Optimizing a Lexical Approach to Instructed

Second Language Acquisition

This page intentionally left blank
Optimizing a Lexical
Approach to Instructed
Second Language Acquisition

Frank Boers
Erasmus University College, Brussels, Belgium

and

Seth Lindstromberg
Hilderstone EFL College, Broadstairs, UK
© Frank Boers and Seth Lindstromberg 2009
All rights reserved. No reproduction, copy or transmission of this
publication may be made without written permission.
No portion of this publication may be reproduced, copied or transmitted
save with written permission or in accordance with the provisions of the
Copyright, Designs and Patents Act 1988, or under the terms of any licence
permitting limited copying issued by the Copyright Licensing Agency,
Saffron House, 6-10 Kirby Street, London EC1N 8TS.
Any person who does any unauthorized act in relation to this publication
may be liable to criminal prosecution and civil claims for damages.
The authors have asserted their rights to be identified as the authors of this
work in accordance with the Copyright, Designs and Patents Act 1988.
First published 2009 by
PALGRAVE MACMILLAN
Palgrave Macmillan in the UK is an imprint of Macmillan Publishers Limited,
registered in England, company number 785998, of Houndmills, Basingstoke,
Hampshire RG21 6XS.
Palgrave Macmillan in the US is a division of St Martin’s Press LLC,
175 Fifth Avenue, New York, NY 10010.
Palgrave Macmillan is the global academic imprint of the above companies
and has companies and representatives throughout the world.
Palgrave® and Macmillan® are registered trademarks in the United States,
the United Kingdom, Europe and other countries.
ISBN: 978–0–230–22234–2 hardback
This book is printed on paper suitable for recycling and made from fully
managed and sustained forest sources. Logging, pulping and manufacturing
processes are expected to conform to the environmental regulations of the
country of origin.
A catalogue record for this book is available from the British Library.
A catalog record for this book is available from the Library of Congress.
10 9 8 7 6 5 4 3 2 1
18 17 16 15 14 13 12 11 10 09
Printed and bound in Great Britain by
CPI Antony Rowe, Chippenham and Eastbourne
Contents

Acknowledgements vii

List of Figures and Tables viii

List of Boxes ix

Abbreviations x

1 Introduction 1
1.1 What parts of language is this book about? 1
1.2 Why have we written this book? 16

2 The Contribution of Chunks to

Acquisition and Proficiency 24
2.1 Comparing the nature and necessity of
L1 and L2 chunk acquisition 24
2.2 Evidence that L2 learners benefit from
chunk knowledge 35

3 Estimating the Chances of Incidental Uptake of L2 Chunks 39

3.1 Fostering learner autonomy through awareness-raising 39
3.2 Estimating the chances of semi-incidental
uptake of L2 chunks 51

4 Selecting Chunks for the L2 Classroom 55

4.1 Utility 55
4.2 Teachability 68

5 Semantic Elaboration 79
5.1 Exploiting imagery 79
5.2 Organizing lexis 96

6 Structural Elaboration 106

6.1 Phonological motivation in the formation and
standardization of chunks 106
6.2 The memorability of phonologically motivated chunks 119

7 Bearing in Mind 126

7.1 The need for width and depth of chunk knowledge 126
7.2 The need for automaticity 134

v
vi Contents

8 Directions 146
8.1 Broadening the scope 146
8.2 Testing chunk knowledge 160

Notes 169

References 174

Index 193
Acknowledgements

Frank would like to thank his collaborators June Eyckmans, Hélène

Stengers, Aline Godfroid and Julie Deconinck for sharing their
research findings and for talking him into believing that his advice
sometimes matters. A very special thanks from Frank also to Murielle,
Tom and Pauline for putting up with a moaning and groaning
husband/daddy-turned-author for three long months (again).
Seth would especially like to thank Tessa Woodward – for all kinds of
reasons. Also, Sergei and Larry – thank you for Google; Richard Lowry –
thank your for ‘VassarStats’; and thanks too to all the authors who have
helped us by making their papers available on the web.
Frank and Seth would like to thank Jill Lake, Melanie Blair and
Priyanka Pathak at Palgrave Macmillan for their guidance and assist-
ance throughout this project.
Last but not least, thanks also to the hundreds of students who partici-
pated in one or the other of our numerous classroom experiments over
the years.

vii
Figures and Tables

Figures

4.1 Priority zone for explicit chunk targeting in the classroom 63

5.1 Example (1) of a collocation box proposed
by Lewis (1997: 78) 102
5.2 Example (2) of a collocation box proposed
by Lewis (1997: 79) 103

Tables

1.1 Word sequences classified by frequency plus MI score;

an abbreviated presentation based on Ellis,
Simpson-Vlach and Maynard (2008: 381) 7
4.1 Exact-word Google hits as an indication of relative
commonness 58
4.2 Joint-frequency data on some verb–noun collocations
(from the Collins Cobuild collocations sampler) 59
6.1 /b/_ + /b/_ vs. /f/_ + /b/_ multiword lexis in the MED 109
6.2 /k/_ + /k/_ vs. /s/_ + /k/_ multiword lexis in the MED 109
6.3 Evidence of the role of alliteration in the formation and
standardization of multiword units (sourced from
the Macmillan English Dictionary for Advanced Learners,
2007, 2nd edn) 110
6.4 Consonant repetition within monomorphemic words
in three 1402-token texts 115

viii
Boxes

1.1 Partial transcription of a BBC Radio 4 programme 3

1.2 Strong collocates of four prepositions, generated by
the Collins Cobuild on-line collocations sampler
at http://www.collins.co.uk. 6
3.1 Strong verb–noun collocations occurring in
the first 120 pages of Val McDermid’s (2007) thriller
Beneath the Bleeding 42
4.1 Idioms encountered in the first 120 pages of
Val MacDermid’s (2007) thriller Beneath the Bleeding 67
5.1 ‘Identify-the-source’ multiple-choice items and feedback 85
5.2 ‘Meaning’ multiple-choice items 86
5.3 ‘Gap-fill’ post-test 86
5.4 Examples of contextualized idioms accompanied
by ‘origin’ hints 90
5.5 Examples of pictorial elucidation of the origins of idioms 91
5.6 Examples of figurative uses of manner-of-motion verbs 94
5.7 Examples of ‘identify-the-informal-idiom’ exercises 95
5.8 Phrasal verbs grouped per metaphor theme 98
5.9 Anger-related expressions grouped according to
metaphor themes (based on Boers, 2000b) 99
5.10 Grouping idioms according to their source domains 101
6.1 Some proverbs, similes and binomials displaying
phonological repetition 117
6.2 Some compounds and collocations displaying
phonological repetition 118
7.1 Reducing blind guessing in matching exercises 130
7.2 Reducing blind guessing in completion exercises 131
8.1 Creating paired associates for phrasal verbs 155
8.2 Examples of colligation and collocation test
items from Hargreaves (2000) 161
8.3 Example items from the Discriminating Collocations Test,
from Eyckmans (in press) 163
8.4 Example of a Deleted Essentials Test 165
8.5 Key to the example Deleted Essentials Test 166

ix
Abbreviations

ACCESS automatization in communicative contexts of essential

speech segments
ACT active control of thought (theory)
ALM audio-lingual method
BNC British National Corpus
CL Cognitive Linguistics
CLT communicative language teaching
DET Deleted Essentials Test
Disco Discriminating Collocations Test
ESS essential speech segment
FLT foreign language teaching
IT instance theory
ISLA instructed second language acquisition
LA lexical approach
L1 first language, mother tongue
L2 additional language
LG derived from Latin or Greek
MED Macmillan English Dictionary for Advanced Learners (2007,
2nd edn.)
MI mutual information
OAL Oxford Advanced Learner’s Dictionary of English (2000,
6th edn.)
ODE Oxford Dictionary of English (2005, 2nd edn.)
PA paired associates
SLA second language acquisition
TBI task-based instruction
TEFL teaching English as a foreign language

x
1
Introduction

1.1 What parts of language is this book about?

One of the things that distinguishes native speakers and highly profi-
cient language learners from less proficient language learners is their
mastery of a large stock of semi-fixed lexical phrases, also known as
‘chunks’ (Pawley and Syder, 1983). It is well known that even advanced
learners who have learned a great many words and ‘grammar rules’
nevertheless often fail to combine words the way native speakers do.
For example, a learner may ‘know’ the words full, total and functional,
and know the grammar rule that adjectives can usually be turned into
adverbs by adding the suffix -ly, but still may not realize that fully
functional, in the sense of ‘in optimal working order’, is much more
idiomatic (that is, native-like) than totally functional.1 This learner may
also know that naked and nude are closely synonymous but not real-
ize that stark naked is idiomatic whereas stark nude is not.2 Given the
fact that most English words have synonyms, the chances of learn-
ers producing non-idiomatic word combinations are statistically quite
considerable. Thus, the common learner strategy of transferring word
partnerships from L1 to L 2 is hardly reliable since combinations that
are idiomatic in one language often sound awkward when translated
word-for-word into another, even when both languages are closely
related. For example, due to L1 interference Dutch-speaking learners
of English may say *do an effort (instead of make an effort), *with other
words (instead of in other words) and *take someone by the nose (instead
of lead someone by the nose). Similarly, L1 interference may lead French-
speaking learners of English to produce *do a mistake (instead of make
a mistake), *realize a survey (instead of conduct a survey), *let’s drink a
glass (instead of let’s have a drink) and *a flat water (instead of a mineral

1
2 Optimizing a Lexical Approach

water). Monoglot native speakers may now and then wonder why a
learner whom they have perceived as fluent and expressive produces
word combinations as deviant as these, but in fact mistakes of this
kind are quite common even in the speech and writing of learners
who are indisputably at an advanced level of proficiency (Nesselhauf,
2003; De Cock, 2004).
It is true that L1 transfer does sometimes happen to generate idiomatic
L2 word combinations, but how is a learner to know in advance when
this will work? How can the learner tell solely on the basis of know-
ledge of single words that a large majority of is, judging by its much
greater frequency, decidedly more native-like than a big majority of ?3
When equivalent L1–L2 word combinations do happen to be available,
this equivalence may be confined only to the ‘content’ words (that is,
nouns, verbs, adjectives and certain adverbs). For example, the Dutch
counterpart for English fall asleep is in slaap vallen. Here, the two content
words in the Dutch expression transfer relatively straightforwardly, but
transferring the preposition as well will result in something like *fall
in sleep. Similarly, the French counterpart of take charge of [something]
is prendre [something] en charge, in which prendre means ‘take’. L1 trans-
fer is therefore likely to lead to take and charge, but carrying en over
into English (as in) cannot lead to a natural result. More seriously, for
a considerable number of L2 chunks there simply will be no easily rec-
ognized L1 counterpart at all. This is why the translation of chunks is
notoriously difficult. So much so that it is often treatment of phrase-
ology which most clearly indicates which of two translations is super-
ior (Colson, 2008). In the same vein, it is often differences in mastery
of phraseology that distinguish an advanced foreign language learner
from one who is upper-intermediate.
The stock of chunks in English (for example) is very diverse. It
includes strong collocations (commit a crime), social-routine formulae
(Have a nice day), discourse markers (on the other hand), compounds (peer
pressure), idioms (take a back seat), standardized similes (clear as crystal),
proverbs (when the cat’s away ...), genre-typical clichés (publish or perish),
exclamations (you must be kidding!) and more. The number of chunks at
the disposal of adult native speakers of English is estimated to be many
thousands (Pawley and Syder, 1983). According to some estimates, about
half of English written text is made up of them and the proportion is
likely to be even larger in spoken discourse (Butler, 2005: 223; Erman
and Warren, 2000). Consider the text in Box 1.1, which we transcribed
from a BBC Radio 4 programme. We have underlined word sequences
which, to us, seem to be chunks.
 Introduction 3

Box 1.1 Partial transcription of a BBC Radio 4 programme

Are the numbers of boys and girls in our families really down to the toss of a
coin? In fact, it’s not quite so simple. You as an individual may actually load
the dice towards a son or a daughter right at conception. Especially the con-
dition of mothers could be playing a part according to some studies. Ruth
Mace was in Ethiopia when that country was hit by a severe food shortage. As
part of a study on nutrition she looked at the birth statistics of women caught
up in the crisis: Mothers that had a higher body-mass index were more likely
to have boys than girls. Why this happens is still open to debate.

Valerie Grant says dominance in personality may also tip the balance towards
male offspring. ‘I’ve come to notice that dominant women tend to have more
boys. The explanation may be that dominant women have higher levels of
the male hormone, testosterone.’ She also proposes an explanation for the
higher ratio of boys born in times of war. It also features the testosterone
that is at the heart of her dominance theory, but it’s all about stress in this
case. ‘Testosterone in women rises under stress. So, women who might be just
below the level for conceiving a son in normal times may be more likely to
conceive sons in times of hardship.’

A little exercise like this illustrates just how densely packed with lexical
phrases an ordinary text can be, especially since we have taken a con-
servative stance in our identification. For instance, we did not under-
line those whose components were separated (is ... down to ...) and those
which allow a small range of substitutes (have boys / girls; above / below
the level; a higher / lower ratio and hit by a (severe) shortage / drought / earth-
quake ...). We could also extend several of the underlined sequences by
including additional elements as part of the chunk: right at conception;
propose an explanation for; in times of war; in times of hardship. Finally,
some authorities would also highlight common combinations such as
tend to and levels of.
As our marking of the excerpt further demonstrates, chunks can come
in a variety of forms and fulfil a variety of functions. This diversity may
help explain why a plethora of terms have been used in the literature
both for chunks in general (for example, multiword units, holophrases,
phrasemes, formulae, set phrases, lexical bundles, prefabricated routines,
ready-made utterances, formulaic sequences and so on) and various sub-
classes of chunk (such as phrasal verbs, multiword verbs, figurative idioms
and dead metaphors; for more, see Wray, 2000).
Alison Wray’s (2002: 9) often-quoted definition of what she calls
‘formulaic sequences’ and what we will generally and interchangeably
call ‘lexical phrases’ or ‘chunks’ is as follows: ‘a sequence, continuous
4 Optimizing a Lexical Approach

or discontinuous, of words or other 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’. This definition covers all the word strings
that are assumed to be processed as holistic units by any given individ-
ual language user and it thus allows for the likelihood that particular
word strings that are formulaic to one individual need not be so for
another. In other words, the repertoire of chunks at the disposal of a
given speaker may be expected to be somewhat idiosyncratic. This is
because, from childhood on, native speakers pick up recurring word
combinations from the language samples they are exposed to, and
these samples will never be exactly the same for any two individuals.
For example, differences in exposure may be due to regional variation
(such as American versus British English), diachronic variation (such
that the chunk repertoires of different generations diverge), variation in
people’s personal interests and/or professional occupations which hap-
pen to involve differing exposures to a particular jargon or text genre,
and so on. As a result, word strings that are experienced as formulaic
(and idiomatic) by members of a particular group of language users (and
which may actually help to mark them as in-group members) will not
always be perceived (or mentally processed or remembered) as such by
people outside that group.
Leaving aside learners of a language for a specific purpose (Business
English, Financial English, English for Law, English for Academic
Purposes, etc.), it must be word strings that are formulaic to most native
speakers that are most relevant for language learners to master, for it is
mastery of these strings that is most likely to increase learners’ chances
of producing the kind of discourse that native speakers and other expert
speakers will feel to be natural. How this ‘common core’ can be deline-
ated is far from self-evident, however. In theory, it requires a compari-
son of the chunk repertoires that are part of the mental lexicons of
all native speakers of a language, and such an enterprise is clearly not
feasible. An indirect way of identifying chunks that are shared by the
majority of native speakers is to resort to corpus linguistics – the study
of words, phrases and other linguistic phenomena in texts, especially in
collections of texts known as ‘corpora’.
Electronic corpora such as the British National Corpus, the Cambridge
International Corpus and the Collins Cobuild Wordbanks are used as
sources of information both about the frequency of occurrence of
given word combinations and about the strength of word partnerships,
known as ‘collocations’ (Sinclair, 1991). Word sequences that occur
 Introduction 5

with high frequency in the corpus (such as in other words) and (near-)
exclusive word partnerships (such as wreak havoc) can be taken as very
good candidates for membership of a class of common chunks shared
by many expert speakers of a language. The degree of bonding into a
partnership is calculated from the number of co-occurrences of two
words in the corpus and is referred to as the Mutual Information (MI)
score or (when the overall relative frequency of the individual words is
incorporated into the calculation of probability of co-occurrence) as the
T-score (a variant of which is called the Z-score). The higher these stat-
istical scores, the stronger the word partnership is. Box 1.2 lists forty of
the top collocates of four English prepositions generated by the free on-
line collocations sampler of Collins Cobuild, which extracts data from
the Wordbanks corpus.
Can you guess on the basis of these lists of collocates which preposi-
tions we typed in as query words in the sampler?
The fact that is it feasible to make educated guesses about the search
items that go with the collocates in Box 1.2 illustrates the truth of John
R. Firth’s famous statement ‘You shall know a word by the company it
keeps’ (1957: 11). Let’s go into this in more detail. You probably found the
task easiest for the left-most column in Box 1.2 (which lists collocates of
over) and most difficult for the right-most column (which lists collocates
of beyond). That is because the statistical probability of co-occurrence of
over with its top forty collocates is much higher overall (T-scores range
between 54 and 9) than it is for beyond and its top forty collocates (where
T-scores are between 4 and 2 only). In other words, the number of times
that you have encountered over in combination with one of its top col-
locates is likely to be greater than the number of times you have encoun-
tered the beyond collocations. The second column (collocates of under)
and the third (collocates of behind) are made up of words whose statis-
tical probability of co-occurrence with their respective prepositions lies
in-between the others (T-scores for the listed collocates of under range
between 30 and 8; those for behind are between 7 and 4).
People pick up chunks from the language they are exposed to. The
more frequently a meaningful word combination occurs in the lan-
guage one is exposed to, the more likely it is that that word combination
will be picked up, that is, remembered as a single chunk. On the assump-
tion that intelligently compiled corpora can be highly representative
of usage in particular languages, corpus data offer an indirect but very
useful way of identifying chunks which are likely to be shared by most
(if not all) adult native speakers. At the same time, one should not lose
sight of the fact that any corpus which consists of a mix of unattributed
6 Optimizing a Lexical Approach

Box 1.2 Strong collocates of four prepositions, generated by the Collins

Cobuild on-line collocations sampler at http://www.collins.co.uk.

