
logical loop to allow the form-meaning relationship to be retained and
possibly accessed as a whole from long-term memory in future. A com-
mon example of this is the mental repetition of a newly heard seven-digit
telephone number in order to facilitate recall later. Robinson (1995)
posits that language input can be converted into language intake through
this type of process, as chunks of input are rehearsed and repeated in
short-term memory. Hulstijn (2001) likewise concludes from his research
that rehearsal is an essential process in lexical acquisition.
A final dichotomy often used to frame discussions of language
processing is that of controlled and automatic processing. Controlled
processing is seen as a conscious process of selecting and manipulating
language knowledge, which is amenable to the use of learned rules and
parameters. Automatic processing, on the other hand, is seen as a fast,
effortless process of retrieving elements from long-term memory, which
occurs without conscious attention or below the level of a speaker’s
awareness (McLaughlin, Rossman, & McLeod, 1983).
62 Formulaic Language and L2 Speech Fluency

A deeper discussion of these current understandings or generally

agreed upon points can provide a framework for understanding first and
second language speech production from an information processing
perspective.

Declarative and procedural knowledge and automatization

One psycholinguistic model with implications for fluency development
is that of Anderson (1983) and his Adaptive Control of Thought (ACT*)
model of cognitive development, later modified as ACT-R (1993). In
this model, there are two types of memory stores, long-term memory and
working memory, the latter having a limited capacity. The long-term
stores deal with the outside world via the working memory. For behavior
such as speech production, which requires rapid performance, conver-
sion of declarative knowledge into procedural knowledge is essential.
This is because working memory has a very limited capacity, and declar-
ative knowledge requires attention and takes up much more “space”
than knowledge which does not require the focal attention of the
speaker. Procedural knowledge, on the other hand, does not require
focal attention and can be processed by working memory in larger units
without running out of working memory capacity.

Automatic and controlled processing

McLaughlin, Rossman, and McLeod (1983) distinguish between the
two mental processes of controlled and automatic processing as these
apply to second language production. Citing Shiffrin and Schneider,
they describe memory as “a large collection of nodes that become
‘complexly interassociated’ through learning.” (McLaughlin, Rossman,
& McLeod, 1983: 139) These nodes can become activated by either
controlled or automatic processing. Controlled processing is described
as a response which is not yet learned, but temporary. In controlled
language use, attention by the subject or speaker is required. Automatic
processing causes certain memory nodes to activate every time certain
appropriate types of stimuli occur. Consistent, repeated activation of a
group of nodes in a certain pattern by the same type of input over time
leads to a learned, automatic response process. This process is extremely
rapid, and requires little or no effort or attention. In contrast, only one
 Cognitive Processing 63

sequence of activation of “nodes” can be controlled at any time. Schmidt

(1992: 360) categorizes controlled processing as slow, inefficient, effort-
ful, limited by short-term memory capacity, under voluntary control,
and at least partly accessible to introspection. It is extremely important,
though, because it allows novelty and flexibility in expression. Auto-
matic processing is described as fast, efficient, effortless, not limited
by short-term memory capacity, not under voluntary control, and
inaccessible to introspection. It is efficient, but not appropriate where
a novel response is required. Controlled processing is particularly
important in situations requiring a high degree of novel construction or
in which sustained or highly complex content is required. Evidently,
both types of processing are in constant interaction in normal fluent
discourse. Once a response sequence, originally requiring attention
and control, has been used repeatedly and become automatized, then
attention and effort can be used to perform other tasks at the same
time. In this view, strengthening of the connections among given nodes
to develop automaticity requires repeated exposure to the input,
together with practice.
Over time, researchers have tended to see controlled versus automatic
processing as being more a spectrum than a dichotomy (DeKeyser, 2001).
Some, such as Kahnemann and Treisman (1984) have conceptualized
the two processing modes as having various levels: a process may be
strongly, partly, or occasionally automatic. For fluency to be achieved in
speech, a learner must use automatic processing for certain elements of
a given utterance, and save attention and control for other elements
such as expression of new or complex content. However, it is not a given
that any particular part of a production will always be produced using
fully automatic processing. A formulaic sequence may be produced auto-
matically in one social, discourse, or linguistic context, partially so in
another, and be assembled using controlled processing in another.

A Model of Speech Production

A highly influential and comprehensive model of speech production is

that of Levelt (1989). Levelt’s model of speech production represents two
kinds of knowledge: declarative and procedural. Declarative knowledge
is knowledge “that,” or knowledge about the world, while procedural
64 Formulaic Language and L2 Speech Fluency

knowledge is knowledge “how,” or that which underlies skilled behavior.

Levelt argues that due to the nature of working memory and the speed
with which speech is usually produced, fluent speech requires proce-
dural knowledge. Production involves three stages: conceptualizing;
formulating; articulating.

Conceptualizing stage: This involves procedural knowledge in develop-

ing the propositional content of the intended message, accessing
declarative knowledge of the information to be expressed, the situa-
tion, and discourse organization. A propositional preverbal message
goes to the formulator.
Formulation stage: The propositional message is given an acceptable
form by a formulator which takes the semantic and pragmatic mean-
ings of the preverbal message and searches the lexicon for a means
of expression. The mental lexicon contains form/meaning pairs
called “lemmas,” which are combinations of the conceptual meanings
of lexical items, linked with their syntactic features. The formulator
selects an appropriate lemma from the lexicon and builds utterances by
combining lemmas in a way relevant to their syntactic requirements and
the meaning to be expressed. The formulator then produces a surface
syntactic form and phonologically encodes the utterance.
Articulation stage: The phonetic plan is then passed on to the articula-
tor which produces explicit speech.

All these speech operations, initially under conscious control, gradually

become proceduralized in the production system of fluent speakers. The
speech production process is continuous, without feedback between the
preverbal message and the phonetic plan, since the procedural stages
involved in formulation run automatically to meet speed requirements.

Theories of How Automatization Occurs

In instance theory, Logan (1988) posits a process by which automatization

occurs, based on memory. Logan’s assumptions about how memories
are stored and retrieved have to do with a process of substitution. Unlike
Anderson, who held that underlying processes do not change, but
simply speed up and form proceduralized chunks, Logan theorizes that
the initial utterance construction process is replaced over time by a
 Cognitive Processing 65

qualitatively different process which is simpler and faster. This process is

one of single-step memory retrieval, driven by storage of exemplars or
key sample items representing rules or concepts. The status of a given
repeated output over time is determined, in this view, by a sort of com-
petition between the original step-by-step declarative sequence, and
memory retrieval. Eventually, if the sequence is produced frequently
enough, memory retrieval dominates, being a more efficient means of
producing strings of language under the time and attention pressures
of spontaneous speech. Thus, in Logan’s view, it is not the internaliza-
tion of rules or procedures which accounts for fluent speech production,
but rather, the shortcut of direct memory retrieval, likely of lexicalized
chunks (Skehan, 1998). Later theorists have said that the process of direct
memory retrieval likely applies to instances which are similar or which
belong to similar categories, and that it does not apply only to situations
in which the same instance is repeated to allow single-step memory
retrieval (DeKeyser, 2001).
A related cognitive psycholinguistic conceptualization of how auto-
matization or proceduralization occurs is restructuring (Cheng, 1985;
McLaughlin, 1990). Restructuring models, like instance theory but in
contrast to the Anderson or Schiffrin and Schneider models, see
improved and faster skilled performance as the result of changes in the
organization of knowledge rather than mere speeding up of the pro-
cesses of storage and retrieval. In restructuring, underlying patterns or
rules are substituted for the entire strings of memory retrieval required
to access declarative knowledge through controlled processing.
It has been posited that frequency of exposure and experience with
language input may be the basis of language acquisition and automatic
processing of language. From such a connectionist perspective, Ellis
(2002: 144) sees language learning as a process of developing a set of
associations of statistical probabilities of how language is used to accom-
plish pragmatic functions:

Language learning is the associative learning of representations that

reflect the probabilities of occurrence of form-function mappings. Fre-
quency is thus a key determinant of acquisition because “rules” of lan-
guage at all levels of analyses (from phonology, through syntax, to
discourse), are structural regularities that emerge from learners’ life-
time analysis of the distributional characteristics of the language input.
66 Formulaic Language and L2 Speech Fluency

The idea is that language production and comprehension is determined

by a huge amount of implicit statistical information about the behavior
of various items of language such as parts of speech, and, especially,
lexical items. Acquisition proceeds through experience with input which
exhibits distributional properties such as the relative frequencies
with which certain verbs appear in different tenses, as active or passive
constructions, and so on.
Ellis (2002: 145) draws a distinction between explicit and implicit
memory, important to understanding how memory-driven exemplar-
based learning can occur. Explicit memory is conscious, implicit
memory less so:

(explicit memory) is a conscious process of remembering a prior

episodic experience or fact such as questions like what did you have for
breakfast? or What is the capital of Andorra? . . . (implicit memory)
involves facilitation of the processing of a stimulus as a function of
encounters with an identical or related stimulus but where the subject
at no point has to consciously recall the prior event.

Logically, the frequency-based development of knowledge of statistical

frequency distribution of form and function mapping theorized by
connectionists is largely based on implicit memory. Humans are con-
scious of communicating, focusing on meaning first and foremost,
but in the course of communicating we naturally acquire the knowledge
of the frequencies of the elements of language and their patterns,
combinations, and mappings.
Connectionist views of language learning are based on cognitive
science models of human mental categorization ability. The basic tenet
of connectionism is that each repetition of a stimulus increases the
strength of memory connections between a particular mental category
to which the stimulus belongs and the features of the stimulus which
allow it to be assigned to the category. In assessing whether an entity
belongs to the same category as other previously encountered entities,
one retrieves the memories of past encounters and assesses their simi-
larity to the new one, which is then classified accordingly. Such learning
through association is seen as the “raw basis of human cognition” (Ellis,
2002: 146). The model applies to language acquisition in that such
learning is seen as exemplar-driven and largely implicit. Although
 Cognitive Processing 67

implicit, the process is not exclusive of conscious control. Schmidt

(1990, 1993) theorizes that attention and noticing of features of lan-
guage input allow learners to make hypotheses about how language
works and even how their own output compares to native-like norms.
Ellis (2002: 145) allows that such conscious focus is important in that it
allows initial registration of a pattern-recognition unit which then can be
the basis of a category or an exemplar to be classified. Bishop (2004)
conducted a research study on noticing of formulaic sequences by
L2 learners in word recognition lists, and found that learners tended
not to notice formulaic sequences which they did not already know as
frequently as they noticed words which they did not already know. This
may be due to the need for L2 learners to identify holistic formulaic
sequences as distinct from similar looking grammatically generated
word sequences, or to match multiword lexemes with a single lemma
(meaning) (Bishop, 2004: 18).
Associations and classifications are made and strengthened through
multiple cue sources, according to research by connectionists. An exam-
ple is the determination of how children come to segment words from
continuous speech which they hear, that is, word boundary perception.
The research evidence is that using phonotactic information alone, a
child can achieve 47% success. Adding utterance-boundary and rela-
tive stress information boosts the success rate to 70% (Ellis, 2002: 140).
There is also evidence of cohort effects in lexical retrieval, in that
retrieval or exposure to the initial phoneme of a word activates other
words in the lexicon with that initial phoneme. As more information is
retrieved we narrow down the possibilities and high frequency words
get much more such activation (Ellis, 2002: 141). For formulaic
sequences, as with all lexical items, it is likely that they are automatized
through repeated exposure and frequency in input due to the
pragmatic requirements of the communication contexts that learners
encounter regularly. They are likely stored and retrieved automatically
over time through a variety of cue sources, including initial phoneme
classification and more.
Phonological memory is a psycholinguistic concept which appears
relevant to formulaic language processing and speech fluency. In the
past 15 to 20 years, a growing body of research has been published
linking phonological memory, a component of working memory, to
language learning. The basic model of working memory, as elaborated
68 Formulaic Language and L2 Speech Fluency

by Baddeley (2000) and others, is that language production involves the

short-term retention of aspects of language in a cognitive loop. The
loop has several components: a visual-spatial sketchpad which holds
visual and spatial information related to the target language items; a
phonological loop which deals with verbal information and is linked to
phonological memory; an episodic buffer which integrates information
from the sketchpad and phonological loop with long-term memory.
Phonological memory is said to facilitate language acquisition through a
process of holding phonological information temporarily over and over
until a permanent or long-term memory representation can be created.
This aspect of the working memory model has been studied quite exten-
sively in laboratory settings, and considerable evidence exists to confirm
the existence of a phonological loop (see Baddeley, Gathercole, &
Papagno, 1998). While studies of adult L2 acquisition and phonological
memory are numerous and have actually indicated that phonological
memory plays a significant role in vocabulary acquisition, a limited body
of work has investigated possible links between phonological memory
and oral L2 fluency.
A key study which focuses on phonological memory and L2 fluency
gain is that of O’Brien et al. (2007). The researchers moved the phono-
logical memory research out of the laboratory and into the real world of
language learning, comparing the development of Spanish L2 fluency
of native speakers of English in a regular university language program
and a study-abroad semester. Psycholinguistic tests of phonological
memory were conducted before and after a semester of study and
examined for correlation with gains in speech fluency over the same
period of time. Results indicated that, independent of the learning
context, phonological memory actually appears to be related to gains
in L2 fluency on measures of speech rate and length of runs. One could
speculate that the effect of phonological memory on the ability to retain
and produce formulaic sequences may be a key component of the
fluency gains of the learners in this study.
While it is not clear which of these explanations, instance theory,
restructuring, connectionist models, or phonological memory best
characterizes how knowledge becomes automatized or proceduralized,
there is agreement on one point which is of crucial relevance to the
study of formulaic sequences in second language speech fluency. The
ability of a speaker to produce given sequences is dramatically speeded
 Cognitive Processing 69

up with time and practice through psycholinguistic mechanisms, and in

the process the sequences change in nature to phonologically coherent
units retrievable as fixed chunks. Therefore, production becomes faster
and there is a qualitative change to the mental procedures underlying
speech production.

Storage and Retrieval of Formulas

One key issue in the psycholinguistic aspect of the examination of

formulaic language is the question of what is meant by “produced or
recalled as a whole.” As suggested above, there are several possible
answers to this question. According to Weinert (1995), the first could be
that formulas are recalled based on the linear surface order of their
parts, or by their phonological units, following a particular type of
lexical recall . . . first and last words of a unit might be most prominent
in memory, and act as triggers for total recall. This is echoed by the
cohort element of recall discussed by Ellis (2002) in which recall of the
initial phoneme of a formulaic string triggers the activation of a mental
network of connections related to all lexical items which contain that
phoneme in the initial position, and in which most frequent strings
are activated the most strongly. Secondly, it could be that formulas are
stored as “cognitive bundles,” retrieved in various ways, depending on
their form, pragmatic aspects, and so on. Indeed, this links to the con-
nectionist view that, based on frequency of input, lexical items such as
formulaic sequences may be retrieved based on various cues and that
such retrieval may be speeded up as repeated exposure and experience
activate the response sequence over time and implicit awareness develops
of how such strings function and are used. It may be that, in accordance
with the variability of automatic and controlled processing suggested by
Kahnemann and Treisman (1984) there is a continuum of creative/
formulaic-holistic processing, for example, different for fixed strings as
opposed to frames which accept different inserts. This could mean that
a formula is sometimes retrieved automatically, sometimes partially so,
and sometimes by controlled processing.
How might a string acquire formulaic status? As Peters (1983) indi-
cates, it may be simply due to a perception of meaning in chunks
extracted from input by a learner, then stored as a whole. Or, according
70 Formulaic Language and L2 Speech Fluency

to Weinert (1995), it might be either due to initial segmentation of the

input, followed by fusion or combination of formulas, or due to lexical-
ization of syntactic strings through frequent production. This latter
process would involve automatization of the string and storing it as a
piece of procedural knowledge, no longer available for analysis. It has
been shown that the brain can use automatization to bypass the process-
ing route that was used to learn an item (Wray & Perkins, 2000). It
may be that consciousness, awareness and noticing of salient formulas
in input (Schmidt, 1990) leads to an initial registration of a pattern-
recognition unit, which then is consolidated by frequency in input
and leads to single-step memory access. Or, it could be that frequent
production of a formula, requiring assembly in working memory, leads
to its automatization over time, or its retrieval in a single-step process of
remembering as posited by instance theory (Logan, 1988).
An intriguing aspect of the storage and retrieval of formulas as
wholes is that they appear to be articulated as wholes as well, exhibiting
phonological coherence. Bolander (1989), remarks that this quality of
the production of formulas could be due to chunk processing, quoting
Ladefoged (1972: 282):

There is a great deal of evidence that muscular movements are

organized in terms of complex, unalterable chunks of at least a
quarter of a second in duration (and often much longer) and nothing
to indicate organization in terms of short simultaneous segments
which require processing with context-restricted rules. (p. 85)

This process of storage and retrieval of sequences of small muscle

movements adds to the evidence for formulaic sequences as a real and
necessary part of human communication, suggesting that humans are,
so to speak, built and wired to use them.
Peters (1983), looking at child L1 acquisition, suggests that formulaic
sequences are a result of identification of meaningful chunks in the
input, and that the resulting multiword chunks are used for specific func-
tions. This is at least partly echoed by adult L2 acquisition research (Ellis,
1996; Schmidt, 1983). Other researchers (Fillmore, Kay, & O’Connor,
1988: 504) remark that there are “. . . things which are larger than words,
which are like words in that they have to be learned separately as indi-
vidual whole facts about pieces of language.” It is clear that formulaic
 Cognitive Processing 71

sequences exist, but the question of how they achieve storage and
retrieval as wholes is still uncertain. It may in fact be a result of a combi-
nation of processes, including segmentation of chunks from input based
on frequency or pragmatic salience, repetition and practice, and an
interplay between automatization and instance retrieval.

Evidence of Faster Processing of Formulaic Sequences

Despite the large amount of theory and logical reasoning behind the
idea that formulaic sequences are processed faster than other stretches
of language, there is a surprising lack of hard evidence that this is in fact
the case. However, psycholinguistic research which seeks to provide a
view of how formulaic sequences are processed has emerged recently.
Jiang and Nekrasova (2007) administered two online grammaticality
judgment tasks to native and nonnative English speakers, using formulaic
and nonformulaic phrases as prompts. Results indicated that both
groups responded significantly faster to the formulaic phrases than the
nonformulaic. Conklin and Schmitt (2008) conducted a similar sort of
study, comparing reading times for formulaic versus nonformulaic
phrases with groups of native and nonnative English speakers. Like
Jiang and Nekrasova (2007), they found both groups read the formulaic
phrases significantly faster than nonformulaic. Both of these studies
indicate a faster processing time for formulaic phrases, but are restricted
to reading or recognition rather than language production. In a multi-
study work, Siyanova and Schmitt (2008) looked at the efficiency and
speed of processing and written production of a range of adjective-noun
collocations, finding that L2 learners tended to produce appropriate
collocations in writing, but that they had rather poor intuitions about
the frequency of collocations and were slower than native speakers
in processing them. While this study deals with both processing and
production of formulaic sequences, it does not address spoken language.
In sum, evidence exists for faster processing of formulaic sequences but
there is little to show that online speech production of formulaic
sequences in L2 is faster or more efficient. The study reported below is
a step in that direction.
 Chapter 4

Social and Cultural Factors

Second Language Acquisition (SLA) theory has tended to draw a

division between the language user and the context in which the
language is being used. In addition to a cognitive or linguistic expla-
nation of L2 performance, there is a need to reconceptualize the
individual and the social in SLA, as the relationship between the
learner and larger social processes has often been overlooked. Speech
is an interactive process, and its performance is integral to identity,
acculturation, and the development of cultural fluency and the
emergence of voice in the target language. Discussions of L2 spoken
language performance must take into account the dynamics of how
the way in which one speaks influences how one is perceived by others,
as well as how the act of communicating in L2 is influenced by a range
of factors beyond the linguistic: first language, culture, identity, and
issues of voice and attitudes, among others.
Four related sociocultural concepts which relate to L2 performance
are anxiety, self-efficacy, voice, and social identity. Discussions of
language anxiety are heavily influenced by early social psychological
work on anxiety which took place within a largely positivist paradigm.
The literature on self-efficacy, voice, and social identity stem from other
psychoanalytic or poststructuralist paradigms. However, all four of
these factors can help inform how sociocultural issues play a role in
speech performance in L2.

Language Anxiety

Language anxiety is one factor which may affect speech fluency, both
in performance and in the development of fluency, which requires
practice. It is linked to the concept of communication anxiety, defined
as “level of fear or anxiety associated with either real or anticipated
 Social and Cultural Factors 73

communication with another person(s)” (Beatty, Balfantz, & Kuwabara,

1989: 278). Researchers studying communication and speech anxiety
have traditionally operated on a strongly positivist, clinical and experi-
mental base. They typically have distinguished between trait and state
anxiety as measures of anxiety. Behnke and Sawyer (1998) discriminate
between the two: state anxiety is viewed as a transitory state or condition
of the organism which varies in intensity and fluctuates over time; trait
anxiety is described as a personality trait or an inherited personality
characteristic which is antecedent to social experience and not learned
or dependent on social or learning experiences. Discussions of state
and trait anxiety have a long history in experimental social psychology
and psychometric tools have been developed to measure them. Com-
munication anxiety may be either state or trait in nature, depending on
the nature of the learner and/or the learning context.
The distinctive type of anxiety connected with language learning has
been connected with the quality of oral language production, making
students appear less fluent than they really are. Perceived lack of ability
can have negative consequences for one’s success in school, social life,
and overall adaptation to culture and society in L2 (Horwitz, Horwitz, &
Cope, 1991). Fear of public embarrassment is key to language anxiety,
and two particular personality variables have been linked to it, perfec-
tionism and fear of public speaking, as well as a belief that one’s skill
level is lower than that of classmates or others (Price, 1991: 106).
While language anxiety may play a role in fluency development in L2,
alternative and more post-structural views of sociocultural factors influ-
encing communication have emerged which can enrich understanding
of how social processes affect communication.

Self-efficacy

Another key social element of fluency is the individual’s sense of self-

efficacy. As developed by Alberto Bandura (1986) as an element of social
cognitive theory, rather than the experimental and positivist paradigm,
self-efficacy refers to the development of perceptions about the self and
one’s abilities and characteristics. These guide behavior by influencing
what one tries to achieve and how much effort to put into performance.
This is linked to Bandura’s notions of reciprocal determinism, in which
the environment, personal factors, and behavior interact. In this,
Bandura drew on Adler’s personality theories which held that behavior
74 Formulaic Language and L2 Speech Fluency

is purposeful, motivated by the pursuit of goals, that perception of and

attitude toward the environment influence behavior, and that thoughts,
feelings, and behavior are transactions with one’s physical and social
surroundings.
While there is no substantial body of literature on the subject, a sense of
self-efficacy could influence and be influenced by speech fluency in L2.
A high degree of self-efficacy built by a history of positive experiences as
a communicator, a sense of belonging to an empowered group, and as a
speaker of L1 and/or L2 could encourage self-efficacy as an L2 speaker.
This could lead to the early emergence of voice and a sense of power in
L2 discourse. Linked as it is with confidence and assertiveness of the self,
self-efficacy could lower the barriers to communication, encourage risk-
taking, and allow expression in the L2 to occur more easily through
practice and experience and exposure to more input. Less reliance on con-
trolled processing and a facility with formulaic language units of relevance
to the speech context could be the result.
This ties in with the influence of anxiety on language ability, in that it
appears that accumulated perceptions of self and one’s relationship to the
world can play an inhibiting role in the motivation to speak, which in turn
limits the development of speech fluency.
Related ideas lie in self-determination theory (Deci & Ryan, 1985; Noels,
Pelletier, & Vallerand, 2000), which distinguishes among several aspects
of motivation. According to self-determination theory, there are two
fundamental categories of motivation: intrinsic motivation, defined as
motivation based on intrinsic interest in an activity, and extrinsic moti-
vation, based on rewards which are external to the activity itself. These
two categories are not mutually exclusive, but are situated on a spectrum
of self-determination. Intrinsic motivation may be based on enjoyment
and satisfaction in engagement in an activity, and the pleasure may be
derived from feelings associated with new ideas and knowledge develop-
ment, from feelings related to mastering a task or achieving a goal, or
from feelings of fun, excitement, or esthetic appreciation.
Extrinsic motivation may relate more closely to self-efficacy. At base,
extrinsic motivation implies that one engages in activity in order to earn
a reward or to avoid punishment, but such motivation can be situated
along a spectrum of integration with the self-concept, or how it is self-
determined. Three general categories of such extrinsic motivation have
been identified: external regulation, introjected regulation, and identified
regulation, in order from lowest to highest level of self-determination.
 Social and Cultural Factors 75

External regulation relates to activity motivated wholly from sources

external to the person. Introjected motivation refers to activity per-
formed due to external pressures which have become internalized,
for example, speaking an L2 out of a feeling of shame associated with
not doing so, or to reduce guilt. The most self-determined is identified
regulation, in which activity is performed for reasons personally rele-
vant, for example, in order to achieve a cherished goal.
It appears that self-determined motivation is useful in explaining the
ways in which social and cultural factors influence L2 learning and per-
formance. In many cases in which learners study abroad, such as those in
the study reported later in this volume, it may be the case that any of the
three categories of extrinsic motivation may apply to their activity. Some
may be motivated purely by the expectations of their parents or family, or
by certain requirements of the education system in their home countries.
In such situations, it is unlikely that a learner would invest in communi-
cating in the L2 often or for sustained periods, thus generating little
input. Consequently, it could be expected that L2 acquisition would be
slow and based largely on classroom experience, and that development
of a repertoire of formulaic sequences could be inhibited. Similarly,
introjected regulation could have inhibitory effects on input generation,
connection with the L2 community, and comfortable use of formulaic
sequences for learners. Participating in a study abroad program out of
guilt or duty, while perhaps requiring more personal investment than
simply because of parental or community expectations, is unlikely to
produce much desire to engage with the L2. Identified regulation may
be a better basis for L2 learning abroad, as in cases in which one studies
the L2 in order to enter degree programs at a university in the L2 coun-
try. All in all, it is likely that extrinsic motivation is a dynamic construct
which may shift over time in individuals due to the effect of living in
an L2 milieu. In any case, self-determination, in conjunction with self-
efficacy, likely has strong and deep influences on the development of
fluency and use of formulaic sequences in learner speech.

Voice

While little research has been conducted into fluency and the concept
of voice in communication, there are links to be made. Voice in the
sense of “voice with authority” (Elbow, 1994) is clearly linked to the
76 Formulaic Language and L2 Speech Fluency

development of fluency in speech, as it can influence others’ perception

of competence in speech or mastery of communication patterns and
elements which may include formulaic sequences. Voice with authority
is the sense of voice that is frequently used in feminist research (Belenky
et al., 1986), and it relates to the possession of the authority to speak or
wield influence in a group. Voice in this sense can also be referred to as
the “speaking consciousness” (Wertsch & Smolka, 1993). The social
learning theories of Russian theorists such as Vygotsky and Bakhtin have
contributed heavily to an understanding of voice and its role in learning
and development of identity and consciousness. Vygotsky (1934/1986)
elaborates a concept he terms “verbal thought” (p. 83), in which external
or outward-directed speech becomes egocentric or self-directed speech,
which then evolves into inner speech, the framework of cognition.
Vygotsky sees language as the set of tools necessary for the shaping of
perceptions and thought, and voice is the outward expression of inner
speech or thought. Similarly, Bakhtin (1986) posits that language is
learned through contextualized social interaction, and that all language
use is dialogic, in that it is contextual and directed toward an audience.
Dialects, idiolects, and genres develop specific to groups, and are loaded
with traces of the values, perspectives, and experiences of those groups.
Voice is the expression of the consciousness which has developed
through group social interaction.
Voice and speech fluency are inextricably linked in a two-way dynamic.
It is through fluent speech that voice can emerge, and fluency can
best be achieved if one’s voice is present in the discourse. The social
context is key to the existence of this interplay; if the power structures
of the situation silence the voice of the group to which a learner belongs,
then the emergence of speech fluency is likely to be inhibited through
low input generation and reduced opportunities for output practice.
It is through speech that one is able to control and make choices in
everyday life, and it is through increasing facility in speech that an L2
learner can gain access to the L2 world and empower himself or herself
therein. In a sense, developing fluency is a process of convergence of
multiple realities, in that becoming fluent in an L2 parallels a process
of altering one’s own reality to more closely fit with that of the L2
community. In fact, spoken language has a strong link to learner
identity, which is, according to Morgan (1997) “. . . a guide with which
ESL students negotiate their place in a new social order and, if need be,
 Social and Cultural Factors 77

challenge it through the meaning-making activities they participate in”

(p. 431). The greater the level of L2 fluency, the greater the ease with
which the learner can achieve a level of comfort in interaction with
L2 interlocutors and make personal connections to enhance life
experience. The sense of voice in the L2 world can emerge through
fluent speech.
Vygotsky’s (1934/1986) model of inner or private speech has implica-
tions for fluent language production and formulaic sequences. The
internalized and reduced language of inner speech can be vocalized
by adults in times of stress, and appears to be used as a metacognitive
aid in planning, guiding, or monitoring the performance of an activity
or as an affective support as in “I did it” upon completion of a task
(McCafferty, 1994). In such cases, one is basically talking aloud to
oneself. Drawing on the earlier work of Wertsh, and Frawley and
Lantolf, McCafferty (1994) outlines three stages of development of
inner or private speech which can also occur in adult L2 learners. The
earliest form is object-regulation, by which the learner describes and names
elements of the environment, for example, naming characters in a
narrative or stating the obvious aspects of a picture such as “it is a boy.”
The second stage is other-regulation, by which a learner uses self-directed
questions such as “what’s this?” to gain control over a task or to substi-
tute for not having assistance from an interlocutor. The third stage is
self-regulation, in which the learner is able to resolve difficulties related
to a task by him or herself, and the private speech shows signs of such
mastery, for example, “OK,” or “now I know.”
McCafferty (1994) conducted a study attempting to show that these
three stages correspond to stages of L2 learner proficiency. Of the 39
participants in the study, whose task was to retell a picture story aloud in
their L2, English, the lowest proficiency group did tend to produce more
object-regulating private speech, but there were no significant differ-
ences among the proficiency levels for the other types of private speech.
In any event, it is worth noting that this phenomenon of private speech
often can consist of formulaic strings such as “I know,” “I don’t under-
stand,” “what is the thing’s name?” and so on. Furthermore, production
of such speech in L2 during an L2 speaking task may indicate a certain
level of automatization of L2 formulas as well as a sense of voice in that
the self is present in the speaking task, despite difficulties or the need to
hesitate to formulate or recall content.
78 Formulaic Language and L2 Speech Fluency

Social Identity

Linked to this, and related to the emergence of voice, is the concept of

social identity. An early theorist, Tajfel (1981), saw social identity as the
product of a process of social categorization, formation of awareness of
social identity, and a search for psychological distinctiveness. In Tajfel’s
view of social identity formation, the social context of learning provides
categories into which individuals, through recognition of linguistic cues,
allocate themselves and others and learn the values and valuation associ-
ated with those categories. Weedon (1987), has a more post-structuralist
concept of social identity, seeing it as being multiple and flexible over
time and contexts, and subject to change. She states that “language is the
place where . . . our sense of ourselves, our subjectivity, is constructed”
(1987: 21). Drawing on this, Peirce (1995) puts forth the idea of learner
investment, arguing that learners have complex social histories and mul-
tiple desires, and may exhibit more or less accurate or fluent language
depending on the context and the interlocutor. As is the case with voice
and self-efficacy, it appears that the role of social identity in speech
fluency is linked to a sensitivity to context, audience, and the speaker
as a social being. Angelil-Carter (1997) found that the learner in her case
study showed a variation in investment in L2 communication over time
and subject to context, and that his sense of the right to speak changed
over time. Context and socially constructed senses of the self and power
make a great difference in fluent speech performance.
We have seen how gender can be a relevant issue in terms of voice,
and an L2 can be an empowering force in women’s lives and can enhance
their abilities to talk fluently—to give them a voice they would not
otherwise easily find in L1. McMahill (1997), dealing with women’s
groups in Japan, found that English language ability was for some a way
to find a voice to talk about otherwise difficult issues:

Fluency in another language, especially English, empowers women

to be more active in the world and thus is seen as congruent with
feminist goals for oneself and for other women . . . native Japanese-
speaking women in particular may perceive English as allowing or
requiring them to express themselves more directly and specifically
than Japanese does and thus undermining prescribed feminine
linguistic roles. (p. 613)
 Social and Cultural Factors 79

Here it would appear that the nature of L1 communication is key to L2

fluency. The social and cultural order of the L1 community and its
effects on voice and social identity can play a heavy role in fluency in L2.
If the nature of the L1 communication is incompatible with the nature of
fluency in L2, there may be a gap in the development of fluent speech in
L2. Still, if the L2 is seen as a liberating and empowering medium because
of its discourse structures, an acceleration in L2 fluency could be expected.
As well, it could be expected that the emergence of a sense of social iden-
tity in a second language would allow learners to include content related
to the self, affect, and perceptions in discourse. This in turn could influ-
ence the range and choice of formulaic language sequences used, with
implications for the development of speech fluency.

Cultural Fluency and the Nature of the First Language

Another key sociocultural point to be considered in discussing L2

speech fluency is cultural fluency (Poyatos, 1984). A culturally fluent
person, according to Poyatos, makes contextually correct behavior
choices from alternatives, developing an ability to move back and forth
between cultures with flexibility. This movement is expressed in linguis-
tic, paralinguistic, kinesic, proxemic, and chronemic forms, and is
context-dependent. According to Barnlund (1987), cultures determine
the content of conversation and the amount of self-disclosure. For
example, fixed patterns of conversation are preferred in Japan, and the
level of spontaneity and self-disclosure there are much more restricted
than in North American conversation. As well, vagueness, ambiguity,
and a slower message delivery format are preferred in Japanese social
discourse as compared to North American conversation.
The linguistic relativity idea of Whorf (in Lee, 1997) supports this
idea of cultural fluency. Whorf sees language capacity as central to the
uniqueness of human cognition, noting that as humans we communi-
cate thoughts through language and, during the process of socialization,
incorporate language into our cognitive processes, which affects the way
the mind functions. As well, Whorf argues that we attend consciously to
certain aspects of our sensory input and impressions of our environment,
relegating others to the background. Part of what drives this process
in any individual is the way that his or her language has encouraged
attention to particular elements of a situation.
80 Formulaic Language and L2 Speech Fluency

This has, of course, implications for fluency in speech. If the first

language is built on different abstractions and principles, it stands to
reason that cultural fluency and therefore speech fluency in L2 are
likely to be inhibited. It may be that learners have perceptions of what a
given speech task involves based on first language cultural norms. This
could influence what they attend to in, for example, narrative retells,
and how much of their own perceptions and opinions are included
in such tasks, or whether they perceive a need to abstract lessons or mor-
als from stories and experiences. Wray (2002) provides an interesting
overview of how oral storytelling is constructed in different cultures
and how such verbal activity plays important social and cultural roles.
She concludes that “. . . the culturally sanctioned forms of words sym-
bolize the identity of the society and mastery of them is a mark of status
and trust”(p. 76).
Furthermore, the structure of a first language could influence the
development of ability to use formulaic language. Native speakers of an
agglutinative and highly inflected language such as Japanese might have
a different way of dealing with formulaic sequences in a word-order
dominant language such as English, whereas speakers of word-order
dominant Chinese or Indo- European Spanish might be able to draw
on first language abilities to deal with English phraseology. There is
evidence that in agglutinative languages which are structurally similar
to Japanese, such as Finnish and Turkish, lengthy polymorphemic words
are used and mentally processed just as multiwords strings are in word-
order dominant languages such as English (Wray, 2002: 268, 269). As for
the use of formulaic sequences in Spanish or Chinese, Nattinger and
DeCarrico (1992: 67–69) provide evidence and examples of multiword
formulaic sequences in both languages. Therefore, while the process of
learning, storing, and retrieving complex units such as formulas may be
similar across language types, the nature of the complex units themselves
may differ, including their orthographic representation in the mind.
This could have implications for the ways in which native speakers of
typologically different languages might deal with formulaic sequences in
English as a second language.
A final cultural and linguistic issue which might have an effect on the
development of fluency in English as a second language is the discourse
and temporal features of first language speech. According to Ho (1993),
spontaneous discourse in spoken Mandarin Chinese has some distinct
 Social and Cultural Factors 81

temporal features which stem from the ways topics are linked and
themes are elaborated. In Mandarin there are various possible syntactic
categories from which the main theme of an utterance may be selected,
unlike in English where the theme, at least in narrative, tends to be
a noun. As a result of this, spoken discourse in Mandarin may be
characterized by temporal patterns which could be perceived as
dysfluent in English:

. . . in spoken Chinese discourse, hesitation and false starts are a noticeable

feature such that a chosen theme is often abandoned or reformulated
the moment it is uttered . . . (compared to English speakers) there are
more disfluencies manifested as premature starts, reformulation,
stammers, long pauses and hedges in the spontaneous response of
Chinese speakers. (Ho, 1993: 61, 62)

In addition to the discourse structure of spoken Mandarin, features of

syntax, lexical chunking, and prosody in Mandarin speech also can help
create a distinctive set of temporal features if transferred into English.
For example, according to Ho (1993), Mandarin tends to be spoken
without sentence boundaries, and in fragments juxtaposed without
distinct boundary markers, in sequences which Ho terms “utterance
clusters.” Ho describes Chinese spoken discourse as “long series of
simply constructed and loosely connected short utterance segments
with the major constituents often suppressed or not suppliable” (p. 84).
The prosody of spoken Mandarin also reflects the syllable-timed nature
of the language, with word stress generally absent but mostly mono-
syllabic words uttered rapidly and tonically (p. 85). If transferred to
English speech, these features of spoken Chinese can produce a
discourse pattern showing a high degree of hedges, false starts, refor-
mulations, frequent short pauses, rapid syllable-timed clusters of
fragmentary utterances.
Spanish is an Indo-European language like English, and so it can be
expected that the two languages share certain key features of syntax and
discourse. However, temporal features of spoken Spanish may differ
from English. While speech rate data for Asian languages are not high-
lighted in the literature, Laver (1994) notes that the range of speech
rate for Spanish is from 276 to 420 syllables per minute (pp. 541–542).
This is fast compared to the 270 to 300 syllables per minute for English
82 Formulaic Language and L2 Speech Fluency

(Pawley & Syder, 1983), and might have a noticeable influence if Spanish
L1 learners of English transfer this high rate of speech to their L2.
Neither Laver nor Pawley and Syder specify whether these speech rates
include filled pauses in the syllable counts.
Japanese has its own set of distinctive discourse features which could
influence how Japanese learners of English perform in their L2.
According to Maynard (1989), phrasal units in Japanese are often
accompanied by pause-warning decreased speed, resulting in variations
in articulation speed over a given stretch of discourse (p. 24). As well, in
Japanese, clauses are often uttered broken into smaller units bounded
by short pauses resulting in a high frequency of pausing overall. There
is also a high frequency of use of hesitations, fillers, and sentence-final
particles, often drawled, for a range of purposes including politeness,
and building of utterances from fragments called bunsetsu, typically
containing one content word and function words and bracketed by
pauses (Maynard, 1989: 24–32). The result of this is that spontaneous
speech in Japanese may be more fragmented and show more speed vari-
ation and more frequent pausing that English. If transferred to English,
this type of speech style could appear disfluent.