Preposition 1 Preposition 2 Preposition 3 Preposition 4

Collocate Collocate Collocate Collocate

years pressure falling boundaries

all control curtain circumstances
period circumstances eyes bounds
last way second hills
control conditions hiding life
took water seconds capacity
take Act reasons range
victory fire fallen confines
year law schedule age
again age screen expectations
months 21 shut understanding
weekend threat ball yards
win attack stayed even
hand section car dispute
million cover chair nothing
taken arrest lagging normal
shoulder come caught horizon
head agreement pulled question
handed stress finished duty
and rules just reproach
world government walls mountains
top system trailing fringe
here supervision locked lies
weeks leadership camera imagination
row influence ears extending
taking protection lagged walls
dispute rule glass ordinary
controversy scheme lag spread
place command place stretch
spread siege weight maturity
country name hid reaches
time terms lags power
presided auspices mask capabilities
came investigation idea realm
concern breath fall way
debate scrutiny façade joke
counter sun scenes above
turned arm tied goes
turn guidance lie pale
bent roof from and
 Introduction 7

contributions from more than one speaker (which is the case with all
large corpora) cannot with clarity reflect the language production of
any particular individual language user (Howarth, 1998). For example,
although data drawn from large eclectic corpora reveal that down the
drain occurs more frequently than down the tube, there might be individ-
ual speakers who use only the latter variant. Likewise, collocations data
from a large corpus may show that make a presentation is a firmer word
partnership in statistical terms than give a presentation, but it is conceiv-
able that some speakers produce only the latter. Importantly, none of
this precludes the possibility that at the level of reception (while reading
or listening), speakers recognize and process many more chunks than
they actively use themselves.
Ellis, Simpson-Vlach and Maynard (2008) argue that it is import-
ant to distinguish between (1) word sequences which are common
merely because their component words happen to be highly frequent
and (2) sequences which (a) comprise words whose frequency of co-
occurrence is greater than chance and (b) show clear semantic coher-
ence by virtue of expressing a relatively identifiable unitary meaning
or function. The following examples of the first type derive from the
five-million-word CANCODE spoken corpus as reported in O’Keeffe,
McCarthy and Carter (2007: 65–6), who provide ‘top 20’ frequency lists
of two-, three-, four- and five-word chunks: in the, do you, and it was, one of
the and you know what I. As examples of the second type, Ellis, Simpson-
Vlach and Maynard (2008) mention: in other words, a wide variety of, it
has been shown and it should be noted that. They use frequency scores and
the MI (mutual information) statistic to sort word sequences into the six
sub-categories shown in Table 1.1. A high MI score, which means that
the words occur together much more often than would be expected by
chance, is used as an indicator of semantic coherence whereas a low MI
score, which indicates that the words occur together substantially by
chance, is taken to indicate low overall meaningfulness.

Table 1.1 Word sequences classified by frequency plus MI score; an abbreviated

presentation based on Ellis, Simpson-Vlach and Maynard (2008: 381)

Frequency per
million ↓ Low MI Medium MI High MI

Low that the only happens is that it has been shown

Medium and at the that may be see for example
High in the case of the a kind of In other words
8 Optimizing a Lexical Approach

It stands to reason that it is word strings which have both a (rela-

tively) high frequency of occurrence and a (relatively) high MI score
that have the best chance of being recognized by many native speakers
as conventionalized multiword units – that is, recognized as ‘chunks’.
If a collection of lexical phrases shared by most native speakers were
to be compiled with the help of corpus data, its size would still be daunt-
ing from the standpoint of a foreign language student. For instance, the
Oxford Collocations Dictionary for Students of English (2002), which the
publisher states is for upper-intermediate and advanced learners, lists
about 150,000 collocations (which is just one type of chunk), and even
this inventory must be far from exhaustive since it is built around a
clearly non-exhaustive selection of just under 9000 single headwords
(only nouns, verbs and adjectives).4 The Collins Cobuild English Language
Dictionary (5th edn, 2006), a general-purpose dictionary for advanced
learners, lists nearly 7000 phrasal expressions while the Oxford Idioms
Dictionary for Learners of English (2001), which of course concerns only a
single type of chunk, includes over 10,000 specimens (although some are
purely British while others are purely North American). According to its
publisher, the very compact Oxford Phrasal Verbs Dictionary for Learners
of English (2006) covers 7000 specimens of another type (again, both
British and North American). Even a very small dictionary of phrasal
verbs or idioms may cover between 4000 and 5000 target expressions.
Such figures are all the more suggestive since it is possible that even for
a language as intensively studied as English, some number of frequent
chunks are likely to have escaped the notice of the compiler(s) of any
given learner’s dictionary.
Given the enormous stock of chunks that could usefully be targeted
in second or foreign language learning – and given too the recognition
that language learning must involve more than learning chunks – it
seems inevitable that some prioritization will need to be carried out,
especially in contexts of classroom-based language learning (which
we shall refer to as instructed second language acquisition, regardless of
whether the target language is a second or third (or nth) language), where
time devoted to the target language very seldom exceeds four or five
teaching hours a week. The question this raises is: what criteria should
be applied in deciding what (kinds of) lexical phrases merit attention in
learning and teaching?
As already illustrated above, chunks make up a very diverse class of
multiword units that differ in function as well as form. Moreover, some
forms are multi-functional (for instance, kind of and sort of ) while a great
many functions are expressible by multiple approximately synonymous
 Introduction 9

forms; for example, anyway, at any rate and anyhow5 may all be used to
signal the end of a digression. As a result, the various taxonomies that
have been proposed in the literature have fuzzy boundaries (see Wray,
2002: 44–6, for a review of attempted taxonomies). Not only that: cat-
egories in one classification tend to cut across categories in other clas-
sifications (Howarth, 1998: 25–9). Given the amount of work already
done in classifying chunks, in this book we will not take a stand on
any particular classification ourselves beyond what might be neces-
sary in order to address the question of whether some types of chunk
might generally be better candidates than others for explicit targeting
in the classroom or in pedagogical materials. We will discuss the issue
of chunk selection in more depth in Chapter 4, but before that, let us
briefly examine whether the criteria that have so far been used to make
distinctions among lexical phrases might be helpful for our purposes.
A first type of classification of chunks involves considering aspects
of their function. For example, some phrases are especially useful dur-
ing face-to-face interaction: social routine formulae (such as Excuse me,
Have a nice day, How are you doing?), conversational fillers (sort of, you
know what I mean, you see and so on), interactional sentence heads (for
example, Shall we ⫹ infinitival phrase and Would you mind ⫹ gerundial
phrase) and situation evaluators (like you must be kidding!, small world!,
when the cat’s away ...). Other phrases organize discourse (by the way, on
the other hand, having said that, last but not least, let’s move on to) and still
others have a ‘referential’, or ‘message oriented’, function (stomach ache,
commit a crime, break up, put on weight). From the vantage point of lan-
guage learners, the immediate relevance of mastering particular chunks
is likely to depend on their functionality. For example, social routine
formulae, which can help learners ‘fit in’ with a group of native speakers,
are likely to be felt as especially useful by learners in contexts of nat-
uralistic language learning, for instance during (the early stages of) an
immersion course. Lexical phrases that help to organize discourse may
be felt to be especially useful as aids in performing composition tasks
(in academic writing, for instance) or in listening comprehension (when
trying to follow lectures, for instance). On the other hand, in contexts of
predominantly classroom-based learning without immediate opportun-
ities for social interaction with groups of L2 speakers, it may be chunks
with a referential function (which help the students communicate about
things in the world) that are felt to be most useful.
A second way of trying to classify chunks is to take their formal
features as a starting point. For example, chunks can be located on a
continuum ranging from ones where the word partnership, from the
10 Optimizing a Lexical Approach

perspective of at least one of the words, is fully fixed (commit suicide) via
chunks that allow for restricted substitution (conduct / do / carry out an
experiment) and on to ‘open slot’ frames such as the _er, the _er (as in the
sooner, the better), as _ as_ (I did it as fast as I could) and it takes/took/will
take (someone)[time] [infinitival phrase] (for example, It took me two hours
to get there). Intuitively, it would seem to be highly fixed word partner-
ships that merit special attention on the part of teachers and learn-
ers, simply because the chance of erring with combinations like these
is statistically greater than is the case with combinations that allow
more variation. However, it must be noted that a learner’s interest in
mastering chunks will not solely be to produce idiomatic language (our
starting point at the beginning of this introductory chapter). An add-
itional major advantage that has been proposed for acquiring a great
many chunks is that chunks (by definition) are word strings that can
be retrieved from memory more quickly than they can be composed
word by word on the basis of one’s knowledge of grammar. (We will
survey the evidence for this claim in Chapter 2.) This extra efficiency
enhances fluency not only in language production but also in reception
since chunks can be recognized and interpreted as wholes, not parsed
word by word. This frees up cognitive processing space for attending
to any less predictable word strings in the same incoming message. For
example, most readers will effortlessly anticipate the missing words
in as a matter of ____; I couldn’t care ____; people breathed a sigh of ____
when witches were burnt at the ____; to tell you the ____; you should take
his stories with a pinch of ____; it was two in the morning and I was still
wide ____; I hereby pronounce you man and ____; and stop beating about
the ____. Readers who are familiar with academic prose will also effort-
lessly anticipate the missing words in there is a growing body of ____ and
the difference was not statistically ____. The fact that we can predict so
well what comes next in many word strings is indicative of their prob-
abilistic nature, and perhaps also of the holistic way we each access
these strings in our mental lexicon.
In order for learners of a second language to benefit from this
fluency-facilitating feature of chunks, chunks that are met, noticed and
learned must then be adequately entrenched in the learners’ long-term
memory. Consequently, with respect to progress in oral production,
the challenge for the learner is to build a repertoire of chunks that not
only has sufficient ‘breadth’ (meaning that it should include many indi-
vidual phrases) but also sufficient ‘depth’ (in the sense that particular
word combinations which are at least fairly frequent in spoken discourse
should be so well and durably entrenched in memory that they can be
 Introduction 11

recalled more or less instantly as needed). A deeply entrenched repertoire

of frequent and fairly frequent chunks must also be conducive to fluency
in composition of (non-literary) texts. With respect to increased profi-
ciency in reading and listening, however, it seems plain that breadth is
more crucial than depth. It is interesting in this connection that it may
be sensible for teachers to stress breadth of learning more than depth,
particularly in the case of word partnerships in which one of the slots
can be filled not by just one word but by a category of them (for instance,
cause trouble / pain / damage ...). Because deeper entrenchment takes more
time, such a shallow but broad approach would contribute to making
a chunk-learning programme more feasible in normal school-based
instruction. In any case, it seems certain that the chunk repertoire of
any individual native speaker will show marked variation in the extent
to which learned chunks are entrenched, with phrases used regularly
being retrievable from memory with greater ease than ones whose mem-
ory traces are consolidated only by sporadic encounter and rarer use.
Thus, some chunks are likely to play a crucial role as facilitators of flu-
ency in an individual’s language production while others that may be
produced less often may nevertheless stay well enough entrenched in
memory to facilitate that person’s fluency of language reception.
A third way of classifying lexical chunks is to locate them on a con-
tinuum between those that are semantically transparent and those that
are semantically opaque. A chunk whose meaning is plain to whoever
understands the words that make it up would count as transparent (for
example, make a presentation and tell a lie), while an expression that you
might find in an idiom dictionary would typically be categorized as
opaque (for example, spill the beans and hit it off with [someone]). The first
sort of chunk is often said to be ‘compositional’, which means that the
overall meaning of the expression can be worked out by combining the
meaning of the separate words it is composed of. In this view of things,
the second sort of chunk is said to be ‘non-compositional’ (or ‘non-
decompositional’) because – so the argument goes – the sum of spill ⫹
the beans does not equal the meaning ‘reveal a secret’. However, recent
studies have cast doubt on this distinction between idioms and non-
idioms on the basis of the criterion of compositionality. Specifically,
even expressions that have long been considered compositional turn
out to convey meanings that exceed the sum of the meanings of the
individual component words. Taylor (2006) gives the example of all
over, which in fact is not typically used to express the meaning ‘com-
pletely covering’ which seems to be suggested by all ⫹ over. Instead, all
over is used to express random, disorderly distribution – which is the
12 Optimizing a Lexical Approach

meaning we see in figurative uses such as your essay is all over the place.
Taylor (2006) goes on to argue that, because many more lexical phrases
than those traditionally called idioms turn out not to be thoroughly
compositional, idiomaticity in language deserves much broader recog-
nition and attention than it has so far been given. It is also important to
realize that descriptive linguists’ estimation of the semantic transpar-
ency of a given word string in their L1 will not necessarily correspond
to the perception of individual learners. For one thing, comprehension
of a so-called compositional word string evidently requires compre-
hension of its constituent words (or, more accurately, of the particular
sense the words have in that particular string). For example, if learners
do not know the words wreak and havoc, they may find cause havoc
quite opaque and wreak havoc thoroughly so – even though a descriptive
linguist with a large vocabulary and a wide knowledge of etymologies
might well characterize these expressions as perfectly compositional.
Or again, even if learners of English as an L2 ‘know’ both of the con-
tent words in make a presentation, the vagueness of make may lead them
to take the expression to mean ‘create or prepare a presentation’ rather
than ‘give a presentation’.
On the other hand, expressions that may at first strike a learner as
opaque can be made transparent by appropriate intervention by a teacher
(or materials writer). For example, it can be possible to make the idiom spill
the beans transparent to a learner simply by explaining the ‘figuration’ (in
this case, the metaphor), namely that the spilled beans (since beans are
normally kept in a container) represent secrets not meant to be ‘spread
out in the open’. It is important to note that this kind of explication
involves assigning a ‘role’ to component words and phrases – the beans
stand for information that was supposed to stay secret and the implica-
tion of a container suggests that at the start of this scenario the beans
are not exposed to view. Thus, an explanation of this kind, to a greater
or lesser extent, shows the target expression to be compositional ‘under
the surface’.6 In any case, whether or not learners initially experience
the chunks that they meet as transparent or opaque is likely to affect the
learning process. For example, word combinations that are experienced
by learners as perfectly transparent will tend to pass unnoticed (since such
combinations, by definition, will cause no obvious, immediate compre-
hension problems). But if learners fail even to notice a particular combin-
ation (and it is a fact that learners are usually intent on grasping meaning
rather than noticing precise wording), they will certainly not pause to
consider whether it is a potentially useful chunk or a rare and fleeting
combination of words. On the other hand, chunks that are experienced
 Introduction 13

as opaque may readily evoke a mental effort on the part of the learner
and/or an intervention from the teacher, and either the effort or the
intervention will have some potential to bring about a correct mapping
of a holistic meaning to the entire word string in question. A consider-
able part of Chapter 5 of this book will be devoted to the classic category
of so-called opaque, non-compositional chunks: idioms.
A fourth criterion for classifying lexical phrases, already touched on
above, is frequency of occurrence. At one extreme are chunks which
occur very frequently in everyday discourse (for example, some dis-
course organizers like and so on and for example but also conversational
‘fillers’ like sort of and you know [what I mean]). At the other extreme
are chunks that are extremely rare in current discourse (for example,
archaic idioms like it’s raining cats and dogs). Focusing on phrases at
the latter extreme is clearly likely to be a sad waste of time for virtu-
ally all learners in whatever pedagogical setting. Intuitively, it is high-
frequency chunks that merit most attention. However, there are several
reasons why frequency should not be the only criterion used in select-
ing chunks for classroom attention.
Firstly, we know from research in vocabulary acquisition that it is
high-frequency items that stand the best chance of being picked up by
learners ‘incidentally’ – that is, acquired in the context of a meaning-
focused activity such as reading for pleasure rather than during a ses-
sion of premeditated vocabulary study – which means that they may
not actually require pedagogical treatment in the first place. Such inci-
dental acquisition is much less likely when it comes to not-so-frequent
items, however. This is because it usually takes repeated encounters with
a lexical item within a relatively short span of time for it to leave a stable
trace in the learner’s memory. Provided a learner gets enough suitable
L2 exposure, the only chunks that easily meet this key condition for
incidental uptake are, by definition, ones that are highly frequent. But
if high-frequency chunks stand a fair chance of being picked up by
learners autonomously (that is, with no direct teacher intervention),
there would seem to be grounds for teaching time to be re-directed
towards chunks of medium-frequency. In considering this possibility, it
is helpful to realize that the radically differing vantage points of learn-
ers and teachers may lead to approaches to chunks which, while also
very different, may be fruitfully complementary. Thus, it may be pre-
cisely because a L 2 word combination occurs time and again in teacher-
managed classroom input even though it is not specifically targeted
that a learner who is exposed to the combination assumes it is useful
enough to merit sufficient attention for it to be learned. Teachers, for
14 Optimizing a Lexical Approach

their part, may sensibly decide to devote classroom time to particu-

lar medium-frequency chunks precisely because, among other reasons,
students are relatively unlikely to learn them on their own. We will
present evidence to back up this claim in Chapter 3.
The second problem with frequency as a criterion for choosing
chunks for pedagogical treatment is that this criterion is very difficult
to implement in a principled way. As one starts positioning chunks on a
(vertical) frequency cline, it soon becomes clear that the class of highly
frequent chunks has relatively few members (Moon, 1998a, b; Shin and
Nation, 2008). Targeting only these would help meet only the most
modest of learning aims. If the aim, though, is for learners to progress
beyond, say, lower-intermediate level, they will need to master many
more chunks than only those that are exceptionally frequent. Below
the small group of highly frequent chunks, frequency distribution rap-
idly levels off, confronting the learner with a great many chunks of
medium frequency. Differences in frequency from one chunk to the
next are so small in this region of our cline that it tends to be wholly
impractical to use frequency as a criterion in deciding which chunk to
teach at the expense of any other. In Chapter 4 we will demonstrate this
as well as other, pedagogical, considerations which can run counter to
a strict application of the frequency criterion.
Finally, as Ellis, Simpson-Vlach and Maynard (2008) have shown,
over-reliance on the criterion of frequency would lead to the inclusion
of word strings which happen to be highly frequent but are of such low
semantic coherence (that the only, for instance) that few if any native
speakers would identify these as semantic units in the first place. They
therefore advocate addition of a more ‘pragmatic’ selection criterion,
namely ‘teaching worth’, which teachers can use to balance the criter-
ion of frequency (p. 392).
In sum, there seem to be firm grounds for moderated application of
the frequency criterion in deciding which lexical phrases to teach and
which to leave for students (maybe!) to pick up on their own. The consid-
erations involved are sufficiently diverse, and in some cases subtle, that
busy teachers cannot possibly give precisely the right weighting to each
of them every time there is a need to decide about a particular chunk.
For example, merely checking on the frequency of a small set of chunks
may take more time than a teacher has to spare. On the other hand,
estimating the ‘teaching worth’ of any given chunk is at present still
largely a matter of intuitive judgement. In Chapter 4 we will therefore
argue for an additional criterial factor, ‘memorability’, which we believe
should be relatively easy for an informed teacher to estimate. What we
 Introduction 15

aim to show is this: many chunks are especially memorable, provided

appropriate pedagogical steps are taken. This ‘extra-ordinary’ memor-
ability of certain chunks is afforded by particular non-arbitrary charac-
teristics which they may have, characteristics which may be semantic
in nature, formal in nature, or both. These non-arbitrary ‘affordances’
invite implementation of teaching strategies and techniques (which we
will describe in detail in Chapters 5 and 6) that have empirically dem-
onstrated potential to lead to durable retention of chunks. Put another
way, these strategies and techniques exploit the fact that many chunks
are ‘motivated’. What we mean by this is as follows.
The ‘motivation’ of a chunk (more precisely, ‘linguistic motivation’)
includes anything about the history of the chunk, its meaning and/or
its form which has influenced the acceptance into conventional usage
of its particular combination of words for the purpose of expressing a
particular meaning. To give an example of one kind of motivation, the
idiomatic meaning of show someone the ropes (⫽ ‘teach someone how to
perform a task’) makes perfect sense – is not accidental or arbitrary – if
you know that this expression was originally used to refer to those occa-
sions when an experienced sailor would show a novice seaman how to
handle the ropes on a sailing ship. Thus, the existence in English of the
idiom show someone the ropes is partly explained, or motivated, by the
fact that seafaring was for centuries a part of British life that had great
social, cultural and economic significance. Another important kind of
motivation, or influence on the acceptance into common usage of a
particular combination of words, is that of sound repetition. We see
this in the chunk it takes two to tango where use of the word tango rather
than, say, waltz or jive has almost certainly been motivated (or influ-
enced) by the catchiness of the alliteration – the appealing repetition of
the ‘t’ sound at the beginnings of four of the words.
In Chapter 5 we will demonstrate more fully how the meaning of
many conventional figurative expressions can be shown to be moti-
vated. To give two more examples (the first concerned show someone
the ropes), a teacher can raise learners’ awareness of underlying meta-
phor themes as follows. Feel up to _ing can be shown to be motivated
with reference to the possibly universal human tendency to metaphor-
ically associate HAPPY/ACTIVE/GOOD with UP. A teacher can also simply
and briefly inform learners about the context in which a particular
chunk was apparently used in its original, literal sense. For instance,
the meaning of jump the gun can be elucidated with reference to the
image of an athlete who starts running before the starting pistol has
been fired.
16 Optimizing a Lexical Approach

In Chapter 6 we will go on to demonstrate that also the form of lex-

ical phrases (that is, their word-by-word composition) can often be
made sense of by looking not only at the interplay of meanings but also
at patterns of sound repetition. For example, we will show that alliter-
ation (along with other phonological patterns) plays a significant part
in the formation of a large number of word partnerships (for example,
health hazard, feeding frenzy, wage war, make a mess, time will tell, through
thick and thin, chop and change, first and foremost, fit as a fiddle, bring _ to
the boil, the more the merrier, he who pays the piper ...).
Additionally in Chapters 5 and 6, we will show with reference to vari-
ous empirical studies that students’ learning can benefit significantly
from a range of teacher interventions that draw their attention to such
non-arbitrary semantic and formal features of chunks as those just
mentioned. The interventions described and assessed tend to be very
brief – an important consideration, given the possibly very large num-
ber of chunks worth learning over the short span of years that make up
the formal, non-intensive learning experience of a typical, reasonably
ambitious second or foreign language student. The basis for our propos-
als are findings which show that the semantic and/or formal (especially
phonological) motivation of many chunks carries considerable mne-
monic potential – provided that teachers do something definite to help,
for it has also been found that learners are not particularly likely to be
able to unlock this potential on their own. All in all, there is good news
here: given the advantages of building a sizeable repertoire of L 2 lexical
phrases and given too the heavy burden on memory that this entails,
any techniques that effectively facilitate long-term retention must be
very welcome. In sum, one of the principal claims we will be making
in this book is that it is by targeting chunks which have non-arbitrary
properties, and by exploiting those properties to help students to remem-
ber what they have learned, that teachers and material writers can make
a big difference in how effectively their students learn chunks.