Social and Cultural Factors in Fluency Development

Clearly, affective factors, be they learner-internal or socially influenced,

can have an effect on speech fluency development and performance. In
order for the automatic processing and spontaneous use of formulaic
sequences to occur, a certain amount of “free head space” is needed.
Feelings of lack of identity and voice, undeveloped L2 social identity,
limited investment in the L2 learning and communication experience,
minimal cultural fluency, restricted self-efficacy and high levels of lan-
guage anxiety can block the development of fluency and the ability
to produce fluent speech.
Given that formulaic sequences retrieved and produced as wholes
are important in fluent speech production, it is logical to assume that
learner language output will exhibit the use of such phrases learned
due to genuine interactional needs (such as apologizing or claiming a
turn in conversation), or due to pressure to reduce mental processing
load at key points in talk. Therefore, it stands to reason that one’s social-
interactional experience will influence fluency and use of formulaic
 Social and Cultural Factors 83

sequences. Language is the way we externalize our innermost sense of

ourselves—including insecurities, hopes and fears, personal perception
of the world and its animate and inanimate aspects, including other
humans. People exchange knowledge to achieve real-life goals or simply
because that knowledge is there to be known. L2 teaching, especially
with regard to fluency, requires an activation of the same socio-interac-
tional, processing, and information management stimuli that stimulate
the use of formulaic sequences in native speakers. It is also important to
remember that the expression of one’s self-identity and group member-
ship may entail the “failure” to sound fully native-like.
In the end, there is a complex relationship among all of these factors
to influence fluency in L2 speech. Certainly, the immediate context
plays a strong role, as social factors all may have varying effects depending
on the nature of the interlocutors, the purpose of the speech produced,
setting, and degree of self- investment required. The underlying psycho-
linguistic processes allow the emergence of voice, cultural fluency, and
self-efficacy. Formulaic sequences, if internalized, allow for the expres-
sion of those aspects of the self in society, and fluent temporal patterns
in speech allow social and cultural contact to bloom and change the
degree of anxiety, self-efficacy, cultural fluency, and so on, and the emer-
gence of voice. In turn, these developments permit ever more social
contact in L2 and result in higher levels of fluency over time.
This page intentionally left blank
 Part II

Evidence
This page intentionally left blank
 Chapter 5

Design of the Study

Collecting and Analyzing Spontaneous Spoken Language

It is interesting to note that human spontaneous speech is a phenomenon

which has been in existence for thousands of years, likely emerging
from an early primate “protolanguage” (Wray, 1998). Speech is the
primordial form of human communication, predating the earliest
efforts at written communication by many thousands of years. In spite of
the ubiquitous and even mundane nature of speech, it has only recently,
since the 1960s, been seriously studied in a scientific way. Before that,
written work on the subject of speech was prescriptive, instructing on
how to speak well or properly, rather than describing what ordinary
people do in communication. The general view was that speech is cha-
otic and unruly, but the introduction of recording equipment enabled
the discovery of definite order in speech, particularly conversation (ten
Have, 1999: 3).
The advent of tape recording technology liberated the study of speech
communication by allowing researchers to venture into the field and
capture samples of real-life speech. Many coding and analysis proce-
dures were developed for analyzing recorded speech in a quantitative
way; formulas for rates and frequencies of speech phenomena were easy
to apply. One of the most studied aspects of spontaneous speech is
the nature of conversation and the discourse analysis of how spoken
discourse “unfolds in time” (Cook, 1989: 52). Many discourse analysts
turned to observation of how people behave and cooperate in managing
discourse. Such analysts are often termed ethnomethodologists, concen-
trating on how pairs of utterances relate to one another, how turn-taking
is managed, how conversational openings and closings occur, and how
face-preservation and politeness are engaged in. The emphasis in such
study is usually on real-life data, not experimentally elicited speech.
88 Formulaic Language and L2 Speech Fluency

Research methods such as ethnomethodology have yielded important

information about spontaneous conversation, but the tradition in
fluency research has been different. As seen in Part I of this volume,
much of the fluency research has used elicited speech and often the
speech samples are monologic. This is generally due to the nature of
the phenomena under study in fluency research, which are temporal
aspects of speech rather than interactive phenomena, as well as a need
to standardize the nature of the speech under consideration. Most flu-
ency research studies have used relatively small cohorts of participants
and relatively small corpora of data, which creates constraints on the
type of speech data needed. If a researcher is examining speech samples
for prosodic features and temporal phenomena, it is useful to ensure
that all participants in a cohort perform under similar cognitive and
affective pressures. This makes conversational data more difficult to
deal with than monologic data. The study presented below follows
this tradition by analyzing elicited monologic speech rather than free
conversation. However, this is not to say that monologic speech is the
only type analyzable for fluency—other researchers such as Boers et al.
(2006) and McGuire (2009) have used controlled conversation data in
studying fluency.
There are a number of issues in data collection which make it essential
to ensure that data be collected unobtrusively and with full consent of the
participants. Ten Have (1999) notes that most people are reluctant to
be taped and analyzed in spontaneous communication:

It is a fact that many people dislike the idea that known or unknown
aspects of their spontaneous actions will be considered in great detail.
This is clearly evident from the reluctance of many to have their
actions recorded for research purposes (. . .) there seems to be a
common-sense association, then, between a detailed consideration of
actions and an unpleasant exposure, or critical assessment, of those
actions. People seem to be afraid of being caught off-guard. (p. 61)

A certain level of awkwardness is evident in the speech actions of

participants who are aware or suspect that they are being recorded (ten
Have, 1999: 66). For L2 learners, this type of affective influence may
compromise the quality of their speech production and lead to speech
samples which are not representative of their true capabilities. To avoid
 Design of the Study 89

this “stiffness” or artificiality of production it might be advisable to have

the participants record themselves with the researcher at a distance, or
have a third party record. While both of these procedures deny the
researcher the opportunity to take field notes or observe paralanguage,
they might reduce the stress load on participants enough to encourage
more authentic and unmonitored speech behavior. Getting informed
consent is a necessary, and, perhaps, difficult task for the researcher in
speech analysis. It is important to be aware of the fact that participants
may withdraw from a study at any point, and that attrition is a given
if the study is longitudinal in nature. An advantage for a researcher
working with L2 learners is the possibility of offering the participants
a detailed analysis of their language, including some error analysis or
advice. If the study is longitudinal, a letter attesting to their improve-
ment in the measures studied may also be an inducement to staying in
the study and may raise motivation levels among participants.
In order to record not only what speakers say, but how they say it, it
is necessary to write down what they say and use a system of symbolic
notations to indicate aspects of the speech production beyond the
words themselves. Such a system preserves the key features of talk, such
as certain types of intonation, pausing, sound stretches, and so on. This
also enables the easy communication of research findings and results.
Psathas (1995: 12) outlines some key features of transcription notation:

Emphasis is noted by underlining or using italics for those parts of an

utterance that are stressed. Sounds that are stretched are indicated by
colons, :::::, for example, so :::: Sounds that are cut off are marked by a
dash, for example, la-bu- Pauses may be noted by timings in seconds and
tenths of a second (e.g., 1.2), with micropauses of less than two tenths of
a second by a dot (.). Brackets ([) indicate speed that is overlapped.
Punctuation is used to indicate features of speech such as pitch. For
example, a question mark (?) indicates rising intonation, a comma (,)
continuing intonation, and a period (.) terminal falling intonation.

Of course, the range and extent of transcription notation goes beyond

this brief list, but the key elements are here. Other points to note
are accelerating and slowing down of intonation, filler sounds, laugh
particles, inaudible parts, and so on. Not all of these are used in the
transcribing which was undertaken in the study described in this section
90 Formulaic Language and L2 Speech Fluency

of this volume; the transcription conventions are explained within the

methodology section below.
In this study as in many others the issue of timing pauses is a trouble-
some aspect of transcription and data analysis. In fluency research,
pauses are of the utmost importance, their timing is essential in deter-
mining fluency levels of different speakers. One type of timing mode
used by some researchers (ten Have, 1999: 84, 85) is “rhythm-sensitive
timing.” This involves a systematically and consistently used metronomic
beat counted by the transcriber while listening, such as the type com-
monly used in counting seconds by subvocalizing “one-one-thousand . . .”
This is a distinctly “relativist” (ten Have, 1999: 85) timing system, and is
clearly not adequate for in-depth research on fluency, in which pause
timing is more vital than in other more meaning-focused analysis. Finely
calibrated stopwatches are another possible tool to use in pause timing.
While they offer a more refined and reliable mode of timing than the
rhythm-sensitive one, ten Have (1999: 85) notes limitations:

I must say I feel more comfortable timing with a stopwatch. But even
there, I use repeated timings, because one has the problem of “catching”
the exact onset and finish “points” (of a pause). By closely monitoring
one’s timing activities, the closeness of the correspondence of one’s
“clickings” with what one hears, one can observe which of the timings
are better than others.

There must be a better way, and there is. Computer software has been
available for a number of years which can produce a visual display of
digital sounds. These allow the acoustic pause length to be interpreted
from a display. Clearly, the data gained from this mode of pause timing
would be more accurate than other more mechanical or subjective ones,
and would be much more valuable for researchers who need to deter-
mine such phenomena as ratio of pause time to total speech time in
production. See below for details on the software and the pause timing
conventions used in the present study.

Studying Formulaic Sequences and Fluency

The study which is described in this section of this book is one of

the very first attempts to explore the connection between the use of
 Design of the Study 91

formulaic sequences in adult English L2 learners and speech fluency

development. Not intended to be the last word on the subject, nor a
perfect model for fluency research methodology, it is exploratory in
intent and designed as a jumping-off point for future work in this area.
It can serve as an initial source of evidence for researchers seeking to
investigate this phenomenon further, as a rough template to be refined
for future research, or as a model of some types of procedures useful in
such research. It is by no means meant to be a definitive final word on
formulaic sequences and L2 fluency.
The methodology of the study is noteworthy for a number of reasons,
including the timing of sampling, the lack of a pedagogical intervention,
the blend of quantitative and qualitative analyses, the means of identifying
formulaic sequences in the speech samples, and the combined analysis of
data from the group and individual participants.
This is a longitudinal study spanning 6 months, during which speech
samples were regularly collected from study abroad participants (N = 11).
This enabled the researcher to track the development of key language
phenomena over time: the temporal variables associated with fluency
(see Part I of this volume), and the possible role of a growing repertoire
of formulaic sequences in the change in temporal variables.
The researcher collected speech samples from the participants
outside of their ESL (English as a Second Language) classroom studies.
It was assumed that consistent and regular classroom study of spoken
English, combined with the study abroad context as a whole, including
homestay experience, would drive language acquisition and skill develop-
ment. Therefore, there was no pedagogical intervention or comparison
of groups etc. in this study.
The study combines both quantitative and qualitative data analysis
methods. The temporal variables of speech associated with fluency are
measured and analyzed statistically using repeated measures analysis of
variance (ANOVA), to establish that fluency gain did in fact occur as
measured empirically. It is important to bear in mind, however, that a
small sample and wide variances in performance make the statistical
determination of significance problematic—the overall trends in the
raw data are as good or better an indicator of change over time as the
statistical treatments. The speech samples are subjected to a discourse
analysis after the fluency gain is established. This allows the researcher
to see how the participants used formulaic language to further their
92 Formulaic Language and L2 Speech Fluency

fluency, what functions the formulaic sequences might have played in

the discourse, and what types of mental processing and sociocultural
factors might have influenced the performances.
The means of identifying formulaic sequences is at the heart of this
study, and the methodology used to do so is innovative. The speech
samples were examined by three expert native speaker judges, well read
in the literature on formulaic sequences and prepared using a group
session in identifying sequences in transcripts. The judges used a set of
criteria drawn from Wray (2002, 2008) and Wray and Namba (2003)
and others to mark formulaic sequences in the transcripts of the speech
samples. This procedure lends itself to further study—one might exam-
ine its validity as a means of analysis, its efficacy in terms of interrater
reliability or other concerns, or ways it can be adapted to work better or
to fit other research designs.
The study presented here is also distinctive in that it presents analyses
of both group data and individual data. After the quantitative and
statistical analysis of the group data, the results for each individual
participant are analyzed and discussed. This allows for an exploration
of the constraints under which the individual participants may have
operated in producing the speech, as well as how individuals developed
their fluency over the 6 months of the study. This gives the data analysis
a finer grained and more qualitative aspect.
I encourage the reader to accept, reject, or adapt the methods
and analyses presented here. The study is certain to be a beginning, an
inspiration, a motivator, a guide, or a case study.

Design of the Study: Conceptual Framework and Hypotheses

This study is framed by two theories—one of language production, and

the other of language development. In addition, the research is guided
by hypotheses drawn from previous temporal research on fluency.

Language production and development

Production
The model of production is that of Levelt (1989), which represents two
kinds of knowledge: declarative and procedural. Production requires
procedural knowledge of three types: conceptualizing; formulating;
 Design of the Study 93

articulating. The conceptualization stage involves generating a “prever-

bal” message. The formulation stage of production involves taking the
preverbal message from the conceptualizer and assigning it an accept-
able grammatical form, mapping on lexical information at the same
time, and developing a phonetic plan, followed by the stage of physical
articulation to produce speech. Speed requirements for fluent speech
can only be met insofar as operations of formulation and articulation
are proceduralized.

Development
The model of language development here, which explains how the
procedural knowledge in the production model is created, is a synthesis
of the information discussed above on formulaic language and mental
processing. The notion of automatization is basically that of Anderson
(1983, 1993) and his Adaptive Control of Thought (ACT*) model of
cognitive development. Procedural knowledge consists of units called
“productions” and takes the form of “if/then” kinds of match and exe-
cution sequences. In this way, an entire previously learned production
sequence can be accessed instantly, reducing greatly the strain on
working memory. However, these units of production are inflexible,
and must be recalled as single units without modification. Basically,
then, it is claimed that proceduralization of the knowledge needed
to produce speech can occur through several means: through auto-
matization, which develops gradually through repetition and practice,
through restructuring leading to single-step memory retrieval, or by
segmentation and rapid automatization of some salient formulaic
sequences from input due to their utility in types of discourse frequently
encountered. This process may be driven by activation of memory
sequences due to input frequency, with the most frequent and salient
lexical strings retrieved more quickly. Hence, the encoding processes
detailed in the Levelt model are simplified or even bypassed, and for-
mulaic sequences can be retrieved as single lemmas, avoiding the need
to assemble phrases word-by-word.

Hypotheses
Hypotheses for this study were generated by integrating the temporal
variables associated with speech fluency with the Levelt (1989) model.
94 Formulaic Language and L2 Speech Fluency

This may be seen by considering speaking rates and interpreting them

in relation to aspects of the model. If overall Speech Rate (SR) increases
over time, it is likely that more knowledge is being proceduralized. It is
also possible to posit that an increase in the mean length of runs (MLR)
over time is an indicator of increased proceduralization at the formula-
tion stage of the model. Although it may be that this increase in MLR is
attributable at least in part to greater time spent planning each utterance,
if this is true, then this planning would have to be occurring outside of
the time during which speech is being produced. This means there
would have to be more pauses, and/or longer pauses between utterances.
Therefore, the temporal pattern expected from increased procedural-
ization of knowledge would be increased MLR, no increase in pause
times, and an increase in Percentage of Time Spent speaking (PTS),
overall. In the present study, location of pauses was not taken into
account in measuring fluency. This is because previous empirical studies
of fluency have not done so, in addition to the fact that it is problematic
to determine where clauses begin and end in spontaneous speech in an
L2 context. In any case, MLR provides a measure of how pauses relate
to stretches of unbroken discourse.
At the surface, empirically observable level, then, it is hypothesized
that speech rate will increase over time, that the percent of time spent
in speech will increase, and that the MLR will increase. These comprise
the set of quantitative hypotheses in the research. Underlying this set of
temporal variables is the psycholinguistic foundation of fluency. Evidence
of change at the surface level will be taken as evidence of increased pro-
ceduralization or automatization of knowledge in the formulation stage
of speech production. The key linguistic element to be automatized is
formulaic sequences, including all multiword units such as collocations
and lexical phrases, which allow for the increase in MLR in particular,
and for a general facility in speech production.
It is also understood that, as Boers et al. (2006: 246, 247) note, a
command of formulaic sequences may aid L2 speakers in three ways:
idiomatic sequences may allow a native-like production because their
meanings are not predictable by rules of grammar or the sum of the
meanings of their component words; formulaic sequences retrieved
directly from memory can facilitate fluency by reducing hesitations
and increasing the length of runs; sequences which are recalled and
stored correctly may “[. . .] help speakers reach a degree of linguistic
accuracy, because these prefabricated chunks constitute ‘zones of safety’
 Design of the Study 95

and appropriate use of them may thus confine the risk of ‘erring’ to the
spaces in between the formulaic sequences in one’s discourse” (p. 247).

Hypotheses:
Over time, with continued learning and experience:

1. L2 speech will exhibit a faster rate of production.

2. L2 speech will exhibit a greater amount of production time spent
speaking as opposed to pausing.
3. L2 speech will exhibit longer runs between pauses.
4. Formulaic sequences will appear more frequently in the longer runs
between pauses.

Furthermore, consideration will be given to whether linguistic, social,

and cultural factors may have played a role in the development of flu-
ency as measured by temporal variables and use of formulaic sequences,
in the context of a given performance.

Methodology: Overview

The study is longitudinal in design, using repeated measures. Speech

samples were collected on tape from participants at regular intervals six
times over the course of a 6-month period and analyzed for changes in
temporal variables and learners’ use of formulaic sequences. The study
was conducted in an intensive ESL program at a Canadian university.
The participants were 11 students enrolled full-time in the program, all
at approximately the same level of oral proficiency as measured by
an interview-based placement test. They were from three L1 groups:
Spanish; Chinese (Mandarin); Japanese. The Spanish L1 group con-
sisted of two females and two males, as did the Japanese L1 group, while
the Chinese L1 group consisted of one female and two males (originally
two Chinese females participated but data from one of them was flawed
and removed from the final analyses). Once per month over the 6-month
period, participants watched three short animated films as speech
prompts, after which they were asked to retell the story of the film.
These retellings were tape-recorded and transcribed. The three films
used were of similar length and narrative complexity, and each was
96 Formulaic Language and L2 Speech Fluency

viewed twice, film one the first and fourth months, film two the second
and fifth months, film three the third and sixth months.
As an initial step, in order to establish that fluency gain occurred,
measures of the key temporal variables of speech rate, articulation rate,
phonation/time ratio, and MLR were calculated for each specific
sample, six per participant. Formulaic sequences of all types were iden-
tified in the samples by three expert native speaker judges. Based on
these, a formula/run ratio was calculated. This ratio was a measure
of the proportion of formulas to the number of runs in a sample, an
indication of how the number of formulas per run may have changed
over the time of the study.
At the heart of the research was a discourse analysis of the samples
which was conducted for each participant to identify comparable
exemplars in which segments of earlier and later speech samples were
compared. This comparison focused on the role of formulaic sequences
in more fluent expression of the same ideas and content, as pauses
reduced and speech and articulation rates increased.

Pilot Research

A pilot research study testing the four hypotheses and the conceptual
basis outlined earlier was conducted prior to the main study. Six high-
beginner level learners from the university’s Intensive ESL program
participated, providing samples of spontaneous spoken narratives at
four intervals over a 13-week period of study. The learners were from
three typologically different languages: Spanish; Chinese; Japanese,
to control for influence of L1 type. Samples were elicited through the
use of two silent animated films, each viewed on two occasions, and
participants were instructed to retell the film narrative spontaneously,
without notes or preparation time, in the university’s language labora-
tory after viewing it. They were instructed not to rewind or stop their
tapes during telling.
The tapes were then transcribed and, using SpeechStation 2 speech
analysis software, the transcripts were marked for pauses and their dura-
tions. The samples were analyzed for rate of speech, length of fluent
runs, and the number of formulaic sequences used and the ratio of for-
mulaic sequences to runs per sample. Formulas in the oral texts were
identified by two native speaker judges using criteria developed from
 Design of the Study 97

the taxonomy of Nattinger and DeCarrico (1992), as well as the criteria

put forth by Coulmas (1979) and Peters (1983), of phonological coher-
ence and greater complexity than other nonformulaic output by the
same speaker. This proved to be a complex task, as many strings of
speech appeared to fit the criteria but not qualify as formulaic sequences.
For example, entire sentences were sometimes uttered with phonologi-
cal coherence, appearing to be more syntactically complex than the
immediately surrounding text, yet not fit with any one pragmatic pur-
pose as in the Nattinger and DeCarrico categories of lexical phrases. On
the other hand, items which appeared to fit with the Nattinger and
DeCarrico taxonomy sometimes did not exhibit phonological coher-
ence. These issues are addressed in the methodology of the larger study
by an expansion of the set of judging criteria.
Results showed trends in the direction of both increased fluency as
measured by temporal variables and greater formulaic sequence use for
all participants. A debriefing session with the participants revealed that
they felt that they had become stronger and more fluent in expressing
ideas and content in producing the samples over time. Although most
were satisfied with the relative difficulty of the two film prompts, they
reported that repeating the same film two times might have affected the
quality of their participation in that they felt that they paid less attention
to the film on the second viewing and might have tended to use their
memory of the previous viewing to construct the narrative on a second
viewing and retelling.
Several important aspects of data collection and analysis were clari-
fied as a result of the pilot study experience: the number of film
prompts—Three film prompts are used; the statistical analysis of data—
The larger study uses repeated measures analysis (ANOVA); the means
of determining what constitutes a lexical phrase—The transcripts and
audio samples in the larger study have been analyzed for formulaic
sequences by three expert judges. The procedures and criteria are
detailed below.

Data Collection

Sample
The study involved native speakers of three typologically different
languages, Japanese, Chinese (Mandarin), and Spanish, to allow
98 Formulaic Language and L2 Speech Fluency

exploration of any influence of L1 typology on the development of

fluency and the acquisition of formulaic sequences. It might be a com-
monsense expectation that Spanish L1 speakers would acquire English
L2 fluency and formulas more readily than would those from typologi-
cally more distant languages such as highly agglutinative and inflected
Japanese or monosyllabic, tonal, word-order dominant Chinese. As well,
having three different L1 groups allows for greater generalizability
of results, bearing in mind the limitations of the small sample studied.
Generalizability of the results is furthered by the fact that the partici-
pants in this study are also representative as to gender. At the beginning
of the data collection, half of the learners from each L1 group were
males, half females, all aged from 20 to 25. The recordings for the twelfth
original participant, a Chinese female, were flawed due to equipment
failure for several samples so that comparisons across recordings could
not be made. Her data was removed from the final data set, leaving two
Chinese males and only one female.

Selection
The participants, all of whom had completed secondary education in
their own countries, were at an intermediate level of English proficiency.
All were students in an intensive ESL program at a Canadian university,
and had been enrolled for at least one 12-week term prior to the study.
The participants remained in the program for the summer and fall
terms as well, meaning that they continued to receive the same rate of
instruction over the 6 months. Students were initially selected at the
beginning of the first summer term based on the results of their oral
proficiency scores on the four-skills placement test used in the program.
Due to weighting of all their component scores on the test, the partici-
pants, all of whom scored within the intermediate range on the oral test,
were enrolled in different classes.
The oral part of the placement test consisted of an interview with an
examiner who was an experienced instructor in the program. Typical
topics of conversation included past English language learning experi-
ence, travel abroad experiences, experience in Canada, and future
plans and motives for studying English. The topics elicited mainly
narrative talk, which influenced the decision to use narrative retell
 Design of the Study 99

as a prompt for the speech samples collected in this study. The oral
proficiency of the students was judged holistically by the examiners and
entered in a band scale corresponding to the three class levels of the
intensive ESL program itself: beginner, intermediate, advanced. The
students selected for participation in the present study were all placed
in the band level corresponding to intermediate oral proficiency.
The intensive ESL program provided 24 hours of language instruc-
tion per week, of which six were specifically focused on spoken lan-
guage. The program consisted of four main components:

Core four-skills course: thematically organized and content-based, with

all four language skills integrated in the teaching plans. 9 hours per
week.
Writing course: largely process oriented. 6 hours per week.
Elective component: selected by students from content or language options
such as TOEFL preparation, vocabulary building, sports themes, cross-
cultural activities, and so on. 3 hours per week.
Speech-focused course: preparation for formal speech presentations,
semi-prepared assignments, and spontaneous speech activity of
various kinds. 6 hours per week.

All courses and all levels of the program followed essentially the same
curriculum, so that the types of classroom tasks experienced by the
participants in the present study were much the same regardless of their
class placement. The acquisition of fluency and formulaic sequences in
such a situation can be assumed to take place in a more or less natural-
istic way. The curriculum for the oral and four-skills components of the
program did not specify any particular focus on fluency or formulaic
language.
The participants in this study also lived in homestay situations with
Canadian families. This provided a naturalistic acquisition environment
with rich and sustained opportunities for English input and communi-
cation outside of the university ESL classrooms. Since fluency and
formulaic sequences were not an explicit part of the ESL program cur-
riculum, the experience of homestay may have played a strong role in
the development of fluency and formulaic language competence in the
participants over the time of the study.
100 Formulaic Language and L2 Speech Fluency

The Participants

Spanish L1 group
Female
Lilia was a 22-year-old woman from Venezuela. She had been enrolled
in the program for one previous term and had studied English at
university in Venezuela as well. At the time of the study she had moved
out of a homestay into an apartment with a friend from her country.
Sally was a 20-year-old woman from Venezuela. She had studied
English in secondary school in Venezuela but had no university study
experience. She had been enrolled in the program for one previous
term and was living in homestay during the time of the study.

Male
Miguel was a 25-year-old man from Venezuela. He had been enrolled
in the program for one previous term and had studied English at
university in Venezuela. At the time of the study he had moved out of
homestay into an apartment with a friend from his country.
Carlos was a 20-year-old man from Venezuela. He had been enrolled
in the program for one previous term and had studied English at
secondary school in Venezuela, but had no university study experience.
He was living in homestay at the time of the study.

Japanese L1 group
Female
Natsuko was a 24-year-old woman from Japan. She had been enrolled in
the program for one previous term and had studied English in univer-
sity in Japan. She was living in homestay during the time of the study.
During part of the summer she left the intensive program to take an
English for specific purposes diploma program in teaching EFL, but
returned to the intensive program in the fall term.
Yuka was a 22-year-old woman from Japan. She had been registered in
the program for one previous term and had studied English in univer-
sity in Japan. She was living in homestay at the time of the study.
 Design of the Study 101

Male
Jun was a 23-year-old man from Japan. He had been enrolled in the
program for one previous term and had studied English at university in
Japan. He was living in homestay at the time of the study.
Isamu was a 23-year-old man from Japan. He had been enrolled in the
program for one previous term and had studied English at university in
Japan. He was living in homestay at the time of the study.

Chinese L1 group
Female
Meiling was a 20-year-old woman from China. She had been enrolled in
the program for one previous term. She left homestay partway through
the study to live with other Chinese speakers. She left for several weeks
during the course of the study to visit family in China, and her speech
samples were collected separately from the other students at one point
because of this. She passed the university’s English language admissions
exam during the time of the study and spent the fall term in credit English
for Academic Purposes (EAP) instead of the intensive program.

Male
Lin was a 20-year-old man from China. He had been enrolled in the pro-
gram for two previous terms. He had studied English at secondary school
in China. He was living in homestay during the time of the study.
Liang was a 20-year-old man from China. He had been enrolled in the
program for two previous terms. He was living in homestay during the
time of the study.

Procedure

Narrative retell
A narrative retell task was used to elicit speech samples. There were
four factors which influenced the decision to choose narrative retelling
to elicit speech: the tradition in fluency research; the nature of the test-
ing used to assemble the sample of participants; the need to standardize
102 Formulaic Language and L2 Speech Fluency

the procedures; the nature of clause chaining in spontaneous narrative

speech.
First, it has been the tradition in both quantitative and qualitative
research into speech fluency to use narrative retell prompts to elicit
speech samples. In many cases, narratives were presented as picture
sequences from which participants were required to create a narrative
(Raupach, 1980; Lennon, 1987, 1990a, 1990b; Hansen, Gardner, &
Pollard, 1998), and sometimes a written narrative was presented first
(Dechert, 1980, 1987). In other cases, viewing a short film was the stim-
ulus for narrative speech (Chafe, 1980; Towell, Hawkins, & Bazergui,
1996). In general, these research designs produced clearer results as
compared to those which used interview or discussion tasks for elicita-
tion (eg. Riggenbach, 1991; Freed, 1995).
Second, the nature of the test used to select the participants and the
desire for standard elicitation procedures also influenced the choice of
instruments in this study. As described above, the sampling procedure
here involved identifying potential participants by looking at the results
of their oral proficiency interviews on the intensive ESL program
Placement Test. This interview consisted largely of questions and discus-
sion likely to elicit narrative responses, such as the nature of previous
English learning, experiences in Canada to date. Therefore, it seemed
that if the participants were selected based on a narrative-type place-
ment task, then requiring narrative speech samples in the actual research
procedure would follow from the initial placement task criteria. It was
thus possible to compare their progression on a measure related to the
initial placement procedure.
Third, having one monologic speech task is a way to standardize the
experience for participants and avoid the complications of determining
reliability among various interlocutors in dialogic tasks. A standard
one-way speech task seemed to be the safest way to ensure equality of
opportunity to speak for all participants. While some researchers insist
that elicited monologic samples are not representative of real-life com-
munication, the exploratory nature of this study meant that a number
of potentially distracting factors needed to be controlled. For example,
the influence of the presence of an interlocutor if using discussion
or interview tasks to elicit spontaneous speech might have caused wide
differences in the nature of the speech produced by participants. It was
felt here that in order to really examine directly how formulaic sequences
 Design of the Study 103

might facilitate fluency, it was best to use elicited monologic narrative

retells to make sure that each participant performed under the same
constraints.
A fourth and important motivation for requiring narrative retells in
this sample elicitation procedure has to do with the nature of spontane-
ous speech with respect to clause chaining. As Pawley and Syder (1983)
note, a great deal of everyday speech consists of narrative discourse. It is
in this narrative discourse that clause chaining and use of formulaic
sequences is most apparent. In other types of tasks such as expository
speech or spontaneous explanations of complex interrelated topics, a
high degree of cognitive load occurs and fluency suffers even for native
speakers. Therefore, it would appear that a narrative retell procedure
would provide optimal opportunities for participants to produce dis-
course exhibiting clause chaining and use of formulaic sequences.

Silent film prompts

It was decided to use short 8 to 10 minute silent films as prompts. Films
with spoken dialog or voiceover could have presented problems related
to listening ability and other language skills. In other words, if partici-
pants had difficulty understanding the speech in such films, the retell
task might have become more of a challenge for them due to factors
other than their speaking ability. Also, silent films allowed the learners
to control what to attend to and what to say on tape, rather than trying
to repeat spoken language heard in the films.
The prompts for the retells were three silent animated short films
from the National Film Board of Canada: Neighbours (Norman McLaren,
1952: 8 minutes 10 seconds), to elicit samples 1 and 4; Strings (Wendy
Tilby, 1991; 10 minutes 18 seconds), to elicit samples 2 and 5, and The
Cat Came Back (Cordell Barker, 1988: 7 minutes 37 seconds), to elicit
samples 3 and 6. The first two of these films had been used in the pilot
research, the third is similar in length and narrative complexity. All
three include roughly the same number of characters, have equally
complex settings, and the same number of plot movements. Having
three films, each used twice over 24 weeks, allowed a 3-month time
interval between viewings of the same film and minimized the chance of
participants remembering the previous viewing and not paying full
attention when viewing and retelling the second time.
104 Formulaic Language and L2 Speech Fluency

The film used to elicit samples 1 and 4, Neighbours (Norman McLaren,

1952: 8 minutes), tells the story of two neighbors who live peacefully and
happily side-by-side, sharing the front lawn, sharing a newspaper, sharing
a light for their pipes. A flower suddenly appears between them one day
and they are overjoyed by its presence, dancing, spinning, and somer-
saulting around the property. Things get nasty when one of them gets
too possessive of it, covering it with his hand and smelling it closely. The
two begin to trace property lines between their homes which place the
flower on one side or the other. They magically create wooden fences
which run on one or the other side of the flower. Pieces of the fence are
used as weapons as punching and pushing give way to outright savagery,
they club each other, rip each other’s clothes. One of them pounds the
other’s face to a pulp while his own face transforms into an evil mask of
twisted rage. They each take a fence stake and smash each other’s houses,
beat their wives senseless, and kick their infant children like footballs.
Eventually they kill each other and end up in side-by-side graves, the
fence stakes serving as crosses, and two little flowers like the original one
marking where they lie. A multilingual message appears on the screen
advising us to “Love your neighbor.”
The film used to elicit samples 2 and 5 is Strings (Wendy Tilby, 1991:
10 minutes), which tells the story of two elderly people who live in the
same apartment building, a woman on the third floor directly above a
man on the second floor. They take the same elevator to go home one
day, both carrying shopping bags. They enter their respective apart-
ments and go about their business: the woman unwraps her parcel,
which contains a piece of a model ship she is building in her living
room, and carefully attaches the piece to the model using string. She
then proceeds to pour a hot bath for herself. Meanwhile the elderly
man is downstairs laying out a meal of canned fish, fruit, and wine while
his cat looks on. The woman upstairs looks out the window and sees a
taxi disgorge several people carrying string instruments. The musicians
arrive at the old man’s apartment and all begin to play chamber music
together. Back upstairs the woman is soaking in her bath and listening
to the music rising from downstairs. However, water is trickling from the
tub through a hole in the floor and dripping down into the old man’s
apartment. He notices it and hastily grabs tools and goes upstairs to the
woman’s place, ringing her bell and entering to try to fix the problem.
While he attempts to repair the problem, with the old woman looking
 Design of the Study 105

wistfully on, the chandelier on the ceiling of his apartment downstairs

lets go from the plaster and crashes to the floor. The musician friends
quickly leave the premises. The old man returns downstairs and plays
his violin alone, while the old woman returns to her hot bath.
The film used to elicit samples 3 and 6 is The Cat Came Back, (Cordell
Barker, 1988: 7 minutes 37 seconds), tells the story of a man tortured by
a stray cat which appears unannounced at the door of his house one
evening. He is playing a horn alone and the cat comes to his door. He
loves it at first and plays with it happily until it destroys a cherished
childhood toy of his. He then makes repeated attempts to get rid of the
cat, to no avail. He takes it to a forest to abandon it but it returns to
wreak havoc in his home. He takes it to the ocean to drown it in a bag
but is outwitted and the cat reappears at home. He is repeatedly foiled
in his attempts to destroy the cat as he tries to throw it from a balloon
and throw it from the top of a mountain. Finally he attempts to blow it
up with dynamite but inadvertently kills himself instead. As his corpse
falls from a height after exploding, it lands on the cat, who dies in turn.
The final scene shows the soul of the man flying to heaven screaming in
frustration as nine souls of the cat follow him into eternity.

Narrative complexity
All three films have two main characters and are of similar length.
As for narrative complexity, it is of course impossible to be certain
that any two films are of exactly the same narrative complexity, but
roughly eight narrative turns are present in each of the three films
used in this study.
In Neighbours, the eight turns are as follows:

1. Neighbors are relaxing together.

2. Flower appears
3. Initial excitement
4. Growing competitiveness
5. Fence building
6. Violence
7. Murder of families
8. Mutual destruction.
106 Formulaic Language and L2 Speech Fluency

In Strings, the eight turns are as follows:

1. Neighbors return home in elevator

2. Man unpacks food
3. Woman prepares bath
4. Man’s guests arrive
5. Water leaks
6. Man goes upstairs
7. Chandelier falls
8. Guests leave, all back to silence.

In The Cat Came Back, the eight narrative moves are as follows:

1. The man discovers the cat at his door

2. The cat destroys a cherished possession
3. Attempt to abandon the cat in the forest
4. Balloon incident
5. Boat incident
6. Train and mountain incident
7. Dynamite explosion and accidental suicide
8. Death of the cat and all ten souls head for heaven.

Viewing and retelling

The participants were shown each film in its entirety without pause only
one time for each sample collection. The viewing took place in a univer-
sity classroom with the entire group present. In some cases individual
participants were unable to attend a group viewing and they were shown
the film in the same room at a different time. The content of the films
was not introduced and no language help was provided, nor were partici-
pants allowed to take notes; they simply watched. After viewing, they were
immediately provided with blank cassette tapes and instructed to retell
the story of the film spontaneously in the university language laboratory.
They were advised not to write a script of their retellings, but to push the
record button on the console and tell what they remembered from the
narrative. They were not permitted to stop, pause, or rewind the tapes.
What resulted was truly spontaneous retelling of the narratives from
the viewings. The samples were raw speech complete with pauses,
 Design of the Study 107

restarts, circumlocutions, self-talk such as “I don’t know the word in

English,” or “I don’t remember exactly, but . . .”

Data Analysis

Transcribing and pause marking

The data recorded on tape in the language laboratory were transcribed
from a Sony handheld tape recorder using Microsoft Word. To locate
pauses in the speech samples, the tapes were then recorded into
SpeechStation2 speech analysis software, and a spectrogram produced
for each. The visual representation of the speech was used to identify
pauses and their duration. In SpeechStation2, a spectrogram of 5 to
10 seconds of speech can be viewed. For all of the speech samples, spec-
trographic segments of this size were examined while the researcher
listened repeatedly to the corresponding speech. Each speech sample
was analyzed from beginning to end in this way, segment by segment.
The pauses observed in this way were marked in the corresponding
transcripts. In determining the lower cutoff point for pauses, 0.3 seconds
was used. Anything less than 0.3 seconds is easily confused in a spectro-
gram with other speech phenomena such as the stop phase of a plosive
sound, and anything longer can omit significant pause phenomena.
Given that native speakers seldom hesitate longer than 0.5 seconds
in mid-clause or 2 seconds at a clause juncture, 0.3 seconds seems a
reasonable cutoff. As well, the tradition in fluency research has been
to use 0.25 to 0.3 seconds as a lower end cutoff (Towell, Hawkins, &
Bazergui, 1996: 91).
Transcription was done at a medium level of granularity (Du Bois, 2006),
in that prosodic lengthening or drawling of syllables was not marked,
intonation contours were not marked, filled pauses were transcribed
but not timed, unfilled pauses were timed and marked but not coded
for features such as breathing. Since the data were samples of monologic
speech, no indication of overlap or other features of interaction was
necessary in the transcriptions. A sample segment of a transcript from
Natsuko contains most features common to the corpus overall:

ah one day one guy was playing the instrument music instrument
and ah (0.6) and ah (0.6) he he he heard some noise (1.0) um (1.2)
108 Formulaic Language and L2 Speech Fluency

somebody somebody knocking (0.5) knocked on the door (0.4) and

ah (0.6) and it really (0.7) it really bugged him (0.5) and the (1.2) firs
first he just ah (1.0) he just ah (0.5) just kep (0.5) kept practicing the
instrument but he he couldn’t stand for the noise and he (1.0) he
went to the door (0.6) and ah (0.5) but nobody nobody was there
(0.5) and he found a (0.5) cat in the small basket (0.3) and (0.5) it
was so cute (0.9) and so (0.7) he was happy and he (0.7) brought it
back (0.3) brought it in the house (0.8) and he tried to (0.7) tease
him and tease the (0.3) cat (0.8)

Features of the transcript include:

z Filled pauses and fillers are represented as a syllable, ah.

z Unfilled pauses are marked in parentheses for duration.
z Repeated syllables or words are represented verbatim, for example, he he.
z Phonological reductions such as reduction of a consonant cluster in firs
instead of first or kep instead of kept were transcribed verbatim.
z Phonological reductions which might indicate formulaicity in the way
Bybee (2002) indicates—for example, wanna instead of want to, were not
included in the transcripts unless they involved the adding or subtracting
of a syllable. The judges were expected to perceive such phonological
reductions by listening to the recordings, rather than being prompted by
transcribing conventions.

Calculation of temporal variables

As a means of establishing fluency gain, four temporal variables of
speech are analyzed in this study, as well as a variable linking MLR with
use of formulaic sequences:

z Phonation/Time Ratio (PTR), calculated by dividing the time spent

speaking (excluding pauses) by the total time of the speech sample. See
above for cutoff criteria for determining pauses.
z Speech Rate (SR), calculated by dividing total number of syllables uttered
by the total number of minutes of recording time, including pauses.
z Articulation Rate (AR), calculated by dividing total number of syllables
uttered by the minutes of speech time only, excluding pauses.
z Mean Length of Run (MLR), calculated by dividing the total number of
syllables uttered by the number of runs between pauses.
 Design of the Study 109

z Formula/Run Ratio (FRR), determined by dividing the number of runs

by the number of formulaic sequences, to see whether the use of formu-
laic sequences corresponds to the changes in the temporal variables.
More specifically, this ratio is a quantitative measure of how the use of
formulas contributed to longer runs (greater MLR).

Statistical treatments
As this is a repeated measures study, involving measurement of continu-
ous variables, statistical tests were employed to determine the significance
of differences in scores on the variables over time. A repeated-measures
one-way ANOVA was used to determine the statistical significance of
changes in the temporal variables of speech over time, as well as the
formula/run ratio. The ANOVA was used to compare the scores on each
variable for all samples; with pair-wise comparisons, sample 1 was com-
pared with samples 2, 3, 4, 5, and 6, sample 2 with samples 3, 4, 5, 6, and
so on. Two-way ANOVA was employed to check for possible effect of L1
on the scores.
It is important to bear in mind that the small cohort of participants
here, combined with the large variances among participants and scores,
makes it very difficult to determine statistical significance. Rather than
adhere to statistical representations of significance, the reader is urged
to observe that the overall trends are a clear indication of what occurred
over the course of the study.

Identifying Formulaic Sequences

Identifying formulaic sequences in the data was a central concern in

this study. Wray (2002) reviews approaches to the issue of what consti-
tutes a formula and how to detect formulaic sequences in corpora.
She notes that use of corpus analysis computer software is one possible
method of identification, but presents some serious concerns:

It seems, on the surface, entirely reasonable to use computer searches

to identify common strings of words, and to establish a certain fre-
quency threshold as the criterion for calling a string “formulaic” . . .
(however) problems regarding the procedures of frequency counts
can be identified. Firstly, corpora are probably unable to capture the
110 Formulaic Language and L2 Speech Fluency

true distribution of certain kinds of formulaic sequences . . . The

second serious problem is that the tools used in corpus analysis are no
more able to help decide where the boundaries between formulaic
sequences fall than native speaker judges are. (pp. 25, 27, 28)

Clearly, especially for a small corpus like that used in the present study,
use of computer corpus analysis software presents problems. First, the
specific nature of the type of speech elicited here and the relatively
small number of samples from each participant mean that mere frequency
cannot suffice as a criterion for determining formulaic sequences. Some
formulas may be used only once or idiosyncratically in such a situation.
But it is Wray’s second major concern which is most serious in this case.
Many formulas are blended into surrounding language in the transcripts
and many have larger fillable slots, which are impossible for machine
analysis to cope with. As well, since the participants in the study are L2
learners, many formulas are nonstandard or idiosyncratic. So we are left
with what Wray terms “the application of common sense” (p. 28) in
determining what constitutes a formulaic sequence.

Native speaker judgment

Native speaker judgment is another possible means of identifying
formulaic sequences in a corpus. However, Wray (2002: 23) identifies
five weaknesses in this method: it has to be restricted to smaller data
sets; inconsistent judgment may occur due to fatigue or alterations in
judgment thresholds over time; there may be variation between judges;
there may not be a single answer as to what to search for; application of
intuition in such a way may occur at the expense of knowledge we do
not have at the surface level of awareness.
In the present study, native speaker judgment was used to determine
what constitutes a formulaic sequence. The small corpus accords
with Wray’s first concern about native speaker judgment. Inconsistent
judgment was addressed by having judges individually listen to as well
as read the transcripts, and variation among judges was addressed by
having a discussion and benchmark identification session before actual
individual judging began. The samples used for the benchmark session
were not included in later judgment processes, but were set aside as
complete after the benchmark session ended. In the benchmark session,
two random transcripts were analyzed individually and judges presented
 Design of the Study 111

the formulaic sequences they had marked. As for knowledge beyond the
surface level of awareness of judges, all judges read the most salient lit-
erature on criteria for identifying formulaic sequences. In the benchmark
sessions, the criteria taken from the background literature were used as
justification for selecting particular items as formulaic sequences in the
transcripts, and features of the recorded speech such as speed and
volume changes were also used as guides. Given the small and very spe-
cific corpus obtained, it is logical to avoid complete reliance on frequency
counts as required when using computer corpus analysis. As mentioned
previously, some formulas might be uttered only once or be highly
idiosyncratic. As well, it would require a great deal of ad hoc judgment
on the part of the researcher as to what is or is not actually a formula
after conducting computer analysis; not all combinations of words can
be deemed formulas simply by combining. And it must be borne in mind
that the participants here were intermediate L2 English speakers with a
tenuous grip on English phraseology.
Perhaps the most compelling reason for using native speaker judg-
ment in the present study was the fact that this was a corpus of spoken
language and the act of listening to speech and noting intonation and
pause patterns cannot be done by machine. In other words, human
judgment was required if all the factors relevant to formulaicity in
speech were to be determined.

Judgment criteria
Five overarching criteria were applied in deciding whether a sequence
was a formula, drawn from previous research on formulaic sequences
(see Chapter Two). No particular criterion or combination of criteria were
deemed as essential for a word combination to be marked as formulaic,
these were guides only:

1. Phonological coherence and reduction. In speech production formulaic

sequences may be uttered with phonological coherence (Coulmas,
1979; Wray, 2002), with no internal pausing and a continuous into-
nation contour. Phonological reduction may be present as well, such
as phonological fusion, reduction of syllables, deletion of schwa, all
common features of the most high frequency phrases in English,
but much less in low frequency or more constructed utterances,
112 Formulaic Language and L2 Speech Fluency

according to Bybee (2002). Phonological reduction can be taken as

evidence that “much of the production of fluent speech proceeds by
selecting prefabricated sequences of words” (Bybee, 2002: 217).
2. The taxonomy used by Nattinger and DeCarrico (1992). This includes syn-
tactic strings such as “NP+Aux+VP”(. . .), collocations such as curry
favor, and lexical phrases such as how do you do? that have pragmatic
functions (. . .) (p. 36). This taxonomy is not necessarily applicable in
every case, it was used as a guide to possible formulaicity. For exam-
ple, if a sequence matched other criteria and fit into a category in this
taxonomy, it might be marked as formulaic.
3. Greater length/complexity than other output. Examples would include using
I would like . . . or I don’t understand, while never using would or nega-
tives using do in other contexts. Judges were able to see and hear the
entire output of a particular participant to help in applying this
criterion.
4. Semantic irregularity, as in idioms and metaphors. Wray and Perkins
(2000: 5) note that formulaic sequences are often composed holisti-
cally, like idioms and metaphors, and not semantically. Examples of
this were apparent in the background literature for the judges, and
many formulas readily match this criterion.
5. Syntactic irregularity. Formulaic sequences tend to be syntactically
irregular. This criterion was readily applied to some sequences, but it
was important to check syntactically irregular sequences against other
criteria on this list.

Features of the recorded speech such as speed and volume changes

were also used as guides.
If two or all three of the judges agreed a sequence was formulaic, it was
marked as such. Idiosyncratic or nonnative-like sequences were accepted,
given that the process of judgment involved juggling various criteria—it
was agreed that nonnative-like productions which met all or most of the
criteria were examples of several phenomena marginally relevant to the
study. A sequence might had been stored and retrieved as a whole in
misperceived form, for example, what’s happened instead of what happened,
or thanks god instead of thank god. The communicative and cognitive
stress of the retell situation also might have caused such utterances,
as participants needed to recall events seen in the film while creating
 Design of the Study 113

a running narrative, causing articulatory slips or gaps and inaccuracies

of some components of sequences. This implies that a sequence could
match the criteria and still be idiosyncratic, misperceived and stored
with errors, or misarticulated due to stress.
Formulaic sequences crossed a broad range, from idioms (love your
neighbor, that’s it, instead of) to 2-word verbs (throw away, come back, let out,
give up, fall down) to repeated prepositional and participial phrases
(living in the same house, taking a bath, started fighting, out of the house, at the
moment, in the middle).

Judgment Procedure

The expert judges were two graduate students in Applied Language

Studies, and the researcher himself. All read Nattinger and DeCarrico
(1992), Coulmas (1979), Peters (1983), Wray and Perkins (2000).
A preliminary discussion session was held in which the criteria were
clarified, and a benchmark session on two transcripts was held to
standardize the overall approach to identification of formulas. Given
that the speech samples were very specific narrative retells, some quite
specific formulas were identified and covered a wide range. Idiosyncratic
and deviant formulaic sequences were accepted along with those more
typical of native speakers. In the end, a high degree of consensus
emerged, with rather surprising results as witnessed by the marked
transcripts. The judges then marked the transcripts individually, in
the sequence of the production of the speech samples, beginning with
sample 1 for a given participant and continuing to sample 2 and on
through sample 6 for the same participant. After the judges had indi-
vidually marked the transcripts for formulaic sequences, marked items
were accepted as formulaic if two or all three of the judges were in agree-
ment. In some cases issues such as location of the boundaries between
formulas and the surrounding language, or judges’ determination that
some items were possibly but not definitely formulaic, were decided by the
researcher.
Given the hypothesis that formulaic sequences would occur with
greater frequency over time in learner speech, and that they play a
facilitative role in fluency as a temporal aspect of speech, qualitative
114 Formulaic Language and L2 Speech Fluency

analysis was conducted. The transcripts were examined for specific

instances in which participants were able to use formulas to produce
longer runs and fewer hesitations or clusters of dysfluencies compared
to earlier samples in which they attempted to express the same element
of the narrative without using formulas. Exemplars of these situations
were sought for each participant across the samples collected.
 Chapter 6

Quantitative Results

Data Analysis

The taped speech samples for all participants were transcribed and the
hesitations marked and timed using the spectrograms in SpeechStation
2 software. Formulaic sequences were identified and marked by three
expert judges who listened to the samples and studied the transcripts.
As mentioned earlier, four temporal variables and a formula/run ratio
were calculated for each speech sample in this corpus:

z Speech Rate (SR): The actual number of syllables uttered, divided by

the total speech time in minutes. This is a gross measure of speed of
speech production, it includes the hesitation time in the total time
spent speaking.
z Articulation Rate (AR): The actual number of syllables uttered, divided
by the total amount of time spent speaking. In this case, the hesitation
time is eliminated from the calculation; this gives a measure of the
speed of actual articulation only.
z Phonation/Time Ratio (PTR): This is determined by totaling the
pause times for each speech sample and calculating it as a percent of
the total speech time. It indicates the amount of hesitation relative to
actual speaking time, a combined measure of pause frequency and
duration.
z Mean Length of Runs (MLR): The mean number of syllables uttered
between hesitations. It indicates the length of utterances between
pauses.
z Formula/Run Ratio (FRR): The ratio between the length of runs and
the number of formulaic sequences in a sample. It indicates how
many formulaic sequences were used relative to the number of runs
in a sample.
116 Formulaic Language and L2 Speech Fluency

The pattern of change over time to indicate increased fluency is increased

SR and AR, reduced PTR, and increased MLR. In addition, a higher FRR
would indicate increased use of formulaic sequences in connection with
increased length of runs, a possible indication that formulas may play a
role in expanding the amount of speech produced between hesitations.
The data were examined from three perspectives:

z First, the temporal variable means for the whole group were analyzed
for change over the six samples. One-way repeated-measures analysis
of variance (ANOVA) was used to test for statistical significance of
changes over time. If significance was indicated by the ANOVA calcu-
lation, pairwise comparisons were examined to locate where signifi-
cance occurred over the six speech samples. The statistical results
here are rather mixed as regards statistical significance, due to large
amounts of variance. The overall trends in scores on the temporal
variables is more informative.
z Second, the means for the temporal variables for the three L1 groups
were compared. Two-way ANOVA was conducted to see if there was a
main effect of L1 or an interactive effect between L1 and time for
each temporal variable.
z Third, changes in the temporal variable scores of each individual
participant were examined.

To determine whether the three different film prompts may have

affected the scores over time, correlations of scores were calculated.
These correlations were calculated between the sets of scores for all film
prompts for each temporal variable, that is, scores for sample 1 were
compared with sample 2, 3, 4, 5, 6, and so on. The correlation coeffi-
cients were compared to see if higher correlations appeared between
samples related to the same film, that is, samples 1 and 4 (Neighbours),
samples 2 and 5 (Strings), and samples 3 and 6 (The Cat Came Back), than
between samples from different film prompts.

Results: Whole Group Data

Speech Rate (SR)

Descriptive statistics for SR across all participants are shown in Table 6.1.
Data are listed by L1 group in ascending rank order based on the first
 Quantitative Results 117

sample, indicated by J (Japanese), mean 75.0; S (Spanish), mean 84.9;

and C (Chinese), mean 117.8. Within each L1 group, individuals are
listed in ascending order based on the first sample. Gender is indicated
for each participant by F (female) or M (male).

Table 6.1 SR Scores by Participants

Sample # 1 2 3 4 5 6

JM Isamu 58.4 59.0 59.5 74.4 96.2 105.9

JF Yuka 61.5 56.6 72.0 77.8 53.9 65.0
JF Natsuko 89.0 73.3 86.9 94.4 88.5 113.9
JM Jun 91.1 79.5 97.3 91.1 87.4 98.5

SF Sally 23.3 88.3 94.3 96.1 85.8 93.3

SM Carlos 75.6 97.7 135.6 110.4 101.1 119.6
SF Lilia 120 87.0 108.9 93.1 101.6 109.4
SM Miguel 120.6 116.5 118.6 110 108.4 119.8

CM Liang 93.3 108.3 106.2 107.4 123.9 126.8

CM Lin 127.8 141.1 153.6 149.4 118.0 144.8
CF Meiling 132.3 116.6 116.5 80.7 116.6 140.0

For SR, the correlation coefficients for earlier and later viewings of the
same film prompts were not systematically higher than correlations
between scores based on different film prompts. This may be taken as
an indication that there was no effect of film prompt on the SR scores
in this study.
The mean and standard deviation for the total sample are shown in
Table 6.2. A repeated-measures analysis of variance test revealed a
significantly higher SR over the six samples, F(5,50) = 3.26, p < 0.05.
Pairwise comparisons revealed significance between the whole group
means for samples 1 and 6; 2 and 3; 2 and 6; 4 and 6; and 5 and 6. While
a significant difference between the means for the first and last samples
indicates growth as measured by SR over the 6 months, no significance
was found between the means for earlier and later performances based
on the same film prompt. That is, for SR, there is no statistically signifi-
cant difference between the whole group means for sample 1 compared
to sample 4 (Neighbours), between samples 2 and 5 (Strings), or between
samples 3 and 6 (The Cat Came Back). However, there are gains for the
means for samples from the same film prompt, a 9.2% gain for sample
4 over sample 1, a 5.6% gain for sample 5 over sample 2, and a 7.7%
gain for sample 6 over sample 3.
118 Formulaic Language and L2 Speech Fluency

Table 6.2 Mean and Standard Deviation of SR

1 2 3 4 5 6 % change

M 90.3 93.1 104.5 98.6 98.3 112.5 24.6

SD 34 26.1 27.0 21.0 19.6 22.3

SR Whole Group
120
100
80
SR

60 M 24.6%
40
20
0
1 2 3 4 5 6
Sample

Discussion
The SR scores show a high degree of variability. The standard deviations
are consistently large and the range of scores is broad in every case. For
example, in sample 1, scores ranged from a low of 23.3 for Sally, to a
high of 132.3 for Meiling, in sample 2 from a low of 56.6 for Yuka to a
high of 141.1 for Lin, in sample 3 a low of 72.0 for Yuka to a high of 135.6
for Carlos. In Sample 4, scores range from a low of 77.8 for Yuka to a
high of 149.4 for Lin, in sample 5 from a low of 53.9 for Yuka to a high
of 123.9 for Liang, and in sample 6 from a low of 65.0 for Yuka to a high
of 144.8 for Lin.
These variations may reflect a range of fluency ability, effects of the
task and context among the participants, or be an indication of effect of
L1 on fluency as measured by this particular variable. In all cases, the
lowest end of the range is the score of a female participant, in five out
of the six samples it is a single participant, Yuka. The highest end of the
range is held by a male participant five out of six times, two of them
Chinese males. Individual participants may have performed inconsis-
tently and in different ways depending on factors unrelated to L1, but
including temporary situational factors.
 Quantitative Results 119

The participants who scored lowest on the SR measure on the first

sample all show increases over the six samples, although the patterns
of development vary among them. The participants with higher SR
scores at the beginning of the study do not appear to share a general
pattern over six samples.
The Chinese participants show relatively high SR scores across
the six samples in this study, due perhaps in part to transfer of
aspects of discourse from their first language. As noted in
Chapter Four, spoken Chinese is characterized by rapid produc-
tion of sequences of fragments, together with a high degree of
false starts, reformulations, stammering, and hedging (Ho, 1993).
The Chinese participants in this study show evidence of transfer
of this speech style to English, and it may be this that allowed
them to show such high speed scores compared to the other two
language groups.
The Spanish L1 participants show lower SR scores overall than the
Chinese participants, but higher than the Japanese group. Given the
high speech rate in spoken Spanish of 276–420 syllables per minute
(Laver, 1994), compared to the 270 to 300 syllables per minute in English
(Pawley & Syder, 1983), the relatively high SR scores for these participants
may be due to L1 influence or transfer.
As for the Japanese participants, their relatively low SR rates may
be due in part to transfer of speech style from their L1. As discussed
earlier, spoken Japanese is characterized by speed variations and a
high degree of fragmentation and frequent pausing, a combination
of features which might mitigate against rapid speech rates (Maynard,
1989).

Articulation Rate (AR)

Descriptive statistics for AR across all participants are shown in Table 6.3.
Data are listed by L1 group in ascending rank order based on the first
sample, indicated by J (Japanese), mean 117; S (Spanish), mean 163.5;
and C (Chinese), mean 177.7. Within each L1 group, individuals are
listed in ascending order based on the first sample. Gender is indicated
for each participant by F (female) or M (male).
120 Formulaic Language and L2 Speech Fluency

Table 6.3 AR Scores by Participants

Sample # 1 2 3 4 5 6

JM Isamu 112.3 121.0 97.6 132.6 154.8 150.9

JF Yuka 134.9 136.9 147.3 158.5 114.9 152.7
JF Natsuko 155.8 121.7 152.1 154.7 152.3 187.3
JM Jun 205.0 175.0 185.2 192.2 181.5 189.2

SM Carlos 161.3 153.6 187.7 162.6 173.3 179.4

SF Sally 122.5 176.7 146.7 179.3 147.8 155.6
SM Miguel 170.0 178.2 176.1 170.0 168.0 198.7
SF Lilia 200.0 146.7 178.2 153.2 162.1 147.3

CM Liang 160.0 191.2 196.5 170.0 188.4 198.7

CM Lin 164.3 206.2 200.6 207.8 168.6 200.7
CF Meiling 208.9 169.5 167.5 80.7 116.6 140.0

For AR, the correlation coefficients for earlier and later viewings of the
same film prompts were not systematically higher than correlations
between scores based on different film prompts. This is likely an indi-
cation that there was no effect of film prompt on the AR scores in
this study.
The mean and standard deviation for the total sample are shown in
Table 6.4. A repeated-measures analysis of variance test revealed no sig-
nificantly higher AR over the six samples, F(5,50) = 0.706, p = 0.621.

Table 6.4 Mean and Standard Deviation of AR

1 2 3 4 5 6 % change

M 163.2 161.5 166.9 160.1 157.1 172.8 5.9

SD 32.2 27.8 29.7 33.2 23.8 23.6

AR Whole Group
175
170
165
AR

160 M 5.9%
155
150
145
1 2 3 4 5 6
Sample
 Quantitative Results 121

Discussion
While there is no statistical evidence of significant development for
the whole group in AR over the 6 months, the data do show some
important patterns. In the case of AR, again we see a high degree of
variability in scores as witnessed by the high standard deviations and
the range of scores.
The mean AR for the whole group dropped considerably for the fifth
sample. It is unclear why this is the case, but it is an effect seen in the
performance of Yuka, Natsuko, Jun, Lin, Sally, and Miguel, more than
half of the group. The same phenomenon occurred with sample 2, the
same film prompt, Strings. In sample 2, AR dropped for Natsuko, Jun,
Meiling, Lilia, and Carlos. Only Natsuko and Jun declined in AR for
both samples. Different individuals show declines in AR for the second
as compared with the fifth sample, making it unlikely that the film
prompt itself is a cause of the phenomenon.
As was the case with SR scores, these results may indicate variation
in fluency ability among participants. The reasons for these shifts may
have to do with L1 effect, individual speech strategies, or with hesitation
phenomena as opposed to speed measures such as SR and AR.
For AR, the participants who began the study with the lowest scores
generally show profiles over the six samples similar to those of their
SR scores. As for the higher AR scorers, for the most part their AR score
patterns over the six samples mirror those of their SR scores, except for
Lilia and Liang, who show clear differences for one particular sample.
Lilia had a difference in AR patterns from sample 5 to sample 6, and
Liang from sample 3 to sample 4.
The language groups fall into the same basic pattern for AR as
they did for SR, with the Chinese group fastest, followed by the
Spanish group, with the Japanese group slowest. Some of the
reason for this may be based on transfer of speech styles from L1, as
discussed earlier.

Phonation/Time Ratio (PTR)

Descriptive statistics for phonation/time ratio (PTR) across all
participants are shown in Table 6.5. Data are listed by L1 group
in descending rank order based on the first sample, indicated by
J (Japanese), mean 50.2; S (Spanish), mean 49.6; and C (Chinese),
122 Formulaic Language and L2 Speech Fluency

mean 36.3. Within each L1 group, individuals are listed in descending

order based on the first sample. Gender is indicated for each partici-
pant by F (female) or M (male).

Table 6.5 PTR Scores by Participants

Sample # 1 2 3 4 5 6

JM Jun 55.1 55.3 47.5 52.4 52 48.6

JF Yuka 54.4 58.7 51.0 51.1 53 57.3
JM Isamu 49.0 49.9 39.4 43.5 37 30.0
JF Natsuko 42.3 39.3 43.0 39.7 42 39.1
SF Sally 82.0 49.2 39.9 45.0 42.5 38.3
SM Carlos 51.8 35.9 27.9 32.8 41.9 33.3
SF Lilia 35.7 41.7 40.2 39.8 38.1 23.9
SM Miguel 29.0 36.0 33.0 36.2 35.4 39.7
CM Liang 43.5 42.6 45.5 37.8 33.5 37.1
CF Meiling 38.2 31.7 30.1 23.1 31.3 27.9
CM Lin 27.1 31.3 24.6 28.0 30.3 28.2

For PTR, the correlation coefficients for earlier and later viewings of the
same film prompts were not systematically higher than correlations
between scores based on different film prompts. This may be a sign that
there was no effect of film prompt on the PTR scores in this study.
The mean and standard deviation for the total sample are shown in
Table 6.6. The one-way ANOVA test revealed marginal significance in the
decrease in the PTR of the participants over the six speech samples,
F (1.99, 19.88) = 3.33, p = 0.057. Since the sphericity assumption was
rejected in this analysis, the degrees of freedom (df’s) were adjusted
using the Greenhouse-Geisser epsilon. Pairwise comparisons showed
that significance occurred between the means of samples 1 and 6; 2 and
3; 2 and 4; and 2 and 6. While a significant difference between the means
for the first and last samples indicates strong development as measured
by PTR over the 6 months, no significance was found between the means
for earlier and later performances based on the same film prompt. That
is, for PTR, there is no statistically significant difference between the
whole group means for sample 1 compared to sample 4 (Neighbours),
between samples 2 and 5 (Strings), or between samples 3 and 6 (The
Cat Came Back). However, an observable reduction in mean PTR of
40.8% occurred between samples 1 and 4, and of 18.3% between
samples 2 and 5.
 Quantitative Results 123

Table 6.6 Mean and Standard Deviation of PTR

1 2 3 4 5 6 % change

M 46.2 42.9 38.4 39 39.7 36.7 −20.6

SD 15.2 9.3 8.5 9 7.6 9.8

PTR Whole Group

50
40
30
PTR

M –20.6%
20
10
0
1 2 3 4 5 6
Sample

Discussion
In the case of PTR, as with SR and AR, considerable variability exists
among participants. PTR is a different type of measure from SR and
AR in that it is related to hesitation in speech as opposed to speed
of delivery. The highest scores here (indicating the most hesitation in
speech) are for Japanese participants, with Yuka, the Japanese female,
the highest in five out of six samples. The lowest scores (indicating
less hesitation) are held by the Chinese participants in five out of six
samples, with Lin, a male, highest in four of these. Lilia, a Spanish
female L1 participant, has the lowest score in one sample only.
For the less fluent participants, looking at PTR patterns in connection
to those for SR and AR, some indications of individual speech process-
ing emerge. For example, Yuka, generally a less fluent speaker on all
three measures throughout the six samples, shows more hesitation but
faster articulation for samples 2 and 6. It may be that she attempted to
speak faster but suffered an increased cognitive load as a result and so
needed to hesitate to formulate utterances more in these cases. On the
other hand, Sally, also generally a less fluent participant, shows slower
articulation in sample 3 but less pausing as well, perhaps indicating an
attempt to slow down speed of utterances with the positive consequence
of less stopping as a result of being able to do real-time formulating of
124 Formulaic Language and L2 Speech Fluency

utterances. Natsuko’s PTR patterns display a similar effect for sample 2.

It may be that she attempted to articulate more slowly in the second
sample to avoid hesitation. Carlos shows a pattern of PTR consistent
with his SR and AR patterns, generally pausing less and speaking faster
over time.
The more fluent participants also show some PTR patterns which can
be linked to SR and AR patterns to indicate how they approached the
speaking tasks and what types of mental processing might have been
used. Liang shows an interesting overall profile on the three measures.
He has relatively high speed measures combined with a high propor-
tion of pausing for the first three samples, as might be expected, since
maintaining speed would require more hesitation for formulating. In
samples 4 and 5, on the other hand, his PTR and AR are lower but SR
higher. It may be that his speech style altered somewhat over the time
of the study to slow articulation somewhat to reduce pausing and allow
formulating of utterances to occur online. This could help to maintain
an illusion of fluency. Lilia shows a similar pattern for sample 6. Lin,
however, shows a rather opposite pattern in several samples, showing a
higher proportion of pauses in sample 2 along with high speed scores.
Meiling also shows this combination in sample 5. It may be that these
two Chinese L1 participants produced relatively rapid speech in these
instances, but had to hesitate more frequently and for longer times as
they were unable to formulate utterances quickly enough or had not
automatized formulaic sequences and other aspects of language to a
degree which would allow such rapid speech. There may also be an
influence of transfer from L1 Chinese, characterized by bursts of frag-
mentary speech and a high degree of reformulation, restarts and fillers
(Ho, 1993).
The three language groups rank similarly for PTR as they do for SR
and AR. Again, transfer of speech styles from L1 may have played a part
in this, particularly for the Japanese participants, whose L1 is character-
ized by a large amount of hesitation in spoken discourse.

Mean Length of Runs (MLR)

Descriptive statistics for mean length of runs (MLR) across all participants
are shown in Table 6.7. Data are listed by L1 group in ascending rank
order based on the first sample, indicated by J (Japanese), mean 3.3;
 Quantitative Results 125

S (Spanish), mean 3.7; and C (Chinese), mean 3.8. Within each L1 group,
individuals are listed in ascending order based on the first sample.
Gender is indicated for each participant by F (female) or M (male).

Table 6.7 MLR Scores by Participants

Sample # 1 2 3 4 5 6

JM Isamu 2.7 2.6 2.5 3.7 3.8 4.2

JF Yuka 2.8 2.5 3.0 3.7 1.8 3.0
JF Natsuko 3.7 3.1 4.0 3.6 3.6 4.1
JM Jun 4.1 3.6 4.1 4.3 4.3 5.1

SF Sally 2.0 2.9 3.4 3.2 3.0 3.3

SM Carlos 2.3 3.6 5.5 4.2 3.5 4.2
SM Miguel 4.9 5.0 4.3 4.1 4.3 4.3
SF Lilia 5.5 3.9 4.8 4.0 4.4 5.6

CM Liang 3.2 3.4 3.4 3.8 4.0 3.9

CM Lin 3.8 4.4 5.0 6.0 4.4 4.7
CF Meiling 4.5 4.8 5.1 5.3 5.2 5.1

For MLR, the correlation coefficients for earlier and later viewings of
the same film prompts were not systematically higher than correlations
between scores based on different film prompts. This may be taken as
an indication that there was no effect of film prompt on the MLR scores
in this study.
The means and standard deviations are shown in Table 6.8. The
one-way ANOVA test revealed a significantly higher MLR over the six
speech samples, F(5,50) = 3.21, p < 0.05. Pairwise comparisons showed
that significance occurred between samples 1 and 6; 2 and 3; 2 and 4;
2 and 6; and 5 and 6. While a significant difference between the means
for the first and last samples indicates strong development as mea-
sured by MLR over the 6 months, no significance was found between
the means for earlier and later performances based on the same film
prompt. That is, for MLR, there is no statistically significant differ-
ence between the whole group means for sample 1 compared to
sample 4 (Neighbours), between samples 2 and 5 (Strings), or between
samples 3 and 6 (The Cat Came Back). However, an increase in MLR of
16.7% occurred between samples 1 and 4, an increase of 5.6%
occurred between samples 2 and 5, and an increase of 4.9% between
samples 3 and 6.
126 Formulaic Language and L2 Speech Fluency

Table 6.8 Mean and Standard Deviation for MLR

1 2 3 4 5 6 % change

M 3.6 3.6 4.1 4.2 3.8 4.3 19.4

SD 1.1 0.8 1.0 0.8 0.9 0.7

MLR Whole Group

4.4
4.2
4
MLR

3.8 M 19.4%
3.6
3.4
3.2
1 2 3 4 5 6
Sample

Discussion
As was the case with the other temporal measures, the data for MLR
show considerable variability as evidenced by standard deviations and
the range of scores. Similar to AR scores, the whole group mean for
MLR drops considerably for the fifth sample. This is largely an effect
of the performance of Yuka, Lin, and Carlos. Unlike the AR pattern,
the MLR drop in sample 5 is not matched by a drop in sample 2, the
same film prompt, Strings. Instead, the whole group mean MLR for
sample 2 is unchanged from that of sample 1. Also, Yuka and Lin show
a drop in sample 5 for both AR and MLR. Therefore, it is unclear why
the drop in performance on these two variables occurs; it is likely not
an effect of the film prompt, since sample 2 scores, the same film
prompt, do not decline as much and for different participants. Further-
more, MLR declines on sample 5 do not occur for the same group of
participants as for AR scores.
If it was the case that AR and/or SR decreased for sample 5, in com-
bination with an increase in PTR and a decrease in MLR, we might
conclude that the film prompt or something in the circumstances
surrounding the retelling negatively influenced fluent speech produc-
tion. However, that is not the profile shown in these data and it is likely
that these somewhat random declines on some variables for some
 Quantitative Results 127

participants do not represent any observable influence of the film

prompt, the gender or L1 characteristics of certain participants, or the
data collection situation.
It is clear that the pattern of development of MLR over the six samples
reflects those for other measures for most participants. As SR, AR, and
PTR rise and fall from sample to sample, MLR follows a similar route.
This is especially the case for Isamu, Lin, Liang, Sally, Lilia, and Carlos.
This indicates that as speed of speech increased and time spent pausing
decreased, length of runs grew, and vice versa.

Formula/Run Ratio (FRR)]

Descriptive statistics for FRR across all participants are shown in Table 6.9.
Data are listed by L1 group in ascending rank order based on the first
sample, indicated by S (Spanish), mean 0.22; C (Chinese), mean 0.27;
and J (Japanese), mean 0.30. Within each L1 group, individuals are
listed in ascending order based on the first sample. Gender is indicated
for each participant by F (female) or M (male).

Table 6.9 Formula/Run Ratio by Participants

Sample # 1 2 3 4 5 6

SF Sally 0.08 0.22 0.39 0.33 0.29 0.36

SM Carlos 0.09 0.26 0.30 0.30 0.24 0.23
SF Lilia 0.32 0.28 0.44 0.40 0.40 0.59
SM Miguel 0.41 0.44 0.45 0.50 0.48 0.41

CM Liang 0.13 0.22 0.32 0.26 0.34 0.37

CM Lin 0.33 0.21 0.41 0.61 0.46 0.53
CF Meiling 0.35 0.47 0.69 0.51 0.61 0.62

JM Isamu 0.06 0.12 0.07 0.22 0.16 0.25

JF Yuka 0.22 0.45 0.33 0.40 0.31 0.38
JF Natsuko 0.36 0.23 0.33 0.31 0.27 0.29
JM Jun 0.56 0.47 0.51 0.58 0.47 0.57

For FRR, the correlation coefficients for earlier and later viewings of the
same film prompts were not systematically higher than correlations
between scores based on different film prompts. This may be inter-
preted as a sign that there was no effect of film prompt on the FRR
scores in this study.
128 Formulaic Language and L2 Speech Fluency

The means and standard deviations are shown in Table 6.10. The one-
way ANOVA test showed a significant increase in FRRs over the 6 months,
F(5,50) = 7.38, p < 0.05. Pairwise comparisons showed that significance
occurred between many of the sample means, including samples 1 and
all the other samples, 2 through 6. As well, significance occurred
between sample 2 and samples 3, 4, and 6, and between samples 5 and
6. A significant difference between the means for samples 1 and 6 indi-
cates strong development of FRR over time for the whole group. On
the other hand, there is no significance in the differences between
means for FRR on samples based on the same film prompt except for
the samples based on Neighbours, samples 1 and 4. FRR means increased
48.1% between samples 1 and 4, 19.4% between samples 2 and 5, and
5.1% between samples 3 and 6.

Table 6.10 Mean and Standard Deviation for Formula/Run Ratio

1 2 3 4 5 6 % Change

M 0.27 0.31 0.39 0.40 0.37 0.41 51.9

SD 0.15 0.12 0.15 0.13 0.13 0.13

FRR Whole Group

0.5
0.4
0.3
FRR

M 51.9%
0.2
0.1
0
1 2 3 4 5 6
Sample

Discussion
Typically, for the temporal measures in this data, the FRR variable shows
a high level of variability and a large range in every sample. As was the
case with other measures, the FRR mean for the group dips slightly for
sample 5, but shows increases in all other samples. From sample 1 to 6,
FRR increased a substantial amount, almost 52%.
The FRR is a measure of how use of formulaic sequences might have
influenced the growth of MLR, a key hypothesis for the present study.
 Quantitative Results 129

An increase as shown here, combined with the increase in MLR shown

earlier, indicates that as MLR increased, use of formulas did also.
The overall picture which emerges from the whole group data on FRR
is that increased formula use went hand in hand with improvement on
other temporal variables, including MLR. This might be a sign that the
formula use facilitated the improvements in other measures. In fact, in
many cases, FRR patterns over the six samples mirror those for MLR,
and the SR, AR, and PTR patterns. On the other hand, in several cases
it appears that increases in MLR and SR, AR, and PTR scores do not
relate to use of formulas as measured by FRR. For example, in sample 2
Lin shows increases in SR and AR, together with MLR, but his PTR score
shows increased hesitation and a lower FRR. This indicates that he used
nonformulaic fillers or repeated individual lexical items to lengthen
runs and increase speed, but not formulaic sequences. In any event, the
picture which emerges from a comparison of patterns of all measures
over time is that in general speed and hesitation changes were matched
by increased length of runs, and that use of formulas followed a similar
path of development.