1.2 Why have we written this book?

Our plea for second and foreign language teachers (and materials
writers) to give more attention to lexical phrases is not new. Almost a
century ago, the importance of chunks in conversation was noted by
Harold E. Palmer (1925) and the learning of chunks – indeed, their
over-learning – was a prominent aim of what became known as the
Audiolingual Method of foreign language teaching (Fries et al., 1958),
although the emphasis in this method lay on frames with ‘open slots’
 Introduction 17

(for example, as _ as _; the _ the _) rather than on fixed word partner-

ships such as strong collocations. In 1983 Pawley and Syder’s seminal
paper appeared, pin-pointing lack of chunk knowledge as an import-
ant but under-investigated deficiency in the English produced by
learners whose knowledge of the language is otherwise very good.
However, the reputation of this paper took years to build. Rather, it is
Nattinger and DeCarrico’s systematic and detailed argument in their
well-known monograph Lexical Phrases and Language Teaching (1992)
that has proved to be the milestone in the growth of appreciation of
the place of chunks in language learning. Virtually all subsequent
research on phraseology in language learning owes a great debt to
the work of these two authors. Still, from our contacts with language
teachers across Europe, we have the impression that the message that
sufficient time should be invested in the teaching and learning of
multiword lexis has been most successfully conveyed to the foreign-
language-teaching (FLT) community (especially the EFL teaching
community) through three books written and published by Michael
Lewis (1993, 1997, 2000). The name Lewis has given to his proposal is
the Lexical Approach; and because we have noticed that it is now well-
known in FLT circles, at least in Europe, it is the term we have chosen
to use in the title of the present book. We want to be clear, though,
that it is not our intention to focus exclusively on Lewis’s work in our
own proposal of ways to optimize a chunk-oriented pedagogy – hence
our use of the indefinite article (‘a lexical approach’) in the title of
this book.
A second reason why we allude in our title to the work of Lewis
rather than Nattinger and DeCarrico is its wider scope. Nattinger and
DeCarrico (1992) focus on lexical phrases that fulfil pragmatic func-
tions in conversation (such as it’s been nice talking to you; how are you
doing and could you possibly ...) or which serve as discourse organizers
(by the way; having said that and in a nutshell). (Lewis would count these
as members of a large class of ‘institutionalised expressions’ [1993: 92]).
It goes without saying that phrases which fulfil pragmatic functions
are of the utmost importance and the ability to use them in socially
appropriate ways can be immensely beneficial to learners. Not only
that – these benefits can be virtually immediate for learners who are in
so-called immersion contexts. Nevertheless, it is undeniable that learn-
ers also benefit from mastering any number of so-called ‘referential’
chunks such as compounds and collocations and, in Lewis’s publica-
tions, these chunks, which he refers to as ‘message-orientated’ (Lewis,
1993: 92; see also Sinclair and Mauranen, 2006: 79–89), do receive
18 Optimizing a Lexical Approach

at least as much attention as those having pragmatic, interaction-

orientated functions. Message-orientated chunks will also feature
prominently in the present book.
The kind of learners we had in mind while writing this book are
teenagers and young adults learning a second or foreign language not
in a naturalistic setting (meaning not immersed in a target-language
environment) but instead predominantly through classroom activ-
ities complemented by additional exposure to the target language via
the media and the internet as well as, possibly, through independ-
ent reading, homework and so on. We will adopt the term instructed
second language acquisition to denote this type of learning situation.
(Incidentally, we use the word second as shorthand for any additional
language acquired after the mother tongue, be it a second, third,
fourth or nth additional language.) Lewis, although he does not say so
explicitly, appears to have had the same kind of learning situation in
mind when he launched his Lexical Approach (henceforth LA). Many of
Nattinger and DeCarrico’s proposals, on the other hand, are more rele-
vant for learners in naturalistic settings, non-native speakers attend-
ing college in an English language environment, for example (1992:
131–52).
But – if highly acclaimed books recommending a lexical-phrase
approach to instructed second language acquisition (henceforth ISLA)
are already available – why have we decided to write the book you are
now holding? The reason is two-fold. Firstly, the challenge that learners
face in building a sizeable repertoire of chunks is so enormous, espe-
cially in terms of the heavy burden on memory entailed, that there
is a pressing need for learners to be given more help than the current
approach offers them. Secondly, in the years that have passed since the
publication of Lewis’s LA books, new findings and insights from applied
and cognitive linguistics research point to ways in which the received
version of the LA can be improved. In short, learners have long needed
better teaching of chunks; we hope to show that this is now feasible,
provided, that is, that three tenets of the LA in its present form are
largely jettisoned.
The first tenet we question is this. Given the inevitability that most
chunk acquisition will have to take place outside the classroom, class
time should be devoted to teaching strategies that foster learner auton-
omy rather than to teaching individual items. Our reaction is as follows:
in regard to acquisition of chunks, over-reliance on learner autonomy
is naïve. We will argue that accelerated chunk-uptake depends in large
part on the initiative of teachers and materials writers.
 Introduction 19

The second tenet we question is the following. Teachers can create

opportunities for learning by providing digestible input, but whether
their students actually commit any of the input to memory is up to
the students themselves. Our reaction to this second tenet is that the
received version of the LA gives far too little attention to memory.
Making students aware of the phraseological dimension of language
and alerting them to the presence of chunks in texts is one thing; tak-
ing steps to help them remember these chunks in quite another. We will
demonstrate effective means of accelerating uptake of lexical phrases
into long-term memory.
The third tenet we want to qualify is as follows. Lexis is arbitrary,
and so there is no point in trying to explain why any particular word
combination has come to be understood, used, and conventionalized as
a chunk. Teachers and learners should refrain from looking for reasons
behind any given word partnership and accept that ‘this is just the way
it is’ in the target language. Our reaction to this tenet is that it is too cat-
egorical about language being fundamentally arbitrary. We will show
that many word partnerships are not arbitrary – that their motivation
can be shown. We will show as well that teachers and materials writers
can easily exploit the motivation of many chunks in ways with strong
potential to result in durable memory traces.
Although we will elaborate on our reservations about these three
tenets in later chapters, a preview of our full rationale may be in order
here. The LA in its present form proposes classroom activities and exer-
cises that raise learners’ awareness of the importance of chunks. The
central strategy is pedagogical chunking (Lewis, 1997: 54–5); its essence
is the encouragement of learners to notice chunks. That is, students
should first of all be alerted to lexical phrases encountered in authentic
texts and then encouraged to make records of these chunks in vocabu-
lary notebooks adapted to accommodate this kind of lexis. Lewis (1993:
122–3; 1997: 45–7) recognizes that the quantity of lexical phrases
that qualify as good targets for learning by far exceeds what can be
acquired on a normal, non-intensive language course. His advice is
to help students develop strategies for the recognition and recording
of chunks in samples of L 2 they encounter not just in the classroom,
but outside it too. In more detail, his recommendation is to expose
students to substantial quantities of listening and reading materi-
als in the classroom, make them conscious of the chunks that occur
in these materials by helping them ‘chunk’ texts ‘correctly’, that is,
notice the authentic chunks they contain (Lewis, 1997: 58). We share
the belief that it is important to use some class time to raise learners’
20 Optimizing a Lexical Approach

awareness of the prevalence and importance of lexical phrases. And it

seems intuitively reasonable to expect that – after their awareness of
chunks has been sufficiently raised through, for instance, practice in
‘chunking’ – students, autonomously, will strategically apply what they
have learned to samples of the target language they encounter outside
the classroom. The problem with this ‘learner autonomy hypothesis’
is that it is rather difficult for learners to identify authentic chunks
without expert help. Specifically, it is hard for learners to recognize
any particular word sequence as a chunk unless they have encoun-
tered – and noticed – it at least a couple of times before. But the vast
majority of chunks that could be useful targets for learning beyond
lower-intermediate level are not at all likely to occur frequently enough
for this to happen inside the time span where an earlier encounter is
still fresh in memory. In Chapter 3 we will show how even quite com-
mon collocations, such as make an effort, are not likely to be encoun-
tered often enough during extensive reading to meet the conditions for
incidental uptake to occur. As for chunks that upper-intermediate or
advanced students might find useful additions to their English reper-
toires, the chances of multiple encounters in a short time span are very
slim indeed. Take, for example, the discourse organizer when it comes
to. In the approximately 8400 words you have now read of this book
chapter this is used only once. Exact-word Google searches suggest that
oddly enough is 15 times less frequent than when it comes to. It should be
easy from this to form a general impression of what large stretches of
text vocabulary – and time in which forgetting may occur – might lie
between individual occurrences of chunks such as take precedence over,
which is 25 times less frequent than when it comes to. It therefore stands
to reason that the most likely benefit to students of heightened aware-
ness of phraseology is an ability to notice word strings which they
already know are chunks, rather than an ability to identify novel word
strings as chunks and add these to their repertoires. The advantage of
such awareness-raising probably lies, therefore, in consolidation rather
than in expansion of knowledge.
Besides the recommendation that students ‘chunk’ texts and record
in notebooks the lexical phrases that they find, the LA involves
stimulating awareness of lexical phrases in more ‘direct’ ways, for
example through the presentation of particular word partnerships
in ‘collocation boxes’ and ‘pattern displays’ (Lewis 1993: 126) as well
as through a variety of matching and sorting exercises (Lewis 1997:
86–106). Use of similar formats is suggested also for ‘rehearsal’, that
is, for testing and/or consolidating students’ ability to recollect the
 Introduction 21

chunks they have previously encountered. Still, Lewis (1997: 47)

emphasizes that:

[...] class time is better spent helping learners develop strategies for
dealing with unknown items they meet when listening or reading
[...] rather than laborious practice aimed at consolidating individual
items. Similarly with the learners’ active lexicon: class time is better
spent raising awareness and encouraging effective recording of pat-
terns, rather than too much concentration on individual items.

While no-one will deny that strategy development can be useful, recent
findings from vocabulary learning research indicate that, unless indi-
vidual lexical items are given a good deal of consideration, they are not
likely to be retained in long-term memory. Relatively fleeting exposure
to large numbers of chunks may leave too ephemeral an impression of
any individual chunk for a satisfactorily increased rate of acquisition to
occur. Yet, surely, a satisfactory increase is what the LA is supposed to
be all about.
In actual fact, the present form of the LA is virtually silent on the
question of how learners are supposed to go from noticing chunks in
texts, and/or viewing and manipulating them in collocation boxes, to
storing these chunks in long-term memory. Management of this final,
crucial step is, ultimately, left to the students themselves. Lewis (1997:
38) says the following:

Expressions may be glossed in dictionary-like fashion with the sugges-

tion that learners should look for equivalents in their own language.
A major departure from traditional methodology is the explicit sug-
gestion that the teacher should not dictate what is then done with
the list. That’s fine, but what do you do with these expressions? How do
you ensure that they learn them? teachers ask. The simple answer is
that you can’t. Writers and teachers should ensure the accuracy of
carefully chosen expressions, grouped in ways likely to aid retention;
writers can ensure suitably sized, well-chosen and well-arranged lists;
teachers can manage the time devoted to the lists, both in class and
by suggestions for out-of-class activity; they can arrange class activ-
ities to generate and maintain interest while meaning is explored,
often through the search for L1 equivalents and the importance of
contextual constraints – then it’s up to individual learners, who may
or may not wish to learn particular expressions, and may or may not
succeed in retaining them. Accepting learner autonomy also means
22 Optimizing a Lexical Approach

accepting that teachers cannot guarantee what is learned. The teacher

must be content and fulfilled by the role of learning-manager.

It is true, of course, in the strictest sense, that there is no way of

controlling precisely what chunks (if any) students will remember nor
any way of making learners adopt stipulated mental processes for doing
so. On the other hand, the experimental evidence we will survey in
Chapters 5 and 6 shows that appropriate interventions by teachers and
materials writers can markedly increase the chances that particular, tar-
geted chunks will be retained in memory. This is important because
vocabulary-learning research tells us that noticing alone seldom results
in uptake. In order for lexical items most reliably to be lifted over the
threshold from short-term to long-term memory, some form of mental
elaboration (or deep, rich processing) is required. That is, the learner
needs to engage in a cognitive activity with regard to a particular lexical
item that goes beyond this item merely being noticed. We will describe
how the non-arbitrary properties of chunks can be exploited in rela-
tively simple ways to put students on paths of elaboration that have
been shown to foster learning.
The main reason why students are largely left to their own devices
when it comes to the challenge of remembering lexical phrases under
the current form of the LA must be that Lewis (along with many others)
has a firm belief in the arbitrariness – the unmotivated nature – of
chunks. This belief leaves too little room for ‘insightful’ learning as an
alternative for rote-learning:

Words, Collocations and Expressions are arbitrarily sanctioned by

particular native speaker communities. That is easily said and briefly
summarized, but it lies at the heart of many individual responses
teachers will make on a daily basis. What is arbitrary cannot be
explained; any apparent explanation is at best a waste of time; at
worst it causes confusion later. [...] teachers must abandon many of
the folk-linguistic ‘explanations’ they sometimes offer learners. Few
of these are true, and most will confuse rather than help. (Lewis,
1997: 196)

We disagree: a sizeable proportion of chunks have non-arbitrary prop-

erties which can be used as entries to insightful, accelerated learning.
These entries have been blocked by acceptance of the tenets of the LA
discussed above; replacement of these tenets by others can open them
up and lead to a radically optimized lexical approach.
 Introduction 23

And this is not all. At the level of method and technique (rather than
that of assumption and approach), recent research into the acquisition
of L2 vocabulary of all types (not just chunks but also single words)
provides grounds for re-appraising the current heavy reliance in the
LA on such means of presentation as collocations boxes and such exer-
cises as item matching. In Chapter 7, we begin by touching on such
thorny issues as the potential confusion caused by the presentation of
negative evidence and the simultaneous presentation of several novel
forms. In addition, we will raise the question of what else is needed to
complement the stages of noticing and elaboration if students wish to
approximate to native speakers’ chunk mastery, for example at the level
of productive fluency.
We hope this book will encourage further research into both descrip-
tive and pedagogical approaches to phraseology. A great deal of work
remains to be done. There are even important issues about which ques-
tions have hardly even been asked, let alone answered. For example,
astonishingly little has been said in the literature about the usefulness
of a lexical-phrase approach beyond EFL. Might English be exception-
ally suited to a chunk-oriented pedagogy? In Chapter 8, our final chap-
ter, we therefore sum up a study carried out with a view to gauging
the applicability of a lexical approach in the teaching of Spanish as a
foreign language. This is a language many of whose lexical phrases are
morphologically more variable (because of inflection) than in English.
Another big, under-addressed question is this: how can learners’ chunk
knowledge be assessed (and formally tested) in a reliable, valid and user-
friendly way? The final sections of the book will explore some of the
work that is being done in this area.
2
The Contribution of Chunks to
Acquisition and Proficiency

2.1 Comparing the nature and necessity of

L1 and L2 chunk acquisition

Young children are exposed to streams of speech in which they try

to discern recurring bits and determine their meaning. These bits are
not necessarily words, since the young child does not yet discern word
boundaries (Peters, 1983). Rather, the bits that young children discern
are of various lengths and often consist of several words. For example,
it is known that meaning-bearing word sequences such as compounds
and noun phrases like cup of tea are learned by young children as unana-
lysed wholes (Cruttenden, 1981). It thus appears that young children
acquire semantic units consisting of two or more words holistically and
realize only at a later stage that these units break down into a certain
number of single words.
According to Wray (2002), the fact that pre-literate children do not
discern word boundaries is an advantage for them precisely because it
leads them to process formulaic sequences holistically. This helps them
build a very large repertoire of chunks from a young age. According to
Wray’s (2002) model (of first language acquisition), the benefits of hol-
istic chunk acquisition are much less accessible for teenagers and adults
trying to learn a second language. This is because, as children become
literate, they adopt a more analytical mode in processing the language
that they are exposed to and so become increasingly inclined to pro-
cess chunks as sequences of single words rather than as (multiword)
units. This development is believed to hamper the storage of chunks in
the mental lexicon as prefabricated wholes (Wray, 2002: 206–10). Even
so, all remotely typical native speakers have accumulated an enormous
stock of these items by the time they reach adulthood, and new ones