Overall Discussion

It is clear from these data that the participants improved significantly

in fluency as measured by the temporal variables, and that they used
more formulaic sequences along with that improvement. The overall
trend in this group fits the pattern outlined in Chapter Five: Increased
speech and articulation rates, although articulation rate increases
were not statistically significant and were irregular, decreased amount
of hesitation, increased length of runs, and increased use of formulas
per run. ANOVA tests show significance in the changes over time for
all measures except for AR, and in all of those cases, significance
appears between samples 1 and 6. As well, correlations of scores on
various film prompts show no evidence of systematically higher corre-
lations for the samples for the same film prompt.
The small sample used in this study makes it difficult to generalize
from these results. Generalizations about L1 are particularly problem-
atic given that the groups consist of only three or four individuals.
As well, it is important to bear in mind that these data are based on
speech samples involving narrative retell, and, as such, are limited
130 Formulaic Language and L2 Speech Fluency

indicators of the real communicative ability of the participants. These

are performance samples, and thus far from perfect mirrors of under-
lying ability.

Influence of First Language

There is little in these results to make us suspect an overall effect of L1,

as members of all three language groups appear at the top and bottom
of the range of scores, the exception being that Japanese participants
never held top place. Yuka again appears at the bottom of the range sev-
eral times, likely an indication of her general fluency level. The Spanish
L1 group holds top place four out of six times here, although in one
sample the lowest score is held by Sally, an L1 Spanish participant.

Language Groups Data

The means by L1 group for each temporal variable by speech sample

are shown in Table 6.11.

Table 6.11 Temporal Variables by L1

SR by L1
Sample Japanese Chinese Spanish

1 75.0 117.8 84.9

2 67.1 122.0 97.4
3 78.9 125.4 114.4
4 84.4 112.5 102.4
5 81.5 119.5 99.2
6 95.8 137.2 110.5
27.8% 16.5% 30.2%

SR by L1
150

100 Japanese
SR

27.8%
50 Chinese
16.5%
0 Spanish
1 2 3 4 5 6 30.2%
Sample
 Quantitative Results 131

Table 6.11 Continued

AR by L1
Sample Japanese Chinese Spanish

1 152.0 177.7 163.5

2 138.7 189.0 188.2
3 145.6 188.2 172.2
4 159.5 152.8 166.3
5 150.9 157.9 162.8
6 170.0 179.8 170.3
11.9% 1.2% 4.2%

AR by L1
200
150 Japanese
11.9%
AR

100
Chinese
50 1.2%
0 Spanish
1 2 3 4 5 6 4.2%
Sample

PTR by L1
Sample Japanese Chinese Spanish

1 50.2 36.3 49.6

2 50.8 35.2 40.7
3 45.2 33.4 35.3
4 46.7 29.6 38.5
5 46.0 31.7 39.5
6 43.8 31.07 33.8
−12.8% −14.3% −68.1%

PTR by L1
60
50
40 Japanese
PTR

30 –12.8%
20 Chinese
10 –14.3%
0 Spanish
1 2 3 4 5 6 –68.1%
Sample
132 Formulaic Language and L2 Speech Fluency

Table 6.11 Continued

MLR by L1
Sample Japanese Chinese Spanish

1 3.3 3.8 3.7

2 3.0 4.2 3.9
3 3.4 4.5 4.5
4 3.8 5.0 3.9
5 3.4 4.5 3.8
6 4.1 4.6 4.4
22.8% 19.3% 18.2%

MLR by L1
6
5
4 Japanese
MLR

3 22.8%
2 Chinese
1 19.3%
0 Spanish
1 2 3 4 5 6 18.2%
Sample

FRR by L1

Sample Japanese Chinese Spanish

1 0.30 0.27 0.23

2 0.32 0.30 0.30
3 0.31 0.47 0.40
4 0.38 0.46 0.38
5 0.30 0.47 0.35
6 0.37 0.51 0.40
23.3% 88.9% 73.9%

FRR by L1
0.6
0.5
0.4 Japanese
FRR

0.3 23.3%
0.2 Chinese
0.1 88.9%
0 Spanish
1 2 3 4 5 6 73.9%
Sample
 Quantitative Results 133

Discussion
Two-way ANOVA showed no main effect for L1 and no interactive effect
of L1 and time for any of the temporal variables over the six speech
samples.
There appears to be little evidence of an effect of L1 on fluency devel-
opment patterns in this group of learners. The general pattern for all
variables appears to be that the Chinese L1 group are generally at higher
levels across the six samples, followed by the Spanish L1 group, with the
Japanese L1 group generally at lower means for all variables.
For the speed variables SR and AR, the Japanese L1 group shows a
somewhat greater rise in mean SR and AR than the other two groups.
Pause time as a percentage of total speech time shows the Spanish L1
group with greater drop in PTR over the six samples. Mean length of
runs increased for all three L1 groups over the six samples, but each
group took a different route. FRR L1 group means show a complex
picture over the six samples. Chinese L1 participants showed great
increases in the first three samples but this increase leveled off there-
after. Japanese and Spanish L1 groups showed different trends for this
variable, as Spanish participants showed an increase for the first three
samples then dropped for the rest, while Japanese learners were little
changed for the first three samples, increased for sample 4, dropped
for sample 5, then increased for sample 6.
As discussed earlier, there are differences in the spoken discourse
styles of the three first languages, and this may have transferred some-
what or influenced the speech style in English for some participants.
Chinese participants may have transferred some of the rapid speech
and tendency to false starts and repairs from their L1, Japanese partici-
pants may have transferred the varying speed of speech and high fre-
quency of pauses from their L1, while Spanish participants may have
attempted to produce speech at a higher rate of speed as in their L1.
Overall, however, comparing group means for the three L1 groups or
looking at L1 discourse styles gives little information that might be
helpful in understanding how English learners from typologically
different L1s acquire fluency as measured quantitatively. The Japanese
learners showed strong steady increases in speed variables over the six
samples, unlike the other two groups, but this may be a result of their
learning circumstances or individual effort in participating in the study,
or related to the fact that as a group they were at the lowest initial rate
134 Formulaic Language and L2 Speech Fluency

of fluency as measured by the temporal variables. Furthermore, they

had no particular cultural or linguistic similarities to English to give
them any advantage as learners. PTR declined for all three groups as
expected, although the Spanish group mean declined more than the
other two. However, in light of the more erratic changes in the speed
variables for the Spanish L1 group, this PTR decline alone tells us little.
MLR changes in L1 group means are complex, with all groups increas-
ing MLR over the 6 months, but showing group-specific declines and
rises from sample to sample. Formula-run ratios increased for all three
groups, with the Spanish and Chinese groups increasing for the first
three samples and leveling off or dropping for the last four samples,
while Japanese participants showed large increases for samples 4 and 6
only. These are complex but not systematic changes for each L1 group
on the variables over the course of the study.
It is likely that many of the differences among the groups are attribut-
able to the circumstances surrounding individual performance on each
sample, combined with initial fluency ability. The Japanese group means
generally indicate lower fluency from the beginning of the study and
this appears to have affected the performance of the group over the 6
months of the study. On some variables, the Chinese group means gen-
erally indicate higher fluency at the beginning, which may largely
explain the more modest and nonlinear trend in their data over the
course of the study in comparison with the Japanese group. For the
Spanish-speaking group, initial proficiency on these variables generally
lies somewhere between the Japanese and Chinese groups, but they may
have been able to progress more rapidly on certain measures as speak-
ers of a language more closely related to English.

Individual Results

As far as gender, L1, and sociocultural influences are concerned, it may

be more informative to look at individual participant performance to
see evidence of L1-related social and cultural factors at play. However, a
full examination of the role of such factors in fluency development
would require a more qualitative research design which is outside of the
scope of the present study. The following is an analysis of the data for
each participant, organized by language group and gender.
 Quantitative Results 135

Japanese females
Yuka

Table 6.12 Temporal Variable Scores for Yuka

1 2 3 4 5 6 % change 1 to 6

SR 61.5 56.6 72.0 77.8 53.9 65.0 5.7

AR 134.9 136.9 147.3 158.5 114.9 152.7 13.2
PTR 54.4 58.7 51.0 51.1 53.0 57.3 −5.3
MLR 2.83 2.52 3.0 3.7 1.78 3.0 7.1
FRR 0.22 0.45 0.33 0.40 0.31 0.38 72.7

Discussion
Yuka’s profile on all variables is complex. She performed poorest on
sample 5 on all variables, and her PTR scores show increased rates of
pausing over time. However, she managed a strong increase in formula-
run ratio over time. While her data are not a model of the pattern of
variables which shows steady development of fluency, she did demon-
strate improvement in some aspects.
It is interesting that she digressed from straightforward narrative retell
at times to comment on other issues, for example, in sample 4 she makes
lengthy reference to the September 11th terrorist attacks in New York.
Sample 4 is brief and she focuses on the actual retell for less than half of
the speech time. In sample 2 she comments several times that she doesn’t
understand. This may account for her decrease in fluency for that sam-
ple, but also, it may account for the fact that she shows a large increase
in FRR for sample 2. Formulas such as “I don’t understand,” “I don’t
know,” and “I’m sorry” add to the number of formulas but do not
facilitate the actual retell of the narrative itself.

Natsuko

Table 6.13 Temporal Variable Scores for Natsuko

1 2 3 4 5 6 % change 1 to 6

SR 89.0 73.3 86.9 94.4 88.5 113.9 28

AR 155.8 121.7 152.1 154.7 152.3 187.3 20.2
PTR 42.3 39.3 43.0 39.7 42.0 39.1 −7.6
MLR 3.74 3.1 4.0 3.6 3.6 4.1 10.8
FRR 0.36 0.23 0.33 0.31 0.27 0.29 −19.4
136 Formulaic Language and L2 Speech Fluency

Discussion
Natsuko also shows a complex pattern of development. Her speed
scores, SR and AR increase steadily, and her mean length of runs also
increases, although not particularly strongly. Her PTR scores are up and
down from sample to sample, to show a slight decrease overall. Her FRR
actually declines over time, indicating perhaps that any increases in her
fluency profile in the 6 months was not due to automatization of formu-
las, but to other factors such as automatization of syntax or strategies for
fluency which involve lexical devices or other language features. Her
MLR increase is quite modest, and her PTR erratic, which would seem
to show that automatization of formulas does not account for the
increase in speed variables.
It is important to note that Natsuko’s speech samples were usually the
longest and most detailed of the group, and that she began the research
project at a relatively high level of fluency as measured by the temporal
variables. By exploring details of the narratives and making an effort to
address some of the complexities of the retell task directly, it is possible
that she overextended her language and fluency ability somewhat. This
would mean that she did not avoid difficult parts of the narrative or
events which might have been difficult for her to express comfortably,
leaving her to struggle and reformulate, repair, and so on, resulting in
clusters of dysfluencies in places. Furthermore, her FRR declined over
the six samples, which may mean that she lacked the appropriate
formulas to express what she wanted to express, or that she became
cognitively overloaded by the task of recalling what she had seen and
could not use automatized chunks which she might otherwise have
retrieved with more ease.

Japanese males
Isamu

Table 6.14 Temporal Variable Scores for Isamu

1 2 3 4 5 6 % change 1 to 6

SR 58.4 59.0 59.5 74.4 96.2 105.9 81.3

AR 112.3 121.0 97.6 132.6 154.8 150.9 34.4
PTR 49.0 49.9 39.4 43.5 37.0 30.0 −38.8
MLR 2.7 2.6 2.5 3.7 3.8 4.2 55.6
FRR 0.06 0.12 0.07 0.22 0.16 0.25 316.7
 Quantitative Results 137

Discussion
Isamu shows development in all variables to fit the profile of increased
fluency and formula automatization. His speed scores and PTR show
good development, especially in the last three samples. However, his
MLR scores level off for those same last three samples. His FRR develop-
ment shows rises and drops over the samples but more than tripled from
sample 1 to sample 6.
Like Yuka, Isamu shows a tendency to talk about issues related to the
topic or themes of the film prompts in addition to direct retelling of the
narratives. For example, in sample 4 he reflects on the September 11th
terrorist attacks at length, and in sample 5 he comments at length on the
unusual floor plans of the apartments depicted in the film and how they
are unlike Japanese apartment layouts. Unlike Yuka, however, he man-
ages to progress on all temporal aspects of fluent speech over all six
samples. His speech samples are all relatively brief, and he is generally
cautious to retell only the main narrative moves without detail.
He avoids conceptually or linguistically challenging content and
injects his own opinions and observations into the task. While he was
clearly among the least fluent participants at the start of the research
project, he showed steady improvement as time passed. Unlike Yuka, he
was able to perform the task without being overwhelmed each time, and
unlike Natsuko, he chose what to express most efficiently.

Jun

Table 6.15 Temporal Variable Scores for Jun

1 2 3 4 5 6 % change 1 to 6

SR 91.1 79.5 97.3 91.1 87.4 98.5 8.1

AR 205 175 185.2 192.2 181.5 189.2 −7.7
PTR 55.1 55.3 47.5 52.4 52.0 48.6 −11.8
MLR 4.1 3.6 4.1 4.3 4.3 5.1 24.4
FRR 0.56 0.47 0.51 0.58 0.47 0.57 1.8

Discussion
Jun shows a complex and contradictory profile on the five variables. His
SR scores are relatively steady over the 6 months, while his AR scores
actually drop, especially in samples 2 and 5. The film prompt for those
samples was Strings, and it may be that he articulated more slowly while
138 Formulaic Language and L2 Speech Fluency

retelling that particular narrative. His PTR scores, however, drop over
time, especially in sample 3, but rise for sample 4. In this case, the film
prompt would not have had any influence on the pause times. His MLR
grows fairly steadily over the samples, dropping for sample 2 and staying
level for sample 5, which were based on the film prompt Strings. His FRR
increases modestly and again we see the possible effect of the film Strings
in his drop in FRR for samples 2 and 5. Overall, Jun may show a film
prompt effect, which makes his general fluency profile a weak fit with
the goal of increased SR, AR, MLR, and FRR, with reduced PTR.
It may be that Jun showed a reduction in AR because he articulated
more slowly to allow himself to plan ahead or retrieve formulas, con-
cepts, and creatively constructed language in the retells. He shows
reduced pause time over the course of the study but little increase in
length of runs. It could be the case that he used slower articulation
instead of hesitation as a strategy to create an illusion of fluency.
He stays on the topic of narrative retell and does not tend to use self-
talk formulas such as “I don’t know,” or “I think.” He does not show
much evidence of avoiding potentially difficult aspects of the narratives,
except in the first two samples, which are brief and cursory general
descriptions of the main thrust of the stories.

Chinese female
Meiling

Table 6.16 Temporal Variable Scores for Meiling

1 2 3 4 5 6 % change 1 to 6

SR 132.2 116.6 116.5 80.7 116.6 140.0 5.82

AR 208.9 169.5 167.5 80.7 116.6 140.0 −33
PTR 38.2 31.7 30.1 23.1 31.3 27.9 −27
MLR 4.5 4.8 5.1 5.3 5.2 5.1 13.3
FRR 0.35 0.47 0.69 0.51 0.61 0.62 77.1

Discussion
Meiling shows a complex pattern of change in the variables over the six
samples, influenced perhaps in part by two changes in her acquisition
environment: she returned to China for several weeks in the summer,
 Quantitative Results 139

between samples 2 and 3, and she entered the university’s credit English
for Academic Purposes (EAP) program just before sample 4, and there-
fore was studying in an environment focused much more on reading and
writing skills than on spoken language. Meiling’s SR scores drop after
sample 3, and her AR declines over the 6 months but the decline accel-
erates after sample 3. She articulated more slowly over time. However,
her PTR declined steadily over the 6 months, with a surprising drop in
sample 4, where we would expect an increase. Her MLR increased
sharply over the first four samples and declined thereafter, and her FRR
increased for the first three samples and dropped and leveled off there-
after. It is likely that changing circumstances and reduced spoken input
and practice caused Meiling’s fluency development to change in the
second part of the study. Perhaps she hesitated less and articulated more
slowly over time, and used more formulaic sequences but also relied
more on filled pauses or lexical repetition strategies to fill silence in her
retelling during the second half of the study.
In terms of content, Meiling’s retells of the stories differ from those of
other participants markedly in several instances. In sample 1 she spends
the last third of the speech focusing on the theme of the story as she
sees it and comparing it to a Chinese proverb. In sample 2 she sees the
theme as having to do with water conservation, and announces her
upcoming trip home to China. In sample 3 she spends time at the end
thanking the researcher for waiting for her return from China. In sam-
ple 4 she spends time at the end saying what she would do if she were in
a situation like that of the protagonists in the film. In sample 5 she again
spends time at the end identifying the theme of the film Strings as hav-
ing to do with water conservation.
Her overall sense of the speaking task in this research seems to be that
she should comment on the moral themes of the films and agree with
them, which may be a cultural or social value she has learned in her
educational background. This could relate to cultural fluency in that
she interprets narrative retelling as more than just storytelling. How-
ever, this also shows a high level of investment in speaking and a sense
of voice, especially as she chats fairly comfortably about her travel plans
and her own cultural values as related to the film themes. Unfortunately,
none of this helps her to show consistent gains in fluency over the 6
months as measured by the temporal variables.
140 Formulaic Language and L2 Speech Fluency

Chinese males
Lin

Table 6.17 Temporal Variable Scores for Lin

1 2 3 4 5 6 % change 1 to 6

SR 127.8 141.1 153.6 149.4 118 144.8 13.3

AR 164.3 206.2 200.6 207.8 168.6 200.7 22.15
PTR 27.1 31.3 24.6 28 30.3 28.2 4.1
MLR 3.8 4.4 5 6 4.4 4.7 23.7
FRR 0.33 0.21 0.41 0.61 0.46 0.53 60.6

Discussion
Lin shows development in some variables but not in others. His SR and
AR scores increased modestly over the six samples with a drop for sam-
ple 5. In PTR, however, there is almost no development, the trend line
is flat. For MLR, Lin shows steady growth for the first four samples but
drops for the last two to manage an overall increase of 23.7%. His FRR
scores show growth, although samples 2 and 5 show drops, perhaps, like
Jun, a result of the effect of the film prompt for these two samples,
Strings. Overall, it appears that Lin may have increased fluency in some
temporal variables in the early part of the study but lost momentum for
some reason thereafter.
Although Lin began the study at a relatively high rate of fluency
as measured by temporal variables, he shows a loss of momentum
in development in the last half of the study. Also, like Natsuko,
Lin produced consistently long and detailed retellings of the films.
He may have pushed his conceptualizing and formulating abilities
beyond the comfort level and tried to express ideas which were
challenging for him. If this was the case, he showed investment but
compromised his speech fluency in the process as measured by the
temporal variables in this study. His speech overall seems to become
less fragmented as time passed, as he shows improvements in PTR,
MLR, and FRR. This may indicate a move away from a Chinese L1
style of speech, with a high degree of fragmentation and strings of
loosely connected short sequences, to a pattern which fits more with
English speech styles.
 Quantitative Results 141

Liang

Table 6.18 Temporal Variable Scores for Liang

1 2 3 4 5 6 % change 1 to 6

SR 93.3 108.3 106.2 107.4 123.9 126.8 36

AR 160 191.2 196.5 170 188.4 198.7 24.2
PTR 43.5 42.6 45.5 37.8 33.5 37.1 −14.7
MLR 3.2 3.4 3.4 3.8 4 3.9 21.9
FRR 0.13 0.22 0.32 0.26 0.34 0.37 184.6

Discussion
Liang has a profile in all five variables which shows steady fluency devel-
opment. His SR and AR scores developed over the six samples, his PTR
declined, MLR and FRR increased. In several samples there were rever-
sals for one or more variables, but in general he appears to have been a
participant who fit the quantitative profile of development well.

Spanish females
Sally

Table 6.19 Temporal Variable Scores for Sally

1 2 3 4 5 6 % change 1 to 6

SR 23.3 88.3 94.3 96.1 85.8 93.3 300.4

AR 122.5 176.7 146.7 179.3 147.8 155.6 27
PTR 82 49.2 39.9 45 42.5 38.3 −52.3
MLR 2 2.9 3.4 3.2 3 3.3 65
FRR 0.08 0.22 0.39 0.33 0.29 0.36 350

Discussion
Sally’s case is complex and unique in this study. Her first speech sample
was extremely disfluent on all measures and she improved strikingly on
the second sample, but then shows little development after that. Her SR
scores from samples 2 through 6 are generally stable, and she shows
erratic and declining AR scores on the same five samples. Her PTR
drops over almost all samples, though, and her MLR and FRR increase
for the first three samples but not the last three. It may be that she was
not a particularly fluent speaker at first and was challenged by the retell
task, but was more comfortable with it after sample 2. In any event, her
142 Formulaic Language and L2 Speech Fluency

profile shows dramatic improvement after a very difficult first perfor-

mance, but with weak quantitative evidence of fluency development
after that except for several samples.
It may be the case that Sally was shocked by the nature of the task at
the beginning of the research process and had a strong sense of lan-
guage anxiety and little sense of voice or self-efficacy as a result. It is
conceivable that this effect remained throughout the course of the 6
months, in spite of a growth in fluency in some aspects as measured by
the temporal variables. Possibly, she invested little in the tasks and did
the minimum necessary; in most cases her retells are general summaries
of the narratives and she appears not inclined to explore details or step
beyond her comfort level. As a result, she may have put as little strain as
possible on her speaking ability by using a sort of avoidance strategy.
This might explain her development, albeit weak and wavering, in quan-
titative fluency measures over the 6 months. However, it might also be
the case that her discomfort with the nature of the task made these
speech samples weak examples of her actual speaking ability.

Lilia

Table 6.20 Temporal Variable Scores for Lilia

1 2 3 4 5 6 % change 1 to 6

SR 120 87 108.9 93.1 101.6 109.4 −8.85

AR 200 146.7 178.2 153.2 162.1 147.3 −26.4
PTR 35.7 41.7 40.2 39.8 38.1 23.9 −33.1
MLR 5.5 3.9 4.8 4 4.4 5.6 1.8
FRR 0.32 0.28 0.44 0.4 0.4 0.59 84.4

Discussion
Lilia also shows a complicated profile of change in all variables over the
six samples. Her speed scores show fluctuating SR for the first three sam-
ples, gentle growth in the last three and her AR scores actually decline
over the six samples. Her PTR scores are level for the first three samples
then drop for the last three. Her MLR scores fluctuate for the first three
samples, then drop and increase for the last three, while her FRR scores
rise by 84.4% over the 6 months but show little development for samples
3, 4, and 5. Lilia actually shows her best fluency performance in the very
first sample as measured by SR and AR, and modest improvement in the
 Quantitative Results 143

last three samples, although she does show improvement in MLR, PTR,
and FRR in the last three samples. This is not a profile which matches the
ideal for increased fluency over time, as some variables show positive
change and others show declining ability. A possible reason for this is
that Lilia left homestay to live with a Spanish-speaking friend over the
time of the study, reducing her exposure to input in English and stalling
her fluency development. Her level of investment and sense of social
identity might have negatively affected her performance on these tasks,
or she had put herself into a situation in which she had less opportunity
for exposure to input in spoken English and had less chance to produce
L2 output on a daily basis.
Lilia, like Yuka, Isamu, and Meiling, tended to expand her retells away
from direct description of what happened in the films toward an interpre-
tation of the themes involved. For example, in samples 1 and 2 she makes
the theme clear at the end. In sample 4 she devotes roughly the last quar-
ter of the speech time to a discussion of the theme of the film, using many
formulas such as “I think,” and “this is the message.” In sample 5 she also
devotes a long stretch of speech time to exploring the theme of the film
and its implications. However, none of this translated into improved scores
on temporal aspects of fluency, perhaps because the task of narrative retell
and monologic samples did not allow her to show her true underlying
speech ability at its best or most genuine.

Spanish males
Carlos

Table 6.21 Temporal Variable Scores for Carlos

1 2 3 4 5 6 % change 1 to 6

SR 75.6 97.7 135.6 110.4 101.1 119.6 58.2

AR 161.3 153.6 187.7 162.6 173.3 179.4 11.2
PTR 51.8 35.9 27.9 32.8 41.9 33.3 −35.7
MLR 2.3 3.6 5.5 4.2 3.5 4.2 82.6
FRR 0.09 0.26 0.3 0.3 0.24 0.23 155.6

Discussion
While not a perfect fit with the ideal quantitative profile of increased flu-
ency, Carlos’s performance over the six samples shows development in a
nonlinear fashion. His SR scores show strong increase over the first three
samples then show a U-shaped profile for the last three. His AR increase
144 Formulaic Language and L2 Speech Fluency

is a modest 11.2% overall, with steady increases over the first three sam-
ples and weaker increases in the last three. PTR shows a strong overall
decrease, with sharp declines in the first three samples, but more erratic
changes over the last three. Similarly, MLR increased a great deal over
the first three samples, and shows a U-shaped curve for the last three.
FRR increases steadily for the first three samples and declines somewhat
over the last three. Sample 5 appears his weakest on most variables,
although the reason for this is not clear.
It may be that Carlos had an attitude toward the narrative retell
process that affected his performance somewhat. He produced short
samples in most cases, staying fairly close to a summary of the main
movement of the narratives. However, he frequently commented at
the beginning about having seen the film earlier in the course of the
research, or that the film was “interesting” or “funny.” So, on one hand
he produced simple and fast summaries of the film contents, but, on the
other hand, he made his attitudes about the films clear. In any case, he
showed progress in development of fluency.

Miguel

Table 6.22 Temporal Variable Scores for Miguel

1 2 3 4 5 6 % change 1 to 6

SR 120.6 116.5 118.6 110 108.4 119.8 −0.7

AR 170 178.2 176.1 170 168 198.7 16.9
PTR 29 36 33 36.2 35.4 39.7 36.9
MLR 4.9 5 4.3 4.1 4.3 4.3 −12.2
FRR 0.41 0.44 0.45 0.5 0.48 0.41 0

Discussion
Of all participants, Miguel is the one with the least evidence of fluency
improvement over the six samples, despite, or perhaps because of, his
high initial performance as measured by the five variables. His speed
scores show declines over the course of the study, although he shows
large increases in the last three samples. His PTR and MLR show reduced
fluency, and his FRR scores remain level throughout. It may be that
Miguel invested less energy in the retellings than did some others,
or the fact that he left homestay to live with a Spanish-speaking friend
 Quantitative Results 145

during the time of the study limited his exposure to English language
input and caused his fluency development to stall.
Miguel produced consistently lengthy and detailed retells of the film
narratives. Furthermore, there is nothing in his speech samples to
indicate that he was overreaching his ability or using any particular strat-
egy to compensate for overloading his memory or ability to formulate
language or retrieve formulas. He displays a confident and competent
attitude, with a calm expressive voice and conversational tone in all
samples. He stays at a high level of ability over the 6 months, showing
little further development of fluency.

Conclusion

In summary, the data presented here show an increase in fluency over

the course of the study for most of the participants on most of the
measures. There is no evidence of a clear effect of gender or L1 on the
speech fluency development of the participants. According to their
speech samples, individual participants took different paths to improve-
ment over the six samples, and only one, Miguel, demonstrated no
overall improvement. The increase in the ratio of formulas to length of
runs in the data for all but two participants is notable, indicating that it
is related to increasing overall proficiency and fluency development.
Some of the participants may have attempted to describe the film narra-
tives in detail and ended up in a situation in which they were challenged
by the task and unable to demonstrate their underlying abilities. On the
other hand, it appears that some participants were anxious as a result of
the task involved and produced very general and simple summaries of
the film content, also not demonstrating their real underlying abilities.
The influence of social and cultural issues on the performances of the
participants is difficult to identify in this study. The purpose of the study
is to look at the effect of formulaic sequences on the development of L2
fluency in terms of established temporal measures, and the methodology
used does not involve interview or introspection, so it is left to speculation
whether or how identity, voice, anxiety, self-efficacy or cultural fluency
played an active role in any of the speech samples of the participants.
The methodology, including the narrative retell procedure used here,
apparently did not provide a vehicle for the effect of such factors to be
observable. However, particularly in the analysis of individual participant
146 Formulaic Language and L2 Speech Fluency

performance, these data do yield some suggestive findings whose nature

and significance will require further research with a different design.
In order to observe how formulas might have facilitated more fluent
speech over time, is useful to compare use of formulas in the perfor-
mances of the participants on earlier and later retells of the same film
prompt. These comparisons are presented in the next chapter.
 Chapter 7

Qualitative Results

While the statistical analyses show fluency development for the group as a
whole over the six speech samples on all temporal variables, supporting
the hypotheses of this study, examination of formula use to further fluency
can only be effective when the same film prompt is used. In other words,
comparing the data qualitatively across samples using different film
prompts is problematic. The events in the three films and the language
needed to describe them are considerably different and meaningful
comparative analysis of participant speech in such cases is doubtful.
Rather, it seems preferable to compare the speech samples which show
development of all temporal variables for the same prompt. This can
ensure isolating instances where later samples show how formulas were
used to facilitate the development of the temporal aspects of fluency.
Table 7.1 shows which participants show patterns of temporal vari-
ables corresponding to the hypotheses for the same-film prompt when
the second viewing is compared to the first. The temporal pattern which
is selected for this is increased speech rate, reduced Phonation/Time
Ratio (PTR), longer Mean Length of Runs (MLR). In addition to this,
increased ratio of formulas to runs is used as a criterion.
After this quantitative analysis of the data was completed, a more
qualitative, discourse-focused analysis was undertaken with the objective
of determining how formula use might have facilitated improvements
in temporal variables. It was hypothesized that, since the temporal
variables showed significant increases in fluency and that length of
runs and formula/run ratios increased, it would be possible to isolate
the role of formulas in the changes over time in the narrative retells.
It was expected that, especially given that participants retold the same
story twice over the course of the data collection, it would be clear where
dysfluency in expressing ideas in an earlier retelling was cleared up by
using formulas in the later retelling. The qualitative analysis was meant
148 Formulaic Language and L2 Speech Fluency

Table 7.1 Same-film Improvement in Fluency Measures (Percent)

Samples 1 and 4: Neighbours
Yuka Isamu Liang Sally Carlos

SR 26.5 27.4 15.1 312.4 46

AR 17.4 18.1 6.3 46.4 0.8
PTR 6.1 11.2 13.1 4.1 36.7
MLR 32.1 37 18.8 60 82.6
FRR 81.8 266.7 100 312.5 233.3

Samples 2 and 5: Strings

Isamu Jun Liang Lilia

SR 63.1 9.9 14.4 16.8

AR 27.9 3.7 No increase 10.5
PTR 25.9 6 21.4 8.6
MLR 46.2 19.4 17.6 12.8
FRR 33.3 No increase 54.4 42.9

Samples 3 and 6: The Cat Came Back

Natsuko Isamu Liang Lilia

SR 31.1 78 19.4 0.5

AR 23.1 54.6 1.1 No increase
PTR 9.9 23.9 18.5 40.5
MLR 2.5 68 11.8 16.7
FRR No increase 257 15.6 34.1

to be a search for exemplars of the facilitating role of formulas in

increased fluency over time in this study.
Initially, it had been envisioned that clusters of dysfluencies might be
reduced by use of formulas. Freed (1995: 131) defines clusters of dysflu-
encies as “the presence of two or more interruptions to the flow of
speech. Examples include filled pauses which occurred in combination
with other speech markers such as repairs and/or unfilled pauses of
more than 0.4 seconds.” However, that type of cluster occurred fre-
quently in both earlier and later retellings of the same film prompt, and
in the comparisons for individuals it was difficult to locate exactly
whether such a cluster was cleared up by use of formulas in the later
retelling, due to other aspects of discourse and the overall different
approach to retelling used in a later instance.
In addition to the issue of isolating instances where formula use
helped clear up clusters of dysfluencies, a second expectation was
proven wrong by the reality of the data. It had been expected that the
 Qualitative Results 149

clause-by-clause or clause chaining principle of narrative oral discourse

described by Pawley and Syder (1983) would become apparent in later
retells of the same prompt, facilitated by use of formulas. Again, the
variation in approach to retelling over time by participants made it dif-
ficult to determine whether the same ideas were being retold using
clause chaining and use of formulas.
These initial expectations were too simplistic; analysis of the transcripts
revealed a much more complex picture than first supposed. It became
apparent that rather than simply retelling the stories in a linear fashion
move by move, the participants often took different approaches to the
retelling of the same story two times. They began their narratives in differ-
ent ways, they focused on different aspects of the stories each time, and
they elaborated more on different parts of the stories. In most cases, they
produced quite different retells of the same film prompt two times, at a
conceptual level as well as at discourse and linguistic levels. In short, it was
virtually impossible in any particular case to simply set the transcripts
side-by-side and compare how particular ideas were expressed each time.
In no case did a participant attempt to express exactly the same idea or
piece of content twice. Therefore, the search for features of fluency and
the role of formulas had to take place in a quite different conceptual,
discourse, and linguistic context for each narrative retell, making a sim-
ple side-by-side or linear search for exemplars a great challenge.
In light of this, a different approach was taken. Two types of broader
analysis were conducted, one focusing on the ways different narrative
moves were expressed, the other focusing on how formulas were used to
facilitate fluency in a broader sense than simply at the level of compari-
son of expression of particular ideas. For the latter analysis, instances in
which MLR and FRR showed improvement from the first retell of a film
to the second were isolated, and particular uses and effects of formulaic
sequences were identified and classified.

Analysis by Narrative Moves

The first part of the qualitative analysis involved examining how each
narrative move in the film stories was expressed by each participant.
The first retelling of a given film prompt was compared to the second
retell, which occurred 3 months later, and they were examined move by
move for improvements in total pause lengths, the number of formulas
used in each, and an increase in MLR. Some clear examples of a facilitat-
150 Formulaic Language and L2 Speech Fluency

ing role for formulas were identified in this way. The following is a
presentation of the clearest examples. The name of the participant is
followed by the speech samples; 1 and 4 are the retells of Neighbours,
2 and 5 are the retells of Strings, 3 and 6 are retells of The Cat Came Back.
After each segment of transcript are presented the total seconds of pause
time in the segment and the total number of formulas and the MLR in
the segment. In each case, a clear pattern exists in that total pause time
is reduced as the number of formulas increases along with MLR.
In these retells of segments of the films, several themes emerge. Partici-
pants used formulas to extend the length of runs and give concise expres-
sion to events in the narratives. By doing so, they were in some cases able
to eliminate all or part of the dysfluency evident in earlier retells of the
same film. As well, pause times and frequencies are reduced in most
instances in the second retell, as use of automatized formulas made
expression smoother. In some instances, the second retell of a film seg-
ment included extra content but nevertheless the participants were able to
navigate the more complex terrain more efficiently by making use of for-
mulas. In several cases, retrieval of a key lexical item in retell number 2 is
followed by a brief pause and a formula containing the key item, uttered
coherently and quickly; this may be an indicator of automatic retrieval as
the lexical item may have triggered retrieval of the whole formula.

Samples 1 and 4: Neighbours

Yuka
Due to the fact that Yuka’s retells of the story differ so greatly in content
and theme, no comparison is possible.

Isamu
The beginning of competition over the flower
z First Attempt
first purpose (1.5) a flower (2.0) um (3.0) beautiful (1.9) um by
(1.3) his (1.0) mine (1.0) their’s mine (1.0) their (1.6) get
14.3 seconds 0 formulas
MLR 1.8
 Qualitative Results 151

z Second Attempt
this case is just flower (2.0) they (0.5) are (0.7) they are their mind
change (1.2) very bad
4.4 seconds 3 formulas
MLR 3.2

The first attempt is one large cluster of dysfluency with long and fre-
quent pauses, no evidence of using formulas.
The second attempt contains one short dysfluency cluster in runs
2 and 3, but a formula is retrieved to extend and complete run 4. This
is followed by a pause and another formula completes the thought
efficiently.

Liang

Setting the scene

z First Attempt

and a little boy (0.7) but ah (0.6) neighbor (0.9) um (0.7) two
homes two chair and two men (0.6) two men (0.5) sitting the two
chair (1.4) um (0.5) with the newspaper ah (0.3) and ah (1.6) smoke
(1.3) and they (0.3) they don’t take a fire (0.5) ah a man may give
me fire (0.7) so they (0.5) is very friendly
11.1 seconds 2 formulas
MLR 3.44

z Second Attempt

This in this movie is (0.5) about two guys (0.9) so ah neighbor (0.3)
they live together (0.7) sit together on the (0.4) grass smoke
together
2.8 seconds 4 formulas
MLR 5.0

The first attempt is quite fragmented, it is difficult to find the lan-

guage to describe the scene. It is almost a long cluster of dysfluency.
152 Formulaic Language and L2 Speech Fluency

Attempt number 2 shows how use of formulas can allow a more concise
and accurate description using less time and with shorter pauses and lon-
ger runs. Three simple formulas all containing the word together allow
Liang to efficiently express what he saw without searching for the lan-
guage to do so. It may be that these three similar formulas were retrieved
in a linked fashion, as the first one was retrieved the other two followed
with relative ease based on the shared lexical component together.

Sally

Conclusion and moral of the story

z First Attempt
the things (2.6) aren’t (6.5) ah (0.6) other (14.0) important (16.3)
how (4.5) love (0.4) your (4.5) said (4.5) more important (8.5)
important (11.0) I think
73.4 seconds 1 formula
MLR 1.83

z Second Attempt
this is the reason of this history (0.3) about the neighbors (0.7) and
(1.0) we know that (0.8) we need to (0.9) to have (0.3) to (0.9) have
a good neighbors (1.8) and this is the end of the story (0.9) bye
7.6 seconds 5 formulas
MLR 4.0

The first attempt is extremely dysfluent with many single word or

single syllable runs and very long pauses.
The second attempt is much smoother and shorter but with added
content. Sally still has a slight cluster of dysfluency in runs 6 to 8 but
otherwise it is quite formulaic and with much shorter pauses. The length
of runs shows a large increase, due to the use of formulas. The longest
run in the first retell is a mere four syllables. The second last run con-
tains a long rhetorical device which closes the narrative effectively and
allows for a long run of nine syllables. The lengthy 1.8 second pause
preceding it is likely a function of the speaker having run out of things
to say and may not be a sign of struggle to formulate the following run,
 Qualitative Results 153

which is uttered quickly and with considerable phonological coherence

compared to other parts of the narrative.

Carlos

Mutual destruction
z First Attempt
yeah fin final (1.0) finally (0.5) ah (0.8) and and (0.5) any anybody
(0.7) ah have (0.5) two flowers ah
4.0 seconds 0 formulas
MLR 3.14

z Second Attempt
the end of history (0.8) everybody’s (0.7) died (1.0) but (0.5) um
(0.5) each man has the rose in your in your (0.9) funeral you
4.4 seconds 2 formulas
MLR 3.70

The first attempt is difficult to understand, fragmented and with

repairs and repetitions. The second attempt is started by a 6-syllable run
with a rhetorical formula. A cluster of dysfluency follows that but the use
of the formula each man seems to help stretch the 6th run somewhat.
There is greater overall pausing but better MLR and evidence of some
use of formulas.

Samples 2 and 5: Strings

Isamu

Preamble to setting the scene

z First Attempt
um no speaking (0.5) ah very very (0.4) difficult (0.6) but (0.4) very
fine (0.3) I like (0.5) this (1.0) video
3.7 seconds 0 formulas
MLR 3.14
154 Formulaic Language and L2 Speech Fluency

z Second Attempt
um before I don’t know this ship ah this ship name (0.5) but now
I (0.3) know because I found Titanic this movie is Titanic (0.8)
so (0.8) maybe I guess
2.4 seconds 3 formulas
MLR 7.0

The first attempt is simple and consists largely of 1 or 2-word runs

separated by substantial pauses.
The second attempt deals with different content. Several formulas
help to extend the length of runs and the pauses are shorter overall.
One of the longest runs, the third one, is nonformulaic. Using the self-
talk formula I don’t know in the first run allows the speaker to produce a
fairly lengthy run which may have given him mental energy to deal with
the subsequent talk with relative ease. The following two or three runs
are not particularly long or formulaic but are uttered comfortably and
quickly. This effect appears to die out after the end of this segment. As
well, at this point he is dealing with his personal interpretation of the
film and not retelling the narrative per se, which may allow for more
relaxed speaking which is reflected in overall fluency features.

Jun

Guests arrive and begin to play music

z First Attempt
came (0.3) his house (1.4) to make music (2.5) and they played
music (0.7) with (1.1) guitar (0.4) violin contrabass
6.4 seconds 2 formulas
MLR 3.0

z Second Attempt
some people will come to his house / to play music (1.5) their instru-
ments are also strings
1.5 seconds 3 formulas
MLR 10.0
 Qualitative Results 155

The second attempt deals more concisely with the content, probably
because Jun avoids trying to recall the names of the particular instru-
ments. The first run is extended by linking two formulas.

Liang

The old woman takes a bath

z First Attempt
prepare (0.9) to wash (1.5) he (1.5) he spend (1.7) many wash in
the (1.2) bath (0.6) bathtub (0.8) in the bathtub

8.2 seconds 0 formulas

MLR 2.38

z Second Attempt
and then the old woman (0.4) is (0.6) washing (0.5) in the wash-
room and just ah (0.4) close (0.6) ah close her eyes (0.3) make a
dream

2.8 seconds 4 formulas

MLR 3.43

The first attempt is characterized by repairs, repetitions, and

short runs with many pauses. It is virtually one large cluster of
dysfluency.
The second attempt contains several useful simple formulas which
make the runs longer and the expression more exact. Only two
dysfluent pause phenomena occur in the second attempt, between
runs 2 and 3, and between runs 5 and 6. The second instance is
interesting in that Liang appears to start to describe the old woman
closing her eyes and retrieves the lexical item close, which leads to
the formula close her eyes after a brief pause and a nonlexical
filler ah. This formula was uttered at high speed, perhaps indicat-
ing automatic retrieval triggered by the initial retrieval of the
verb close.
156 Formulaic Language and L2 Speech Fluency

The old woman returns to her bath

z First Attempt
the the woman (1.0) do (0.6) wa (0.3) wash (0.3) wash in the bath
(0.3) bathtub (0.5) and ah dream again
3.0 seconds 1 formula
MLR 2.57

z Second Attempt
the old woman (0.6) washing (0.5) ah keep washing (0.8) and the
(0.4) make her dream by herself
2.3 seconds 2 formulas
MLR 3.60

Similar to the above, here Liang struggles in the first attempt and
produces a cluster of dysfuency in runs 2 to 4 while trying to describe
the return to the bathtub. He produces 1-word runs and many pauses
throughout.
The second attempt is shorter and much more efficient in expressing
basically the same content. He produces a more sophisticated formula
to describe her actions, keep washing, possibly triggered by the retrieval
of the verb washing in the previous run, followed by a short pause and a
nonlexical filler, ah. Again, this formula was uttered very quickly, per-
haps indicating automatic retrieval after the uttering of the core verb
washing a moment earlier.

Lilia

Setting the scene

z First Attempt
Today we (0.8) show a film ah (0.4) about two neighbor (1.3) they
living in / the same building
2.5 seconds 2 formulas
MLR 5.0
 Qualitative Results 157

z Second Attempt
this film is about two neighbors (0.8) they are living in the same
building but (1.0) they are living alone / each other (0.4) alone each
ah of them
2.2 seconds 5 formulas
MLR 8.25

In this case, the second attempt deals with more content, the notion
that the two people are living separately and alone. In spite of this, Lilia
manages to use more formulas, reduce pausing, and lengthen runs. She
uses multiple formulas which are uttered rapidly and coherently in runs
2 and 3.

The old man goes upstairs to fix the leak

z First Attempt
he (0.3) has to (2.1) go to the (0.5) third floor to (0.7) talk with the
neighbor (1.4) um (0.8) she’s (1.3) she was (0.4) she repaired the
(1.0) ah problem in the bathroom
8.5 seconds 4 formulas
MLR 2.90

z Second Attempt
and he decided to (0.4) go to the (1.0) neighbor’s the old (0.4)
woman house (0.4) and (0.7) she leave to (0.8) she leave him to
enter (0.5) to fix the problem in the
4.2 seconds 4 formulas
MLR 4.13

In this case, Lilia produces a cluster of dysfluency in the first attempt

in runs 6 to 9. In the second attempt, she deals with the situation some-
what differently in terms of content, stating that the woman allowed
the old man to enter. She uses an idiosyncratic formula she leave him to
158 Formulaic Language and L2 Speech Fluency

enter, preceded by a pause of 0.8 seconds duration. Interestingly, this

formula follows an initial attempt, a short run consisting of she leave to.
As was the case with Liang, discussed earlier, it may be that the retrieval
of the verb leave triggered the automatic retrieval of the formula,
following a short pause.

The message of the film

z First Attempt
but he (1.3) live alone (0.8) and she live alone (2.8) maybe this is
the message of this ah (1.7) ah short (0.7) film (0.9) they are very
alone but (1.2) they don’t

9.4 seconds 2 formulas

MLR 4.0

z Second Attempt
I don’t know maybe (0.3) is ridiculous but is (0.4) very common
situation in the world of today (1.0) of today in the world (0.3)
everybody (0.3) thought their own (1.0) way and (0.3) for (1.5) one
(2.0) maybe we forgot (0.4) that the people (0.8) need to share with
us and we have many (0.7) many things to share with the other
people (1.2) to learn (1.0) to share (3.4) to stay (0.6) with other
(0.3) with another person (1.2) I don’t (0.8) is a good example of
the (3.0) reality of this of the reality (0.7) I think so (2.0) is no good
being alone

23.2 seconds 20 formulas

MLR 5.22

Here, Lilia goes into greater detail in her second attempt, producing
a lengthy and, at the same time, more fluent reflection on the perceived
message of the story of the film. She uses multiple formulas to create
quite lengthy runs and she runs into dysfluency only in runs 7 to 9.
Because of this, her MLR score for the segment increases greatly, from
4 to 5.22.
 Qualitative Results 159

Samples 3 and 6: The Cat Came Back

Natsuko

Taking the cat to the forest

z First Attempt
and (1.0) first he (1.0) um (1.5) took the cat tried to (0.5) out the
cat (0.4) in the forest (1.2) by car (0.3) but the cat (0.5) um returned
the house (2.2) came back the house (2.6) faster than him
11.2 seconds 5 formulas
MLR 3.09

z Second Attempt
I forget / I forget the order but maybe the f he went to the forest
first (0.6) and ah (0.3) to put it (0.4) put it (0.3) in the forest, leave
it (0.7) but (0.6) ah (1.0) he couldn’t make it (0.3) cause the (0.9)
um (2.8) before he (0.6) he went back to his car (0.4) the cat already
came back to his car and ah
8.9 seconds 8 formulas
MLR 5.0

In this case, Natsuko produces a much more fluent description of the

episode in the second attempt while adding a comment about her diffi-
culty recalling it. While she still is rather dysfluent in runs 2 to 4 and 6 and
7 and 9 to 11, she uses formulas to good effect to extend runs and express
more efficiently. Pausing is reduced and MLR increased. The formulas
used are simple and the last two are similar in that they contain a com-
mon lexical element back. This is necessary to effectively relate the events
and is perhaps triggered by the title of the film, The Cat Came Back.

Isamu

Accidental suicide and pursuit by nine souls of the cat

z First Attempt
she dead (1.0) and (0.5) then (0.5) they happy because (0.3) they
(0.5) separate (0.7) ah (1.0) by (1.0) cat (0.7) cat (1.0) die dead
160 Formulaic Language and L2 Speech Fluency

(0.9) and she unhappy (2.5) ah (0.5) then (0.5) many cat (1.0) she
go (1.5) he (0.5) go to (1.3) heaven with (0.5) many cat (0.4) she
very cry
16.8 seconds 0 formulas
MLR 2.10

z Second Attempt
and then he dead (1.0) yeah (0.5) and then but ah (0.8) next non-
sense (0.5) why cat (1.0) cat dead but (0.8) cat spirit is (0.5) just
nine (0.5) ah a lot of ni it’s ah nine (0.5) spirits (1.0) so (1.0) terri-
ble he grow up
8.1 seconds 4 formulas
MLR 3.33

Here, Isamu’s initial attempt to describe the events consists of 1 or

2-word runs and many lengthy hesitations, almost one large cluster of
dysfluency. In the second attempt, however, he is able to improve fluency
by use of several simple formulas and more direct and concise descrip-
tion. This helps him to increase MLR and reduce pausing significantly.
He makes effective use of the rhetorical device and then to lengthen runs
and mark the sequence of events and it appears to help him to buy time
in articulation because the formula is repeated after a long pause and he
may be using it to create an illusion of fluency as he tries to recall the next
event or formulate the next stretch of language.

Liang

Taking the cat to the sea by boat

z First Attempt
ah in the sea (2.2) and (0.5) at (0.4) he lo’he (0.4) ah he’s in
the sea
3.5 seconds 2 formulas
MLR 2.80
 Qualitative Results 161

z Second Attempt
go to sea (0.7) he he drive his boat (0.9) in the sea (0.3) and ah
want put down (0.5) the cat (0.5) on the sea
2.9 seconds 5 formulas
MLR 3.50

In the second attempt Liang adds the idea that the man wanted to
lose the cat in the sea. He uses simple formulas to express the incident
concisely and he reduces pausing. In the first attempt he appears to get
lost in runs 2 to 4, a cluster of dysfluency following a long 2.2 second
pause, perhaps due to trouble recalling the events in order or due to
lack of language to express what he recalls. In the second attempt, how-
ever, he manages to avoid this problem by using formulas like drive his
boat and in the sea.

Lilia

Taking the cat to the sea by boat

z First Attempt
he tried to (0.4) lose it (0.8) but ah have many problems he tried for
many way (0.7) to lose the cat (0.3) and is (0.9) always (0.3) almost
impossible for he
3.4 seconds 3 formulas
MLR 4.86

z Second Attempt
try to / far away / in the ocean in the in um (0.6) for many ways
(0.6) throw away the cat of his life
1.2 seconds 6 formulas
MLR 8.33

Lilia is more exact in her second rendering of the event, using three
consecutive formulas in the first run to extend it and several others
in subsequent runs. Pausing is reduced and runs lengthened in the
process.
162 Formulaic Language and L2 Speech Fluency

Particular Uses and Effects of Formulaic Sequences

In a further attempt to locate instances of use of formulas facilitating

fluency, retells of the same film prompt by each participant were
examined to determine categories of formula use. For this analysis,
there was a particular focus on transcripts in which MLR and FRR
showed improvement from the first retell to the second. In most cases,
it was clear that the increased use of formulas facilitated the increase in
MLR, and a pattern of categories of formula use emerged.
In most transcripts it is clear that there are also features of discourse
which facilitate effective speaking that are unrelated to the use of for-
mulaic sequences. For instance, participants in some cases went into
much more detail in their narration in the second attempt, rendering
any comparison between two retellings moot. As well, in several cases
participants moved away from straightforward narration in their second
attempt, instead including comments on the events in the narrative,
reflection on the theme of the film, and even connection to world events
such as the September 11, 2001 terrorist attack on the World Trade
Center in New York, which occurred the day before the data collection
number four. In addition to this, some longer runs in later transcripts
are not formula-based, but rather, involve more complex language and
vocabulary. Two further non-formula-based features of more fluent later
retells are the repetition of nonformulaic pieces of language and the
repetition of single words in longer runs.
Overall, however, it is clear that five broad categories of formula
use contributed to increased length of runs in later retellings of a
film prompt:

1. Use of self-talk and fillers.

2. Repetition of formulas in a run.
3. Use of multiple formulas to extend a run.
4. Use of formulas as rhetorical devices.
5. Reliance on one formula or filler repeatedly.

Use of self-talk and fillers

A particularly noticeable feature of the speech samples in this data set
is the increased use over the six samples of self-talk and filler formulas
 Qualitative Results 163

by participants. Such formulas include self-referential collocations as

I know, or I think or I guess. Also included in this category are long strings
used for self-talk or circumlocution such as I don’t know, or I don’t know
the thing’s name.
The following are some examples of how participants used self-talk
and filler formulas to lengthen runs:

Samples 1 and 4: Neighbours

Yuka
In sample 1 Yuka uses few formulaic sequences, and uses the sequence
I guess only once. In sample 4 she uses I think a total of four times, length-
ening the runs by several syllables in the process.

Isamu
In sample 1 Isamu uses very few formulaic sequences of any type, but
uses many self-talk sequences in sample 4, for example, I don’t believe,
I don’t know, I don’t understand. By doing so he lengthens the runs or
produces fluent 2 to 5-syllable runs.

Carlos
Like Isamu, Carlos tends not to use any formulaic sequences in his first
quite dysfluent sample. However, in sample 4 he uses I think and
I don’t know, lengthening runs by several syllables or creating runs of
that length.

Samples 2 and 5: Strings

Isamu
In sample 2 Isamu tends not to use many formulas of any type but
in sample 5 he produces some substantial runs partly with the help
of using self-talk and filler formulas such as “I forgot this word so
but it’s OK.” Similarly, “um before I don’t know this ship ah this
ship name.”
164 Formulaic Language and L2 Speech Fluency

Liang
Liang produces a full-clause long self-talk formula which is a self-
contained run in sample 5:

I don’t know how to say the machine

Lilia
While discussing the moral of the story in depth and at great length in
sample 5, Lilia uses a large number of formulaic sequences not present
in her sample 2. Among these are the self-talk sequences I don’t know,
and I think so. The latter is particularly noteworthy as a formulaic
sequence because she means to say I think (it is . . .) but produces
I think so (it is . . .) This is likely a sign that she has acquired the 3-word
I think so sequence as a whole and is not using I think that, or I think, as
syntactic rules would require.

Samples 3 and 6: The Cat Came Back

Liang
As was the case in other later samples, Liang uses the self- talk formulaic
sequence I think to his fluency advantage in sample 6 as compared to
sample 3. In fact, he uses it twice in two adjacent runs at one point,
extending both runs by two syllables likely with minimal effort:

because all of the souls follow his ah souls I think (1.2) I think a cat’s
souls always to follow

It may be that the 1.2-second pause between these runs is a reflection of

the need to formulate the last part of the second run, perhaps Liang has
difficulty with this and has a syntax error in it. The 1.2-second pause
combined with the formula I think before and after the pause may have
given him enough time in real-time speech to compose the last part of
the second run.

Repetition of formulas in a run

In some cases, participants were able to extend the length of runs by
simply repeating a particular formula consecutively within one run. For
 Qualitative Results 165

example, in one case the formula go upstairs was repeated consecutively

several times in one run by a participant in the second retelling of a film
narrative. This increased the length of a run considerably and gave an
impression of increased fluency as measured by temporal variables such
as MLR.

Samples 3 and 6: The Cat Came Back

Natsuko
Natsuko is more fluent in sample 6 than in sample 3 on all measures
except formula-run ratio. This may be evidence that she produced
longer runs and increased other temporal measures of fluency by
means other than use of formulaic sequences. However, she does
show some use of a strategy of repeating a formulaic sequence within a
run to extend it. For example, when describing one of the cat’s many
returns to the house in the film she uses the formula came back two
times in one run:

And he came back the cat came back to the his house and ah

This results in a run of 13 syllables, only one of which is a filler non-

lexical item ah.
She also repeats a self-talk formula later:

I forget I forget the order but maybe the f he went to the forest

Here she appears to be thinking aloud while buying time to recall

the next event in the narrative and uses a very simple subject + verb
formula to repeat her lack of clear recall. It helps her to produce a
19-syllable run.

Lilia
In sample 3, when describing the man’s unintended suicide by dynamite,
Lilia only manages to produce two short nonformulaic runs:

finally he (2.3) he died

166 Formulaic Language and L2 Speech Fluency

In sample 6, however, she extends the description considerably and

helps to extend one run and create a third one by repeating an adverb/
adjective formula:

the dynamite ah (0.8) ah caused a (0.5) motion where the man

died (0.4) and when he’s really happy he’s hysteric he’s really happy
because (0.5) he’s really happy

Perhaps it is the nature of formulaic sequence retrieval that a single

short formula can be uttered several times in sequence more or less
effortlessly, allowing a large string of discourse to be produced with min-
imal strain on formulating capacity.

Use of multiple formulas to extend a run

A typical feature of many of the longer runs in later narrative retells in
the data was the linking of formulas to create a longer run. Generally,
these were 2-word, shorter collocations strung together or used mixed
with nonformulaic talk to lengthen runs by several syllables. For instance,
in the film Strings several participants described the old man in the story
making music by himself in his room, a combination of three short 2-word
formulas making music, by himself, and in his room. This produces a very
convenient and fluent 10-syllable run.

Samples 2 and 5: Strings

Jun
When describing the old man going upstairs to investigate the water
leak in sample 5, Jun strings together two short formulaic sequences
and produces a 10-syllable run:

. . . found (1.2) some water and went upstairs to her house

In sample 2 he also uses formulaic sequences to describe this situation

but separates them with substantial pauses:

Water (0.3) spoiled from the (1.2) upper stairs (3.7) the old man was
angry and went to the (0.4) upstairs (2.6) he rang the ring (1.2) and
went into (0.4) the old woman’s (0.5) room
 Qualitative Results 167

In the second sample he explored the situation in greater depth and

perhaps was struggling to recall exactly the sequence of events, and was
engaged in a search for appropriate vocabulary to express the events. In
sample 5 he chose a much simpler way of summarizing the action and
produced a brief and much more fluent description.

Liang
Liang manages to produce a long run using two connected formulas in
sample 5, while to describe the same event in sample 2 he uses two for-
mulas but needs to separate them with a short pause:

Sample 2: cross the floor (0.5) to second floor

Sample 5: from her room to his room

It is not clear why the hesitation occurs in sample 2, but it may be that
he was caught trying to recall the destination of the water crossing the
floor . . . the old man’s room? The other floor, which number was it? In
sample 5 he avoids the dilemma by simply stating the water moved from
one place to another and he utters one long and fluent run.

Samples 3 and 6: The Cat Came Back

Natsuko
Natsuko strings two formulaic sequences back-to-back in a run in her
sample 6 which helps to extend a run rather effortlessly while describ-
ing one of the cat’s many returns to the house in the film:

And the cat came back to the house

This is an efficient way to express what occurred and she is able to then
link that with the fact that the owner came back after and discovered the
cat already there, while avoiding a cluster of dysfluency.

Liang
Liang makes productive use of the practice of stringing multiple formu-
las in a single run. When describing the solitary music making of the
man in the film, in sample 3 he remarks only:

And make a music

168 Formulaic Language and L2 Speech Fluency

Instead of this very simple and syntactically flawed single formula run,
in sample 6 he expands the picture and manages an 11-syllable run of
considerably greater sophistication:

He’s make music by himself in his room

It appears that after 3 months Liang was sufficiently more fluent and/or
confident to take a risk and describe in a richer way, using formulaic
sequences to produce longer runs.

Lilia
Lilia also was able to use several formulas in a single run more effectively
in sample 6:

He put outside and ah close the door

Always try to far away the cat but he couldn’t do it

In both cases, Lilia is able to keep the discourse flowing in more com-
plete units and avoid pausing, while producing runs of 9 and 14 syllables
in these cases.

Use of formulas as rhetorical devices

In some cases, participants were able to organize the information in later
retellings by using rhetorical marker formulas. They used beginning for-
mulas such as at the beginning, narrative move markers such as when the
story is go ahead, and endings such as that is the end of the story. All of these
add greatly to the length of runs as well as to the effectiveness of the
storytelling.

Samples 1 and 4: Neighbours

Sally
In sample 1 Sally produces quite dysfluent speech with few formulaic
sequences of any type. However, in sample 4 she uses two common
rhetorical formulas to mark the movement of the narrative effectively,
beginning of the . . ., and and this is the end of the story. She may have had
greater control of the task and the speech requires and was able to
 Qualitative Results 169

retrieve these rhetorical formulas to frame the beginning and end of

the narrative.

Carlos
Like Sally, Carlos used few formulas in sample 1 and produced choppy
text of very short runs. Three months later in sample 4, however, he uses
a starter formula the start the history, and later when the history (.3) go
ahead.
As was the case with Sally, Carlos appears to have begun to acquire a
repertoire of formulaic sequences to mark the progress of the narrative.
They are simple and, in the first case, syntactically inaccurate, and in the
second case preceded by a short pause, but they show the beginning of
an ability to use such units of speech with some effectiveness.

Reliance on one formula or filler repeatedly

In some instances, the participants created an illusion of increased
fluency by relying heavily on one simple formula throughout a later
narrative retell. To introduce the next action in the story, for example,
it was common to use and then, or and next.

Conclusion

Clearly, a qualitative analysis of the film narrative retells yields a com-

plex and somewhat irregular picture. Participants did use formulaic
sequences more frequently in later retells, and they used them to facili-
tate more fluent speech. However, the nature of the retell task makes
direct comparison of the earlier and later film viewings difficult.
From the exemplars presented here we have seen that participants
used formulaic sequences in a variety of ways and to several effects
related to increased fluency of speech. They used formulas to more
concisely express aspects of the narratives. They used them to reduce
hesitation and to lengthen the runs. They altered their focus in the
retells and used correspondingly appropriate formulas to describe and
explain, similar to Dechert’s (1980: 240) observation that more fluent
speakers tend to retell narratives more fluently by establishing “islands
of reliability” where they are confident and pausing or being dysfluent
170 Formulaic Language and L2 Speech Fluency

outside or between these fluent stretches. In the present study, partici-

pants may have also recalled formulas in some cases based on single
lexical trigger items. Furthermore, there is some evidence here that
formulas were used as fillers to buy time to formulate the next stretch
of speech or to recall events in the narratives. They were also strung
together at times based on a common lexical element, perhaps as the
first formula was retrieved, subsequent ones containing a similar lexical
item were triggered. In these cases the articulation was markedly
coherent and rapid.
In addition to the fluency effects of using formulaic sequences,
participants used particular categories of formulas to allow for more
appropriate expression. They used formulas in self-talk and as fillers
to create an illusion of fluency. They repeated formulas within runs
and used multiple formulas within the same run to lengthen runs. They
used formulas as rhetorical devices to piece together and mark segments
of the narratives. In some cases, they relied heavily on one or several
formulas throughout a retell to add to an illusion of fluency, in a sense
establishing lexical “islands of reliability.”
 Chapter 8

Conclusions of the Research Study

The present study has shown that speech fluency development in

English as a second language is related to and facilitated by the use of
formulaic language. This study has dealt with some basic questions
hitherto unasked or unanswered, and the richness of the speech data
analyzed here has provided a valuable source of information about the
development of fluency over time as measured quantitatively and by
discourse analysis. Furthermore, analysis of this data has yielded a wealth
of information about how formulaic language is used by L2 learners to
create more fluent narrative discourse.
At the same time, the complex and shifting nature of the develop-
ment tracked in this study has highlighted issues around the analysis of
spontaneous spoken language data. This study has made it clear that
such development is by no means linear, nor is it readily captured and
examined in empirical research. Indeed, it has raised some important
questions about the nature of formulaicity, the complexity of language
in discourse, and the difficulty of applying cognitive science theory to
the study of extended language performance.

Quantitative Results

The research was framed by four main hypotheses. The hypotheses were
focused mainly on quantitative aspects of speech fluency development
previously established in empirical research:
Over time, with continued learning and experience:

1. L2 speech will exhibit a faster rate of production.

2. L2 speech will exhibit a greater amount of production time spent
speaking as opposed to pausing.
172 Formulaic Language and L2 Speech Fluency

3. L2 speech will exhibit longer runs between pauses.

4. Formulaic sequences will appear more frequently in the longer runs
between pauses.

Hypotheses one to three relate directly to previous research on fluency.

Increased speed, less hesitation, and longer runs are hallmarks of speech
fluency. Moreover, increased speed and reduced pausing indicate faster
language processing and can be taken as indicative of increased auto-
maticity of production.
As the first three hypotheses establish fluency development at a sur-
face, observable level, the fourth hypothesis is at the heart of this
research, a quantitative measure of a facilitating role of formulaic
sequences in fluency development. As the speed and hesitation and run
length develop over time and show increased automaticity, it is expected
that more formulaic sequences will be used in the longer speech runs.
Indeed, all four hypotheses were supported by the data in this study,
as evidenced by the analyses presented in Chapter Six. The whole group
of participants improved on all measures to a statistically significant
degree. Especially relevant to this study were the results for mean length
of runs and the formula/run ratio, in which a high degree of develop-
ment was shown. Taken together, the quantitative results in this study
clearly confirm the hypotheses and indicate a role for formulaic
sequences in the development of fluent L2 speech.
Results by L1 group showed no significant effect. It could be conjec-
tured that cultural issues or the nature of the L1 might influence
fluency development in some ways, but again, the quantitative data
analysis does not bear that out. The very small cohort of participants
here also set a high bar for statistical analyses in terms of between-
subjects factors such as L1. There were only four subjects in each
language group, three in the Chinese group, while there were six
subjects in the male gender group and five in the female group. Even if
effects exist, these small numbers make it unlikely that they would be
statistically significant.
On the other hand, an examination of the development of individual
participants for each dependent variable showed some evidence of
possible influence of gender, L1, and sociocultural factors. Several
participants appeared to take a particular approach to storytelling
and narrative which could be based on L1 or cross-cultural factors.
 Conclusions of the Research Study 173

The fluctuations in performance on some variables over time for some

participants seemed to indicate an effect of issues of self-efficacy, voice,
anxiety, or cultural fluency. However, the structure and methodology of
this study made it impossible to draw definite conclusions about such
interactions and effects. With narrative retell procedures and no inter-
view or self-revelatory discussions with participants, one is left to rely
on supposition and conjecture in an attempt to determine why some
subjects may have performed in certain ways at certain times.

Qualitative Results

As the measurable and quantitative analyses in this study yielded clear

evidence of a possible facilitating role for formulaic sequences in L2 flu-
ency development, so the qualitative discourse analyses provided a rich
view of the workings of the interrelationships between fluency and for-
mulaic language. In fact, the analysis of how formulaic sequences fur-
thered fluent speech and the ways the participants used them is the
greatest contribution of this study. The development of L2 fluency as
measured by change in temporal aspects of speech over time had been
well established prior to the present study. The evidence that increased
use of formulaic sequences paralleled the changes in temporal measures
in this study is a new contribution to knowledge about fluency. The quali-
tative analysis in this study serves to deepen and enrich this knowledge
with evidence of how and for what purposes formulas were used.
The analysis was conducted on speech samples based on the same
film prompt and produced by participants who showed strong improve-
ment in temporal measures between the first and second viewings.
A direct comparison between the samples was impossible due to the
fact that the participants often took quite different approaches to
the retells at different times. However, comparison between chunks of
the narratives, or the moves in the stories, yielded clear and abundant
evidence of how use of formulaic sequences reduced pause times and
increased length of runs.
The use of formulaic sequences appeared to occur in certain classifi-
able ways in these data. Participants used formulas in some instances for
self-talk or fillers in the discourse, perhaps showing what McCafferty
(1994) calls a tendency to convert outer or social speech to inner speech.
174 Formulaic Language and L2 Speech Fluency

As well, participants often repeated formulaic sequences to extend the

length of runs, or they strung together multiple formulas to extend
runs. In some cases it may be that recall of one lexical item at the begin-
ning of a sequence triggered the recall of several more sequences
beginning with the same lexical item, perhaps evidence of a cohort
effect in lexical retrieval, in which retrieval or exposure to the initial
phoneme of a word activates words in the lexicon with that initial
phoneme (Ellis, 2002). Formulas were often used by participants as
rhetorical devices to organize the discourse of their narrative retells,
perhaps sometimes serving as anchors for the subsequent language and
content of the following chunk of the story, similar to Chafe’s (1980)
notion of “islands of reliability.” A final category of formula use involved
the use of one particular formula repeatedly, perhaps to allow for some
controlled processing or conceptualizing or recall of film content while
continuing to speak. These categories of formula use expand somewhat
on those of Wray (2002). Wray’s summary of the value of formulaic
sequences in speech production focuses mainly on their roles in con-
trolling the flow of information, allowing time for mental processing of
other aspects of speech and signalling the organization of speech. The
analysis in the present study adds to this the functions of self-talk and
extending the length of runs of fluent speech.

Methodological Implications

The methodology used in the present study was distinctive to the exami-
nation of the evolution of spontaneous speech data over time, and to
the investigation of formulaic language use as it links with quantitative
measurement of fluency development. It also included participants
from three typologically different language groups, a shift from the
types of L1 groups used to study fluency development in English L2 in
the past.
Since the participants in the study were learning English in a natural-
istic context, it was important to use a method of speech sampling which
would allow the researcher to capture their speaking ability in a way
which was both standardized and flexible, and which would capture a
type of discourse which would be representative of real-life performance.
The use of silent film prompts to elicit narrative was highly effective in
 Conclusions of the Research Study 175

striking this balance. For one, the lack of a soundtrack for the films
ensured that no other language proficiency-related factors could inter-
fere with the sampling of speaking ability. Using films with spoken
dialog or voice-over narration could have produced different results
for different participants depending on their listening comprehension
ability. Furthermore, since narrative is such a common genre of speech
in everyday spontaneous speech, eliciting narrative speech allowed for
stronger data with relevance to real-life daily performance. On the other
hand, selection of such film prompts has to be made carefully, with a
standard number of characters, narrative moves, overall length, and
degree of repetition of actions.
The use of the two quantitative measures of Mean Length of Run
(MLR) and Formula Run Ratio (FRR) in this study helped produce rich
data. MLR had been used in previous studies of temporal aspects of
fluency, and in the present study it yielded findings which were more
complex and informative than the rougher speed measures of Speech
Rate (SR) and Articulation Rate (AR). The FRR was unique to the pres-
ent study and was intended as a quantitative measure of the link between
formula use and MLR, and, by implication, fluency. The FRR results by
group and individual showed complex patterns and led to a deeper and
more sophisticated analysis of the link between formula use and fluency
than could have been done otherwise. The FRR data also led to the
qualitative analysis in this study, which in turn produced a comprehen-
sive picture of the ways in which formula use facilitated more fluent
speech production. Therefore, the FRR is a contribution to the method-
ology of studies of fluency and formulaic language, and bridges the gap
between quantitative analysis of temporal aspects of speech and dis-
course analysis of how formulas are used in speech.
The inclusion of L1 speakers of three typologically different languages
makes this study distinctive in comparison to the tradition in English L2
fluency research. It has been much more common in such research to
use participants from one L1 background or of European language L1
groups. Lennon’s studies (1984, 1989, 1990a and 1990b) used single
participants or small groups of the same European L1. Möhle (1984),
Möhle and Raupach (1989), and other European researchers also
used single participants or small cohorts of a shared European L1
background. Freed (1995) used a group of English L1 learners of
French as a second language, and Riggenbach (1991) a group of
176 Formulaic Language and L2 Speech Fluency

Chinese L1 learners of English. The present study goes beyond this

narrow tradition to include participants from three L1 backgrounds,
which enriches the data in that it allows for insight into how L1 may
have influenced fluency development in English.

Complexity of Development

One overarching theme which has emerged from the present study is
the complexity of human speech and the varying routes which can be
taken to arrive at the same speech goal. In these data the developments
in terms of quantitative measures are not linear, but show a varying and
changing route to improved fluency for the whole group and for indi-
vidual participants. Similarly, the discourse itself displays a wide range of
approaches and methods to achieve a common goal: to retell the stories
seen in the films. The variability seen in the results for both types of data
analysis likely have to do with the fact that the actual data here are sam-
ples of real-time, real-life performance under the constraints of cogni-
tive load, external situational factors, and sociocultural issues.
The nature of the speech task in this study, and the circumstances under
which the speech was produced, likely influenced the results. A certain
amount of cognitive load no doubt occurred as participants attempted or
struggled with the task of recall of the events in the films, while retelling
the narratives at the same time. As well, producing spontaneous speech
under the pressure of being recorded must have influenced the quality
of the results in both quantitative and qualitative terms. In some cases,
participants appeared to differ on their interpretation of what was
expected of them, which influenced their performance.
Circumstances such as the time of day, the season, and events in the
lives of the participants probably also played a role in the complex
picture which these data produce. Participants may have felt anxiety,
frustration, strength, confidence, and a host of other emotions and affec-
tive dispositions to varying degrees over the 6 months of this study. Some
may have experienced homesickness, the stress of moving, changes in
class and teachers, the abrupt and rapid changes of weather in Canada,
and more. All of these things may have had an effect, either positive or
negative, on their performance at the time of a given speech sample.
The sociocultural variables introduced in Chapter Four may have
influenced the route and level of development of the participants.
 Conclusions of the Research Study 177

Language anxiety, especially for the least proficient participants, may

have influenced their earlier performances at least. Similarly, self-
efficacy and a sense of voice may have developed over time for some
participants more than others, depending on their experiences in the
class and the world outside of the study. Given the integral roles that
gender and culture play in most aspects of life and especially in patterns
of communication, it is likely that gender, culture, and L1 all influenced
the types of speech samples produced here. However, it remains
unknown how and to what degree.

Formulaicity and Automatization

The body of research on formulaic sequences and the mental process-

ing involved in fluent speech production shaped the present study.
Formulaic sequences and their nature and functions are the subject of
growing interest in applied linguistics, and the present study takes the
state of knowledge a step ahead on one front. As for the large area of
knowledge centering around cognitive processing in language learning
and use, this study explores notions of automaticity in new ways.
It is in the area of formulaic language and language production that
this research contributes most strongly. Until now, the research on
formulaic sequences and their relation to language production, particu-
larly fluency, has been largely speculative and theoretical in nature.
While the nature of fluency as measured by temporal variables has been
established, the role of formulaic language in fluency has not been
empirically tested to any great extent. The present study is an effort to
conduct that empirical testing, and the results here are encouraging. It
appears from this research that formulaic sequences indeed do facilitate
fluent language production and development in a variety of ways.
The present study was framed in terms of cognitive theory, as it relates
to language production and acquisition. Most previous research on
automatic processing, short and long-term memory, frequency effects
in language acquisition, and other key notions, was conducted in con-
trolled laboratory-style experimental circumstances. The present study
attempts to apply some of these notions to the study of fluency in real
discourse. While it remains unclear how the speech phenomena studied
in this research link to automaticity, or whether the development of flu-
ency here is due to frequency effects, instance development, and so on,
178 Formulaic Language and L2 Speech Fluency

the use of such concepts as a theoretical underpinning for the research

has been important in the research design and the interpretation of
findings. However, it is not possible to show evidence of the workings of
such concepts in fluency development in a study with such a large cor-
pus of language produced under such naturalistic circumstances. This
style of research does not fit closely with the type of methodology
required to examine the effects of specific mental processes on fluency
development, so that any conclusions about automatization or memory
or frequency effects can be speculative at best.