24
 The Contribution of Chunks 25

continue to be learned even then. There is now little dispute that these
chunks play certain beneficial roles. Let us look at three, of which the
first two are particularly uncontroversial.
Firstly, messages couched in conventional forms are especially likely
to be deemed appropriate by members of the speech community in
question. For example, children tend to learn very early on that Could
I have ... is usually much more effective than I want ... as a speech act for-
mula if they want to get a biscuit or a present.
Secondly, prefabricated chunks facilitate L1 and L2 fluency, both in
language production and in comprehension. Near the end of this chap-
ter, we present recent evidence for this statement with respect to L2.
Before that we need to say, with respect to the role of chunks in oral
fluency, that we are not going to be concerned with all of the several
kinds or facets of oral fluency that have been proposed – the ability,
for instance, to speak at length in a witty and entertaining fashion or
to engage in confident, uniformly coherent and lengthy explication of
deep intellectual problems (Fillmore, 1979/2000). Our two main con-
cerns lie elsewhere.
One concern is the likelihood that speakers enhance their fluency in
normal conversation by planning out what they are going to say next
while they are still speaking, and that this is very difficult to do with-
out wholesale use of previously memorized chunks (Pawley and Syder,
1976/2000; Wray, 2002: Chapter 5).
A second matter of interest is the vital component of both recep-
tive and productive fluency known as ‘automaticity’ (Schmidt, 1992;
DeKeyser, 2001; Segalowitz, 2003; Segalowitz and Hulstijn, 2005).
With respect to listening, this consists in the ability to recognize and
understand words and phrases with extreme rapidity. (The ability of
individuals to do the same when reading, of course, varies greatly,
with complete illiterates not being able to do it at all.) With respect
to speaking (and, to a degree, writing), automaticity is manifest in the
everyday discourse of ordinary native and other expert speakers who
‘smoothly, rapidly and accurately’ (Segalowitz, 2003: 384) produce
streams of words and phrases in which there are definitely pauses but
ones which, in general, seem neither too long nor out of place. It is
important, when thinking about oral fluency (and, to a degree, flu-
ency in writing), to distinguish between performance fluency which, as
shown further below, can be measured by surface features such as the
density of false starts, and cognitive fluency, which has to do with the
‘efficiency of the operation of the cognitive mechanisms underlying
performance’ (Segalowitz, 2000: 202).
26 Optimizing a Lexical Approach

In the study of cognitive fluency, there has been on-going controversy

about whether it is fundamentally a matter of automatized processing
(in particular, accelerated on-line composition of words into phrases) or
a matter of swift retrieval of multi-morphemic/multi-lexical ensembles
from memory, with automatization being the process whereby strings
of individual words become stored as ensembles in the first place. We
will return to the issue of automatization in Chapter 7. Now though,
it might be useful to consider how the asymmetry of production and
reception is relevant to L2 fluency.
Suppose, for example, that two or more common chunks are effect-
ively equivalent in meaning (play a part and play a role, for instance). For
receptive purposes it seems a good idea to learn them both sufficiently
well that they can be recognized and understood in the, possibly rather
distant, future. For productive purposes, on the other hand, it seems
efficient for only one of the chunks to be learned so well that it can
be retrieved more or less instantly when it is needed for fluent produc-
tion. The time that might be spent in learning both chunks to this lat-
ter degree could be better spent. In any case, it seems likely that truly
holistic retrieval for language production is influenced by the degree
to which a chunk can vary, such that those chunks which vary least
are most susceptible to holistic processing. Of course, the least variable
chunks of all are those which are wholly free from inflection, substitu-
tional variation and alteration of word order – for example, such as, in
fact, because of, cut and dried, far-fetched and in terms of.
At the level of language reception, however, at least one kind of vari-
ation – that resulting from inflection – is much less likely to affect a learn-
er’s success rates in using the beginnings of chunks to predict how they
continue. For example, it does not appear to us that any particular one
of the underlined alternatives makes guessing the bracketed words in the
following examples either harder or easier: never put/puts a foot [wrong];
is/was/will be time to throw in the [towel]. What does facilitate guessing are
clues from the mutual constraints of word partnerships (as in the police
suspect that Peter committed __ [suicide; or some other word for a crime or
blunder]) and clues from preceding text and the situation of use.
Although Wray (2002) suggests that (young) children acquire L1
chunks more readily than (relatively) older learners acquire L2 chunks,
she acknowledges that L2 learners may also reap limited fluency benefits
from having a profound procedural knowledge of word chunks. In this
case, Wray argues, chunks are not retrieved from memory as prefabri-
cated chunks; rather, they are assembled by the learner at high speed as
‘proceduralized strings’ (2002: 189–90). This position resembles one of
 The Contribution of Chunks 27

the two strands in theorizing about automaticity we mentioned above,

one which sees fluent use of L2 lexical phrases as being brought about
by accelerated on-line composition of the phrases rather than by their
retrieval as complete ensembles. If this is so, language teachers have
two options: (a) to encourage their students to switch off their analyt-
ical mode of processing so that they can take in chunks as unanalysed
wholes as in early childhood L1 acquisition, or (b) to help their students
make the best of the analytical processing mode by attaining highly
consolidated procedural knowledge of known chunks. The first option
seems to be recommended by Michael Lewis (1997: 78), but little or
no practical advice is currently available for teachers who wish to pre-
vent their students from ‘breaking up’ the chunks they meet (see also
Fitzpatrick and Wray, 2006, on the difficulties involved). Accordingly, in
Chapters 5 and 6 of our book we demonstrate ways of ‘making the best’
of option two: if it is true that adult L 2 learners are drawn to ‘analysing’
novel lexical phrases rather than taking them in holistically, then let us
find ways of channelling this inclination of learners so that at least the
benefits at the level of procedural knowledge are maximized. Moreover,
one may call into question whether holistic retrieval of L2 phrases actu-
ally is an unattainable objective. What if the second strand in automa-
ticity theory (that is, the strand according to which holistic retrieval
of a chunk could be the end product of a process of often repeated
co-retrieval of its basically individually stored components) were cor-
rect? In that case, could increasingly profound procedural knowledge
of a phrase (reflected in accelerated co-retrieval of its components) not
eventually lead to its storage as an ensemble?
In recent models of L1 acquisition, chunks have a third function, which
is to provide the raw material out of which children build their ‘grammar’.
Although generative linguists have a different view (see the responses of
two generativists – Major, 1996 and Ioup, 1996 – to N. Ellis, 1996a; and
see N. Ellis, 1996b, for his reply), psycholinguists now generally believe
L1 acquisition happens as follows. The phrases young children acquire
serve as exemplars for the development of patterns, or ‘constructions’,
through processes of analogy and extension. In other words, the gram-
mar ‘emerges’ from those formulae that children encounter, and imitate,
first (Tomasello, 2003). These seem to be, almost exclusively, formulae
which children hear most often. Nick Ellis (2008a) sums up the research
findings as follows: ‘The usual developmental sequence is from formula
to low-scope slot-and-frame pattern, to creative construction.’ One can
imagine, for example, a child first acquiring as soon as daddy comes home;
then recognizing analogies with as soon as you’ve tidied up and as soon as
28 Optimizing a Lexical Approach

you can; then, on hearing as fast as I can, recognizing a quite different

possibility for extension; and then eventually using the as ... as ... pat-
tern in a flexible fashion. Some of Nattinger and DeCarrico’s (1992)
techniques for teaching L2 lexical phrases follow this same pattern. For
example, students are first given formally simple exemplars of speech
acts (Could you ...) and are then taught how these can be expanded (Could
you possibly ... → I was wondering if you could possibly ... → I was wondering
if you couldn’t possibly ...).
Usage-based models of language (Barlow and Kemmer, 2000; Croft,
2001; Goldberg, 2006; Langacker, 2000) argue that it goes against both
logic and evidence to postulate either of the following: a top-down
grammar such as Chomsky’s Universal Grammar and a ‘dedicated’ lan-
guage acquisition device, as distinct components of a child’s general
cognitive apparatus. It is now believed by many, especially in the para-
digm of Cognitive Linguistics (CL) that children acquire their mother
tongue ‘bottom-up’ or, more precisely, ‘from chunks up’. In this view of
the matter, a large portion of the acquisition process is simply a matter
of imitation and generalization of what is encountered time and again
in the input (Tomasello, 2003). Again, it follows that it is the highest-
frequency elements that stand the best chance of being noticed, under-
stood, learned, remembered, and employed as exemplars (see N. Ellis,
2002, for a survey of frequency effects in language and language acqui-
sition). N. Ellis and Larsen-Freeman (2006: 565) sum up this theory as
follows:

Language acquisition, and language representation too, is exemplar-

based. [...] The knowledge underlying fluent, systematic, apparently
rule-governed use of language is the learner’s entire collection of
memories of previously experienced utterances. These exemplars are
linked, with like kinds being related in such a way as to resonate as
abstract linguistic categories, schema, and prototypes. [...] Linguistic
regularities emerge as central-tendencies in the conspiracy of this
data-base.

Wray (2002) has pointed out an apparent problem with this ‘emer-
gentist’ view: some formulaic sequences do not seem likely to serve as
exemplars for analogy and generalization. The following examples of
chunks may demonstrate the problem. The two adjectives (or rather,
parts of chunks which the learner has now interpreted as adjectives)
thick and thin in through thick and thin are not followed by a noun. The
noun ball in the idiom the ball is in your court is never pluralized. The
 The Contribution of Chunks 29

noun tables in the idiom turn the tables on someone has a plural form for
no apparent reason. The genitival for heaven’s sake! has no correspond-
ing, natural-sounding of-phrase (*for the sake of heaven) unlike other
such genitives. In the idiom play second fiddle, the article in the noun
phrase is missing and in its figurative usage the expression does not
seem natural in the passive voice (?the second fiddle in this firm is played
by Miss Marple). And so on.
One explanation for this apparent paradox is offered by Bybee and
Thompson (2000), who make a distinction between ‘type frequency’
and ‘token frequency’ in the samples of language a learner is exposed
to. For example, the plural marker -s is encountered in countless ‘tokens’
(i.e. individual uses) of many different nouns and is thus highly fre-
quent as a ‘type’. The plural marker -en, by contrast, is used in just a cou-
ple of nouns and, although a young child is likely to hear many tokens
of children (but not oxen), the -en plural marker as a type is not frequent
enough for a child to accept as a good form to generalize from. That is,
it is high type frequency which determines the emergence of regular
patterns even though some irregular forms (such as children) may sur-
vive by virtue of their high token frequency (while less common ones
are likely to be replaced by forms that are regular). We are aware of no
reason why this ‘preservative effect’ of high token frequency would not
apply to chunks as well – viz. the moderately frequent, grammatically
irregular be that as it may (2,940,000 Google hits in October 2008).
In any case, Wray offers an additional explanation for the fact that
children appear not to generalize language patterns from ‘deviant’
forms: the primary mode of children in processing formulaic sequences
is the holistic mode and consequently the analytical mode is switched
on by children only when there is a need for it. That is,

[...] children will simply analyze whichever string needs analyzing, to

the extent that they need to, and no further. [...] Needs-only analysis
entirely turns the tables on the question of irregularity in formulaic
sequences. The sequences that are irregular in the adult language
are the ones whose usage is such that they do not invite analysis.
Because they are unanalyzed, they are able to retain obsolete vocabu-
lary and structures. (Wray 2002: 131)

As a result, adult native speakers are not normally conscious of the pres-
ence of irregularities in the expressions they use. To be clear, Wray’s
explanation applies mostly to what Lewis (1993) calls ‘institution-
alized expressions’, a class of chunks that includes idioms, proverbs,
30 Optimizing a Lexical Approach

standardized similes, clichés and social routine formulae, all of which

serve pragmatic purposes. But it is message-orientated chunks that are
most likely to provide the raw material for exemplar-based grammar
acquisition. Because this class includes such high-frequency types of lex-
ical phrases as compounds (such as noun compounds), verb–noun col-
locations, adjective–noun collocations and slot-frames (for instance, _er
than), children, by analogizing from such highly common types, are
well able to work out the conventional L1 patterns of, for instance, com-
pounds, conjugations and adjective–noun agreement (if any; adjective–
noun agreement is of course absent in English) as well as discover which
elements fill such frames as ‘it ⫹ take (⫹ N) ⫹ time (⫹ to-inf.)’, as in It
took us only three months to write this book.
This does not mean that native speakers are never, in any circum-
stance, capable of analysing an institutionalized expression. While
Wray (2002: 132) stresses that young children cannot and adults gen-
erally do not do this, she observes that an adult might deploy ‘analytic
mechanisms ... if, for instance, the individual had an interest in etymol-
ogy or the origin of idioms’. For instance, it is not so difficult to imagine
a situation in which an adult native speaker becomes aware of the fig-
uration in the idiom the ball is in your court and then creatively extends
it, as in I was hoping my wife would decide, but she’s cleverly dropped the ball
back into my court; and it’s not just one ball, when you stop to think about
it. We may safely assume, though, that this kind of utterance is rare, for
the reason that assembling such an extended version of the idiom is
bound to take longer than simply retrieving the canonical form of the
idiom from memory as a prefabricated chunk.
The evidence that acquisition of L1 chunks is what enables acquisition
of patterns of L1 grammar may tempt one to hope that things work simi-
larly in second language acquisition (SLA), including ISLA: learn chunks
and get grammar for free (Hill, 2000: 52; Lewis, 1997: 90). A good deal
of optimism seems warranted in the case of naturalistic SLA, for example
in the context of long-term immersion programmes, although it is by
now well known that such forms of naturalistic SLA may need to be
supplemented with non-naturalistic, form-focused instruction when,
besides fluency, highly native-like (accurate) expression is a programme
goal (N. Ellis, 2008b; Doughty and Williams, 1998; Swain, 1985). But,
in the case of instructed SLA (in non-intensive classroom-based situa-
tions, for instance), there are at least three grounds for broad scepticism.
Firstly, the amount of exposure to the target language that students
receive does not come close to that experienced by native speakers, with
the consequence that useful exemplars tend to be too scarce to induce
 The Contribution of Chunks 31

generalized patterns. Secondly, L2 students have already mastered the

grammar of their mother tongue, which is likely to cause interference
during ISLA, especially if so-called ‘negative evidence’ (of the non-use
of a given L1 pattern in the target language) is lacking.7 Thirdly, if we
follow Wray (2002) – but below we explain why we feel hesitant about
following her all the way – we would expect L2 students (except for quite
young ones) to have left their primary, holistic processing mode behind.
Conceivable consequences of this include, for instance, a tendency to
employ grammatically irregular chunks as exemplars from which to
generalize or a tendency to ignore repeated evidence in input that gen-
eralizations which have been made are faulty. In short, bottom-up proc-
esses that many now believe sufficient for L1 acquisition may need to
be complemented by some form of grammar instruction in contexts of
ISLA. The amount of investment in whatever kind of grammar instruc-
tion plainly ought to depend on learner needs, learning objectives and
learning opportunities. Given the fact that many chunks appear well-
suited for service as exemplars of grammar patterns to which that we
may wish to draw our students’ attention, there is no obvious reason
why students’ needs to learn grammar and to learn chunks cannot
often be addressed at the same time.
Very largely, we accept Wray’s (2002) model, our main reservation
being that we believe that the difference between the acquisition of
L1 and L2 chunks may not be as great as she claims. Firstly, it seems
that native speakers continue to expand their stock of lexical phrases
beyond the age when they learn to read and write. If they do, and if
adult language processing is strongly analytical, then strongly analyt-
ical processing does not remove the possibility of learning chunks. If all
this is true, then doubt is cast on Wray’s (2002) contention that adop-
tion of the analytical mode is the principal reason why chunk learning
is very difficult in SLA. Put another way, if it is literacy that triggers
a more analytical mode in chunk processing, then the acquisition of
chunks by teenagers and adults should not be so very different in L2 than
in L1. What is more likely to differ between teenagers’ and adults’ chunk
learning in L1 and L2 are the opportunities to learn chunks in terms of
amount of exposure one gets to the language in question and in terms
of the opportunities for interaction with expert speakers.
Secondly, our experience with using listening input in our language
classes tells us that it is not at all uncommon for students (who are
already perfectly literate in their mother tongue) to substitute (non-
existent) single words for unfamiliar multiword expressions during tran-
scription exercises (a.k.a., dictations). Additionally, we find that some
32 Optimizing a Lexical Approach

of our students (Dutch native speakers, for instance) tend to write com-
pounds that are written as two separate words in English (such as traf-
fic lights, door handle and phone number) as single words (possibly due
to interference from L1 spelling conventions). Both these categories of
observation suggest that students have construed the English chunks
as single semantic units and are unlikely to mentally break them down
without prompting.
The result of Wray’s (2002) model of chunk acquisition is a dual
system in which holistic and analytical processing co-habit. The hol-
istic mode caters for routine interactional situations while the analyt-
ical mode allows for the more creative exploitation of one’s linguistic
resources (Wray, 2002: 278). This chimes well with the model proposed
by Skehan (1998), who distinguishes between a mode of language pro-
cessing that facilitates real-time language use and a mode that caters for
planned language use. In production, language users can either take the
fast lane, where fluency is fuelled by unanalysed chunks readily avail-
able in memory, or they can take the slow lane, where they have time to
reflect and plan what they are going to say or write. It also chimes well
with a ‘memory-based’ view of automaticity.
Let us now return to the topic of memory – not, this time, in relation
to automaticity and storage format but rather in connection with the
conception of memory in linguistic theory. Given the complexity and
seeming intangibility of memory – especially before the quite recent
invention of a panoply of hi-tech instruments for making the tangibil-
ity of these operations increasingly evident – linguists have been little
different from laypeople in the extent to which their broadest accounts
of memory have been metaphorical. One metaphor employed particu-
larly often portrays human memory as a container in which things
may be stored and from which they may be retrieved (Draaisma, 2000;
Roediger, 1980). Where words are concerned, the common term for the
container is (mental) lexicon. This lexicon seems all too easy to visual-
ize as a rather ordinary dictionary that happens to contain a large part
of a speaker’s knowledge of a language – that is, not just all its words
but also information about how they are pronounced. The only thing
deemed utterly necessary in order to complete this store of knowledge is
another book that contains the rules for combining words into phrases,
clauses and so on. This ‘dictionary ⫹ grammar book’ conception of lan-
guage is clearly a folk-model, but Taylor (2002, 2006) demonstrates that
it has exerted a powerful influence on linguistic theory to this very
day.8 Certainly, this conception is straightforwardly consistent with a
belief that the mental dictionary, besides single words, lists a relatively
 The Contribution of Chunks 33

small number of word combinations that behave almost exactly like

single words (such as jack-in-the-box) but does not list word combina-
tions that show much more variation than, say, a typical noun (cup/
cups) or verb (fall/ falls/ falling/ fallen/ fell). Such a dictionary would lack
an entry for any of the huge number of chunks that are relatively vari-
able (for example, fall/ falls ... into disrepute and as soon/big ... as). It was
therefore probably not at all accidental that the generative theories
of Chomsky and his successors for decades took little or no account
of such expressions, generative theories being ‘grammar book ⫹ dic-
tionary’ theories par excellence. But even now that the importance of
chunks is acknowledged by (some) generativists of long-standing (see
Jackendoff, 2003: 178), the dictionary-in-the-mind metaphor is still
far from dead. Contemporary discussion of the LA is likely to include
portrayal of the memory of the native speaker as a very fat dictionary
replete with entries both for words and for chunks as opposed to the
thinner dictionary of the L2 learner which includes mainly words rather
than chunks.
Even though it obscures important aspects of both memory and lan-
guage, this dictionary metaphor raises interesting questions. What
could the chunk entries in the native speaker’s lexicon look like? As
has been explained above, many chunks are not fully fixed. Many are
subject to rules of inflection (make/ makes/ making/ made a mistake/
mistakes) or word-order changes (the biggest mistake I’ve ever made) and
many others allow for substitutions of constituent words (conduct/ carry
out/ do a study). Would the mental lexicon list as the prefabricated ver-
sion of a chunk only the version that a speaker might recognize as the
so-called ‘canonical’, or basic, form (bake a cake), leaving the other forms
(bakes/ baked/ baking a cake) to be generated ‘procedurally’ rather than
ever retrieved as wholes? Alternatively, would all the various forms be
included so as to make it possible for each to be holistically retrieved?
Either way, it seems that the acquisition and usage of chunks must be
more problematic in the case of languages that have a considerably
richer inflectional grammar than English does.
Having broadly surveyed the ways in which knowledge of L2 chunks
has been claimed to be beneficial, let us return to the benefit of improved
fluency, for of all the apparent benefits of learning L2 chunks touted by
the LA, improved fluency may be the most attractive for students. Let us
turn first to the matter of how speech can be judged for level and char-
acter of fluency. Drawing on R. Ellis (1990), Lennon (1990), Robinson,
Ting and Unwin (1995), Skehan (1998), Skehan and Foster (1999) and
Wiese (1984), R. Ellis and Barkhuizen (2005: 156–8) propose a classifica-
34 Optimizing a Lexical Approach