Suggestions for Future Research

The present study takes the body of knowledge about speech fluency
development a step ahead. The role of formulaic sequences in L2 flu-
ency development has been noted and the ways in which that role is
executed have been seen. For the future, both broader and more
specific data and analysis are necessary to consolidate and build on this
initial step.
One important aspect of fluency development which needs examina-
tion is the role of sociocultural and other language-external factors in
speech production and acquisition. A more qualitative type of research
is necessary for this.
As for cognitive science issues in speech study, it appears that perhaps
more experimental, controlled studies are needed. It is complex and
difficult to apply cognitive science concepts to large corpora of language
produced in real-time communication. However, in order to under-
stand fluency and formulaic language and their development and
relationship, it is necessary to take a cognitive science stance at least
in part. For future research smaller corpora than the one used in
the present study, with more constrained task conditions, may yield
firmer results.
Future research on formulaic sequences and fluency development
needs to establish as clearly as possible what constitutes a formulaic
sequence. The current state of knowledge about this is fluid. Corpus
analysis software can assist in the case of written corpora or when deal-
ing with native speaker data. For formulas in spoken data we have the
criteria set forth over the years by Coulmas (1979), Peters (1983),
Nattinger and DeCarrico (1992), Wray and Perkins (2000), Wray and
 Conclusions of the Research Study 179

Namba (2003) and others, but they still lack precision and we are left
dependent on listener judgment rather than a firm set of standards.
Future research needs to address this important issue head on and insist
on a clear definition and set of criteria for identifying particular subsets
of formulaic sequences in speech, especially speech produced by non-
native speakers.
In sum, then, the present study has confirmed several important
hypotheses about formulaic language and fluency development in L2
speech. In addition, it explores how and to what purposes formulaic
language may be used by nonnative speakers of English to create more
fluent narrative monologic discourse. The complexity and richness of
human spoken communication are very present in these data, creating
some challenging issues around data analysis and interpretation. The
exact roles of cognitive processes and sociocultural issues in fluency
development and performance remain uncertain, but the heart of the
research has been confirmed: as fluency in L2 speech develops over
time, the use of formulaic language sequences also develops, playing a
facilitative role.
This page intentionally left blank
 Part III

Applications
This page intentionally left blank
 Chapter 9

Fluency and Formulaic Language in the

Classroom

In language teaching materials and programs, speech fluency has usu-

ally been treated in quite a general manner, despite the fact that it is
often mentioned on the dust jackets of textbooks, in course syllabuses,
and in assessment criteria. Classroom activity centered on fluency has
tended to encourage discussion and more or less unstructured talk,
rather than specific types of activities designed around any clear set of
principles. Our background knowledge about fluency and its foundations
in formulaic language has yet to filter into classroom practice in any
significant way, and creating a disciplined, informed, and focused peda-
gogy of fluency is still a challenge. An awareness of how psycholinguistic
processes contribute to fluency, and how the structure and organization
of language itself leads to fluency is enlightening in general, but what
exactly are the implications for second language teaching and learning?
Or, more specifically, how does knowledge about formulaic language
and its role in fluency development translate into curriculum and class-
room practice?
It is interesting to note that, while the link between formulaic
sequences and speech fluency makes logical sense, it has not been
empirically investigated in much depth, nor has there been much effort
to apply this knowledge in teaching materials development. Besides
the study reported in Part II of this volume, few research projects have
investigated whether training in formulaic sequences has positive effects
on oral language proficiency. One such study is that of Boers et al.
(2006). The researchers conducted a small-scale experiment in which
a pedagogical intervention involving tasks encouraging the noticing
of formulaic sequences was used. The participants’ oral proficiency
was then judged by two blind judges in comparison to a control group
which had not experienced the intervention, and results showed that the
184 Formulaic Language and L2 Speech Fluency

judgments correlated well with counts of formulaic sequences in the

utterances of the participants. The authors conclude that “the use of
formulaic sequences (. . .) was shown to be especially beneficial to per-
ceptions of learners’ fluency and range of expression” (Boers et al.,
2006: 257). Another more recent study investigating a link between
instruction in formulaic sequences and increased speech fluency is that
of McGuire (2009). McGuire divided conversation class learners into an
experimental and a control group. The experimental group experienced
instruction in formulaic sequences, highlighting them in model listening
texts, practicing their use in role plays, and being provided additional
examples of useful sequences taken from the Corpus of Contemporary
American English (COCA). In comparison, the control group experi-
enced a vocabulary and grammar-based conversation class. After 5 weeks,
the experimental group scored higher than the control group on mea-
sures of speech rate and length of runs. As well, they used more formulaic
sequences than the control group, and 16 native speaker judges evalu-
ated their fluency as higher than that of the control group. The body of
research into formulaic language and second language fluency remains
very small at present, but the work which has been conducted shows that
increased facility with formulaic language is linked to fluency.
It is interesting to note that, while there has been a growth in knowl-
edge and awareness of formulaic language in general among research-
ers and academics, there has been little effort to utilize this knowledge
in language teaching methodology at all, let alone any specific focus
on fluency. Some pedagogical plans have been developed which are
based on formulaic sequences, aiming at overall language proficiency
or English for specific purposes or English for academic purposes.
These are not specifically focused on formulaic language and fluency,
however. Nattinger and DeCarrico (1992) outline some general prin-
ciples for the teaching of lexical phrases. Willis (1990) integrates
awareness of formulaic sequences into a set of general teaching mate-
rials based in large part on the COBUILD corpus, and Lewis (1997)
also presents a set of principles and a syllabus design for using knowl-
edge of formulaic sequences in language teaching. In English for
academic purposes, authors such as Mudraya (2005) propose using
corpus-based knowledge of formulaic sequences to elaborate curricula
for engineering English, for example, and researchers such as Biber
and Conrad (1999), Biber, Conrad and Cortes (2004), and Cortes
 Fluency and Formulaic Language in the Classroom 185

(2004) take a corpus-based approach to dealing with issues of discipline-

specific language teaching.
There is a small group of ESL-focused textbooks with a focus on teach-
ing formulaic language, such as a corpus-inspired series by McCarthy
and O’Dell (2002, 2004, 2006) dealing separately with idioms, colloca-
tions, and phrasal verbs. These resources represent a starting point
for teachers wishing to deal with formulaic language, but they do not
concentrate on oral language, and the methodology they incorporate
is largely of the present-practice-produce type, not a task-based or
communicative approach. A number of useful dictionaries of formulaic
sequences exist as well, including those of Oxford University Press (2002,
2009) Hill and Lewis (1997), Benson, Benson and Ilson (1997) and
Spears, Birner and Kleinelder (1994), all serving as excellent sources of
lists of formulaic units, but, by definition, dictionaries are not teaching
material in and of themselves.
Probably the most concerted effort to develop a pedagogy based on
knowledge of formulaic language has been the lexical approach of
Lewis (1997; 2008). Lewis proposes a new approach to syllabus and
curriculum design based on lexis rather than grammar, building on
words, collocations, fixed and semifixed expressions, and so on, with
less attention paid to sentence grammar and uncollocated nouns (Lewis,
2008: 15). His approach emphasizes repetition, noticing, consciousness-
raising. In Implementing the Lexical Approach (2008), Lewis presents
classroom reports from teachers who have used the lexical approach
in various ways in their classrooms.
Two relevant classroom reports in Lewis (2008) are by Powell and
Marks, who used principles of the lexical approach to help learners
with spoken language skills. Powells’s technique involves having learn-
ers mark pause boundaries and syllable and word stress in a sample
written text, followed by marking which words and phrases might be
lengthened if spoken. The marked text is then read aloud, and learners
then prepare texts of their own to mark and read aloud and record.
In his classroom application of the lexical approach, Marks integrates
pronunciation with lexis by encouraging teachers not to introduce
new vocabulary in isolation, but as part of a collocation, and to have
learners mark written texts or talk scripts for chunking of words and
phrases, read them aloud, and use them as models for more spontaneous
speaking activity.
186 Formulaic Language and L2 Speech Fluency

Formulaic Language and Fluency

in the Classroom: Principles

In fact, it is possible to take knowledge of automaticity, formulaic

sequences, and the clause structure of spoken language and integrate
them into classroom second language instruction. Such classroom
activity need not stray from the current state-of-the-art of task-based, inter-
active, communicative, or focus-on-form pedagogy either. What would be
basic elements of a pedagogy of fluency? The fundamentals would be
interaction, production, attending to input, and, of course, attending to
formulaic sequences, all operating in an integrated way.

Attending to input
Consistent exposure to fluent input has been established as key to lan-
guage development, especially when paired with interaction and output.
The vital role played by input in language acquisition has been promoted
by Krashen and Terrell (1983) in the input hypothesis, and VanPatten
(2004) in input processing theory. For Krashen, input is the driving
force behind language acquisition and if learners are provided with
large amounts of comprehensible input which contains language
elements which are new to them, they will acquire language. Krashen
used the formula I+1 to illustrate this, I being the input which is
comprehensible, and the +1 being the new elements. Vanpatten’s model
is more complex than this, and rather than a theory of language acquisi-
tion, it is a model of what occurs during L2 comprehension. VanPatten &
Williams (2007) puts forth four claims about what guides learner pro-
cessing of input data:

1. Learners are driven to get meaning while comprehending.

2. Comprehension is effortful for L2 learners in terms of cognitive
processing and working memory, so they will pay attention to some
but not all features of input.
3. Learners have limited processing capacity and cannot process or
store the same amount of information as native speakers in spontane-
ous communication.
4. Learners make use of some universals of input processing. (VanPatten,
2007: 116)
 Fluency and Formulaic Language in the Classroom 187

Input which is intended to aid in the development of speech proficiency

should be spoken input rather than written. Not only is speech different
from written language in a number of important ways including gram-
mar (see, for example, Carter & McCarthy, 1995), it also can be expected
to exemplify the temporal aspects of fluency and a range of formulaic
sequences appropriate to particular genres and contexts. The input to
which learners are exposed needs to be fluent and natural, which
implies native or native-like speech, rich in formulaic language. The
best such input would be unscripted, more or less spontaneous discus-
sion, whether live or taped. It could therefore be expected to display a
normal rate of speed, and a normal degree of pausing. As well, it could
be assumed that such speech would generally contain strings of chained
clauses as illustrated by Pawley and Syder (1983), and pauses at the junc-
tures of those clauses. Learners should be required to attend to the
content and the context of the input, the use of formulaic language,
and also to focus on features of fluency it exemplifies, for example,
marking pause breaks or clause-juncture intonation contours, noting
the use of appropriate formulaic sequences.

Interaction
Interaction in the second language is an axiom of communicative
language teaching, and the focus has often been on practice and
conversation. Student-to-student interaction is the norm in second
language classrooms, but student-to-native speaker interaction is some-
times encouraged as well, if contextual constraints permit access to
native speakers.
The study of interaction and its benefits to language acquisition has a
long history, dating largely from the early 1980s with Long’s (1983)
introduction of the idea of conversational adjustment. Conversational
adjustments refer to the modifications native speakers make to their
speech when communicating with nonnative speakers, in order to
provide comprehensible input. These modifications include shorter
utterances, simplified syntax, and avoidance of lower frequency lexical
items and formulaic sequences. This view of the beneficial nature of
interaction was later refined into the interaction hypothesis (Long,
1996), which moved the discussion beyond the native-nonnative model
of interaction to a broader one in which student-student interaction
188 Formulaic Language and L2 Speech Fluency

and negotiation is seen as having positive benefits. Long suggests that

interaction has intrinsic benefits for acquisition in various ways:

Negotiation for meaning, especially negotiation work which triggers

interactional adjustments by the NS or more competent interlocutor,
facilitates acquisition because it connects input, internal learner
capacities, particularly selective attention, and output in productive
ways. (1996: 451, 452, emphasis in original).

Various types of interaction can prove beneficial to acquisition, and the

development of a repertoire of formulaic sequences can be a part of the
process. As outlined in Part I of this volume, formulaic language can
make production more fluent and comprehensible, and formulaic lan-
guage facilitates the expression of the various functions of discourse.
The negotiation of meaning which Long and others have examined has
important implications for the development of facility with formulaic
language.

Student to student
Pair or group work in the second language classroom has been shown
to have, in and of itself, a positive impact on the accuracy of student
speech. Research has discovered that, especially in cases in which mean-
ing is negotiated in student-to-student exchanges where information
transfer is the purpose, students tend to reformulate and rephrase
utterances in the direction of improved accuracy of syntax, lexis, and
pronunciation (Chaudron, 1988). We could expect that student-to-
student interaction would have similar positive benefits for fluency.
For one thing, extensive interaction with peers in a theme-based,
meaning-focused context could give students the maximum number of
opportunities to produce speech in real time. The nature of student-
student interaction is more stress-free than is communication with
native speakers, since the fear of misunderstanding or of being stigma-
tized as a speaker of “broken English” is reduced in classroom interaction.
As well, student-student interaction can likely help to produce more
fluent speech by a similar dynamic to how it helps to produce more
accurate speech. In reformulating speech while negotiating meaning,
it could be expected that students would reduce or eliminate excessive
 Fluency and Formulaic Language in the Classroom 189

or inappropriate pauses, pay more attention to clause junctures, and

produce more appropriate formulaic sequences to make themselves
understood more readily. In cases of heterogeneous classrooms, it could
also be expected that students would reduce the amount of transfer
of first language procedural knowledge during interactive tasks, and
reformulate in the direction of native speaker temporal patterns.

Student-to-native speaker
If possible, students should interact with native speakers to develop
second language fluency. This provides a student with opportunities
to produce speech in real time, as well as integrating with the speech
patterns and discourse rhythms of genuine native speaker production.
While native speakers may tend to be forgiving of gaps and lapses
in student accuracy and fluency, unlike student peers, they bring a
different set of expectations of discourse to the communication. Native
speakers bring first language procedural knowledge to the situation,
and display a feel for rhythm, pausing, clause structure, and so on that
student peers do not. As well, their speech serves as a context-specific
model of second language fluency for the student.

Production
For automatization to occur, it is essential that students be required to
produce extended stretches of discourse. Production of this sort pushes
the process of automatization in that the juggling of planning, process-
ing, and encoding all need to take place in real time as the production
tasks are performed.
The importance of output has been brought to the foreground of
language acquisition research by, among others, Swain (1985, 1995),
who pointed out that meaningful output can be just as powerful a driver
of acquisition as input. When learners are engaged in communicating
in L2 and encounter difficulties, they are pushed into making the out-
put more precise, coherent, and accurate. It stands to reason that the
adjustments learners are pushed to make when engaging in interaction
or pushed output involves the use of formulaic language.
Presentations, dramatizations, and role plays are samples of productions
which can be used in connection with virtually any thematic context.
190 Formulaic Language and L2 Speech Fluency

For the production to be most effective, though, there are some compo-
nents of the process which need to be present:

Preparation
Students may need to be given time and support in preparing for a pro-
duction task. Ideas, words and formulaic sequences, timing, articulation,
and semantic and syntactic aspects of the talk may need to be worked
out in advance. Research on task-based language teaching and the value
of integrating learner planning has shown that both pre-task planning
and online or within-task planning have positive results for spoken
language development. Ellis (2005) distinguishes among a variety of
planning types, pointing out that pre-task planning may be either
rehearsal of production or strategic planning—in which learners
take time to prepare how to express the content they need to encode.
Within-task planning may be pressured or unpressured, and generally
involves the manipulation of the time needed to complete a given task.
In general, it can be expected that allowing preparation or within-task
time may facilitate the noticing process and aid in the transition from
controlled processing to more automatic processing. All stages of lan-
guage production, from conceptualizing to formulation, to utterance
may be aided by building planning into task execution by learners, and
the processes of input, interaction, and output can be boosted. The
benefits of within-task planning for fluent production have been studied
by Ellis and Yuan (2005), who examined the fluency, complexity, and
accuracy of learner speech when allowed online processing time as
compared to being pressured to speak rapidly. The planning had little
effect on fluency in terms of speech rate, and was linked somewhat to
disfluencies, but boosted the complexity and accuracy of their speech.

Practice
Learners need to have opportunities to practice their production tasks
in order to increase automatization of the language it contains. Practice
with peers can help, as students check comprehensibility of the output,
organization of ideas, and try to pick up speed as the number of
practices increases. In the end, a polished production should allow the
learner to produce extended speech which goes beyond his usual level
of fluency. Nation (1989) describes the positive benefits of a fluency
 Fluency and Formulaic Language in the Classroom 191

workshop technique which uses peer practice. The 4/3/2 technique he

studies requires students to prepare a talk and deliver it to a peer. The
first talk period is allotted 4 minutes. The listener does not intervene,
but merely serves as a passive audience. The speaker moves to a second
peer and delivers the talk again, this time restricted to a 3-minute time
limit. Speakers change partners a third time, and deliver the talk within
a 2-minute limit. Nation analyzed the fluency variables of students, com-
paring their third delivery to their first:

In all except one case study there was an increase in the rate of speak-
ing from the first to the third delivery . . . the number of false starts,
hesitations . . . and repeated words . . . decreased in each case study
from the first to the third delivery of each talk. (Nation, 1989: 379)

It seems, then, that focused practice like this can help push spoken flu-
ency beyond previous levels.
In practice, repetition is important if formulaic sequences are to
be automatized. Repetition can be built into tasks in three main ways:
task-based, involving the repetition of a particular task in its entirety,
such as a presentation or a role play; meaning-focused, whereby the
particular meaning inherent in a task is repeated in different ways, such
as converting written language to speech; form-focused, with a focus
on improving particular points of language such a formulaic sequences.
Any and all of these types of repetition can be helpful for fluency
development with a focus on developing an automatized repertoire of
formulaic sequences.
A recent and intriguing development in the study of formulaic
sequences and the ways they may be beneficial to fluent production
centers around memorization as a tool for pushing competence. The
power of memory is highlighted in research with a specific focus on
teaching methodology. Gatbonton and Segalowitz (1988, 2005) elabo-
rate sets of principles for encouraging memory and automatization with
implications for dealing with formulaic sequences. These are dealt with
in more depth below. Some phonemic aspects of formulaic sequences,
including alliteration and assonance (Lindstromberg & Boers, 2008a,
2008b) have been shown to facilitate the learning or memorization of
formulas, and an emerging theme in research centers around the posi-
tive effects of harnessing memory as a means of pushing acquisition.
192 Formulaic Language and L2 Speech Fluency

Wray reports on several studies which deal with memorized formulaic

language and its effects on effective communication. In one study (Wray,
2004, 2008), a beginner learner of Welsh memorized phrases and
sentences necessary in order to provide a cooking demonstration broad-
cast on television, all within a 1-week period. The learner conducted
a competent and quite fluent demonstration and 9 months later still
recalled a significant amount of the language. In another study (Wray,
2008; Wray & Fitzpatrick, 2008), more advanced learners were required
to memorize native-like formulaic ways of expressing useful ideas in
their everyday encounters with native speakers. After a week or so of
practice and rehearsal, the learners recorded themselves in real life
encounters using the memorized material. While not all participants
produced the memorized utterances completely accurately in their
real-life encounters, they did report that the memorized language aided
them in confidence, satisfaction, and feeling like native-like interlocutors.
This result lends support to the assertion of Boers et al. (2006) that use
of formulaic sequences can help L2 learners by providing native-like
idiomaticity, a native-like temporal pattern of speech, and multiword
strings of accurate speech.
Other research has shown a measurable perceived positive effect of
memorization. A fascinating study by Ding (2007) of the effects of mem-
orization of large amounts of text by university students in China found
positive effects. The learners in the study had extensive experience mem-
orizing lengthy texts in English, and reported that the practice had made
them better communicators in English by enhancing their fluency, focus-
ing attention on collocations and formulas, and enabling the transfer of
these to real-life communication. Similarly, Walker and Utsumi (2006)
found that memorizing dialogs in Japanese as a second language was
valued by learners as a learning technique and as a boost to fluency and
transferable to real-life communication. In another recent study, Dai and
Ding (2010) found that Chinese L2 learners of English who engaged in
text memorization activities used more formulaic sequences in their L2
writing than those who did not, and that their writing proficiency and
ability outstripped that of learners who did not memorize texts.

Feedback
In current research and discussion in language teaching there is an
emphasis on focus on form (FonF) approaches. This refers to teacher
 Fluency and Formulaic Language in the Classroom 193

intervention in the acquisition process by directing learner attention to

language form and structure while they are engaged in activity with a
primary focus on meaning. Some researchers stress that acquisition can
only take place when there is an attention to form in this way, probably
due to the limitations of working memory—as VanPatten (1990) points
out, learners will attend to meaning first due to memory constraints,
and need to have attention drawn to structural or lexical features which
may be outside their window of attention.
FonF pedagogy has a heavy emphasis on corrective feedback, often
requiring that teachers provide learners with online feedback in the
form of repetition, recasts, or other strategies while they are in the
process of expression (Nassaji, 1999; Ellis, Basturkmen, & Loewen,
2001, 2002). The focus of the feedback in the research is usually gram-
mar, but there can be benefits of this type of feedback on formulaic
language as well, with implications for fluency. A teacher might, for
example, recast a learner’s utterance with a more appropriate or more
fluency-enhancing formulaic sequence, or provide alternatives.
In any case, when students have performed, it is useful for them to
have qualitative feedback on fluency aspects of their production. If the
preparation and practice stages of the process elicited aspects of fluency
such as pausing, speed, false starts, and formulaic sequences, it is neces-
sary to discuss them after the final performance, perhaps using a flu-
ency checklist. This feedback can be accompanied by recorded teacher
modeling. By drawing attention to aspects of student performance, pro-
ceduralized problems of fluency in speech can be returned to the realm
of declarative or explicit knowledge, ready for further practice to
become automatized appropriately.

Formulaic sequences
Key to fluency development, of course, would be the development of a
stock of automatized formulaic sequences. One rather primitive
approach to this would be to present learners with lists of formulaic
sequences to suit specific functions and contexts, then have them use
them in tasks such as role plays. Other possibilities would involve having
learners pick such phrases from realistic input and try to incorporate
them into tasks such as Nation’s (1989) 4/3/2 procedure. Such activities
are sensitive to learner and context variables, and foster automatization
of a repertoire of formulaic sequences suitable to particular learner
194 Formulaic Language and L2 Speech Fluency

needs. They would also foster acquisition in a way consistent with

current pedagogy.
It is important to note that even if formulaic sequences are presented
in commercially available teaching materials, they may not be particularly
relevant to learner needs or even particularly useful in communication in
general. Koprowski (2005), in a survey of contemporary coursebooks,
found that, while many textbook authors pay attention to formulaic
sequences in their syllabuses, they may rely on intuition and not be
grounded in language data such as corpora. Furthermore, they may
present formulaic sequences as linked to particular grammatical struc-
tures. This may serve as cautionary information to teachers who wish to
facilitate the acquisition of formulaic sequences and who rely on com-
mercial materials to provide the best input and tasks for the job.
Corpus research on textbook language in English for academic
purposes (EAP) and English for specific purposes (ESP) has shed
light on the nature of their treatment of formulaic language. Chen
(2010), examining ESP materials, focused on lexical bundles, defined
as “sequences of word forms that commonly go together in natural dis-
course” (Biber et al., 1999: 990) and identified mainly by the statistical
feature of frequency in a corpus. She examined a corpus of introduc-
tory textbook language from electrical engineering and compared the
lexical bundles located there with those found in a corpus of English for
electrical engineering textbook material. She discovered that only
roughly a third of the bundles commonly used in the regular textbooks
were present in the ESP materials, and that they covered a much nar-
rower range of discourse functions in the ESP materials. Wood (2010),
in an investigation of lexical clusters in a corpus of EAP textbooks, found
that the highest frequency of clusters was in instructional language and
not the texts used for input, and that the textbook material paid scant
attention to formulaic language at all.
The gaps between the language in teaching material and in target
language discourse raise concerns about the nature of authenticity in
language input in the classroom. In most language teaching materials
which are commercially available, the covers, website promotions, and
prefaces proclaim that the texts used in the materials are authentic.
However, the nature of the authenticity of texts in teaching materials
is qualitatively different from that of texts used for more real-life
purposes–the teaching material includes a selection of texts which are
authentic from various sources such as newspapers, magazines, recorded
 Fluency and Formulaic Language in the Classroom 195

clips from audio or audiovisual media and so on. The teaching material
then presents the texts according to certain pedagogical criteria. The
overarching question remains: how authentic are ESL texts when com-
pared to real language use in the discourse which is important for the
needs of learners? To determine the authenticity of text to be used in
language teaching, one might look for a match between the discourse
of the text and the discourse of the target registers. It has been shown
above that formulaic language is strongly linked to the construction of
discourse. In corpus research in academic discourse, for example, for-
mulaic sequences (lexical bundles) have been shown to make up a large
proportion of text and spoken communication—in the 40-million-word
Longman Spoken and Written English (LSWE) Corpus, 28% of words
in conversation and 20% of words in academic prose occur in 3- and
4-word lexical bundles (Biber et al., 1999). Lexical bundles have also
been shown to be channel and discipline specific. The discourse func-
tions of lexical bundles have been categorized in a finely grained way by
a number of researchers, giving us a valuable template to apply in test-
ing the authenticity of materials used for language teaching. Such test-
ing would involve determining if there is a match between the lexical
bundles in target discourse register texts and in pedagogical materials,
and whether the bundles are used with the same discourse function
in both. This type of research is quite urgently needed in the area of
spoken language if we are to select spoken input which is relevant to
student needs.
There are apparent gaps between the types of formulaic language
required in real-life language use and that found in commercially
available materials, and it also appears that the materials do not attend
to this aspect of language in meaningful ways. Teachers who wish to use
commercially available material should be aware of how inauthentic
some of it may be, and would be well advised to turn elsewhere for peda-
gogical ideas relevant to the teaching of formulaic language.

Sample Fluency Activities

Classroom activity with a fluency focus must take into account the key
elements of automatization and proceduralization, as well as provide
learners with large amounts of naturalistic input and opportunities to
produce and monitor their own speech. Attention to formulaic
196 Formulaic Language and L2 Speech Fluency

sequences and their acquisition is a key element of the teaching of flu-

ency. A fluency activity must pay attention to the continuous input and
context stimuli which will activate and build the nodes in memory which
McLaughlin, Rossman, & McLeod (1983) see as key to learning and
automatization. As well, the replacement of proceduralized knowledge
from the first language by proceduralized knowledge in the second
language is a goal of fluency activity. Pre-task and within-task planning
time, an awareness of the need to balance input, interaction, and feed-
back can boost the fluency-promoting potential of many activities, as
can use of repetition, context sensitivity, and monitoring.

Formulaic Sequences as Vocabulary

To begin a discussion of activities and activity types which may benefit

fluency through developing facility with formulaic sequences, it is
important to note that formulaic sequences are lexical items first and
foremost. As such, they lend themselves to a range of vocabulary activi-
ties which may or may not be interactive, meaning-focused, and so on.
Some macro strategies for dealing with formulas as vocabulary, adapted
from Liu (2007), are presented below, along with specific associated
activity types.

Macro strategies
A range of fluency-enhancing benefits of formulaic sequences can be
harnessed by dealing with formulaic sequences as vocabulary or lexical
items. Some general strategies for this might include the following:

z Watch for the use of formulaic sequences in daily life encounters:

Learners can be encouraged or assigned to make note of how formu-
laic sequences may be used in their own encounters in the commu-
nity, or in overheard exchanges between others. In cases in which
they are unsure of what they have heard, they may attempt to com-
bine a general phonetic image of the sequence with a sense of its
possible function based on context clues. Teachers might assign par-
ticular sequences to learners to listen for in the world outside the
classroom. See below for more applications of this ethnographic
approach to growing a repertoire of formulaic sequences.
 Fluency and Formulaic Language in the Classroom 197

z Make a list or notebook of formulaic sequences heard:

Learners might be encouraged to list the sequences they hear in the
community or elsewhere. The list could be organized or reorganized
as to function and so on (see below).
z Listen to the media and note how recurring formulaic sequences are
used:
As can be the case with sequences heard or encountered in daily life
encounters, learners may be encouraged to listen for sequences in the
media or to listen for specific assigned sequences in the media. They
may also be encouraged to use phonetic and context cues to deter-
mine what the sequence is.
z Interpret functions and meaning of sequences from the context or
from analyzing their component parts:
Learners may be encouraged to use phonetic, context, or compo-
nent analysis to determine the meaning, function, or structure of a
sequence encountered in the community or in the media.
z Explore the cultural metaphors underlying sequences:
Learners may benefit from examining the cultural concepts and met-
aphors underlying the use and structure of some formulaic sequences.
Lakoff and Johnson (1980) note that English tends to work with
images of time as money (don’t waste a minute), argument as war
(you can never win an argument with her), or ideas as tangible objects
(can you give me an idea?). This type of classification can aid in the
linking of a sequence with a mental image and a category, potentially
facilitating its processing and production.
z Use mnemonics to aid in the storage of sequences:
Use of mnemonic supports such as alliteration and assonance
(Lindstromberg & Boers, 2008a, 2008b), or mental imaging may
help in the mental processing, storage, and production of formulaic
sequences as well.

Specific activities
Some specific activity types which might aid in the acquisition of
sequences might include the following:

z Listen to spoken texts and dialogs and mark sequences on a transcript:

Detailed descriptions of how this might work and how to integrate it
into a sequence of activity are presented below.
198 Formulaic Language and L2 Speech Fluency

z Compare L1 and L2 ways of expressing a particular function or

meaning:
As is the case with classifying formulaic sequences as to function,
meaning, and cultural metaphor, making comparisons to L1 means
of expression can potentially aid in mental processing.
z Search corpora for concordances of sequences:
Corpora available online are often easily searched for specific formu-
laic sequences, and concordance lines can illustrate many ways in
which a sequence can be used or integrated with surrounding text.
z Replace single words with sequences:
This is a classic vocabulary activity, as are the following three.
{ Fill in blanks in transcripts with sequences
{ Complete a transcript with a sequence
{ Use sequences in a narrative or other monolog.
See below for more detail on how this and the following activity may
be integrated with a sequence of activity.
z Use sequences in a role play or dramatization/simulation
z Retell a spoken text
z Use mnemonics
{ Catalog by phonological features
{ Make a semantic web
{ Catalog by key lexical item
{ Link to a mental image
z Describe a picture or picture sequence using sequences
z Explain sequences
{ Define them
{ Elaborate on their meanings/functions
{ Paraphrase a text with or without sequences.

Specific Activities for Fluency

1. Shadowing and tracking

Shadowing and tracking are two imitation activities which have long
been used as a weapon in the battle to fight the fossilization of second
language pronunciation (Ricard, 1986), but which have significant
potential for furthering second language fluency as well.
 Fluency and Formulaic Language in the Classroom 199

The procedure involves collecting spontaneous speech from a native

speaker on tape and transcribing it. Learners then follow the transcript
and shadow it, speaking along with the recorded voice, practicing
repeatedly until they feel they have control over the speed, rhythm, pro-
nunciation features, pause profile, and lexis of the talk. Tracking can
follow, in which the learners repeat the talk without reference to the
transcript, a few syllables after the model, while listening. Then, they
record their best effort after the native speaker sample, on the same
tape. The instructor listens to the recording and responds, giving
feedback on the elements of the talk which remain problematic.
Learners then can make more attempts at shadowing and recording,
using the instructor’s feedback as a guide.
Before dealing with the transcript, learners need to listen to it and
have aspects of the fluency features drawn to their awareness. The
location and duration of pauses, breathing pauses, repair phenomena,
are several such features which a class could look at as a whole, before
doing the individual recording work. Formulaic sequences in the
model could be highlighted as well, with learners adding them to their
repertoire through repetition and monitoring.
The recorded native speakers could be discussing any topic or in
any context-specific way. The formulaic sequences used would be
appropriate, then, for the theme, functions, or contexts involved.
Learners would have an opportunity to routinize and automatize these
phrases and to work on uttering them at a fluent rate with appropriately
placed pauses.

2. Mingle jigsaw
Similar in process to the 4/3/2 technique discussed by Nation (1989) is
the “mingle jigsaw” described by Wood (1998) as an information-sharing
technique. The procedure involves repeated information delivery by
learners to peers, while listening to the varied information the peers
have to convey. Wood instructs students to mingle and share specific
information:

z Divide (six pieces of information) equally among class members.

z Each group should carefully prepare to explain their assigned
(information) as simply and clearly as possible.
z All class members should rise from their seats and mingle as at a party.
200 Formulaic Language and L2 Speech Fluency

z No notes or readings should be carried around the room.

z Each student should explain his or her assigned information to other
students, and listen to classmates explaining theirs.
z After each brief session, students may return to their seats when ready
and jot down a few notes in the appropriate boxes in the table below.
Return to the party to continue sharing your information.
z Continue mixing and returning to your seat until everyone has
completed all the boxes in the table. (Wood, 1998: 59)

The mingle jigsaw technique could be used with as many pieces of infor-
mation as there are members of the class, and does not necessarily need
to be reading-based. Opinions and personal experiences could also be
shared in this way.
Repetition of the same information many times is the key to this
technique. Repetition of this type may help to automatize procedural
knowledge and encourage increased speech rate, more appropriate
pausing, and, of course, the building and consolidation of a large reper-
toire of formulaic sequences.

3. Class photo/Family tree

This basic repetition of information or types of speech is also a focus in
the activities described by Gatbonton and Segalowitz (1988), “Class
Photo,” and “Family Tree.”
In “Class Photo,” learners are to arrange themselves for a photo-
graphic group portrait. They follow teacher modeling of instructions of
where to pose, then take turns instructing their classmates about poses.
The student instructions require that they repeat specific phrases used
by the teacher. They then engage in follow-up activity requiring more
use of these phrases to, for example, suggest how individuals in pictures
could be combined to create a good group photograph.
In “Family Tree,” the class forms two groups to pretend to be members
of a family and describe their kinship relationships with each other.
Individuals from each group then pair up and interview each other,
then rejoin their original “families” to pool information and create an
accurate description of the “family tree” of the other group (Gatbonton &
Segalowitz, 1988: 481–484).
 Fluency and Formulaic Language in the Classroom 201

These activities are communicative and interactive, and, like the other
activities described above, require following models and repetition
within an authentic communicative, purposeful framework. It can be
assumed that automatization of the repeated language and its delivery
constraints would be encouraged.

4. 4/3/2
Nation (1989) describes the positive benefits of the 4/3/2 technique,
which requires students to prepare a talk and deliver it to a peer, first
with a 4-minute deadline, then a 3-minute limit, and finally a 2-minute
limit. Students should change partners for each delivery of the talk.
Among many other variations on this, a somewhat similar speedup
procedure is that described by Schloff and Yudkin (1991), in which
speakers are asked to select a 180-word passage from a written source
and practice reading it aloud with a 60-second time limit, without
sacrificing clarity.

5. Marketplace/Messengers
Nation (1989) describes several activities which go slightly beyond the
4/3/2 procedure discussed above. Both “Marketplace” and “Messengers”
require repetition of language within a purposeful context.
Nation describes the two activities, involving repetition and change of
audience, like the 4/3/2 procedure:

In Marketplace, the learners are divided into buyers and sellers. The
sellers are told what they are selling. It might be holidays, furniture,
books, or anything else. Each seller then prepares a sales talk to deliver
to the buyers. The buyers then circulate around the various sellers
listening to the sales talks and finally making a decision about which
holiday they are going to buy. Each seller has to deliver their sales talk
several times to the different buyers. In Messengers, the learners are
divided into describers, messengers, and makers. The describers have
a model, a tangram, or a diagram to describe. It cannot be seen by
the messengers or the makers. A messenger listens to a describer and
then goes to a maker and tells the maker what to do. Because the task
is complicated, the messenger will need to return to the describer
202 Formulaic Language and L2 Speech Fluency

several times for the same information. One describer can work with
two or three messengers. (Nation, 1989: 383)

Again, repetition is the key to fostering automatization and procedural-

ization of language and knowledge here. Clearly, the possibilities for
classroom fluency-focused activity are numerous.

6. Chain dictations
The repetition and automatization of formulaic sequences implicit in
the above activities can also be used in the technique of chain dictation.
There are many variations on the basic activity of dictation (see Davis &
Rinvolucri, 1989 for a resource). In the chain type of dictation, the
teacher groups learners in groups of three or four and instructs them to
write down the dictation text she will read aloud. The students in the
groups are assigned a number, and the teacher moves to the corridor
outside the classroom, or into a quiet corner of the room. She reads the
dictation text aloud to the students numbered one, who then return to
their group and share what they recall from the dictation. The group
transcribes what they hear, and then students numbered two go to meet
the teacher, who reads the dictation text aloud again. These students
return and the group repairs or completes their written text, and the
students numbered three, four, and so on meet the teacher, listen and
return to their groups to report and work on the transcription.
This type of dictation task can especially be useful in encouraging
noticing of formulaic sequences and the role they play in facilitating
effective communication. The groups of learners who listen to the dicta-
tion and return to their groups need to retain the text in short-term
memory, presumably using a phonological loop. However, the text is
too lengthy for them to retain in a word-by-word fashion, and so they
need to attend to the chunks or formulas it contains in order to
minimize processing overload. Back at the main group, the dictation
text will grow and come into focus as chunks or formulas are fixed
and clarified.

7. Student dictations
Student-to-student dictation is also a potential source of benefit in
dealing with formulaic language and effective communication. In this
 Fluency and Formulaic Language in the Classroom 203

activity, learners are given half a dictation text and assigned to dictate it
to a partner who has the other half. The texts may be the first half and
second half of a whole, or two fragmentary versions of the same whole
text. By carefully reading aloud, the learners encourage and help each
other to complete the entire text. This may be done face-to-face or with
a barrier or even back-to-back, to avoid the overuse of nonverbal cues.
If the dictation text is rich in target formulaic sequences, the learners
in this task are forced to notice them and retain them in working mem-
ory in order to complete the gap-filling mutual dictation. As well, they
need to negotiate often word-by-word in order to complete a formulaic
sequence, which forces a focus on its structure and component parts.

8. Chat circles
Like the “Marketplace” activity of Nation (1989), chat circles involve
dividing a class into two large groups. The groups are to stand in two
concentric circles, the inner circle facing out, the outer circle facing in.
Each of the resulting face-to-face pairs should take turns talking sponta-
neously for 2 or 3 minutes on a simple, everyday topic assigned by the
instructor. When the time is up and both have spoken, the outer or
inner circles step one partner to the right or left. Now facing new part-
ners, again a topic is assigned and learners talk spontaneously for the
same amount of time as in the previous round. This continues until
every outer circle member has spoken with every inner circle member,
always on a different topic. The assigning of topics by the instructor
should move from the immediate and personal or familiar, to the more
abstract and opinion-oriented as the rotation of pairs progresses.
This activity again requires continuous spontaneous speech, but dif-
fers in that the topics to be discussed increase in complexity. This
should further automatization of patterns and procedural knowledge
as the comfort level with simpler topics is raised higher and higher with
the requirement for more complex and abstract ideas and language
delivered in the same brief length of time.

A Speech Syllabus: Oral Process

In teaching spoken language and planning a speech syllabus for any

language course, a sound overall structure is necessary. One such
204 Formulaic Language and L2 Speech Fluency

syllabus type is oral process, a way of developing oral skills in a manner

similar to the way writing skills develop in writing process workshops.
The process allows for a range of task types, including a variety of pre-
task and within-task planning. The oral process approach divides speech
into three broad categories: spontaneous; shaped; formal.

1. Spontaneous speech: This is real-time, fluent speech or conversation,

involving no preparation or planning before delivery. Classroom activity
in this realm can take the form of chat circles or shadowing activities
(see above), or a variety of other activities requiring learners to produce
second language speech under the constraints of real-time production.
Integral to this is the maintenance of a talk journal, which is a dialog
journal of recorded spontaneous speech. The teacher should keep the
recordings and return them to the learners on a weekly basis. They are
to listen to the teacher’s comments from previous recordings and talk
for several minutes to further the discourse. The first such recording
could be on the topic of learner self-introduction, to which the teacher
would respond with some self-disclosures and commentary to stimulate
the next learner speech segment. Feedback should be content-focused;
since so little preparation time is allowed to the learners, it is assumed
they have little time to monitor their production. It is useful for the
teacher responses to model useful formulaic sequences, and the con-
tent of each recording should go in various directions, depending on
student concerns and topics.

2. Shaped speech: This is speech which has been semiplanned or planned

in part, at the level of topic, lexis, and some broad concerns of syntax. It
is generally focused on a genre of speech, for example, narration, pro-
cess description, or comparisons. The activities in a cycle of shaped
speech production culminate in the production of a recorded “portfo-
lio” entry for teacher feedback, and involves the five stages of modeling,
preparation, practice, performance, and feedback:
Modeling —The instructor or a recorded model serves as a sample of
the type of speech involved. Learners are encouraged to attend to
formulaic sequences used.
Preparation—The learners prepare their own spoken texts. Topic gen-
eration and attention to formulaic sequences, other vocabulary,
and some preliminary grammatical advice are guided. A rough
outline is prepared.
 Fluency and Formulaic Language in the Classroom 205

Practice—Learners practice their speech using several peers as

audience. Listeners may give feedback on comprehensibility
and aspects of fluency, or the process may be speeded up with
each delivery as in the 4/3/2 procedure described by Nation
(1989) (see above).
Performance—Learners produce a recorded version of their texts.
Feedback—The instructor gives feedback on comprehensibility, aspects
of fluency, and formulaic sequences, modeling if necessary. Learners
may be given an opportunity to redo the text with particular
attention to the elements of feedback.

An example of such a shaped speech cycle for narratives involves having

learners listen to a traditional story such as a fable. This type of story
is more or less universal across cultures, with some plot twists. Then,
learners are presented with three picture prompts of other such fables.
They prepare to tell one of the three to a partner, taking time to
prepare briefly and clarify vocabulary and pronunciation with the
teacher or peers. After this, they are required to prepare to tell a
similar type of tale from their own cultural background and record it
for teacher feedback.

3. Formal speech: This is a cycle of preparation and practice similar to that

of shaped speech, but taking place over a longer time and requiring a
“staged” presentation in front of the class. Conferencing and peer
feedback sheets can help to zero in on aspects of fluency at various
stages of preparation, and topic and format are student-selected. A
taped rehearsal may be given feedback by the instructor at any stage as
well. The final performance is evaluated by instructor and peers. The
performance could be an individual, group, or pair effort.

Oral process holds great promise as a way of organizing oral skills

courses to allow a systematic focus on fluency, and allows for a wide
range of types of activity. Each component can involve input which is
fluent, native-like, and naturalistic, and permits attending to any aspect
of fluency, including pause profiles, rhythm, clause chaining, and for-
mulaic sequences. Learners get a workout not just in talking spontane-
ously, but in planning speech which is likely beyond their current level
of proficiency and fluency. Preparations and rehearsals push them to
repeat and refine or “polish” their production, the idea being that they
206 Formulaic Language and L2 Speech Fluency

will automatize some of the lexis and procedural knowledge in their

work. It also allows for ample instructor feedback on every aspect of
their speech, an important element of learning.

A Fluency Course

The following model of a fluency program is based on fluency

workshops conducted with intermediate levels in an intensive English
as a second language program in a university setting, with students of
varying L1s. The workshops consisted of three blocks of 6 hours each,
with each block following the sequence outline below.

Input stage
The learners listened to a 10-minute audiotape of a native speaker
engaged in a spontaneous discussion of personal interest. The content
of the speech was discussed, along with the speaker’s attitudes and
feelings toward the topic. The group then listened to the recording
again while following along on a transcript and clarifying compre-
hension. During a third listening, the learners marked significant
hesitations on the transcript. On an overhead screen, the instructor
then drew the attention of the group to formulaic sequences which
occurred between the marked hesitations and commented on the
linguistic and discourse functions of the sequences.

Automatization stage
After the input and analysis were finished, the learners spent time
shadowing the recording along with the transcript in the language
laboratory. First, the entire group practiced reading aloud with the
recording, then they went to the laboratory and shadowed at least eight
times. They were encouraged to pay close attention to the formulaic
sequences and hesitation patterns, with instructions to repeat the more
challenging stretches of discourse as many times as they felt necessary.
After the laboratory practice, the learners participated in two classroom
activities designed to further automatization of the formulaic sequences.
First, students listened to a dictogloss of sentences containing key formu-
laic sequences, taken from the input text. Dictogloss (Wajnryb, 1990) is
 Fluency and Formulaic Language in the Classroom 207

a procedure which has rich potential for fluency teaching. Originally

developed as a grammar awareness activity, it requires that a brief text be
read aloud at normal speed to the class, with students taking notes and
jotting down key words. They work together in teams to reconstruct the
entire text as heard, compare their reconstruction with the original text,
and note differences in structure and phrasing. Dictogloss texts rich in
formulaic sequences could be used to raise awareness of the phrases and
their functions in speech, a step in the direction of mastery or automatiza-
tion of them.
A mingle jigsaw (Wood, 1998) was the next technique used to further
fluency through automatization, followed by a chat circle to consolidate
the experience gained. In the chat circle, a topic from a brainstormed
list related to the taped model was used for each spontaneous talk
period in the circle. The partners took time to comment on their
production and reflected on the speed, hesitations, and “rough spots.”

Practice and production stage

After exposure to the phrase patterns and formulaic sequences of a
native speaker model, the learners were given a chance to prepare a
brief talk of their own, based loosely on the topic area of the model.
In preparation for this talk, they were guided through Nation’s (1989)
4/3/2 procedure, after which they went to the language laboratory to
record their talk without using notes or other support. The recordings
were collected, and, at the end of the program, the learners reviewed
their performances and commented on aspects which they felt had
shown development from the first to the third production.

Free talk stage

At the end of the cycle outlined above, learners formed groups and
generated topics related to the theme and genre of the original native
speaker model. The topics were then distributed randomly to individual
class members, and small groups took turns listening to individuals
speaking spontaneously about the topics they had drawn. They com-
mented on the productions and reflected on the speed, hesitations, and
“rough spots” in their own productions.
208 Formulaic Language and L2 Speech Fluency

Evidence

Wood (2009a) presents some case study evidence of positive results of

such an approach to focused instruction of learners to use formulaic
sequences to advance their fluency. The participant produced narratives
before and after a 6-week set of fluency workshops following the input—
automatization—practice and production—free talk sequence outlined
above. Her monologic speech was analyzed with respect to the length of
runs between pauses and the speech rate, as well as the use of formulaic
sequences. A discourse analysis of the productions focused on the
possible effect of the pedagogical intervention of the fluency workshop
on developments in her performance, specifically, whether the work-
shop provided her with a repertoire of formulaic sequences which
could have contributed to improved fluency.
Sachie (pseudonym) was the participant, a female Japanese learner of
English in a university intensive study abroad context. She was enrolled
in intermediate-level classes in the program, had been in the program
already for a previous 12-week period, and lived in a homestay situation
with native speakers or native-like speakers of English.
As part of the study and the workshops, she was required to produce
spontaneous narratives on topics of personal relevance, with no
preparation time and no use of notes to prepare. She produced her first
speech sample on the first day of the fluency workshops, before the
start of the cycles of activities, and she produced her second sample a bit
longer than 6 weeks later—a week following the end of the workshops.
Sachie’s recordings were transcribed and the hesitations were
marked in the transcripts. Her Speech Rate (SR) was calculated as the
number of syllables uttered per minute, and her Mean Length of
Runs (MLR) was calculated as total number of syllables uttered divided
by the number of runs in a sample. Single word runs were counted
as runs, and nonlexical utterances or filled pauses such as mm and ah
were all counted as runs too.

Identifying Formulaic Sequences in Native Speaker and

L2 Learner Speech
Any multiword phenomena which appeared to serve fluency or dis-
course functions in Sachie’s speech samples were marked as formulaic
 Fluency and Formulaic Language in the Classroom 209

sequences. In addition to this initial intuition-based coding, a set of five

general criteria were used as judgment criteria for locating formulaic
sequences in the transcripts. These are the criteria described in detail in
the previous “Evidence” section of this volume. No particular criterion
or combination of criteria were essential for a word combination to be
marked as a formulaic sequence, and judgments were made based on
one, several, or all of these.

1. Phonological coherence and reduction.

2. The taxonomy used by Nattinger and DeCarrico (1992).
3. Greater length/complexity than other output.
4. Semantic irregularity, as in idioms and metaphors.
5. Syntactic irregularity.

The criteria were used holistically and no one criterion was crucial for
the judgment of formulaicity, nor were all criteria necessarily applicable
to all cases of multiword sequences.
Sequences which were idiosyncratic and noncanonical were accepted,
as it was decided that it was probable that the cognitive stresses of the
speech sample production, including recalling events to be retold, could
lead to articulation slips or glitches in expression. Prodromou (2007),
and Wray and Fitzpatrick (2008) have examined how L2 learners may
deviate from native speaker norms in expressing formulaic sequences.
They found that the deviations tend to link very weakly to fluency mea-
sures, but are instead linked to pragmatic intentions (Prodromou,
2007), or gaps in lexical, morphological, or phrasal competence (Wray
& Fitzpatrick, 2008).

Results
The recorded narrative monologs were analyzed for temporal measures
of fluency: SR measured as syllables uttered per minute; MLR measured
as mean number of syllables uttered between hesitations. Analysis of the
data reveals some trends in the development of SR and MLR, the nature
of learner use of formulaic sequences, and the efficacy of focused
instruction in formulaic sequences.
Sachie’s first narrative was about her experience attending a concert in
Osaka by Celine Dion, and her second narrative was a reminiscence on
her summer vacations as a child in Kumamoto.
210 Formulaic Language and L2 Speech Fluency

Table 9.1 displays the strong gains in fluency measures from the first
sample to the second, after the 6-week fluency workshop. MLR showed
a 26.3% increase in the second speech sample, and SR an improvement
of 13.8%.

Table 9.1 Speech Rate and Mean Length of Runs

Sample # 1 2 % change

SR 123.2 140.2 13.8

MLR 5.10 6.44 26.3

The strong increase of MLR and SR clearly indicates that Sachie was
more readily able to speak faster and produce longer runs between hesi-
tations after 6 weeks.

Table 9.2 Use of Formulaic Sequences

Sample # 1 2

Formulaic Sequences 18 50
Formulaic Sequences from NS Models 2 18
Syllables 530 760
% Syllables from Formulaic Sequences 11.3 12.5

As shown in Table 9.2, in Sachie’s first narrative retell, before the fluency
workshop, she produced 18 formulaic sequences, 2 of which were also in
the native speaker models of the workshop. Her second narrative, 6 weeks
later, contained 50 formulaic sequences, 18 of which were present in the
native speaker models. In other words, 11.8% of the formulaic sequences
in the first speech sample were from the workshop, and 36% of those in
the second speech sample were from the workshop activities. The first
sample consisted of 530 syllables overall, 60 (11.3%) of which were in for-
mulaic sequences. The second sample consisted of 760 syllables overall, 95
(12.5%) of which were in formulaic sequences.
Judging from the numerical data, Sachie was able to speak with
increased fluency after the workshop, and she was able to produce a
greater quantity of speech. There were more formulaic sequences in the
second sample, a number of which came from the native speaker models
in the fluency workshop. An examination of the sequences taken from
the models can illuminate how they may have facilitated her improved
performance in sample 2.
 Fluency and Formulaic Language in the Classroom 211

Formulaic sequences taken from the models

All the formulaic sequences in Sachie’s speech samples are listed below.
Those which are also in the native speaker models are bolded and
highlighted.

Formulaic sequences used in first speech sample

My name is
I want to
Talk about
I think
Her name is
And then
And then
I am wondering
Almost the all
Almost all
You know
Most of them
Give up
For more than 2 hours
Came back to
The end of this year
I don’t know
I will give up

Formulaic sequences used in second speech sample

When I was a little girl
Until I graduate
Every summer
South part of Japan
A lot of
Almost all
Of course
Around there
Instead of
Lots of
The interesting thing is that
212 Formulaic Language and L2 Speech Fluency

It took about 10 minutes

Almost every day
Every summer
Almost every day
Would go swimming
In the daytime
In the nighttime
Stayed up late
Kept talking
Had to eat
Tons of
Every day
Tons of
One of my cousins
A lot
Broken down
In summer
Lots of
By the way
Until very late
That’s why
Complained about
After my grandma died
Most of us
Things like that
Two years ago
In some ways very much the same
Try to get out
You know
That’s why
Some of them
Very sad story
That’s why
In that place
Still now
One of my most vivid memories
Little by little
The same things
Kind of
 Fluency and Formulaic Language in the Classroom 213

I want to go
I want to go
One of my most important memories

A striking characteristic of the formulaic sequences which are also in

the native speaker models is their relative length and, in many cases,
their complexity and native-like semantic aspects. The mean length of
the formulaic sequences in sample 2 is 4.46, while that of the first sam-
ple is 3.17—an increase of 40.7%. This exceeds the overall increase in
length of runs from sample 1 to sample 2, and very likely plays a role in
that overall increase. It is noteworthy that the longest formulaic
sequences in the second sample come from the fluency workshop, since
this may be a sign that Sachie was able to recall and use these longer
sequences to facilitate more fluent expression.
The formulaic sequences in the first sample are mostly two or three
word collocations with straightforward functions:

Temporal marking: and then, for more than 2 hours, the end of this year
Quantity marking: most of them, almost all, almost the all
Spatial marking: came back to
Fluency devices: I think, you know
Personal stance markers: I don’t know, I am wondering, give up
Textual functions in the narrative: my name is. I want to, talk about
Phrasal verb/verb+preposition: give up (2)
Cause and effect: none
Comparison and contrast: none
Sentence builders: none
Other: her name is

Only two sequences match those in the fluency workshop models,

I think and you know, both of which are fluency devices which are
neither means novel nor specific to this type of discourse.
In Sachie’s second speech sample, there is a broader range of
functions and types of formulaic sequences used, particularly those
which are also in the workshop models:

Temporal marking: 16 in total, with 8 taken from the models. These

show more length, detail, and complexity than in the first sample,
214 Formulaic Language and L2 Speech Fluency

particularly those also in the models – when I was a little girl, it took
about ten minutes, in the daytime/nighttime, still now. These are more
native-like, add textual bulk to promote longer runs and fluency, and
show greater range than those in the first sample.
Quantity marking: 9 in total, with 3 taken from the models.
Spatial marking: 2 in total.
Fluency devices: 1 in total.
Personal stance markers: 3 in total.
Textual functions: 2 in total, 1 taken from the models: the interesting
thing is that 8 syllables and native-like.
Phrasal verbs/verb+preposition: 4 in total, none from the models.
Cause and effect functions: 3 in total, all the same, not taken from
models.
Comparison and contrast functions: 2, one lengthy one taken from
models: in some ways very much the same 8 syllables and native-like.
Sentence builders: 2, both taken from the models: one of my most vivid
memories, one of my most important memories.
Other: 4, none taken from models.

Sachie’s second sample did not have proportionally any more formulaic
sequences than her first, but it did have longer sequences on average,
by 16.2%.
The majority of the formulaic sequences which Sachie appears to have
borrowed from the native speaker models were taken from the second
model in the workshop. Of 18 formulaic sequences taken from work-
shop models used in sample 2, 2 were from model #1, 15 from model
#2, which was in a similar genre. The second native speaker model was
a reminiscence on childhood memories, which is the narrative genre
Sachie used in her second sample. It may be that the topic Sachie chose
for the second sample was more immediately relevant to her life and
interests, and she may have engaged more with the ideas and observa-
tions she conceptualized and formulated. She appears to have been a
much more comfortable communicator in the second sample, produc-
ing speech at a faster rate, with more complex use of formulaic sequences,
and a greater range of emotion than her first sample.
It seems that Sachie was able to recall the formulaic sequences from
the workshop models, integrate them into her repertoire, and use them
in her own narrative quite effectively. She did not reproduce large
 Fluency and Formulaic Language in the Classroom 215

chunks of narrative from the workshop models, but she does appear to
have worked some useful sequences from the workshops into her own
speech. This resulted in increased fluency, particularly in longer length
of runs between pauses.
A substantial fluency gain is present from the first to the second sam-
ple, and Sachie used a greater quantity of and more complex formulaic
sequences in the second sample. This is evidence that the fluency work-
shops provided her with formulaic language which she added to her
repertoire and used to boost her fluency. Future research in this area
would certainly need a large cohort of participants and a longer time-
line between focused instruction and production of speech samples. As
well, using dialogic tasks rather than narrative monologs, together with
some ethnographic investigation of real-life language use, would help
clarify whether this type of fluency training is generally transferable to
L2 speech performance outside of a controlled classroom or research
environment.

Discourse Analysis Activity

Discourse analysis using an ethnographic approach is a research tradi-

tion with rich potential for applications to language teaching, particu-
larly as regards formulaic language and speech. Riggenbach (1999)
outlines a comprehensive approach to teaching spoken language which
involves having learners conduct focused discourse analysis of native
speaker speech. She presents a paradigm for designing discourse analy-
sis activities in the classroom:

Step 1: Predict Learners make predictions about the target structure

Step 2: Plan Learners set up a research plan that will produce samples
of the target structure
Step 3: Collect data Learners observe and/or record the target struc-
ture in its discourse environment
Step 4: Analyze Learners analyze the data and explain results/make
conclusions
Step 5: Generate Learners discuss the target structure or produce the
target structure in its appropriate context
216 Formulaic Language and L2 Speech Fluency

Step 6: Review Learners summarize their findings or reanalyze the

data that they produced, asking whether the data conform to their
conclusions in Step 4. (Riggenbach, 1999: 45, 46)

A series of activity cycles are outlined by Riggenbach (1999), each focus-

ing on a particular speech phenomenon. For example, in a focus on
fillers and repairs, learners would first discuss how this works in their
L1, and what they may have observed in English speakers. Then they
record native speakers speaking in spontaneous unplanned situations,
and/or interview native speakers about how they use fillers and repairs.
Groups of students then pool their data, and analyze it for particular
phenomena or trends. They then analyze their own recorded speech in
a similar way, comparing the fillers and repairs to the native speaker
“norms” which they have uncovered (Riggenbach, 1999: 73–77). A simi-
lar type of activity sequence would work well for determining the func-
tions and meanings of formulaic sequences in spontaneous speech, and
the fluency-enhancing role which they play.
Wood (2009b) describes a course designed for engineering coopera-
tive education students in which training in formulaic sequences
was undertaken using an ethnographic and discourse analysis approach
largely based on that of Riggenbach (1999). Learners collected
speech samples within a particular register by interviewing native
speakers and conducting analysis of the pooled transcripts of the
encounters. The course consisted of 24 hours of instruction, 6 hours a
week for 4 weeks.
The learners were placed into the spoken English course using an
assessment instrument comprised of a set of three simulation tasks
related to the types of speech situations engineering cooperative
education students might typically meet in their job interviews and
placements. To complete the assessment, the candidates worked in a
language laboratory, listening to the instructions and input through a
headset, and recording their responses and output through a micro-
phone. The first task required candidates to describe their interests in
their field of study, within a 2-minute time limit. They were instructed
to talk about their motivation to enter the field, what contributions
they could make, and the benefits they expected from their coopera-
tive work placement. The second task was a simulated meeting in
 Fluency and Formulaic Language in the Classroom 217

which the candidate was to listen to a manager and introduce a change

in the work environment, with three main points: what the planned
change was; reasons for the change; contributions expected from
employees. The candidate was to relay the information to a colleague.
The third task required candidates to read a memo and answer oral
questions from a customer.
The simulation was scored using the rubric below, which covered
content, organized presentation of material, overall comprehensibility,
grammatical and phonetic accuracy, and fluency. In the rubric, 1 repre-
sents the lowest score and 5 the highest. The learners who scored below
4 overall were required to enroll in the spoken English course.

Rubric
Low High
Clarity of Expression
Pronunciation 1 2 3 4 5
Comprehensibility 1 2 3 4 5
Speed 1 2 3 4 5
Hesitations 1 2 3 4 5
Intonation, rhythm 1 2 3 4 5
Facial expression 1 2 3 4 5
Eye contact 1 2 3 4 5
Clarity of Meaning
Vocabulary 1 2 3 4 5
Phrases 1 2 3 4 5
Grammar 1 2 3 4 5
Overall Evaluation 1 2 3 4 5

In the first section of the course, learners followed a sequence of activity

similar to that of the fluency course described earlier in this section.
They listened to and analyzed native speaker models talking about work,
career choice, and making changes to work plans. The candidates lis-
tened to and analyzed three samples of Native Speaker (NS) speech
related to the themes of the placement assessment. The first model was
a short monolog from a language teacher, in which he explains why he
chose this field of work, what benefits he gets from it, what he expects
to contribute to it. The second was from a manager of language teach-
ing programs, in which she explains upcoming changes to the programs
and what the implications are for the work of the teachers. The third NS
218 Formulaic Language and L2 Speech Fluency

model was from a manager of software design and programming in a

government department, in which he discusses how lack of funding has
had an impact on particular services and how this can be addressed to
meet a particular deadline. With each model, the candidates went
through the automatization and free talk stages as explained above.
In the ethnographic section of the course, learners were paired with
teacher education students from a university education faculty. The
main purpose of the contact activity in this situation was to expose the
participants to native or native-like spontaneous speech in English in an
interview context, independent of discipline-specific language. They
were required to conduct an interview with the education students
on topics related to those of the simulations in the initial cooperative
education placement. They communicated with the education students
using the following template:

You will meet with several native speaker visitors from another
university. They will have some research questions to ask you.
You should prepare to discuss with them the following points:

1. Why they chose teaching as a career

2. What contributions they feel they can make to the teaching
profession
3. What benefits they expect to gain from a teaching career
4. When to schedule another appointment to meet
5. The history of their studies until now
6. An example of changes in their studies or work and what they
had to do to cope
7. An example of a problem with meeting a deadline and how they
dealt with it.

You will need to schedule another meeting to accomplish

everything.
You must record the responses of the participants to each of the
above points.
Before next class make a transcript of their answers. Bring the
transcript and recording to class.
 Fluency and Formulaic Language in the Classroom 219

At the next class meeting, learners worked to pool their transcripts,

recordings, and findings, using the following instructions:

Using your transcripts and referring to the tapes you made from
discussion with the university teacher trainees, join with a partner
and find some sample phrases, words, and expressions used for
these purposes:

1. Explaining a personal choice

2. Listing personal characteristics and/or qualifications
3. Telling about life experiences
4. Introducing the idea of contributions
5. Discussing benefits

Share your phrases, words and expressions with the rest of the class.
Use the recording to illustrate your findings.
Go back to your taped sample of yourself describing your interest,
contributions and benefits. Use the following three sources to help
you to modify and improve it:

z The instructor’s feedback

z The experience with the teacher trainees, including the recorded
samples and transcripts and our in-class discussions about it.
z Our sample recording with transcript.
z Tape yourself again.

Record yourself talking about a change in a workplace. Explain what

the change is, why it is happening, and what the contributions are
expected from employees.

Learners used all of the resources, including instructor feedback, the

ethnographic data they collected from the university teacher trainees,
and in-class experiences and peer feedback to refine their own speech
on the three themes related to the test of speaking: discussing why they
chose a particular career and what they expected to contribute to it
and benefits they expected from it; workplace changes and their causes
and the expected contributions of employees; challenges in meeting
deadlines, reasons for that, and plans to rectify the situation.
220 Formulaic Language and L2 Speech Fluency

Throughout, learners were required to pay careful attention to

fluency factors and formulaic sequences which were relevant to the
speaking tasks.

Final Presentation
The entire sequence of activity ended with an in-class performance
which was a repeat of simulation tasks used initially to place the learners
in the course. Student peers and three English teachers all completed
the evaluation template with Likert scales of 1 (poor) to 5 (excellent) for
various components of speech for each learner.
The learners generally scored higher in criteria related to speed, hesi-
tation, intonation and rhythm, and vocabulary, phrases and grammar in
the final presentation. At the beginning, overall scores for six of the
learners were 3, with the other ten scoring 4. In the final presentation,
all scores were in the 4 range, with six of them advancing from 4 to 4.5.
Key formulaic sequences used in the NS models in class appeared in the
final presentation performances, such as:

We would have to make changes

We’ve noticed that
That’s sort of my main reason for
Have left some very deep impressions
Trying to get through different hoops
To figure out how to use
There are a number of contributions I could make
To weave together my interests and skills.

This shows some improvements in expression after only 3 weeks of

biweekly class meetings, along with transcribing and analysis outside of
class. The oral proficiency of these learners improved, and they showed
evidence of having used a larger repertoire of formulaic language and
chunks taken from the speech models of native speakers.

Summary

The course designs, principles, and activities presented here represent

a beginning step in the direction of a methodology of fluency. From the
 Fluency and Formulaic Language in the Classroom 221

current state of knowledge about fluency, it is apparent that formulaic

sequences are key elements in fluency development, along with automa-
tization of processing. Classroom practice and course design need to be
approached with these two elements in mind. Activities like 4/3/2,
shadowing and tracking can help to further automatization by imposing
speed constraints on performance, and requiring repetition, as well
as forcing the production of speech in chunks, or formulaically. “Market-
place”, “Messengers”, and the mingle jigsaw impose more context-
sensitive, purposeful speed constraints and repetition requirements.
“Class Photo” and “Family Tree” take this realism of constraints a step
further by allowing still more creativity of message and form.
The oral process approach to oral skills development encourages a
broad range of activity to further spoken fluency, including pre-task and
within-task planning, preparation of performances, ongoing student
and instructor monitoring of production, and a blend of spontaneous
and planned speech. In all cases, attention to formulaic sequences is
combined with automatization activity. 4/3/2, shadowing, and tracking
all provide ample opportunity to focus on formulaic sequences, by
requiring their use in the performance and providing naturalistic input
to serve as models of the phrases and other procedural aspects of
fluency such as speed, pause profiles, and clause chaining. Similarly,
“Marketplace”, “Messengers”, and the mingle jigsaw permit instructors
to identify appropriate formulaic sequences for completion of the
activities, and to have repetition facilitate their automatization. Oral
process allows a focus on formulaic sequences in all components; as
instructors assist learners to prepare and polish speech and performances,
appropriate formulaic sequences and the other, temporal elements of
fluency can be integrated.
The fluency workshop sequence of activities is an adaptable and
flexible tool for improving speech fluency with attention to formulaic
language. It can be applied to any set of genres, topics, or speech situa-
tions which might be relevant to the needs of any class or program.
It has been shown to be effective in the two learning contexts described
above—the case study involving Sachie, and the cooperative engineer-
ing student program.
A great deal of further research on these and other types of fluency-
focused activities is needed. One large gap in the fluency-focused
material relates to ways to facilitate fluency in interactive language use
222 Formulaic Language and L2 Speech Fluency

such as conversation and discussion. The research study presented in

this volume is an analysis of learner monologs, and the fluency-focused
course plans and activities presented here are also largely monologic.
Even the engineering cooperative student workshop is geared toward
preparing learners for monologic speech, even though it is essentially
preparation for a dialogic situation, an employment interview. It is nec-
essary for researchers and materials developers to seek the linkages
between formulaic sequences and fluency and conversation analysis
and discourse analysis, and elaborate ways to facilitate and teach
L2 learners to communicate fluently in a range of conversation and
discussion contexts. A starting point might be to take Riggenbach’s
(1999) ethnographic approach and have learners record and analyze
samples of conversation and discussion relevant to their needs, or to
take a cue from Wray (2004, 2008) and provide learners with sequences
relevant to particular contexts and have them try them out and record
their efforts.
Fluency can be facilitated by attention to formulaic language, and an
assortment of activities which more or less incorporate this exists already
in materials and course designs, but much more work needs to be done
on building fluency-based courses and integrating existing knowledge
about fluency into all elements of language teaching programs. As well,
there is a need for materials and program development which can foster
automatization and formulaic language competence with the current
state-of-the art task, process, and focus-on-form teaching methods. As
for the input used to help learners improve spoken language fluency,
the discussion of authenticity of text is important. If we are to provide
learners with authentic input, we must attend to the formulaic sequences
relevant to the target genres, and use these as a syllabus for teaching flu-
ency. Research is needed into the efficacy of some of the pedagogical
techniques used to promote fluency; it is important to ask questions about
whether or not the elements of repetition, attention to input, practice,
tuned output, interaction, and attention to formulaic language can lead
to transfer of abilities and skills outside of the classroom.
One main point remains: it is likely not enough to have students talk to
each other in an unstructured or semistructured way and call it “fluency”
activity. Our knowledge about formulaic language and its automatiza-
tion has taken us at least one step beyond that. Informed teachers and
 Fluency and Formulaic Language in the Classroom 223

course developers can base their fluency-focused efforts on some basic

principles of focused instruction of formulaic sequences:

z Use of spoken rather than written input, and use of models of fluent
speakers
z Ethnographic or discovery tasks which engage learners with spoken
language in use and encourage analysis of the use and functions of
formulaic sequences
z Highlighting and encouragement of noticing of formulaic sequences
in the input and models, and their discourse functions, meanings,
and effects on temporal variables
z Repetition and rehearsal of key formulaic sequences in a range of
practice tasks, often involving time constraints and possibly including
memorization
z Variety in speech tasks, which may include pre-task and within-task
planning time and plenty of interaction
z Peer and teacher feedback with particular attention to formulaic lan-
guage and temporal aspects of fluent speech

Over time, we can expect a growth in the range and depth of investiga-
tion into the link between formulaic language and fluent speech, and a
corresponding development of programs, tasks, and classroom activities
which make use of this knowledge.
 References

Anderson, J. R. (1983). The architecture of cognition. Cambridge, MA: Harvard

University Press.
Anderson, J. R. (1993). Rules of the mind. Hillsdale, NJ: Erlbaum.
Angelil-Carter, S. (1997). Second language acquisition of spoken and written
English: Acquiring the skeptron. TESOL Quarterly, 31 (2), 263–287.
Bachman, L. F. (1990). Fundamental considerations in language testing. Oxford: Oxford
University Press.
Baddeley, A. (1986). Working memory. Oxford: Clarendon Press.
Baddeley, A. D. (1988). Working memory. Oxford: Oxford University Press.
Baddeley, A. D. (2000). The episodic buffer: a new component of working memory?
Trends in Cognitive Science, 4, 417–423.
Baddeley, A., Gathercole, S., & Papagno, C. (1998). The phonological loop as a
language learning device. Psychological Review, 105 (1), 158–173.
Bahns, J., Burmeister, H., & Vogel, T. (1986). The pragmatics of formulas in
L2 learner speech: Use and development. Journal of Pragmatics, 10, 693–723.
Bakhtin, M. (1986). The problem of speech genres (Trans. V. McGee). In
C. Emerson & M. Holquist (Eds.), Speech genres and other late essays (pp. 60–102).
Austin, TX: University of Texas Press.
Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory.
Englewoods Cliffs, NJ: Prentice-Hall.
Barnlund, D. C. (1987). Verbal self-disclosure: Topics, targets, depth. In F. L. Luce
& E. C. Smith (Eds.), Toward internationalism (pp. 147–165). Cambridge, MA:
Newbury House.
Beatty, M. J., Balfantz, G., & Kuwabara, A. Y. (1989). Trait-like qualities of selected
variables assumed to be transient causes of performance state anxiety. Communi-
cation Education, 38 (3), 277–288.
Behnke, R. R. & Sawyer, C. R. (1998). Milestones of anticipatory public speaking
anxiety. Communication Education, 48 (2), 165–171.
Belenky, M. F., Clinchy, B. M., Goldberger, N. R., & Tarule, J. M. (1986). Women’s
ways of knowing: The development of self, voice, and mind. New York: Basic Books.
Benson, M., Benson, E., & Ilson, R. (1997). The BBI dictionary of English word combina-
tions. Amsterdam: John Benjamins.
Biber, D., Johansson, S., Leech, G., Conrad, S., & Finegan, E. (1999). Longman gram-
mar of spoken and written English. Essex: Pearson Education Limited.
Biber, D. & Conrad, S. (1999). Lexical bundles in conversation and academic prose.
In H. Hasselgard & S. Oksefjell (Eds.), Out of corpora: Studies in honour of Stig
Johansson (pp. 181–189). Amsterdam: Rodopi.
 References 225

Biber, D., Conrad, S., & Cortes, V. (2004). If you look at . . . lexical bundles in
academic lectures and textbooks. Applied Linguistics, 25, 371–405.
Bishop, H. (2004). Noticing formulaic sequences – a problem of measuring the
subjective. LSO Working Papers in Linguistics, 4, 15–19.
Boers, F., Eyckmans, J., Kappel, J., Stengers, H., & Demecheleer, M. (2006). Formulaic
sequences and perceived oral proficiency: Putting a Lexical Approach to the test.
Language Teaching Research, 10 (3), 245–261.
Bolander, M. (1989). Prefabs, patterns and rules in interaction? Formulaic speech
in adult learners’ L2 Swedish. In K. Hyltenstam & L. K. Obler (Eds.), Bilingual-
ism across the lifespan: Aspects of acquisition, maturity, and loss (pp. 73–86).
Cambridge: Cambridge University Press.
Butler, C. S. (2003). Multi-word sequences and their relevance for recent models of
functional grammar. Functions of Language, 10 (2), 179–208.
Bybee, J. (2002). Phonological evidence for exemplar storage of multiword units.
Studeis in Second Language Acquisition, 24, 215–221.
Bygate, M. (1988). Units of oral expression and language learning in small group
interaction. Applied Linguistics, 9 (1), 59–82.
Canale, M. & Swain, M. (1980). Theoretical bases of communicative approaches to
second language teaching and testing. Applied Linguistics, 1, 1–47.
Carter, R. & McCarthy, M. (1995). Spoken grammar: What is it and how can we
teach it? ELT Journal, 49 (3), 207–218.
Chafe, W. L. (1980) Some reasons for hesitating. In H. W. Dechert & M. Raupach
(Eds.), Temporal variables in speech (pp. 169–180). The Hague: Mouton.
Chambers, F. (1998). What do we mean by fluency? System, 25 (4), 535–544.
Chaudron, C. (1988). Second language classrooms: Research on teaching and learning.
Cambridge: Cambridge University Press.
Chen, L. (2010). An investigation of lexical bundles in ESP textbooks and electrical
engineering introductory textbooks. In D. Wood (Ed.), Perspectives on formulaic
language: Acquisition and communication (pp. 107–125). London/New York:
Continuum.
Cheng, P. W. (1985). Restructuring versus automaticity: Alternative accounts of skill
acquisition. Psychological Review, 92, 414–423.
Clahsen, H. (1987). Natural language development: Acquisitional processes
leading to fluency in speech production. In H. W. Dechert & M. Raupach (Eds.),
Psycholinguistic models of production (pp. 67–75). Norwood, NJ: Ablex.
Collentine, J. (2004). The effects of learning contexts on morphosyntactic and
lexical development. Studies in Second Language Acquisition, 26 (2), 227–248.
Conklin, K. & Schmitt, N. (2008). Formulaic sequences: Are they processed more
quickly than nonformulaic language by native and nonnative speakers? Applied
Linguistics, 29 (1), 72–89.
Cook, G. (1989). Discourse. Oxford: Oxford university Press.
Cortes, V. (2004). Lexical bundles in published and student disciplinary
writing: Examples from history and biology. English for Specific Purposes, 23,
397–423.
Coulmas, F. (1979). On the sociolinguistic relevance of routine formulae. Journal of
Pragmatics, 3, 239–266.
226 References

Cowie, A. P. (1988). Stable and creative aspects of vocabulary use. In R. Carter & M.
McCarthy (Eds.), Vocabulary and language teaching (pp. 126–139). New York:
Longman.
Dai, Z. & Ding, Y. (2010). Effectiveness of text memorization in EFL learning of
Chinese students. In D. Wood (Ed.), Perspectives on formulaic language: Acquisition
and communication (pp. 71–87). London/New York: Continuum.
Davis, P. & Rinvolucri, M. (1989). Dictation: New methods, new possibilities. Cambridge:
Cambridge University Press.
Dechert, H. W. (1980). Pauses and intonation as indicators of verbal planning
in second-language speech productions: Two examples from a case study. In
H. W. Dechert & M. Raupach (Eds.), Temporal variables in speech (pp. 271–285).
The Hague: Mouton.
Dechert, H. W. (1984a). Individual variation in speech. In H. W. Dechert, D. Mohle,
& M. Raupach (Eds.), Second language productions (pp. 156–185). Tubingen:
Gunter Narr Verlag.
Dechert, H. W. (1984b). Second language production: Six hypotheses. In
H. W. Dechert, D. Mohle, & M. Raupach (Eds.), Second language productions
(pp. 211–230). Tubingen: Gunter Narr Verlag.
Dechert, H. W. (1987). Analysing language processing through verbal protocols. In
C. Faerch & G. Kasper (Eds.), Introspection in second language research (pp. 96–112).
Clevedon, Avon: Multilingual Matters.
Deci, E. L. & Ryan, R. M. (1985). Intrinsic motivation and self-determination in human
behavior. New York: Plenum.
DeKeyser, R. M. (2001). Automaticity and automatization. In P. Robinson (Ed.),
Cognition and second language instruction (pp. 125–151). Cambridge: Cambridge
University Press.
Deschamps, A. (1980). The syntactical distribution of pauses in English spoken as a
second language by French students. In H. W. Dechert & M. Raupach (Eds.),
Temporal variables in speech (pp. 255–262). The Hague: Mouton.
Ding, Y. (2007). Text memorization and imitation: The practices of successful
Chinese learners of English. System, 35, 271–280.
Du Bois, J. W. (2006, January). Transcription and the delicacy hierarchy: What is to
be represented? Paper presented at the Linguistic Society of America 80th Annual
Meeting. Albuquerque, New Mexico.
Elbow, P. (1994). Introduction. In P. Elbow (Ed.), Landmark essays on voice and writing
(pp. xi–xlvii). Davis, CA: Hermagoras Press.
Ellis, N. C. (1996). Sequencing in SLA: Phonological memory, chunking, and points
of order. Studies in Second Language Acquisition, 18, 91–126.
Ellis, N. C. (2002). Frequency effects in language processing. Studies in Second
Language Acquisition, 24, 143–188.
Ellis, R. (2005). Planning and task-based performance: Theory and research. In
R. Ellis (Ed.), Planning and task performance in a second language (pp. 3–36).
Amsterdam: John Benjamins.
Ellis, R., Basturkmen, H., & Loewen, S. (2001).Learner uptake in communicative
ESL lessons. Language Learning, 51, 281–318.
Ellis, R., Basturkmen, H., & Loewen, S. (2002). Doing focus-on-form. System, 30,
419–432.
 References 227