tion of factors in oral fluency along two dimensions: temporal variables

and hesitation phenomena. Temporal variables include speech rate,
number of pauses, pause length and length of run (between pauses).
Hesitation phenomena include false starts, repetitions, reformulations
and replacements (when a word is immediately replaced with another).
Since the impact of L2 chunk knowledge on fluency has been so lit-
tle investigated, it is not surprising that specific indicators of fluency
such as the above are seldom singled out in connection with chunk use.
Usually it is a rather impressionistic appreciation of fluency that is used
to estimate the contribution of chunks. One exception is a study by
Dechert (1983), who asked L 2 learners to re-tell a story and who found
some evidence that use of chunks was associated specifically with
longer runs of speech.
Indirect evidence of the fluency-facilitating role of chunks in oral lan-
guage production comes from Kuiper (1996), who compared the use of
formulaic language across different spoken genres. Some of these genres
(such as horse-racing commentaries) derive from situations that require
a high speech rate to keep up with an often rapid cavalcade of events,
while others (such as cricket-game commentaries) originate in situa-
tions where the speaker is under much less time pressure. Kuiper found
substantial numbers of formulaic phrases in all genres, but the fast-
paced genres contained by far the most. Further, it was when relatively
hectic phases were being commented on that chunks occurred with the
greatest density, something that was observed even in the most ‘laid
back’ genres (for example, cricket commentary on times when players
were running was found to have the greatest density of chunks). This
suggests that speakers tend to rely on chunks especially when there is a
pressing demand for fluency.
Additional evidence that chunks facilitate fluency in speech is found
in studies of the phonology of chunks (reviewed in Bybee, 2002). Firstly,
chunks are typically produced in a hesitation-free fashion; that is, when
hesitations and pauses occur in speech, they tend to occur between
chunks and not inside them. Once a chunk is started there is no hesi-
tation in completing it. This is consistent with the view that recall and
production is both automatized and holistic. Secondly, chunks are spo-
ken faster than non-chunks, with high-frequency chunks tending to
be articulated with the most ‘levelling’ (that is, with vowel reduction
and consonant assimilation). These apparent frequency-effects, accord-
ing to Bybee (2002: 217), suggest that ‘previously experienced colloca-
tions are processed as a single neuromotor unit and that the phonetic
reduction evident in such collocations is due to the automatization
 The Contribution of Chunks 35

that comes from repetition of neuromotor sequences’. Native speakers

should generally be able to recognize these chunks despite their phono-
logical streamlining because, again, it is precisely these chunks that
they should know best on account of their high frequency. Second lan-
guage learners, though, are unlikely to be so lucky. Since it is especially
function words (such as articles and prepositions) that are reduced most
in authentic articulation of these chunks, it may take extra effort (and
perhaps multiple repetitions) for learners to be able to extract these
chunks from heard discourse in a complete and accurate form (see also
N. Ellis, 2008b).

2.2 Evidence that L2 learners benefit from

chunk knowledge

A series of investigations of a possible correlation between oral fluency

and L2 chunk knowledge have been reported in Boers et al. (2006),
Eyckmans (2007) and Stengers (2009). In each of these studies, one
group of upper-intermediate students was given EFL instruction that
included a large amount of ‘pedagogical chunking’, as recommended by
Lewis (1993, 1997), while a same-level parallel group explored the same
course materials but with a focus on single words rather than chunks.
After several months of this differential treatment, the students were
asked to do a speaking task, which was recorded. The students’ oral pro-
ficiency generally and their fluency in particular was then assessed by
‘blind’ judges (fellow EFL teachers who were not aware of the experimen-
tal set-up). Other blind judges were asked to count the lexical phrases
used by each of the students. The resulting ranking of the students in
terms of their oral proficiency scores was compared with their ranking
in terms of the numbers of chunks they had used. For each experi-
mental data set, the strength of association between the students’ oral
proficiency ratings and their chunk use was determined by use of the
Spearman Rank Correlations test. The principal aim of these experi-
ments was to estimate the effectiveness of a lexical-phrase oriented
pedagogy (see Chapter 3). What we are interested in at this point is
whether these studies cast light on the question of whether L2 chunk
use raises the likelihood that learners will be perceived as proficient and
fluent L2 users.
The speaking task given to students in the Boers et al. (2006) experi-
ment was a semi-structured conversation in English (about seven min-
utes long) with their teacher. The topic for conversation was a text in
English the students had been given to read in preparation for the
36 Optimizing a Lexical Approach

talk. The students’ fluency as perceived by the blind judges was found
to be positively associated with the numbers of chunks they used
(correlation coefficient was .45). However, given the nature of the
task, students could easily ‘recycle’ bits of English language from the
text, including chunks. As a result, many chunks from the text reap-
peared in the discourse of all of the students, which of course reduced
the margin for discrimination between ‘rich’ and ‘poor’ chunk-users.
Stengers (2009) therefore opted for a speaking task where students
were given a text in their mother tongue. The students were asked
to read the text and to re-tell its contents in English. To aid memory,
they were given a sheet with vertically arranged single keywords in
the same order as the paragraphs of the text that they related to. By
denying the students the possibility of recycling complete chunks
from the input text (apart from chunks which happened to have one-
to-one equivalents in the students’ L1 and L 2, that is), the researcher
reasoned that she would get a better impression of the students’
true mastery of lexical phrases. Again the ranking of the students
in terms of their perceived fluency and their ranking in terms of the
numbers of chunks they used were found to be positively associated.
In this case, the correlation coefficient was .60, a more significant
result than that obtained from the first study. In her study, Eyckmans
(2007) was especially interested in the use of chunks as an aid to
fluency when heavy demands are made on cognitive processing.
Students were asked to perform a sight translation task. They were
given a text in their mother tongue and were asked to translate it
orally, after very little preparation time. Again, students who used
the most L 2 chunks were judged to be the most fluent. However, the
correlation coefficient of .35, though still significant,9 was lower than
in the two previous experiments. As it happens, some of the students
used an above-average number of lexical phrases, but were not awarded
correspondingly high scores for fluency because several of these expres-
sions were produced hesitantly, as if these students had to make an
extra effort to recall the precise wording of the chunks they were try-
ing to use. Thus, it seems to be when chunks are used with confidence
that they contribute most to perceived fluency. The lesson from this is
that a lexical-phrase oriented pedagogy whose aim is to help learners
become proficient L2 speakers will have to invest time and effort in con-
solidating knowledge of chunks in long-term memory. When it comes
to recognizing chunks during language reception, width of knowledge
may suffice, but for fast retrieval of chunks for purposes of real-time
 The Contribution of Chunks 37

language production, depth of knowledge (including automatization)

is also crucial.
Neither is evidence lacking that knowledge of chunks facilitates
fluency in reception, in L2 as well as L1. Conklin and Schmitt (2008)
and Underwood, Schmitt and Galpin (2004) carried out experiments
in which native- and non-native-speaking participants were asked to
silently read stretches of English discourse on a screen while their eye-
movements were recorded by means of eye-tracking instruments. The
stretches of discourse contained either chunks (such as idioms) or syn-
onymous word sequences without chunk status. Both categories of par-
ticipants were found to read the chunks significantly faster than the
non-chunks. This difference emerged both in the case of chunks used
literally and ones used figuratively. Eye-movement data suggest that
participants recognized the chunks almost instantly after seeing their
onset, so that attention could then be rapidly shifted to the words fol-
lowing the chunks. These results clearly indicate that learners’ reading
pace can benefit from good chunk knowledge too.
Two of the above studies also considered the potential contribution of
students’ chunk use to other parameters in the assessment of language
proficiency. Boers et al. (2006) and Stengers (2009) both found positive
correlations between the number of lexical phrases used by the students
and the scores they were awarded by blind judges for the parameter of
‘range of expression’ (that is, lexical richness and syntactic complex-
ity). The correlation coefficient was especially high in Stengers’ experi-
ment (.65), where the students could not simply recycle bits of language
from an English text but instead had to rely on their own repertoires
of English to re-tell the contents of an L1 source text. Evidently, the
broader that repertoire is (in terms of chunks as well as single words),
the greater the likelihood that the student will be able to retrieve from
it the lexis that appropriately and precisely conveys the message wait-
ing to be conveyed. In addition, the use of chunks can help students to
be perceived as idiomatic language users (see Chapter 1), disposing of a
relatively impressive lexical richness and syntactic complexity.
It has also been suggested that prefabricated chunks – at least if they
have been stored in memory correctly – can serve as ‘islands of safety’
(Dechert, 1983). That is, students can insert chunks into their spoken
discourse as a way of ‘buying time’ so as to free up processing space
so that they can better plan the more creative and thus more ‘risky’
stretches of their discourse. This is especially likely to be a key benefit
of using conversational ‘fillers’, such as kind of, you know (what I mean),
I guess, and those kinds of things. Data from Boers et al. (2006) and Stengers
38 Optimizing a Lexical Approach

(2009) lend support to this possibility, although the correlation was not
statistically significant in the first of these two studies. The task in that
experiment (a teacher–student conversation having to do with written
text) necessitated impromptu reactions on the part of the students in
response to the teacher’s questions and comments. It was during these
unplanned moments of interaction that the students’ use of chunks
tended to be restricted to high-frequency conversational fillers. The use
of these chunks did not, however, prevent them from making errors (for
example, at the level of grammar and pronunciation) in the stretches of
discourse in-between them. The islands of safety created by these few
expressions seemed small and far apart – suggesting again that learners
who want to become both fluent and proficient need not only a reper-
toire of chunks they can retrieve quickly but one that is very large. It
is only when the repertoire is large enough that we can start replacing
the islands-of-safety metaphor for the role of prefabricated phrases by
the metaphor of the stepping stones which is suggested by the cover
of this book.
We believe the evidence cited so far makes it clear that good chunk
knowledge does contribute to proficiency in L2 as well as in L1. What
remains is the question of whether it is reasonable to try to help L2
learners to attain a mastery of chunks that bears even remote compari-
son with the mastery shown by typical native speakers of the same age.
As we will see in the next chapter, proponents of the LA have tended
to put their faith in in-class awareness-raising activities as a means of
enabling learners to autonomously acquire lexical phrases outside of
class. Is this expectation realistic?
3
Estimating the Chances of
Incidental Uptake of L2 Chunks

3.1 Fostering learner autonomy through

awareness-raising

Native speakers build up an impressive repertoire of thousands of lex-

ical phrases, but in doing so they can rely on floods of authentic input
and innumerable opportunities both to use and play with their mother
tongue. Time is hardly an issue: uptake of chunks begins in early child-
hood and seems to carry on more or less until life’s end. It is clear that
the input and time available to teenagers and adults learning a foreign
language in a classroom setting cannot amount to anything like what
is available to native speakers. This is why it is so widely agreed that it is
necessary for students to learn outside the classroom as well as in, and
to do so independently. It is also why advocates of the current, stand-
ard version of the LA argue that the number of chunks worth learning
is so vast that it makes little sense for teachers to spend much precious
class-time teaching individual specimens. Rather, it has been stated,
teachers should set aside time in class not so much for teaching chunks
as for teaching strategies that learners can adopt in order to be able to
learn chunks on their own. Michael Lewis, for instance, is very clear
about this:

You won’t be there outside the classroom. Your whole purpose is your
learners’ autonomy and your own redundancy. Encourage strategies
which help learners to help themselves. (1993: 193)

Chunk-oriented classroom activities should therefore aim to raise

students’ awareness of the importance of phraseology in general (in
contrast to rule-based patterning) and to leave them more inclined

39
40 Optimizing a Lexical Approach

than before to attend on their own to chunks in the L2 discourse they

encounter outside the classroom. Thus:

What is essential is that the teacher equips the students with search
skills which will enable them to discover significant collocations for
themselves, in both the language they meet in the classroom and,
more importantly, in the language they meet outside the classroom.
(Woolard 2000: 33–4)

To be clear, the standard version of the LA is founded on the assumption

that autonomous learners are likely to acquire a significant number of
chunks through incidental uptake while reading and, to a lesser extent,
while listening, provided, that is, such learners have received appropriate
instruction about the nature of chunks, about the roles of chunks, and
about strategies for chunk learning. By way of justifying this assump-
tion, an analogy is drawn with L1 acquisition:

We know from studies of L1 acquisition that children aged around

seven may be acquiring 4/5000 words per year [...]. That happens
without them being directly taught vocabulary. The implication is
clear – it is exposure to enough suitable input, not formal teaching,
which is the key to increasing the learner’s lexicon; most vocabulary
is acquired, not taught. (Lewis 1997: 196–7)

The force of this analogy is that we should be optimistic about inci-

dental acquisition of L2. But should we? First of all, it is a matter of
debate whether all L1 vocabulary acquisition truly is incidental. Parents,
carers and other guardians may not, after all, effectively refrain from
behaviour resembling explicit teaching of language forms, as assumed
by Chomsky (1980) and many others (see Pullum and Scholz, 2002,
for a review of the wider literature on the alleged ‘poverty of stimulus’
confronting the L1 learner). With respect to the role of L1 schooling, Jan
Hulstijn has made the following pertinent observation:

It may well be that the explosive growth of vocabulary between

the ages of 6 to 16 (in countries where people go to school during
these years) ... stems from a variety of oral and written tasks which
not only expose students to new words and concepts, but also force
them to process this lexical information repeatedly. In other words,
instructional programmes of most subject matter, knowledge and
skill, avail themselves of language, and hence vocabulary, as their
 Incidental Uptake of L2 Chunks 41

primary vehicle. Although this vocabulary is learned ‘incidentally’,

i.e., through the performance of subject-related tasks, these tasks
often require learners to process words elaborately and repeatedly.
(2001: 273)

Secondly, ISLA takes place under very different conditions than does
acquisition of L1, so different in fact – as remarked at the beginning of
this chapter – that optimism about incidental uptake of L2 vocabulary,
including lexical phrases, may be far from fully warranted.
Over the past quarter of a century there have been numerous inves-
tigations of incidental L2 vocabulary learning, of which the vast major-
ity have focused on the rate at which single words are acquired during
independent reading (for a review of this literature, see Schmitt, 2008).
One matter on which there is agreement is that incidental acquisition
of L2 lexis is a cumulative process (Nation and Meara, 2002: 40) requir-
ing a new(ish) lexeme to be met repeatedly in different contexts for it
to be learned (Nation, 2001). A number of variables play a part here,
among which are the character of the discourse matrix, affective fac-
tors, and multiple kinds of learner difference such as literacy, working-
memory capacity and knowledge base. Further, since a lexeme can be
understood and remembered to different degrees, there is the matter of
deciding at what degree the lexeme counts as being learned rather than
unlearned. Consequently, estimates vary as to the number of encoun-
ters typically necessary for uptake. Optimistic estimates with regard to
single words range from six to ten encounters; but that is on condition
that these encounters occur within a relatively short span of time (Waring
and Nation, 2004). If a new memory trace is not either strengthened
or added to relatively quickly by new encounters with the lexeme, the
trace is likely to fade way. Once the memory trace has become some-
what stable, the intervals between encounters can gradually become
longer but even then Nation (2001), who applies Pimsleur’s (1967) mem-
ory schedule, recommends at least six repetitions in the course of a
five-hour period. These are troubling findings because they mean that
lexemes outside the highest frequency bands are unlikely to present
themselves to the learner often enough and close enough together for
uptake to occur. This is one of the reasons why it is increasingly believed
that L2 vocabulary expansion through incidental acquisition is a slower
and more problematic process (Huckin and Coady, 1999) than was
once claimed by many advocates of Communicative Language Teaching
(CLT) and all advocates of The Natural Approach (Krashen, 1985, 1989).
According to Laufer (1997: 106), ‘it is unrealistic to expect vocabulary
42 Optimizing a Lexical Approach

learning to be a by-product of limited quantities of reading, such as

they are in a foreign language situation’.
Because most chunks occur relatively infrequently, they are impli-
cated in any reduced expectations concerning incidental acquisition.
Let us, for example, imagine a motivated learner of English picking up
a popular novel, such as Val McDermid’s thriller Beneath the Bleeding
(2007) – a plausible scenario given that the cover of the book touts the
author as writer of the scripts for the BBC television series Wire in the
Blood, an internationally successful series. Let us also imagine that our
reader, carried away by the plot, reads the first 120 pages in the course,
say, of one rainy Sunday afternoon. With a couple of interruptions this
could amount to an extensive reading activity spread over about five
hours. What are the chances that our learner will encounter any given
chunk several times during this time? When we canvass those first
120 pages of the novel for, for example, strong verb–noun collocations,
we find that our imaginary reader would encounter most of these collo-
cations only once in this part of the book (see Box 3.1) and certainly not
the multiple times recommended by knowledgeable researchers (such as
Huckin and Coady, 1999: 185).
Note that this trawl nets a miniscule proportion of the enormous
stock of English verb–noun collocations. Many common ones (such as
make an effort) occur nowhere in this sample while a number of those

Box 3.1 Strong verb–noun collocations occurring in the first 120 pages of
Val McDermid’s (2007) thriller Beneath the Bleeding

Verb–noun collocations occurring more than once:

make a point (pp. 10, 61, 90); make a move (pp. 26, 32, 78); make sense
(pp. 47, 73, 107); make a decision (pp. 39, 50); spend time (pp. 71, 88); pay
attention (pp. 91, 119); tell the truth (pp. 28, 119).