Ellis, R. & Yuan, F. (2005). The effects of careful within-task planning on oral and
written task performance. In R. Ellis (Ed.), Planning and task performance in a sec-
ond language (pp. 167–192). Amsterdam: John Benjamins.
Faerch, K. & Kasper, G. (1989). Transfer in production: Some implications for the
interlanguage hypothesis. In H. W. Dechert & M. Raupach (Eds.), Transfer in
language production (pp. 173–193). Norwood, NJ: Ablex.
Fillmore, C. J., Kay, P., & O’Connor, C. (1988). Regularity and idiomaticity in
grammatical constructions. Language, 64, 501–538.
Fillmore, L. (1976), The Second Time Around: Cognitive and Social Strategies in Second
Language Acquisition, Ph.D. dissertation, Stanford University.
Freed, B. F. (1995). What makes us think that students who study abroad become
fluent? In B. F. Freed (Ed.), Second language acquisition in a study abroad context
(pp. 123–148). Philadelphia, PA: John Benjamins.
Fulcher, G. (1996). Testing tasks: Issues in task design and the group oral. Language
Testing, 13 (1), 23–51.
Gatbonton, E. & Segalowitz, N. (1988). Creative automatization: Principles for
promoting fluency within a communicative framework. TESOL Quarterly,
22 (3), 473–492.
Gatbonton, E. & Segalowitz, N. (2005). Rethinking communicative language teach-
ing: A focus on access to fluency. The Canadian Modern Language Review, 61 (3),
325–353.
Gathercole, S. & Baddeley, A. (1993). Working memory and language. Hove: LEA.
Hakuta, K. (1974). Prefabricated patterns and the emergence of structure in second
language acquisition. Language Learning, 24 (2), 287–297.
Hansen, L., Gardner, J., & Pollard, J. (1998). The measurement of fluency in a second
language: Evidence from the acquisition and attrition of Japanese. In B. Visgatis
(Ed.), On JALT ‘97: Trends and transitions – Proceedings of the JALT 1997 Conference on
Language Teaching and Learning (pp. 37–45). Tokyo: Japan Association of Language
Teachers.
Have, P. ten (1999). Doing conversation analysis: A practical guide. London: Sage.
Hickey, T. (1993). Identifying formulas in first language acquisition. Journal of Child
Language, 20, 27–41.
Hieke, A. (1981). A content-processing view of hesitation phenomena. Language
and Speech, 24 (2), 147–160.
Hill, J. & Lewis, M. (1997). LTP dictionary of selected collocations. EMEA British English.
Ho, Y. (1993). Aspects of discourse structure in Mandarin Chinese. Liangton, NY: Edwin
Mellen.
Horwitz, E. K., Horwitz, M. B., & Cope, J. A. (1991). Foreign language classroom
anxiety. In E. K. Horwitz & D. J. Young (Eds.), Language anxiety: From theory
and research to classroom implications (pp. 27–36). Englewood Cliffs, NJ:
Prentice-Hall.
Hulstijn, J. H. (2001). Intentional and incidental second language vocabulary
learning: A reappraisal of elaboration, rehearsal and automaticity. In P. Robinson
(Ed.), Cognition and second language instruction (pp. 258–286). Cambridge:
Cambridge University Press.
Jiang, N. & Nekrasova, T. M. (2007). The processing of formulaic sequences by
second language speakers. Modern language Journal, 91 (3), 433–445.
228 References

Kahnemann, D. & Treisman, A. (1984). Changing views of attention and automatic-

ity. In R. Parasuraman & D. R. Davies (Eds.), Varieties of attention (pp. 29–61). New
York: Academic Press.
Koprowski, M. (2005). Investigating the usefulness of lexical phrases in contemporary
coursebooks. ELT Journal, 4, 322–332.
Krashen, S. D. & Terrell, T. (1983). The natural approach: Language acquisition in the
classroom. New York: Pergamon.
Ladefoged, P. (1972). Preliminaries to linguistic phonetics. Chicago, IL: University of
Chicago Press.
Lakoff, G. & Johnson, M. (1980). Metaphors we live by. Chicago, IL: University of
Chicago Press.
Laver, J. (1994). Principles of phonetics. Cambridge: Cambridge University Press.
Lee, P. (1997). Language in thinking and learning: Pedagogy and the new Whorfian
framework. Harvard Educational Review, 67 (3), 430–471.
Lennon, P. (1984). Retelling a story in English. In H. W. Dechert, D. Möhle, &
M. Raupach (Eds.), Second language productions (pp. 50–68). Tubingen: Gunter
Narr Verlag.
Lennon, P. (1989). Introspection and intentionality in advanced second-language
acquisition. Language Learning 39 (3), 376–396.
Lennon, P. (1990a). The advanced learner at large in the L2 community: Develop-
ments in spoken performance. International Review of Applied Linguistics in
Language Teaching, 28, 309–321.
Lennon, P. (1990b). Investigating fluency in EFL: A quantitative approach.
Language Learning, 40 (3), 387–417.
Levelt, W. J. M. (1989). Speaking: From intention to articulation. Cambridge, MA: MIT
Press.
Lewis, M. (1997). Pedagogical implications of the lexical approach. In J. Coady &
T. Huckin (Eds.), Second language vocabulary acquisition (pp. 255–270).
Cambridge: Cambridge University Press.
Lewis, M. (2008). Implementing the lexical approach. London: Heinle.
Lewis, M. (1997). Pedagogical implications of the lexical approach. In J. Coady &
T. Huckin (Eds.), Second language vocabulary acquisition (pp. 255–270).
Cambridge: Cambridge University Press.
Lin, P. M. S. (2010). The phonology of formulaic sequences: A review. In D. Wood
(Ed.), Perspectives on formulaic language: Acquisition and communication
(pp. 174–193). London/New York: Continuum.
Lindstromberg, S. & Boers, F. (2008a). The mnemonic effect of noticing alliteration
in lexical chunks. Applied Linguistics, 29 (2), 200–222.
Lindstromberg, S. & Boers, F. (2008b). Phonemic repetition and the learning
of lexical chunks: The power of assonance. System, 36, 423–436.
Liu, D. (2007). Idioms: Description, comprehension, acquisition, and pedagogy. New York:
Rouledge.
Logan, G. D. (1988). Towards an instance theory of automatisation. Psychological
Review, 95, 492–527.
Long, M. (1983). Native speaker/nonnative speaker conversation and the
negotiation of comprehensible input. Applied Linguistics, 4 (2), 126–141.
 References 229

Long, M. (1996). The role of the linguistics environment in second language

acquisition. In W. C. Ritchie & T. K. Bhatia (Eds.), Handbook of second language
acquisition (pp. 413–468). New York: Academic press.
Maynard, S. K. (1989). Japanese conversation: Self-contextualization through structure and
interactional management. Norwood, NJ: Ablex.
McCafferty, S. G. (1994). The use of private speech by adult ESL learners
at different levels of proficiency. In J. P. Lantolf & G. Appel (Eds.),
Vygotskian approaches to second language research (pp. 117–134). Norwood,
NJ: Ablex.
McCarthy, M. & O’Dell, F. (2002). English idioms in use. Cambridge: Cambridge
University Press.
McCarthy, M. & O’Dell, F. (2004). English phrasal verbs in use intermediate. Cambridge:
Cambridge University Press.
McCarthy, M. & O’Dell, F. (2006). English collocations in use. Cambridge: Cambridge
University Press.
McLaughlin, B. (1990). Restructuring. Applied Linguistics, 11, 113–128.
McLaughlin, B., Rossman, T., & McLeod, B. (1983). Second language learning: An
information-processing perspective. Language Learning, 33 (2), 135–158.
McMahill, C. (1997). Communities of resistance: A case study of two feminist
English classes in Japan. TESOL Quarterly, 31 (3), 612–621.
McGuire, M. (2009, October). Teaching formulaic sequences in the classroom:
Effects on spoken fluency. Paper presented at the Second Language Research
Forum, Michigan State University, East Lansing, Michigan, U.S.A.
Melĉuk, I. (1998). Collocations and lexical functions. In A. P. Cowie (Ed.) Phraseol-
ogy: Theory, analysis and applications (pp. 23–54). Oxford: Clarendon Press.
Miller, J. & Weinert, R. (1998). Spontaneous spoken language: Syntax and discourse.
Oxford: Clarendon Press.
Möhle, D. (1984). A comparison of the second language speech production of dif-
ferent native speakers. In H. W. Dechert, D. Möhle, & M. Raupach (Eds.), Second
language productions (pp. 26–49). Tubingen: Gunter Narr Verlag.
Möhle, D. & Raupach, M. (1989). Language transfer of procedural knowledge. In
H. W. Dechert & M. Raupach (Eds.), Transfer in language production (pp. 195–216).
Norwood, NJ: Ablex.
Morgan, B. (1997). Identity and intonation; Linking dynamic processes in an ESL
classroom. TESOL Quarterly, 31 (3), 431–450.
Mudraya, O. (2005). Engineering English: A lexical frequency instruction model.
English for Specific Purposes, 25, 235–256.
Myles, F., Hooper, J., & Mitchell, R. (1998). Rote or rule? Exploring the role of
formulaic language in classroom foreign language learning. Language Learning,
48 (3), 323–363.
Myles, F., Mitchell, R., & Hooper, J. (1999). Interrogative chunks in French L2:
A basis for creative construction? Studies in Second Language Acquisition 21,
49–80.
Nassaji, H. (1999). Towards integrating form-focused instruction and communi-
cative interaction in the second language classroom: Some pedagogical
possibilities. Canadian Modern Language Review, 55 (3), 386–404.
230 References

Nation, Paul (1989). Improving speaking fluency. System, 17 (3), 377–384.

Nattinger, J. R. & DeCarrico, J. S. (1992). Formulaic sequences and language teaching.
Oxford: Oxford University Press.
Noels, K. A., Pelletier, L. G., & Vallerand, R. J. (2000). Why are you learning a sec-
ond language? Motivational orientations and self-determination theory. Language
Learning, 50 (1), 57–85.
O’Brien, I., Segalowitz, N., Freed, B., & Collentine, J. (2007). Phonological memory
predicts second language oral fluency gains in adults. Studies in Second Language
Acquisition, 29, 557–582.
Pawley, A. & Syder, F. H. (1983). Two puzzles for linguistic theory: Nativelike selec-
tion and nativelike fluency. In J. C. Richards & R. W. Schmidt (Eds.), Language
and communication (pp. 191–226). New York; Longman.
Peirce, B. N. (1995). Social identity, investment, and language learning. TESOL
Quarterly, 29 (1), 9–31.
Peters, A. M. (1983). Units of language acquisition. Cambridge: Cambridge University
Press.
Pica, T. (1994). Questions from the language classroom: Research perspectives.
TESOL Quarterly, 28 (2), 49–80.
Poyatos, F. (1984). Linguistic fluency and non-verbal cultural fluency. In
A. Wolfgang (Ed.), Nonverbal behavior: Perspectives, applications, intercultural
insights (pp. 431–460). Liangton, NY: C. J. Hogrefe.
Price, M. L. (1991). The subjective experience of foreign language anxiety:
Interviews with highly anxious students. In E. K. Horwitz & D. J. Young (Eds.),
Language anxiety: From theory and research to classroom implications (pp. 101–108).
Englewood Cliffs, NJ: Prentice-Hall.
Prodromou, L. (2007). Bumping into creative idiomaticity. English Today, 23,
14–25.
Psathas, G. (1995). Conversation analysis: The study of talk-in-action. London: Sage.
Raupach, M. (1980). Temporal variables in first and second language speech
production. In H. W. Dechert & M. Raupach (Eds.), Temporal variables in speech
(pp. 263–270). The Hague: Mouton.
Raupach, M. (1984). Formulae in second language speech production. In
H. W. Dechert, D. Möhle, & M. Raupach (Eds.), Second language productions
(pp. 114–137). Tubingen: Gunter Narr Verlag.
Rehbein, J. (1987). On fluency in second language speech production. In
H. W. Dechert & M. Raupach (Eds.), Psycholinguistic models of language production
(pp. 97–105). Norwood, NJ: Ablex.
Ricard, E. (1986). Beyond fossilization: A course on strategies and techniques in
pronunciation for advanced adult learners. TESL Canada Journal Special Edition,
1, 243–253.
Riggenbach, H. (1991). Toward an understanding of fluency: A microanalysis of
nonnative speaker conversations. Discourse Processes, 14, 423–441.
Riggenbach, H. (1999). Discourse analysis in the language classroom. Ann Arbor, MI:
University of Michigan Press.
Robinson, P. (1995). Attention, memory and the “noticing” hypothesis. Language
Learning, 45, 283–331.
 References 231

Rumelhart, D. E., Hinton, G. E., & McClelland, J. L. (1996). A general framework for
parallel distributed processing. In J. L. McClelland & D. E. Rumelhart (Eds.),
Parallel distributed processing: Explorations in the microstructure of cognition (pp. 45–76).
Cambridge, MA: MIT Press.
Sajavaara, K. (1987). Second language speech production: Factors affecting fluency.
In H. W. Dechert & M. Raupach (Eds.), Psycholinguistic models of production
(pp. 45–65). Norwood, NJ: Ablex.
Schloff, L. & Yudkin, M. (1991). Smart speaking: Sixty-second strategies. New York:
Henry Holt and Company.
Schmidt, R. W. (1983). Interaction, acculturation, and the acquisition of communi-
cative competence: A case study of an adult. In N. Wolfson & E. Judd (Eds.),
Sociolinguistics and language acquisition (pp. 137–174). Rowley, MA: Newbury
House.
Schmidt, R. W. (1990). The role of consciousness in second language learning.
Applied Linguistics, 11 (2), 129–158.
Schmidt, R. (1992). Psycholinguistic mechanisms underlying second language
fluency. Studies in Second Language Acquisition 14, 357–385.
Schmidt, R. (1993). Awareness and second language acquisition. Annual Review of
Applied Linguistics 13, 206–226.
Segalowitz, N. & Freed, B. F. (2004). Context, contact, and cognition in oral fluency
acquisition: Learning Spanish in at home and study abroad contexts. Studies in
Second Language Acquisition, 26 (2), 173–199.
Sinclair, J. & Renouf, A. (1988). A lexical syllabus for language learning. In R. Carter
& M. McCarthy (Eds.), Vocabulary and language teaching (pp. 140–160). New York:
Longman.
Siyanova, A. & Schmitt, N. (2008). L2 learner production and processing of
collocation: A multi-study perspective. Canadian Modern language Review, 64 (3),
429–458.
Skehan, P. (1998). A cognitive approach to language learning. Oxford: Oxford Univer-
sity Press.
Spears, R. A., Birner, B., & Kleinelder, S. (1994). NTC’s dictionary of everyday American
English expressions. New York: McGraw-Hill.
Swain, M. (1985). Communicative competence: Some roles of comprehensible
input and comprehensible output in its development. In S. Gass & C. G. Madden
(Eds.), Input in second language acquisition (pp. 235–253). New York: Newbury
House.
Swain, M. (1995). Three functions of output in second language learning. In
G. Cook & B. Seidlhofer (Eds.), Principles and practice in the study of language
(pp. 125–144). Oxford: Oxford University Press.
Tajfel, H. (1981). Human groups and social categories. Cambridge: Cambridge
University Press.
Towell, R. (1987). Variability and progress in the language development of advanced
learners of a foreign language. In R. Ellis (Ed.), Second language acquisition in
context (pp. 113–127). Toronto: Prentice-Hall.
Towell, R., Hawkins, R., & Bazergui, N. (1996). The development of fluency in
advanced learners of French. Applied Linguistics, 17 (1), 84–119.
232 References

Tucker, G. (2005). Extending the lexicogrammar: Towards a more comprehensive

account of extraclausal, partially clausal and non-clausal expressions in spoken
discourse. Language Sciences, 27, 679–709.
VanPatten, B. (1990). Attending to form and content in the input: An experiment
in consciousness. Studies in Second Language Acquisition, 12 (3), 287–301.
VanPatten, B. (2004). Input processing in SLA. In B. VanPatten (Ed.), Processing
instruction: Theory, research, and commentary (pp. 5–32). Mahwah, NJ: Lawrence
Erlbaum.
VanPatten, B. & Williams, J. (Eds.) (2007). Theories in second language acquisition.
Mahwah, NJ: Lawrence Erlbaum.
Vygotsky, L. (1986). Thought and language (Trans. A. Kozulin). Cambridge, MA: The
MIT Press. (Original work published, 1934.)
Wajnryb, R. (1990). Grammar dictation. Oxford: Oxford University Press.
Walker, I. & Utsumi, T. (2006). Memorizing dialogues: The case for “performative
exercises.” In W. M. Can, K. N. Chin, & T. Suthiwan (Eds.), Foreign language
teaching in Asia and beyond: Current perspectives and future directions (pp. 243– 269).
Singapore: Centre for Language Studies.
Weedon, C. (1987). Feminist practice and poststructuralist theory. London: Blackwell.
Weinert, R. (1995). The role of formulaic language in second language acquisition:
A review. Applied Linguistics, 16 (2), 180–205.
Wertsh, J. & Smolka, A. L. (1993). Continuing the dialogue: Vygotsky, Bakhtin, and
Lotman. In H. Miguels (Ed.), Charting the agenda: Educational activity after Vygotsky
(pp. 69–92). London: Routledge.
Willis, D. (1990). The lexical syllabus: A new approach to language teaching. London:
Harper Collins.
Wood, D. (1998). Making the grade: An interactive course in English for academic purposes.
Toronto: Prentice Hall Allyn and Bacon.
Wood, D. (2009a). Effects of focused instruction of formulaic sequences on fluent
expression in second language narratives: A case study. Canadian Journal of Applied
Linguistics, 12 (1), 39–57.
Wood, D. (2009b). Preparing ESP learners for workplace placement. ELT Journal,
63 (4), 323–331.
Wood, D. (2010). Lexical clusters in an EAP textbook corpus. In D. Wood (Ed.),
Perspectives on formulaic language: Acquisition and communication (pp. 88–106).
London/New York: Continuum.
Wray, A. (1998). Protolanguage as a holistic system for social interaction. Language
and Communication, 18, 47–67.
Wray, A. (2002). Formulaic language and the lexicon. Cambridge: Cambridge
University Press.
Wray, A. (2004). “Here’s one I prepared earlier”: Formulaic language learning on
television. In N. Schmitt (Ed.), Formulaic sequences: Acquisition, processing and use
(pp. 249–268). Amsterdam/Philadelphia, PA: John Benjamins.
Wray, A. (2008). Formulaic language: Pushing the boundaries. Oxford: Oxford
University Press.
Wray, A. & Fitzpatrick, T. (2008). Why can’t you just leave it alone? Deviations from
memorized language as a gauge of nativelike competence. In F. Meunier &
 References 233

S. Granger (Eds.), Phraseology in language learning and teaching (pp. 123–148).

Amsterdam: John Benjamins.
Wray, A. & Namba, K. (2003). Formulaic language in a Japanese- English bilingual
child: A practical approach to data analysis. Japan Journal for Multilingualism and
Multiculturalism, 9 (1), 24–51.
Wray, A. & Perkins, M. R. (2000). The functions of formulaic language: An
integrated model. Language and Communication, 20, 1–28.
Yorio, C. (1980). Conventionalized language forms and the development of
communicative competence. TESOL Quarterly, 14, 433–442.
Yorio, C. (1989). Idiomaticity as an indicator of second language proficiency.
In K. Hyltenstam & L. K. Obler (Eds.), Bilingualism across the lifespan: aspects of
acquisition, maturity, and loss (pp. 55–71). Cambridge: Cambridge University Press.
This page intentionally left blank
 Index

Note: Page references in italics refer to tables.

4/3/2 technique 191, 199, 201 Bachman, L. F.

model of communicative
ACT model of cognitive development competence 10
see adaptive control of thought Baddeley, A.
model of cognitive development model of working memory 61, 67
adaptive control of thought (ACT) Bahns, J. 48, 52
model of cognitive Bakhtin, M.
development 62 on voice 76
agglutinative languages 80 Bandura, A.
amount of speech 16 notion of reciprocal determinism 73
research studies 18–20 Barnlund, D. C. 79
analysis of variance (ANOVA) 91, 97, Bazergui, N. 19, 37, 102
116, 117, 120, 122, 125, 129, 133 Behnke, R. R. 73
Anderson, J. R. Boers, F. 58, 88, 183–4, 192
model of language Bolander, M. 50, 70
automatization 3, 62 Burmeister, H. 48, 52
ANOVA see analysis of variance Butler, C. S. 46
AR see articulation rate Bybee, J. 59
articulation rate (AR) 14, 15, 108, 115, Bygate, M. 52
119, 129
by first language 131 calibrated stopwatches 90
Japanese language 82 Canale, M. 10
notion 14, 15 The Cat Came Back (film) 103–4
authenticity of pedagogical narrative complexity 105–6
materials 195 narrative moves analysis 159–61
automatic language processing 62–3 plot 105
notion 4 Chafe, W. L. 29
automatic transfer 35 on focus clusters 27–8
automatization of language 2, 3–4, 5, chain dictations (fluency promoting
60–1, 93, 177–8 activity) 202
Anderson’s model 3, 62, 93 chat circles (fluency promoting
fluency course model 207 activity) 203
L1 transfer and 34–7 Chen, L. 194
theories of 64–9 child language acquisition
value of 56–9 formulaic sequences and 46–9
avoidance 35–6 through multiple cue sources 67
236 Index

Chinese L1 research participants 95, 101 pause marking 107–8

articulation rate 121 transcription 107–8
individual quantitative data collection in fluency
performance 138–41 research 87–9, 97–9
language transfer 118, 123–4, 130–4 procedure 101–6
speech rate 118–9 DeCarrico, J. S. 80, 184
chunking 3–4, 69–70 definition of formulaic sequences 39
Clahsen, H. 35 notion of composites 43
class photo (fluency promoting on pragmatic competence 51–2
activity) 200–1 taxonomy of lexical phrases 44–5,
classroom practices 183–5 96–7, 111–12
discourse analysis activity 215–20 Dechert, H. W. 25, 55
principles 186 declarative knowledge 60–1, 61, 63
sample activities 195–6 definition 2
clause chaining 56–7, 148–9 Deschamps, A. 25
research studies 30–3 dictionaries of formulaic
clustered pauses 25–7 sequences 185
clusters of dysfluencies 151, 152, 153, dictogloss 206
154, 155, 156 dilexicality 43–4
definition 148–9 Ding, Y. 192
cognitive theory 2, 13, 60 directive formulas 49
communication anxiety 72–3 discourse analysis activity 215–20
communicative competence 10
composites 43 EAP/ESP textbooks see English for
computer software analysis 90, 109, academic purposes/special
115, 178 purposes textbooks
Conklin, K. 71 elicited speech 12, 13, 87, 174, 175
connectionism 65–8 sampling procedure 101–3
controlled language processing 61, Ellis, R. 50
62–3 connectionist view of language
notion 4 learning 65–7
conversational adjustments 188 planning types 190
corpus studies 40 on recall 69
Coulmas, F. 51 English for academic purposes/special
formula for identification of purposes (EAP/ESP) textbooks
formulaic sequences 40, 97 treatment of formulaic
Cowie, A. P. sequences 194–5
categorization of formulaic ethnography 196, 215, 218, 221–2
sequences 42–3 ethnomethodology 87–8
cultural fluency 79–80 explicit memory 66
cultural metaphors 197 expressive formulas 48
curriculum design external regulation 74–5
lexical approach 185 extrinsic motivation 74

Dai, Z. 192 Faerch, K. 35

data analysis in fluency research 115 false starts see untraced restarts
calculation of temporal family tree (fluency promoting
variables 108–9, 115–6 activity) 200
 Index 237

feedback 192–3 uses and effects 162, 169–70, 173–4

filled pauses 16, 148, 162–3 value 54–9
notion 17 as vocabulary 196–8
fillers see filled pauses formula/run ratio (FRR) 109, 115,
first language (L1) 127–9, 172, 175
discourse and temporal features by first language 133
of 80–2 Freed, B. F. 13, 14, 18, 22, 24, 26, 37
first language (L1) speech fluency definition of clusters of
clause chaining 30–3 dysfluencies 148
pauses and 23–4, 25–6, 27–8 fluency variables 17
see also fluency frequency-based approach 40
first language (L1) typology frequency-based development of
impact on L2 fluency 79–80, 97–8, knowledge 66
130–4, 145, 172–3, 175–6 FRR see formula/run ratio
Fitzpatrick, T. 209
fluency game formulas 49
notion and definition 9–12 Gardner, J. 20, 24
as performance-based Gatbonton, E. 191, 200
phenomenon 10–11 gender differentials
procedural knowledge and 12, 63–4 L2 fluency 78–9, 145, 173
see also first language speech fluency; speech rate 118
second language speech fluency grammar acquisition
fluency course model 206–7, 221 children’s 48
evidence of fluency gain 208–15 grammatically specialized formulas 43
focus clusters 27–8 greater length/complexity than other
focus on form (FonF) pedagogy 192–3 output 112
FonF pedagogy see focus on form
pedagogy Hakuta, K. 46
formal speech 205 Hansen, L. 20, 24
form-focused repetition 191 Hawkins, R. 19, 36, 37
formulaic sequences 2, 38 hesitation phenomena 14, 16, 24
approaches to study of 40–1 Hickey, T. 39, 47, 50
categorization 42–5, 46 Hieke, A. 28
definition 38–40 Ho, Y. 80, 81
double role in child language Hooper, J. 47, 48
acquisition 48–9 Hulstijn, J. H. 61
evidence for faster processing of 71
fostering repertoire of 192–5 identified regulation 74, 75
functions 58 implicit memory 66
identification criteria 41–2, 44–6, inner speech 76, 173
109–12, 208–9 stages 77
memory and 3 input hypothesis 186
pragmatic aspects 51–3 input processing theory 186
protolanguage and 53–4 instance theory 4, 64–5
as rhetorical devices 168–9, 174 institutionalized expressions 44
role in fluency 5, 162, 170, 171, 177 interaction hypothesis 187–8
semantic and syntactic interaction in second language 187–9
irregularities 45, 112 interactive features 16–17
238 Index

interactive phenomena 16 length of fluent runs 15, 28, 35, 56

intrinsic motivation 74 repetition of formulas in a
introjected motivation 75 run 164–5
use of multiple formulas to extend a
Japanese L1 research run 166
participants 95, 100 Lennon, P. 15, 19, 21, 23, 25, 27,
articulation rate 121 30, 35
individual quantitative characterization of fluency 11
performance 135–8 Levelt, W. J. M.
language transfer 124, 133–4 model of speech production 12,
speech rate 119 63, 93
Japanese language Lewis, M. 184
discourse features 79, 82 lexical approach 185
Jiang, N. 71 lexical bundles 40
Johnson, M. 197 definition 194
in pedagogical materials 194, 195
Kahnemann, D. 63, 69 see also lexical phrases
Kasper, G. 34 lexical clusters
Koprowski, M. 194 in EAP textbooks 194
Krashen, S. D. 186 lexical phrases 37, 39
classification 44
L1 see first language pragmatic aspects 51
L2 see second language see also lexical bundles
Ladefoged, P. 70 lexical planning 30, 35
Lakoff, G. 197 linguistic relativity 79
language accuracy 10–11 Logan, G. D.
language acquisition theory of automatization 3, 64
input and 4, 65–6, 185–6 Long, M.
language anxiety 72, 176–7 interaction hypothesis 187
language input long-term memory 2, 49–50, 61
in classroom 186
fluency and 6, 65 Mandarin Chinese
fluency course model 206 discourse and temporal
phonological loop and 61 features 80–1
language output 189 marketplace (fluency promoting
language processing activity) 201–2
psycholinguistic aspects of 2–6, 62 Marks, J. 185
types 3, 62 Maynard, S. K. 82
language transfer 36–7, 145, 172 McCafferty, S. G.
individual data 134 on stages of inner speech 77
language group data 134 McCarthy, M. 185
phonation/time ratio and 124, 125 McGuire, M. 88, 184
speech rate and 118–9 McLaughlin, B. 62, 196
types 34 McLeod, B. 62, 196
Laver, J. 81 McMahill, C. 78
learner investment 78 meaning-focused repetition 191
lemmas 64 meaning negotiation 188
 Index 239

mean length of runs (MLR) 93–4, 108, NOA see needs only analysis
115, 124–7, 152–3, 154, 172, 175 noticing 4, 67, 183
by first language 132
impact of increased use of formulaic object-regulation stage of inner
sequences 162 speech 77
memorization 192 O’Brien, I. 68
memory O’Dell, F. 185
description of 62–3 oral process approach 204, 221
role in language processing 2–4, oral proficiency
61–3 fluency as distinct from other aspects
mental lexicon 64 of 11
messengers (fluency promoting other-regulation stage of inner
activity) 201 speech 77
methodology of fluency research 91–2,
95–7 pause(s) 23, 56
implications 174–6 clustered 26
MEU see morpheme equivalent unit filled 16, 108
micropauses 24 micropause 24
mingle jigsaw 199–200, 207 research studies 23–8
Mitchell, R. 48 timing of 89–91, 107–8
mnemonic aids 197 unfilled 16, 17, 23–6
Möhle, D. 14, 15, 18, 20, 23, 29, 33–4 pause frequencies 23–5
morpheme equivalent unit (MEU) 40 pause location 25, 26, 36
motivation 73–5 pause times 15, 23–5, 36
Mudraya, O. 184 Pawley, A. 30, 33, 56, 57, 58, 81, 187
multiple formulas in a run 166–8 definition of formulaic sequences 39
Myles, F. 47, 48 pedagogy of second language speech
fluency 6, 183–5, 220–3
narrative moves analysis 149–51 Peirce, B. N.
The Cat Came Back 159–61 idea of learner investment 78
Neighbours 150–3 Perkins, M. R.
Strings 153–8 definition of formulaic sequences 38
narrative retell 98, 101–3, 106–7, 174 on semantic and syntactic
Nation, P. 190, 201 irregularities 45
native speaker judgment 92, 110 Peters, A. M. 41, 47, 50, 54, 57, 69, 70
Nattinger, J. R. 80, 184 formula for identification of
definition of lexical phrases 39 formulaic sequences 41, 91
notion of composites 43 phatic formulas 49
on pragmatic competence 51 phonation/time ratio (PTR) 108,
taxonomy of lexical phrases 44, 115, 121
97, 112 by first language 130
needs only analysis (NOA) 45 phonological coherence 111
Neighbours (film) 103–4 phonological loop 61, 68
narrative complexity 105 phonological memory 67, 68
narrative moves analysis 150–3 phonological reduction 59, 108
plot 103 phrasal constraints 44
Nekrasova, T. 71 phraseological approach 40
240 Index

Pollard, J. 20, 24 sample activities 199–203

polyfunctional formulas 49 repetitions of formulas in a run 164–5
polywords 44 research on second language (L2)
Powell, M. 185 speech fluency 1–2
Poyatos, F. issues in data collection and
on cultural fluency 79 analysis 87–90
practice of production tasks 189–90 themes 12–17
pragmatically specialized research on second language (L2)
formulas 43–4 speech fluency and formulaic
pragmatics 51–3 sequences 90–2, 145–6
prefabricated sequences 45, 55 data analysis 107–9, 117–18
adults’ use 49, 50 data collection 97–8
children’s use 46, 47 design 92–5
preparation for a production task 190 future directions 178
fluency course model 206 hypotheses 93–4, 171–2
pre-task planning 190 identification of formulaic
proceduralization of language see sequences 109–13
automatization of language methodology 91–2, 95–6
procedural knowledge 12, 60–1 participants 100–1
definition 2 pilot study 97
speech production and 63–4, 92 procedure 101–6
transfer from L1 to L2 33–7 qualitative results 147–9, 169–70,
Prodromou, L. 209 173–4
production tasks 189–90 quantitative results 116, 145, 171–3
fluency course model 206 research studies 20–3
protolanguage 53–4 restructuring model of
pruning of words 15 automatization 65–6
Psathas, G. 89 results of fluency research
psycholinguistics 9, 60–3 qualitative 147–9, 169–70, 173–4
PTR see phonation/time ratio quantitative 115–16, 145, 171–3
retraced restarts 21
qualitative fluency research 147–9, retrieval of formulaic sequences 56,
169–70, 173–4 68–9, 70
quantitative fluency research 116, 145, rhetorical marker formulas 168,
171–3 173–74
question formulas 49 rhythm-sensitive timing 90
Riggenbach, H. 12, 17, 19, 20, 24, 26,
Raupach, M. 28, 33–4, 36 215, 221
on types of formulaic sequences 55 fluency variables 16–17
reciprocal determinism 73 Robinson, P.
Rehbein, J. on automatization of language 3–4,
on fluency 11–12, 30 61–62
rehearsal 61 Rossman, T. 62, 195–6
Renouf, A. routines 55
on dilexicality 43–4 routinized speaking plans 30, 36–7
repair phenomena 16–17, 155 levels of 30
repetition (practice) 191 runs of speech see length of fluent runs
 Index 241

Sajavaara, K. 55 silent film prompts 103–5, 174

satellite units 52 see also The Cat Came Back (film);
Sawyer, C. R. 73 Neighbours (film); Strings (film)
Schmidt, R. 57, 50, 51, 71 simplification and task-management
categorization of controlled strategy 34–5
processing 63 Sinclair, J.
on fluency 12 on dilexicality 43–4
on noticing 4–5, 67 situational formulas 42
second language (L2) pedagogy Siyanova, A. 71
formulaic language and 5–6, 183–5, social identity 78–9
220–3 Spanish L1 research participants 95,
second language (L2) speech 100
fluency 1 articulation rate 121
gender and 78, 145, 172 individual quantitative
interplay between voice and 75–7 performance 141–5
language input and 4 language transfer 119, 133–4
nature of L1 and 79–82 speech rate 119
promotion activities 196, 220–1 Spanish language
sociocultural variables 72, 82–3, 134, temporal features 81
145–6, 172, 176–7 speech fluency see fluency
see also fluency; research on second speech production
language (L2) speech fluency Levelt’s model 12, 63–4, 92
second language (L2) speech fluency speech rate (SR) 17, 94, 107, 108,
development/gain 145–6, 115–16, 133
210–15 by first language 130
complexities 176–7 notion 14, 15
formulaic sequences effects on 5–6, research studies 18–20
162, 170, 171, 177 Spanish language 81
pattern of change, by language speech syllabus 203–6
group 116 spontaneous speech 5–6, 87, 204
pattern of change of SR see speech rate
individuals 129–34 state anxiety 73
phonological memory and 67–8 statistical tests 109
in target language milieu 13–14 individual data 134–45
Segalowitz, N. 13, 14, 191, 200 language group data 130–4
self-correction see repair phenomena whole group data 116–30
self-determined motivation 73–5 storage of formulaic sequences
self-efficacy 73–5, 177 69–70
self-regulation stage of inner strategic planning 190
speech 72 strategic transfer 34
self-talk 162–4 Strings (film) 103–4
semantic irregularity 45, 112 narrative complexity 105
sentence builders 44 narrative moves analysis 153–8
shadowing and tracking (fluency plot 103–4
promoting activity) 198–9, 206–7 student-to-native speaker
shaped speech 204–5 interaction 189
short-term memory 2–3, 61 student-to-student dictations 202–3
242 Index

student-to-student interaction 187, VanPatten, B. 186, 193

188–9 verbal thought 76, 77
study abroad context Vogel, T. 48, 52
extrinsic motivation 74–5 voice 75–7, 177
fluency development in 13–14 Vygotsky, L.
subsidiary transfer 35 notion of inner speech 76–7
Swain, M. 10, 189
Syder, F. H. 27, 30, 56, 57, 58, 82, 187 Walker, I. 192
definition of formulaic Weedon, C.
sequences 38–9 post-structuralist concept of social
syntactic irregularity 45, 112 identity 78
syntactic planning 30, 36 Weinert, R. 69, 70
Whorf, B.
Tajfel, H. linguistic relativity of 79
view of social identity Williams, J. 186
formulation 78 Willis, D. 184
talk journal 204 within-task planning 196
task-based repetition 191 Wong-Fillmore, L. 46
temporal variables of speech 9, 11, Wood, D. 194, 199, 208, 216
94–5 word boundary perception 67
calculation of 108–9, 115–16 word-order dominant languages 80
research studies 12–17 working memory 2–3, 61
statistical significance of changes Wray, A. 209
in 109 definition of formulaic
Ten Have, P. 88 sequences 38–9
on timing pauses 90 definition of morpheme equivalent
Terrell, T. 186 unit 39
textbook language on evolutionary origins of formulaic
in EAP/ESP materials 194 sequences 53
Towell, R. 14, 19, 23, 29, 36, 37 on formulaic sequence
trait anxiety 73 identification 37–40, 115
transcription 107–8 on functions of formulaic
judgement procedure 113 sequences 57–8
shadowing and tracking 198–9 on memorized formulaic
transcription notation 89–90 language 192
Treisman, A. 63, 69 on needs only analysis 45
Tucker, G. 46 on prefabricated formulas
t-units identification 49
notion 15–16 on semantic and syntactic
irregularities 45
unfilled pauses 24
length of runs between 16 Yorio, C. 49
untraced restarts 21, 22 categorization of formulaic
Utsumi, T. 192 sequences 42
utterance clusters 81 Yuan, F. 190