Verb–noun collocations occurring just once:

complete a mission (p. 3); fulfil a task (p. 3); bend the truth (p. 6); spend the
night (p. 15); lose your mind (p. 18); see the point (p. 21); clear your throat
(p. 22); speak your mind (p. 23); make conversation (p. 26); do your duty
(p. 28); shake hands (p. 32); practise a religion (p. 41); commit suicide (p. 44);
waste time (p. 48); climb stairs (p. 52); pay a price (p. 54); take notice (p. 59);
having a laugh (p. 63); do the right thing (p. 63); read your mind (p. 75); make
a start (p. 82); give pause (p. 85); make an impression (p. 90); do your best
(p. 92); shed light (p. 94); serve a purpose (p. 94); make a statement (p. 100);
make no difference (p. 101); pay tribute (p. 102); spend the evening (p. 103);
watch TV (p. 105); have a drink (p. 107); crack a joke (p. 112); take a look
(p. 119); take a picture (p. 119).
 Incidental Uptake of L2 Chunks 43

that do appear (for example, waste time) are likely already to be known
by our (presumably) somewhat bookish learner – she is, after all, read-
ing for pleasure in a foreign language. The chunks which appear that
she might not be fully, or at all, familiar with (for example, pay tribute,
raise the stakes, crack a joke) – because they occur only once – will prob-
ably leave too fleeting an impression for long-term storage in memory
to occur. In contrast, some ‘slot-frame’ chunks do occur in the sample
rather often – but in a variety of guises. For example: it wouldn’t take
long to __ (p. 8), it had taken twenty minutes to __ (p. 18), it’ll take me ten
minutes to __ (p. 20), it took a while for the paramedics to __ (p. 22), it took
him the best part of an hour to __ (p. 40), it would take a lot longer than that
(p. 54), it took us a couple of days to __ (p. 66), how long the symptoms of
ricin poisoning took to develop (p. 71), it didn’t take him long to __ (p. 94), it
still took three CDs to __ (p. 94), it had taken him that long to __ (p. 110) and
a poison that takes days to __ (p. 117). Incidental acquisition of this chunk
seems more plausible on account of its frequency, if our learner manages
to induce the frame pattern from these diverse instantiations.
This brings us to reasons why we believe chunks may be even harder
to acquire incidentally than single words. Firstly, many chunks – per-
haps most – are idiomatic, which (all else being equal) makes a lexical
item relatively hard to learn (Laufer, 1997). Secondly, at least in writ-
ing, a word is generally typographically set off from surrounding words
so that readers can readily see where it begins and ends. Words can
be counted by counting spaces. Chunks, however, are typographically
signalled, mainly by hyphens and commas, far less often (e.g., a jack-in-
the-box; on the other hand, ...). Thus many chunks may be relatively dif-
ficult to acquire from reading because, compared to words, they are less
marked out. Thirdly, as we have seen, chunks tend to be spoken rela-
tively quickly and with relatively more phonological streamlining; this
can hardly fail to make them harder for learners to discern in speech.
Fourthly, some chunks are discontinuous (as ... as, for example) while
others have the potential to appear so as the result of an unusually long
insertion and/or they may appear in altered order (as in the cat, as they
say, has been let out of the bag). It is at least possible that such complexity
and mutability also makes learning harder, given evidence that inflec-
tional and derivational complexity make it harder to learn single words
(Laufer, 1997).
Additionally, whether an encounter with an unfamiliar word or
phrase will leave a lasting trace in memory will obviously depend on
how it is processed. According to Schmidt (1990, 1993, 2001), Robinson
(1995, 2003) and many others, noticing is the crucial first step towards
44 Optimizing a Lexical Approach

learning a feature or element of language. Once noticed in the input

stream, the element/feature is temporarily held in working memory
(also known as short-term memory). This result, which is called intake,
is thought to be a necessary (although not sufficient) condition for
uptake into long-term memory to occur. This ‘noticing hypothesis’ –
in which there is no place for subliminal learning – is the origin of
a much-discussed strand in language-teaching methodology known
as ‘focus on form’ (Long, 1991). In instruction which is intended to
engender focus on form, moments of enhanced language awareness
are incorporated within task-based communicative activities (R. Ellis,
2003). The focus-on-form movement was initially concerned with the
acquisition of grammar features (hence the term ‘form’), but in recent
years its rationale has been extended to vocabulary learning as well. For
example, Laufer (2005) reviews the evidence that tasks which overtly
lead students to notice unfamiliar words result in much better learning
gains than tasks whereby vocabulary uptake is expected to occur as a
by-product of purely meaning-oriented communication.
Even if one day we were to discover that noticing is after all not an
absolute prerequisite for vocabulary uptake, it is hard to imagine that it
could ever be shown to be unimportant for learning to occur. In fact,
it seems that one of the reasons why extensive reading may result in
disappointing rates of incidental vocabulary acquisition is that learners
frequently fail to attend to unfamiliar words that they encounter. For
one thing, when learners feel they do not need to (fully) understand a
particular word in order to comprehend the overall message of a passage
or whole text, they may simply skip over it (see further below).
Advocates of purely meaning-focused language instruction (such
as Krashen and Terrell, 1983) argue that students should be taught to
tolerate any temporary vagueness that may arise when an unfamiliar
word is met in a text. If a word really is important for text compre-
hension, then – so it is argued – it is likely to recur in the text later on
and so contextual clues necessary for successful inference of meaning
will gradually accumulate. This position is sensible if what one wants
is for students to become independent readers and listeners. But if the
hoped for outcome is learners who have significantly expanded their
repertoires of L2 lexis, then exhorting students to tolerate temporary
vagueness is unlikely to be helpful. Firstly, not many words, except for
the most frequent, are likely to recur often enough in one text for a stu-
dent to feel the need to make an effort to infer their meaning. Secondly,
even if the learner does make the effort, available clues may be insuf-
ficient, or even misleading (Bensoussan and Laufer, 1984; Huckin and
 Incidental Uptake of L2 Chunks 45

Coady, 1999; Laufer, 1989; Laufer and Sim, 1985). Thirdly, the meaning-
guessing abilities of learners have been overrated; even when appar-
ently good clues are present, learners may fail to profit from them. In
fact, learners frequently mis-guess the meaning of an unfamiliar word
whether in L2 (Kelly, 1990) or L1 (Pressley, Levin and McDaniel, 1987).
Fourthly, a significant proportion of learners tend to overestimate their
vocabulary comprehension (owing, for example, to the presence of
deceptive L1 cognates such as English demand which a French learner
may construe as being synonymous with demander ⫽ ‘ask for’) and fail
to notice that their interpretation of words occurring in a text is actu-
ally inaccurate (Laufer and Nation, 2001). Finally, the encouragement
not to bother about words that seem inessential for overall text compre-
hension runs counter to the now widely accepted belief we referred to
above, that noticing is a prerequisite for learning.
On the other hand, it is undeniable that some incidental uptake does
occur (Nagy, 1997; Sőkmen, 1997; Huckin and Coady, 1999). What could
be the role of noticing in successful cases? Unfortunately, investigating
learners’ noticing behaviour is fraught with complexity. Some studies
have relied on questionnaires and stimulated recall protocols to elicit
data from learners after they performed a task designed to bring about
instances of noticing (Robinson, 1997; Mackey et al., 2002). Others have
collected data by means of think-aloud protocols collected from learn-
ers in the middle of doing a task (Leow, 2000; Rosa and O’Neill, 1999;
Rosa and Leow, 2004). Both approaches have shortcomings. The former
involves a time-lapse during which learners may re-interpret what they
think went though their minds as they were performing the task. The
latter involves verbalization of cognitive processes concurrently with
the performance of the task, which may influence or even interfere
with the very cognitive processes one is aiming to describe. A recently
proposed alternative technique for gauging noticing is eye-tracking:
learners are asked to silently read short texts on a screen and, while they
are reading, an eye-tracking instrument records their eye-movements.
If particular elements in the texts are paused at longer than the others
and/or if the learner’s eyes backtrack more often to these elements, then
this signals the learner’s momentarily heightened awareness of the ele-
ments in question.
Godfroid, Housen and Boers (2009) report an eye-tracking study set
up to investigate whether learners notice unknown words in short read-
ing texts. The target words in the experiment were pseudo-words, that
is, words invented so as to look like plausible English words. The pseudo-
words were substituted into the texts for words of the same class and of
46 Optimizing a Lexical Approach

the same phonemic length. For example, push boundaries was changed
into push paniplines. Apart from the pseudo-words (which were by def-
inition unknown to the students), all the other words that made up the
texts were common words judged to be familiar to the participating
students. These students, who were unaware of the true purpose of the
experiment, were asked to try to read the short texts in the same man-
ner in which they read texts for leisure. Immediately after the reading
activity, the participants were presented with an unannounced post-test
to measure whether the unknown words had left any memory traces.
The eye-tracking data reveal that the participants did tend to fixate
unknown words more than familiar words in these short texts, which
suggests that they did indeed notice them. This is good news, although
the results of the post-tests clearly confirm previous findings that a sin-
gle instance of noticing a new word is very seldom enough for uptake to
occur, beyond – in some cases – the ability to recognize the encountered
word when it is re-presented shortly after the first encounter.
The question now is whether learners are as likely to notice words
occurring in the immediate vicinity of unknown items. That is: do learn-
ers’ eyes also pause at, and/or backtrack to, the potential collocates of
unknown words? The eye-tracking data in the above study show no evi-
dence that they do. The students did not fixate the words immediately
preceding or following the pseudo-words any longer than they did other
(presumably familiar) words in the text; nor did they regress to them
any more often. For instance, they did not pause longer at the word push
which preceded paniplines. And this is the second reason why we fear
that the rate of incidental uptake of lexical phrases is bound to be even
slower than is so for single words: in many cases, learners will simply
not notice the chunks at all (see also Arnaud and Savignon, 1997: 168;
Bishop, 2004). Failure to notice the chunk status of a word string is per-
haps especially likely in the case of ‘compositional’ collocations – that
is, ones which learners take to be semantically transparent because they
consist entirely of familiar words whose joint meaning seems to emerge
straightforwardly. For example, a learner may come across a strong col-
location such as tell a lie, fully understand it, and – consequently – fail
to take notice of the word combination as such. Due to L1 transfer, lack
of negative evidence in the L2 input and/or lack of an explicit error
correction, the learner may then continue to use their un-idiomatic
substitutes (*say a lie, for instance). When readers do come across an
unknown word (tribute or confinement, say) and take notice of it, the
eye-tracking data show there is no guarantee at all that they will take
any notice of its immediate surroundings (for example, pay ← tribute or
 Incidental Uptake of L2 Chunks 47

close ← confinement). The chances of a reader autonomously noticing a

chunk as such are probably greatest when the word string as a whole
is unfamiliar. This may happen in the case of collocations when both
content words are unknown (wreak havoc, for example), in which case
the learner may indeed contemplate the two words in combination
(cf. Howarth 1998: 29). Autonomous noticing is also likely to happen
when the learner encounters a non-compositional chunk (typically a
figurative idiom, such as at the end of [my] tether and get/give someone
short shrift), even if all the words it consists of seem familiar (as in show
someone the ropes and the gloves are off ).
Advocates of the standard version of the LA could argue that such
difficulties are addressable by means of standard awareness-raising and
chunking exercises which are, after all, supposed to instil the habit of
paying attention to the syntagmatic dimension of read text (Lewis,
1997: 47–53). The hope is then that once such exercises have been done
in the classroom, students will be more on the lookout for recurrent
word combinations during independent reading in L2. Because they
will be more on the lookout for common combinations, they will both
notice and learn more chunks. But is this what happens?
Eyckmans, Boers and Stengers (2007) report an experiment conducted
to estimate whether students who have had plenty of teacher-guided
chunking practice in the classroom do in fact become better at autono-
mous chunk recognition, given experimental evidence that recognition
of chunks (in English) can be very difficult even for advanced learn-
ers (Granger 1998). During a 26-hour proficiency course, one group of
upper-intermediate English majors regularly engaged in teacher-guided
chunking practice; that is, they were given the task of looking for the
presence of collocations, and for multiword lexis generally, in reading
texts and in transcripts of listening texts and then they compared their
outcome with the sets of chunks the teacher had identified. A parallel
group of students used exactly the same materials, but was not taught
to pay special attention to multiword lexis. At the end of the course all
these students were presented with a new reading text (a newspaper
article of about 500 words about a general-interest topic similar in kind
to the texts included in the main course materials) and they were asked
to underline all the word strings they considered to be chunks (defined
for the students as ‘combinations of words that are idiomatic in English,
that is, words that belong together in the natural language of a native
speaker of English’). The same task was also given to four native-speaker
teachers, who were themselves LA practitioners and thus quite familiar
with chunking. The students whose awareness of phraseology had been
48 Optimizing a Lexical Approach

raised through regular teacher-led chunking activities underlined on

average 19.7 word sequences in the text, which was significantly more
than the average of 15.5 word sequences underlined by their peers who
had not received that training. At first sight this is a promising result,
which seems to indicate that the experimental group had, as a result of
LA-type instruction, become better at identifying L2 chunks and thus
better equipped to notice them. The problem is that most of the word
strings that were signalled as chunks only by the LA group were not
identified as chunks by any of the four native speakers (an example
of word sequence mistakenly identified as a collocation by LA-trained
students was to nurture work).
This suggests that while a standard LA activity of the kind just
described may lead students to pay more attention to the syntagmatic
dimension of vocabulary, it does not help students – or does not help
them enough – to autonomously recognize genuine chunks such as strong
collocations. One highly probable reason for this is that unless a learner
has encountered and then often noticed a word (for instance, shopping)
in a particular combination with some other word(s) (such as do the ___),
it is difficult for the learner to appreciate any collocational typicality.
But if this has not been appreciated, the learner is not especially likely to
show a marked propensity to notice the neighbouring words in the first
place. There is something here of what is colloquially termed a catch-22
situation. Given enough exposure, L2 learners are likely eventually to
become sensitive to frequency effects of word sequences. However, as
we mentioned in Chapter 1 with reference to Table 1.1, not all high-
frequency word combinations qualify as chunks, because their MI scores
are low. On the other hand, a lot of word combinations that do have
high MI scores are not highly frequent. The autonomous recognition
of L2 chunks is likely to be impeded since learners may under-attend to
less frequent but high-MI strings. It is the latter, though, which are most
likely to count as chunks in the minds of expert L2 speakers (see N. Ellis
et al., 2008, for evidence and further discussion, including the recom-
mendation to compile genre-sensitive high-MI chunk lists and to refer to
these for teaching purposes). It goes without saying that there is nothing
wrong with learners (also) acquiring high-frequency word strings of low
MI. However, by definition such strings lack easily identifiable semantic
character and so are bound to be relatively difficult to remember com-
pared to more or less equally frequent strings of higher MI.
Finally, there are reasons to doubt that students will apply outside
the classroom the noticing strategies regarding phraseology which they
have been trained to apply in the classroom. For one thing, it is well
 Incidental Uptake of L2 Chunks 49

known that people ordinarily pay much more attention to the meaning
of a message than to its exact wording. We know too that language users
generally do not allocate attentional resources to meaning and form
at the same time (Skehan, 1998; VanPatten, 1990). It is normally only
when one re-reads or re-hears a text that has already been understood
that one attends more closely to such formal features as lexical selec-
tion. Yet we know of no evidence that learners commonly engage in re-
reading or re-hearing of any material that is unrelated either to school
work or to preparation for a professional examination. It is true that one
might encourage learners to give due attention to (true) lexical phrases
during independent reading by giving them texts in which collocations
and other chunks are highlighted, since there is some evidence that this
can make them more willing to seek glosses (Bishop, 2004), which may
in turn increase uptake (but see further below). Currently, though, such
typographically enhanced materials are in very short supply and are, in
any case, unlikely ever to figure prominently among the authentic texts
that students might read outside the classroom.
This is not to say that an L2 reader will never notice any feature of
lexical or phraseological form on first reading or hearing. For example,
words may be noticed if they are used in an unexpected way (as in
word play), but – given habitual allocation of processing capacity –
noticing exact wording must strongly tend to interrupt text interpret-
ation, momentarily at least. Learners’ attention may also be drawn to
linguistic elements that they do not understand. Among all kinds of
chunks, this is perhaps most likely to happen in the case of idioms.
Boers, Eyckmans and Stengers (2007) report an experiment in which
language majors were presented with 15 idioms they had not learned
yet (including no-holds-barred, put __ through their paces, without breaking
the bank, throw in your hand, a white elephant and red tape) embedded in
example sentences copied from the Collins Cobuild Dictionary of Idioms
(2002). These were ‘rich’ contexts that the researchers felt illustrated the
usage of the idioms well. (The lexicographers who selected them from
among all the others in their corpus presumably thought so too.) What
we wanted to find out was the extent to which advanced students might
be able to infer the meaning of an unknown idiom occurring in rich
contexts of this sort. The students were thus asked to read the material
and to write a paraphrase of each idiom, which had been underlined.
Overall, the results were rather disappointing: only 39% of the para-
phrases were considered by a blind judge to be correct. This supports
other warnings that the success of context-guided meaning-guessing
should not be overestimated.
50 Optimizing a Lexical Approach

In reality, of course, learners might consult a dictionary when they

experience a comprehension problem during independent reading, but
Hulstijn (1993) found that their tendency to do so is rather limited. If a
comprehension problem is deemed to be ‘local’ (in the sense that word
comprehension is not felt to be necessary for comprehension of the text
or the message as a whole), it is common for learners to continue reading
and leave it unresolved. They may not even have a dictionary to hand in
the first place. Accordingly, a number of researchers have turned their
attention to ways of prompting potentially fruitful look-ups. Hulstijn,
Hollander and Greidanus (1996), for example, marked particular words
and expressions so as to indicate that their meaning was clarified in a
glossary accompanying the text being worked with. In other studies (for
example, De Ridder, 2002) texts were presented in electronic formats
which afford glosses at the click of a mouse on a hyperlink. An obvi-
ous aim of such techniques is to provide easy access to correct form–
meaning mappings, it being widely believed that this is highly likely to
facilitate uptake. However, experimental results are inconclusive about
whether students do in fact accelerate uptake into long-term memory
by looking words up (Hulstijn, 1992). According to Knight (1994), for
example, there is a strong positive effect of looking words up, but many
other studies (Laufer and Hill, 2000; Mondria, 2003; Peters, 2007a) are
less sanguine. For one thing, readers seem typically to consult a gloss in
order to resolve a local comprehension problem and once it is resolved,
they tend quite quickly to carry on reading. Contemplation of a gloss
seems often to be too fleeting and shallow for the target item to leave a
lasting impression in memory. It seems, in fact, that is not a lookup per
se that determines the chance of uptake, but rather the kind of cogni-
tive processing that comes into play when a gloss is read (Peters, 2007b).
With respect to vocabulary acquisition generally, Laufer and Hulstijn
(2001) emphasize the importance of affective involvement, cognitive
effort and depth of processing. They encompass this all under the con-
struct of ‘task-induced involvement (load)’, their conclusion being that
‘tasks with a higher involvement load will be more effective for vocabu-
lary retention than tasks with a lower involvement load’ (2001: 17). One
technique that has been experimented with in order to stimulate effect-
ive learner involvement is to design the glossary so that learners find a
gloss by doing a task. For example, instead of simply clarifying target
items via paraphrase or translation, the glossary may present clarifi-
cation in the form of a multiple-choice exercise (Rott and Williams,
2003; Watanabe, 1997). Another example is described by Boers (2000a),
who targeted figurative expressions such as wean infant companies off
 Incidental Uptake of L2 Chunks 51

state support and overcome many hurdles but included only the literal
construals of these terms in the glossary. Both of these techniques are
intended to encourage the learner to contemplate the meaning of tar-
get items for longer and in greater semantic depth. The jury is still out
about whether such techniques do in fact lead to significantly greater
vocabulary uptake (see Han, Park and Combs, 2008, for a recent review
of studies of enhancement techniques). Results of early experiments are
generally positive, but ‘enhanced’ glossaries must be delicately tuned in
order to forefend inaccurate interpretations on the part of the students,
and it takes a good deal of time to produce such materials.
By moving on to the use of text-enhancement techniques (such as the
highlighting of lexical phrases) and glossaries, we have left the zone of
incidental acquisition, strictly speaking, and arrived in a zone between
incidental and intentional vocabulary learning, a zone populated with
activities likely to foster what we will call semi-incidental acquisition.

3.2 Estimating the chances of semi-incidental

uptake of L2 chunks

By ‘semi-incidental vocabulary uptake’ we mean acquisition that is still

a by-product of meaning-oriented activities (reading and listening for
pleasure, and communication generally) but which differs from ‘pure’
incidental acquisition in the sense that any object materials or the tasks
in which they are embedded have been designed or manipulated so
as to stimulate the learners’ attention to particular vocabulary items
(Paribakth and Wesche, 1997). Highlighting and glossing figure among
these techniques, and there are others which require more active
involvement with targeted vocabulary (Hulstijn and Laufer, 2001),
translating a text with the aid of a dictionary, for example. In order to
find out how well vocabulary uptake was fostered by a translation-with-
dictionary task, Petitjean (2008) conducted an experiment with Dutch-
speaking students doing an MA in Spanish. They were asked to translate
a short text from Spanish into their mother tongue, with the aid of
a Spanish–Dutch dictionary. The students were subsequently given an
immediate post-test in which they were presented with a number of
words from the Dutch translation they had just produced and asked
to write down the Spanish counterpart of the words they remembered
having looked up in the dictionary. On average, students managed to
reproduce only half of the Spanish lexemes they had felt the need to
look up in the dictionary only a short while before (and this score was
the result of lenient marking whereby, for instance, misspelled words
52 Optimizing a Lexical Approach

were counted as correct). The low retention rate can hardly be explained
by an excess of novel lexis, since the students had felt the need to look
up only three words on average. To us, this suggests that even when the
involvement load is considerable, vocabulary uptake as a by-product of
meaning-focused text-based tasks should not be overestimated.
To our knowledge, similar experiments have not yet been conducted
specifically regarding L2 multiword lexis. However, the pedagogical
chunking activities that are at the heart of the LA can also be taken as
activities meant to foster semi-incidental vocabulary acquisition. After
all, in that approach relatively little time is devoted to the teaching of
particular items; instead students are alerted to the presence of chunks
in sample texts. We have seen above that this approach risks falling short
of its aim, which is to accelerate learners’ incidental uptake of lexical
phrases outside the classroom. So then, is there evidence that ‘chunk-
noticing’ exercises promote chunk uptake inside the classroom? In order
to answer that question a series of three similar experiments were car-
ried out by Boers et al. (2006), Eyckmans (2007) and Stengers (2009). In
each of these experiments, one group of English majors received tuition
inspired by the LA in the sense that text-chunking (by which we mean
identifying lexical phrases in authentic texts) was the core activity in a
26-hour course spread out over eight months. A parallel group of same-
level students worked with exactly the same materials, but the teacher
refrained from drawing these students’ attention to chunks. At the
end of the treatment, students were given a speaking task (which was
recorded) in order to determine whether the text-chunking treatment
had led to a better use of multiword lexis and whether any greater use of
multiword lexis was correlated with the degree to which students were
judged to be proficient L2 speakers. In Chapter 2, we reported that in
all three experiments the correlation was significant. The question we
are interested in now is whether the students who had been trained in
chunking used more chunks than their control peers when performing
later speaking tasks. Blind judges were asked to listen to the recordings
of the participants doing these tasks and to count what they considered
to be chunks in each student’s discourse. It was then ascertained that
the blind judges’ chunk counts were correlated in a statistically signifi-
cant way.
In the Boers et al. (2006) study, the speaking task was a guided con-
versation with the teacher about an English text the student had just
read. This was followed by a more spontaneous interaction about the
student’s hobbies, holiday plans, and the like. Analysis of the first part
of the task revealed that the students whose appreciation of chunks
 Incidental Uptake of L2 Chunks 53

had been raised used significantly more chunks than their controls.
However, the difference between both groups lay in the number of
phrases they had ‘recycled’ from the text they had just read. This is in
accordance with the above-mentioned finding that learners who are
trained in chunking are likely to become more appreciative of the syn-
tagmatic properties of L 2 texts. It is also possible that these students
tried to make a good impression on their teacher by replicating what
she had taught them to do in the course. Be that as it may, the second
part of the conversation, which was not text-based, showed no differ-
ential use of chunks between the two groups of students. The chunk
counts dropped drastically from the levels observed in first part of the
conversation and the few chunks that were used by either group were
typically of the high-frequency filler kind (I mean, kind of and stuff
like that).
In the follow-up experiments, conducted with the participation of
new cohorts of English majors, the students were denied the opportun-
ity to simply recycle words and phrases from an L 2 text. Instead, the
input for the re-tell task in Stengers (2009) and the sight translation task
in Eyckmans (2007) were texts in the students’ mother tongue. In both
experiments, the students who had received extensive chunking prac-
tice in the course were found to use only slightly more lexical phrases
than their controls in the post-test, and the difference did not approach
statistical significance.
In these experiments it is very likely that the amount of LA-inspired
treatment and the period of implementation were insufficient for the
measurement of longer-term effects. The observation that students
whose chunk awareness had been raised appeared keen to recycle
phrases from recent input points to the possibility that they will, in the
long run, acquire more chunks than these three experiments attest. It
is also conceivable that in-class chunk-noticing exercises had helped
them part-learn more chunks than their control peers; that is, perhaps
they had learned more chunks but not so well that they could produce
them in a challenging speaking task. In any case, none of the three
experiments yielded hard evidence that regular chunk noticing is suf-
ficient to bring about a differential uptake of lexical phrases. Thus, it
looks as though teacher-guided chunking alone does little to accelerate
uptake of chunks either inside or outside the classroom.
We agree that chunking instruction can enhance students’ appreci-
ation of phraseology in general and help them notice individual chunks
that figure in the exercises. However, subsequent incidental acquisition
of chunks from independent L 2 reading and listening is, for many
54 Optimizing a Lexical Approach

learners, likely to fall well short of the most optimistic forecasts that
have been made. In general, incidental acquisition of chunks is a slow
process. We also agree about the fruitfulness of teacher-guided noticing
as a stimulus for semi-incidental acquisition. But this form of interven-
tion is similarly likely to be insufficiently effective, at least when the
goal is durable recollection of what has been noticed and understood.
It is an important first step, but others must follow.
If both incidental and semi-incidental uptake are likely to result in
low rates of long-term uptake of chunks, and if it is a learner’s wish
to acquire a lot of lexical phrases, then it must be reasonable to try
to ensure that a considerable number of the chunks presented for in-
class noticing are, also, targeted for long-term retention. Despite his
faith in out-of-class incidental learning, Lewis (1997: 86–106) proposes
a number of explicit chunk-targeting exercises with some potential to
complement noticing in just this sense. Plainly, though, class time is
generally – perhaps always – so limited that explicit chunk teaching
cannot be an option every time a new one is encountered. The question
then is, which chunks are worth trying to teach in such a way (and with
such an investment of time) that learners are relatively likely to remem-
ber them? This issue of selection is taken up in the next chapter.
4
Selecting Chunks for
the L2 Classroom

4.1 Utility

As we demonstrated in Chapter 1, natural discourse abounds with

chunks. Given the time-constraints of classroom-based language learn-
ing, teachers face the difficulty of deciding which chunks merit atten-
tion at the expense of many others. The LA in its present form offers
little help in this regard, apart from emphasizing that it is ‘useful’ lan-
guage that should be selected for classroom use:

Teachers have to select both in and out. It is not sufficient for some-
thing to be unknown and useful; the ideal is to select what is most
useful. (Lewis, 1997: 45)

In this chapter we address the question of what kinds of chunks that

are encountered in classroom materials invite teacher-led intervention
most of all. The first criterion that comes to mind for giving priority in
teaching to some chunks over others is, indeed, their estimated useful-
ness for the students.
Which chunks are particularly useful for a given group of students
depends ultimately on their level of proficiency and on the objectives
of the course they are taking. Chunks whose mastery may be very bene-
ficial to students taking a Language for Specific Purposes (LSP) course
(English for Business, English for Law, English for Academic Purposes
[EAP], and so on), may well be of little use to students taking a general
language course. In this book we will be concerned with usefulness for
the intermediate-to-advanced general language learner.
Let us imagine a non-native teacher using the recording of the BBC
Radio 4 programme, the transcript of which we presented in Chapter 1,

55
56 Optimizing a Lexical Approach

Box 1.1. After a communicative activity designed to elucidate the con-

tent of the programme, a teacher might want to draw students’ atten-
tion to a few useful chunks used in it and take steps to help the students
store these chunks in memory. Given time constraints, and in order to
ensure a manageable learning load for the students, the teacher will
need to be selective. After all, the text abounds with chunks, some of
which the students may already be familiar with, others which they
may need to consolidate their knowledge of, and still others that the
students may not have known at all.
As it happens, usefulness generally correlates with frequency of
occurrence; that is, elements that occur highly frequently (at least
in the genre or type of discourse the learner is aiming to master) are
likely to be encountered, and needed for production, more often than
low-frequency elements. Accordingly, it seems commonsense to sug-
gest that it is highly frequent chunks that should be given priority in
the classroom (and in pedagogical materials too), whereas – because of
the time constraints – learning of less frequent chunks should be left
to the vagaries of incidental acquisition (mostly) outside the language
classroom (Kennedy, 2008: 39). Language corpora are now available for
researchers to verify which chunks actually are highly frequent and
which are seldom used, although it must be acknowledged that auto-
matic identification and counting of chunks in electronic corpora is
far from straightforward (Moon, 1998a, b). A list of the most frequent
chunks occurring in spoken English has recently been compiled by Shin
and Nation (2008), but this is explicitly intended to inform teaching at
beginner level. A considerable number of the high-frequency colloca-
tions identified by Shin and Nation would even qualify for inclusion in
the most frequent 2000 words in English (see also O’Keeffe, McCarthy
and Carter, 2007: 46–7).
Unfortunately for teachers whose students have progressed beyond
beginner level, the currently available chunk-oriented dictionaries,
such as the Oxford Collocations Dictionary for Students of English (2002)
and most idiom dictionaries – while corpus-informed – do not pro-
vide the user with information about the relative frequencies of their
entry items. An exception is the second edition of the Collins Cobuild
Dictionary of Idioms (2002), which signals the idioms that have been
found to be comparatively common (although no information is pro-
vided as to precisely how common), but this only covers a small sub-
set of the ubiquitous class of chunks. The question then is how our
imaginary non-native teacher can check her intuitions about the rela-
tive commonness of the chunks that happen to occur in the teaching
 Selecting Chunks for the L2 Classroom 57

materials. One option is to consult one of the corpora that have been
compiled by corpus linguists for lexicographers and others to draw on.
Probably the best known of these, as far as English is concerned, are the
British National Corpus (BNC) and the Collins Cobuild Wordbanks. While
both are accessible on-line, they do require a subscription, and it is
doubtful whether many (non-research oriented) teachers are prepared
to pay for this. However, a sub-part of the Wordbanks corpus is access-
ible for free through the Concordance and Collocations Sampler on the
Collins Cobuild website. The concordance sampler generates 40 corpus
lines in which the search item is used. The collocations sampler gen-
erates a list of the 100 words that occur most often in the immediate
vicinity of the search item (see also Chapter 1, Box 1.2, for examples).
Neither produces straightforward information about relative frequen-
cies of chunks, although the collocations sampler includes ‘joint fre-
quency’ figures, which may help the user estimate the commonness
of a particular word combination. For example, the joint frequency
of load and dice is much lower in the sample than that of play and
part, which confirms the intuition that load the dice is a less common
chunk than play a part. Another option for a zealous teacher is to treat
the World Wide Web as a gigantic corpus and to compare the number
of ‘hits’ that are generated by ‘exact-word searches’ in Google. This is
a straightforward procedure in the case of chunks that are invariable
(such as in fact, according to and open to debate). However, a great many
chunks are variable, being subject to inflection (play/ played/ playing
a part, load/ loaded/ loading the dice) and/or word-order changes (a part
played by, the dice are/ were loaded). Obtaining a somewhat accurate
impression of their frequency of occurrence through Google exact-word
searches requires several searches per chunk and the adding up of the
number of hits generated per search. In the case of open-slot phrases
or chunks that allow the insertion of various lexical elements (play
a crucial/ important/ major/ minor/ negligible part), the query becomes
impracticable. Still, let us assume that our imaginary teacher is zealous
enough to check the World Wide Web in order to compare the frequen-
cies of occurrence of the chunks that occur in the 220-word excerpt of
the radio programme. Table 4.1 lists the chunks for which the search
is feasible (albeit sometimes intricate) and the number of Google hits
obtained (in November 2008).
Table 4.1 illustrates how the frequency of occurrence of chunks
falls dramatically after a small set of the commonest ones. Given the
nature of the text at hand, our imaginary teacher is dealing with a fairly
advanced group of students, and so the highest-frequency chunks are
58 Optimizing a Lexical Approach

Table 4.1 Exact-word Google hits as an indication of relative commonness

Chunk Hits Chunk Hits

according to 363,000,000 at conception 917,000

in fact 170,000,000 tip the balance 790,000
all about 159,000,000 open to debate 555,000
likely to 140,000,000 in normal times 330,000
at the heart of 22,000,000 birth statistics 259,000
caught up in 8,710,000 come/came to notice 98,600
food shortage 7,920,000 load the dice 52,000
body-mass index 7,130,000 the toss of a coin 26,000
in times of 2,450,000 as part of a study on 21,600

unlikely to require any further teaching. By way of comparison, the

numbers of Google hits generated by according to and by all about are
similar to the numbers of hits generated by the single words ship and
desk, respectively, that is, vocabulary that is undoubtedly part of the
repertoire of an intermediate-to-advanced learner. The numbers of hits
generated by in times of and open to debate are similar to the numbers of
hits generated by gullible and earlobe. Discarding the highest-frequency
chunks still leaves the teacher spoilt for choice as to which of the less-
frequent chunks she could target – and this set of chunks is drawn from
just one, very short text excerpt. She may reasonably choose to target a
couple of chunks that she suspects the students are already half-familiar
with (at the heart of and tip the balance, for instance) and a couple that
are likely to be new (come to notice and load the dice, for instance). The
idioms load the dice (52,000 hits) and by the toss of a coin (26,000 hits)
appear to be low-frequency chunks, but a comparison with the exact-
word Google hits generated by non-chunk equivalents puts this into per-
spective: weigh down the balance generates only 210 hits and the throw of
a coin a meagre 85 hits.
The call to prioritize high-frequency chunks echoes voices from the
dominant strand of research in second-language vocabulary acquisition
in the 1980s and 1990s. The point of departure then was the question
of how large a learner’s vocabulary knowledge needed to be to permit
independent reading of non-specialized texts. When it was calculated
that the 2000 most frequent word families (that is, ‘base’ words plus
their derivations such as read, reading, reader, readership) in English cover
about 80% of general texts, it was deemed pedagogically advisable to
cover at least these high-frequency items in pedagogical materials and,
when classroom time was insufficient, to rely on the learners’ ability
 Selecting Chunks for the L2 Classroom 59

to pick up additional words autonomously, especially through exten-

sive reading. This premise has not been without its problems, though.
For one thing, 80% lexical coverage of general texts has been found
to be much less than is necessary for adequate comprehension of such
texts (Hu and Nation, 2000). Laufer (1997) even suggests that know-
ledge of no fewer than 5000 word families is necessary for 98% coverage
of general texts, which is the coverage now believed to be required for
adequate text comprehension. In other words, the vocabulary-learning
task the student faces is much more daunting than was once believed,
and this is simply with respect to reading. Vocabulary knowledge for
productive purposes presumably needs to be deeper and more automa-
tized than receptive knowledge (see Hulstijn, 2001, for a much fuller
discussion). To that challenge is now added that of mastering countless
multiword units.
While resorting to frequency of occurrence as a principal benchmark
for the inclusion (and exclusion) of chunks for teaching is evidently a
sensible choice, it does not come without its methodological difficul-
ties. As already mentioned, beyond the very-highest-frequency items
the frequencies of words and chunks in a corpus rapidly level off (see
also Shin and Nation, 2008), and so the frequency criterion for selection
soon becomes hard to apply. Let us look at some examples of this in the
form of joint-frequency data on verb–noun collocations thrown up by
the Collins Cobuild on-line collocations sampler and given in Table 4.2.
From Table 4.2 it appears again that some chunks (cause ⫹ damage and
commit ⫹ crime) stand out in terms of frequency and would undoubt-
edly merit attention according to a strict application of the frequency
criterion. These are typically followed by clusters of chunks whose fre-
quencies are so similar that inclusion of some of them and not others
may seem arbitrary. Also in the table, in each column in fact, are exam-
ples of what is known as ‘semantic prosody’ (Hoey, 2005). That is, we

Table 4.2 Joint-frequency data on some verb–noun collocations (from the

Collins Cobuild collocations sampler)

Cause ⫹ Freq. Commit ⫹ Freq. Conduct ⫹ Freq. Perform ⫹ Freq.

damage 457 crime 260 study 83 task 74

trouble 173 offence 100 research 73 function 72
pain 167 murder 90 survey 70 music 51
cancer 153 adultery 20 interview 57 dance 44
injury 137 rape 19 investigation 48 song 37
harm 124 fraud 14 experiment 44 role 29
60 Optimizing a Lexical Approach

see that cause typically collocates with negative outcomes; commit col-
locates with nouns denoting crimes; conduct collocates with nouns
denoting some kind of structured collection of information; and per-
form collocates with nouns for actions which are normally positively
valued, four of which often involve an audience or spectators (music,
dance, song, role) and three of which have to do with useful activity in
general (task, function, role). One could arguably treat these groups of
collocations as ‘collocation families’ (analogously to ‘word families’),
and if an instance of such a family (for example, commit fraud) came up
in a text, exploiting this to alert students to the family at large may be
time well spent, even though – strictly speaking – the collocation per
se might not occur with high-frequency. In a similar vein, the chunk
by the toss of a coin in the simulation (see Table 4.1) is relatively infre-
quent and so, while it seems unworthy of explicit treatment on that
account, its occurrence in the text may be exploited as a pathway for
also teaching the more frequent expression argue the toss (42,500 Google
hits) because both expressions are derived from the same domain and
thus share the same imagery.
It must also be mentioned that the 100 top collocates that are offered
by the free on-line collocations sampler of Collins Cobuild will often
be insufficient for the user who is looking for confirmation about the
chunk-status of a particular word string. For instance, neither the query
word tell nor the query word time generated evidence of their word part-
nerships (tell the time, time will tell) when we performed our searches.
There are several other reasons why strong collocates are sometimes
missing from the lists. Firstly, almost any query word co-occurs fre-
quently with function words such as articles and pronouns simply
because the latter have such a high overall frequency. These func-
tion words tend to push content words down (and maybe off) the list.
Secondly, if the sampler gives all inflected forms of verbs (tell, tells, tell-
ing, told) and nouns (lie, lies) separately rather than in a lemmatized
form (see Box 1.2 in Chapter 1 for examples), that too means that a lot
of room is lost in the list for slightly weaker collocates. Thirdly, while
the query syntax allows the user to distinguish between word class (play
as verb versus play as noun, for example), this does not entirely solve
the problem of homonymy. For example, the query word party will gen-
erate many collocates associated with the word in its sense of political
party, and these push the collocates of the homonym with the festive
sense (which occurs less frequently in the corpus) down (and off) the
list again. For instance, we did not find confirmation of the chunk-
status of throw a party by typing in party. Even querying throw failed
 Selecting Chunks for the L2 Classroom 61

to turn up this expression, suggesting that a list of more than 100 col-
locates is desirable.
Apart from these methodological and technical difficulties, there is a
second problem with blanket prioritization of high-frequency chunks
at the expense of ones that are less frequent. As already mentioned, the
highest-frequency phrases are the ones which, by simple virtue of being
very common, stand the best chance of being learned incidentally. Thus,
teacher intervention is most likely to be fruitful when directed towards
not-so-frequent chunks. Naturally, we are not advocating that teachers
shift their attention to chunks that are rare. But if one of the aims of a
teaching programme is to accelerate the expansion of the students’ rep-
ertoires of lexical phrases, then what we will call ‘medium-frequency’
chunks have to be considered as prime candidates for explicit teaching.
Besides, among the highest-frequency chunks in informal speech are
fillers (you know what I mean and stuff like that) and, in some genres,
colloquial-to-vulgar expressions (no shit) which teachers may not wish
to focus on for, shall we say, aesthetic reasons.
So suppose, then, that we decide to concentrate instruction on chunks
in the middle rather than highest and lowest frequency bands. This still
leaves us with a very large number of chunks, certainly far too many to
deal with in any remotely realistic classroom setting. Of course, this
decision to focus on medium-frequency chunks is at odds with using
frequency of occurrence as sole criterion for selection, but, as we have
seen, exclusive reliance on this criterion is problematic anyway.
Let us return, then, to the broad criterion of utility, but this time
think of it from a processing perspective. We saw in Chapter 2 that
fluent, idiomatic and accurate use of L2 chunks helps learners to be per-
ceived as proficient. We also saw that a mental stock of lexical phrases
contributes most to fluency in real-time conditions when it is not only
large but also when it can furnish chunks for production more or less
instantaneously. For chunks to be so readily available in situations of
real-time use, they must be deeply entrenched in memory – ideally,
automatized. As mentioned in Chapter 1, Section 1.1, lexical phrases
vary in the fixedness of their lexical composition from being com-
pletely fixed (of course), to being nearly fixed (when it comes/ came to ...),
to allowing for very restricted lexical substitution (play a part/ role), and
so on. One advantage of learning relatively fixed types is that commit-
ting them to memory is economical in comparison with learning the
lexical variations of chunks which lie well away from the fixed end of
the spectrum. As far as fostering chunk knowledge for real-time pro-
ductive purposes is concerned, it may even help students if (initially)
62 Optimizing a Lexical Approach

priority is given to just one of any multiple variants – for instance, play
a part rather than play a role. An additional reason for devoting a dispro-
portionate amount of teaching time to relatively fixed chunks is that
deviation from their lexical makeup is, by definition, especially likely
to result in un-idiomatic phrasing. For example, learners who – due per-
haps to L1 interference – say *this boy can already read the time (instead
of tell the time) or *time will say (instead of time will tell) reduce their
chances of being perceived as proficient speakers. The risk of erring at
the level of phraseology is statistically smaller when it comes to phrases
that allow for a wider range of lexical substitution (such as carry out / do /
fulfil / perform a task) simply because the learner has a better chance of
picking a word that happens to be collocationally acceptable, although,
to be sure, L1 interference may lead to such un-idiomatic utterances as
*make a task (from French faire une tâche).
The debate about whether the strength of inter-word bonds should be
criterial in deciding which chunks to target has centred on collocations.
Kennedy (2008), for example, who explicitly favours the frequency cri-
terion, would prioritize conduct/ do/ carry out research over play truant
(which is rigid) simply because the latter chunk is less frequent. Hill
(2000) and Nesselhauf (2003), on the other hand, advocate taking col-
locational strength into account. Which position is pedagogically most
sensible may well depend on the learning goals in the case at hand. As
we mentioned above, students’ language production is likely to bene-
fit most immediately from their mastery of chunks that are stored in
a relatively fixed form, as automatization of these chunks should be
relatively easier to achieve. On the other hand, knowledge of frequent
chunks must especially facilitate (receptive) text processing, irrespect-
ive of the strength of the inter-word bonds. Again, it is conceivable that
chunk knowledge which benefits receptive fluency may have to be very
wide, but not as deep as that which benefits production.
Summing up so far, we have argued that frequency of occurrence
cannot be the only criterion used in selecting chunks for instruction,
because helping students to progress beyond beginner level requires the
targeting of chunks between the highest and the lowest frequency bands.
Because the class of medium-frequency chunks is very large, at least
one additional selection criterion must be applied. We have conjectured
that among sets of same-frequency chunks, it may be the ones that are
fairly fixed (in terms of lexical substitutability) which learners will find
especially useful because good procedural knowledge of fixed chunks is
likely not only to support fluent production, but also to diminish the
likelihood of collocational error. Figure 4.1 is a broad-brush attempt to
 Selecting Chunks for the L2 Classroom 63

demarcate the zone of phraseology which we believe it would be most

beneficial for teachers and materials writers explicitly to target. It sug-
gests a need for a principled trade-off between frequency and degree of
fixedness as targeting criteria. But even with this demarcation, the pri-
ority zone still includes several thousand chunks. Teachers then need to
make further decisions as to which in particular of these chunks they
can most effectively ‘do something with’ as they occur – in instruc-
tional materials, for instance. The little stars in Figure 4.1 are meant
to depict kinds of chunks within the priority zone that – for reasons
we will outline further below (in Section 4.2) – are especially inviting
instructional targets.
One class of chunks that is situated towards the fixed end of the con-
tinuum in Figure 4.1 is that of ‘idioms’. This class illustrates particu-
larly well that is not only fluency which is promoted when a string of
words becomes understood and remembered as a chunk but, very often,
comprehension as well. As is well known, there are many chunks whose
meaning is either completely or virtually opaque (by and large, for
instance) or, at least, somewhat different than the sum of its parts (all
at once in the sense of ‘suddenly’). These chunks – commonly referred
to by the umbrella term idioms – cannot be (fully) understood unless a

Less fixed More fixed

Highest frequency band

Priority zone

Middle frequency band

Lowest frequency band

Less fixed More fixed

Figure 4.1 Priority zone for explicit chunk targeting in the classroom
64 Optimizing a Lexical Approach

particular conventional meaning or pragmatic function is mapped onto

the whole assembly of words (Fillmore, Kay and O’Connor, 1988). This
also holds for idioms that are figurative (in the sense that their literal
origin can still be traced), such as a rule of thumb and pass the buck.10
When the aim is for the learners to be able to insert idioms in their L2
output, the learners are confronted with an additional problem, that
of knowing from the words that make up the idioms whether these
expressions fit a particular communicative (especially pragmatic) need.
For instance, a learner might find the phrase Pleased to meet you under-
standable but still have no way of knowing in advance that people rou-
tinely use this phrase when formally introduced to each other. Again,
the ensemble has a sense (or rather, function) that is not evident from
the individual words. But what a learner most obviously gains from
being able to produce the chunk is native-like phrasing.
What makes many figurative idioms difficult or impossible to fully
interpret is that the speech community has forgotten they were once
used in a literal sense. Hit the bull’s-eye is a combination of words that
we can still use and understand either literally or idiomatically (that is,
in a figurative sense that has become conventionalized). But this is not
so for pull it off (⫽ approx. ‘succeed’). Our speech community has lost
knowledge of what ‘it’ was and what ‘it’ could be pulled off of. Without
that knowledge, full understanding of the idiom is not possible. We
can only infer, perhaps, that ‘it’ was a prize or something like that. We
shall describe the pedagogical usefulness of relating idiomatic and ori-
ginal literal meanings in Chapter 5, for we indeed believe that figura-
tive idioms (at least those of medium frequency) merit a place among
the chunks that invite targeting in ISLA.
We realize, though, that some may object to spending any teaching
time on figurative idioms in the first place, given their ‘icing-on-the-
cake-of-proficiency’ reputation. It is known that learners tend to avoid
using them anyhow (Irujo, 1993). Still, drawing especially on McCarthy
(1998: 129–49) and O’Keeffe, McCarthy and Carter (2007: 80–99), we
adduce at least three grounds for teaching figurative idioms.
Firstly, as we have seen, students have trouble decoding unfamiliar
figurative idioms. This is likely to hamper text comprehension dispro-
portionately as idioms are particularly likely to occur at those important
points (known as codas) in colloquial narrative when a speaker or writer
offers an evaluation of an event or state of affairs (McCarthy, 1998:
131–49). Indeed, idioms fulfil an important, evaluative function (see
also Moon, 1998a, b), which learners often fail to grasp, especially when
the L2 cultural background is very different from that of L1 (Littlemore,
 Selecting Chunks for the L2 Classroom 65

2001). It has in fact been argued that ‘idioms are never just neutral
alternatives to literal, transparent, semantically equivalent alternatives’
(McCarthy, 1998: 145; see also Gibbs, 1994: Chapter 6). If this is so,
then there must be a great many meanings that students will be unable
to express, or even recognize, unless they learn a lot of idioms. On the
other hand, it has also been noted that, because of their evaluative
function, many idioms, especially if used in direct address, pose threats
to face (Strässler, 1982). Consider, for instance, the difference in impli-
cation between Where did you buy/ get that outfit? and Where did you come
by that outfit? The prominence of idioms in the potentially high-stakes
arena of expressing evaluations – a facet of language use for which a
command of nuance and a large repertoire of expressions are bound
to be useful, not least in conversation and correspondence – is actually
grounds not only for inclusion of idioms among the chunks which are
selected for teaching, but also for teaching them well rather than in a
cursory fashion. A similar obstacle to appropriate interpretation and use
of many idioms by learners lies in the difficulty of inferring their pre-
cise ‘usage restrictions’. Just how, for instance, does pull it off differ from
succeed? Part of the answer is that pull off is less formal than succeed, but
we cannot take it for granted that this difference will spring at all read-
ily to the mind of the language learner.
Secondly, it has been pointed out that idioms serve as ‘communal
tokens that enable speakers to express cultural and social solidarity’
(McCarthy, 1998: 145). Here, one thing that L2 learners may find elusive
is an appreciation of the overall function itself rather than an under-
standing of any one idiom in particular. But, if McCarthy is right, part
of fully understanding the use of an idiom in a particular situation
is appreciating it as an emblem of cultural and group identification.
Students who (will) need to make sense of native-speaker discourse
either at some remove (for instance in print literature or films) or face-to-
face, could therefore find learning idioms beneficial for this additional
reason. Moreover, idioms are commonly exploited by native speakers
for wordplay (in newspaper headlines, for example) in ways that can be
particularly hard for learners to understand on their own. For instance,
appreciation of a pun obviously requires familiarity with (the evalu-
ative and cultural connotations of) the absent expression(s) that a pun
is intended to evoke and comprehension is likely to be doubly difficult
if the absent expression is an idiom. Much the same can be said of
cases in which an idiom is only partly used, as in this title of a British
newspaper article about the risks involved in buying property in France
Look before you leap across the Channel. Incidentally, the fact that most
66 Optimizing a Lexical Approach

native speakers do get such puns and partial uses is one of the pieces
of evidence that knowledge of the ‘canonical’, or basic, form of even
rather infrequent idioms can be very widely shared. In J. K. Rowling’s
Harry Potter saga, for example, idioms are occasionally modified to suit
the world of magic. Get off your high horse becomes Get off your high
Hippogriff, for instance (Rowling, 2007: 34). And yet, the substitution of
one of the key words will not prevent the proficient reader from recog-
nizing the idiom. Because of the cultural knowledge which idioms can
display, natural use of idioms can overtly demonstrate participation in a
realm of shared cultural knowledge and interests, and so help a learner
gain social acceptance – at an informal gathering, say, or at work. This
implies a further pertinent criterion having to do with group integra-
tion, one which, it should be noted, may conceivably apply to other
infrequent chunks besides idioms.
Thirdly, although individual idioms may be relatively infrequent,11
as a class they actually do occur rather often, at least in some genres.
Idiomatic expressions are in fact quite numerous both absolutely and
proportionally within the larger population of chunks in general. Let
us illustrate this by returning to the Val McDermid thriller Beneath the
Bleeding (see Chapter 3) and to our imaginary learner-reader to see how
often she would encounter an idiom. A screening of the first 120 pages
of the novel yields the inventory presented in Box 4.1.
All the phrases included in Box 4.1 are idioms listed in the Collins
Cobuild Dictionary of Idioms (2002), which, in comparison with some
other idiom dictionaries, covers but a narrow selection of idiomatic
expressions.12 In other words, the inventory in Box 4.1 should be taken
as a ‘conservative’ one. At any rate, from a comparison with Box 3.1 in
Chapter 3 (the verb–noun collocations inventory from the same novel)
it looks as though the probability of encountering figurative idioms in
this kind of text genre is not at all lower than the probability of encoun-
tering strong verb–noun collocations. This text, which balances narra-
tive and dialogue, contains, on average, one idiom per page and a half.
There is more evidence that idiom usage may be more common in
general than is often assumed. We used the Collins Cobuild on-line con-
cordance sampler to call up 40 random concordance lines for each of
the following, basically spatial, prepositions: above, across, after, against,
among, around, at, before, behind, below, beneath, between, beyond, down,
in, into, on, onto, over, out, through, under, underneath, and up. We then
screened the concordance lines for instances where the preposition
occurred as part of an idiom that is listed in the (relatively slim) Collins
Cobuild Dictionary of Idioms (2002). From this it appears that these
 Selecting Chunks for the L2 Classroom 67

Box 4.1 Idioms encountered in the first 120 pages of Val MacDermid’s
(2007) thriller Beneath the Bleeding

Occurring more than once:

keep _ at bay (pp. 21, 28, 31, 55); on the same wavelength (p. 19, twice); _ up
to speed (pp. 19, 98); take the piss (pp. 19, 72, 76); caught on the wrong foot
(pp. 31, 97); keep you on your toes (pp. 41, 78); call it a day (p. 72, twice); cut
the mustard (pp. 42, 108).

Occurring once:
laid at the door of _ (p. 1); the rough and tumble of life (p. 1); He knew it in
his bones (p. 4); a king’s ransom (p. 5); gone head to head (p. 15); the nuts
and bolts (p. 19); thin on the ground (p. 20); stopped in her tracks (p. 21); off
the hook (p. 32); make a face (p. 32); water off a duck’s back (p. 37); reaping
what she’d sown (p. 37); for good measure (p. 39); hammer home a message
(p. 42); gone far out on a limb (p. 42); run the gauntlet (p. 45); keeping me
posted (p. 45); off the wall (p. 46); set the wheels in motion (p. 47); down the
line (p. 47); screaming from the rooftops (p. 48); put my reputation on the line
(p. 48); eyeball to eyeball (p. 52); the bottom line (p. 53); by the skin of his teeth
(p. 62); for a song (p. 62); for peanuts (p. 62); rub shoulders with (p. 62); on
track (p. 63); fit the bill (p. 67); on a platter (p. 68); run out of steam (p. 69);
keeping tabs on (p. 70); keep on a tight leash (p. 70); at sea (p. 71); play the field
(p. 72); a stay of execution (p. 75); up your street (p. 75); cut both ways (p. 79);
hot on their heels (p. 79); raise the stakes (p. 80); make the grade (p. 85); put
your foot in it (p. 88); chopping and changing (p. 91); a bone to chew on (p. 93);
make headway (p. 94); rattling their sabres (p. 97); get up to speed (p. 98); at
face value (p. 98); hang out to dry (p. 98); hidden agenda (p. 98); on the ground
(p. 99); on the page (p. 99); cover your back (p. 99); run yourself into the ground
(p. 103); fire on all cylinders (p. 105); have your wits about you (p. 105); foot-
loose and fancy free (p. 106); carry a torch for someone (p. 106); make a dent in
something (p. 107); look for a needle in a haystack (p. 108); get your hands on
something (p. 108); keep _ at arm’s length (p. 110); not give a toss (p. 113); get
in on the act (p. 113); shoot your mouth off (p. 113); put your oar in (p. 115);
not miss a trick (p. 116); get your head around something (p. 117).

prepositions are used as part of an idiom on average at least once per

40 occurrences (whether this finding can be extrapolated to other word
classes is a matter for further research). Even if the ratio of idiomatic
to literal uses were found in general to be half our estimate, preposi-
tions are so frequent as a class (and in some cases individually) that
merely taking prepositional idioms into account affords a sound basis
for supposing that learners will regularly encounter idioms in authentic
English discourse.
In addition to the above three lines of argument for the instructional
targeting of figurative idioms, there is a fourth reason, and that is that fig-
urative idioms have properties that make them especially ‘teachable’.
68 Optimizing a Lexical Approach

4.2 Teachability

At the end of Chapter 3 we argued that it would be naïve to count on

students being able to acquire large numbers of chunks incidentally,
even after they have participated in appropriate awareness-raising exer-
cises. We also offered evidence that noticing activities, while vital as a
first step, are relatively ineffective in furthering uptake of chunks in
long-term memory. We therefore contended that it is part of a teacher’s
role to take steps in the classroom to help students not just to notice par-
ticular chunks in the course materials but also to commit these chunks
to memory. In the present chapter, we have also argued that teachers
should take such steps principally when they are most likely to make a
difference, when, for instance, a chunk is neither rare nor so frequent
that it is likely to be learned incidentally. Particular chunks in the broad
middle band might be worth targeting for a variety of reasons, such as
that they seem relatively easy to produce accurately or that they are
necessary for good understanding of a text being worked with in class.
What we wish to argue now is that teachers should at some point,
whenever possible, do something to make targeted chunks more mem-
orable. This means encouraging students to engage in elaboration. This is
an umbrella term for diverse mental operations, beyond mere noticing,
that a learner may perform with regard to the meaning and/or the form
of words and phrases (Barcroft, 2002). Elaboration can, for instance,
consist in thinking about a term’s spelling, pronunciation, grammatical
category, meaning, and associations with other words (Hulstijn 2001:
270) as well as thinking which involves the formation of visual and
motoric images related to the meaning of the term. The more of these
dimensions that are involved, the more likely it is that the term will be
entrenched in long-term memory (Hulstijn, ibid.).
However, it seems that some chunks can be rendered memorable more
easily than others. Put differently, some chunks are inherently more
interesting than others in ways that readily lead to better retention in
memory. This is obviously true of chunks (and single words) which
have a high affective value for a particular student population. Thus,
for students in higher education in Europe, hangover can probably be
embedded in an amusing anecdote more easily than migraine. Hangover
also lends itself better to the mnemonic use of imagery (because it can
be analysed as hang ⫹ over which then raises the image-generating ques-
tion of what hangs over what). Other words or phrases can easily be
made more memorable because students can be reminded of a popular
song, film title or catch phrase in which the words or phrases happen
 Selecting Chunks for the L2 Classroom 69

to occur (such as the bare necessities, die hard, make my day). Still others
may have (semi-)cognates in the students’ mother tongue which invite
comparison leading to insights that boost memorability of the L2 target.
For example, the Dutch cognate for wide awake is klaarwakker in which
the adjective klaar denotes clarity or light; it thus exploits a different
metonymy from the English phrase where the adjective wide refers to
the opened eyelids of the sleepless person. The L1 associates need not be
such close counterparts as wide awake and klaarwakker, as is evidenced
by the well-attested effectiveness of the keyword method (Atkinson, 1975;
Avila and Sadoski, 1996; Shapiro and Waters, 2005). By this method
a learner can enhance the memorability of an L2 target word – for
example, bleak or puncture – by linking it with an L1 word which has
a phonological resemblance to the L2 target and a meaning which in
some, usually roundabout, way can be related to that of the target – for
example, Dutch bleek [‘pale’] for bleak since a landscape of nothing but
snow (which is certainly very pale) can be bleak, or puntje [‘sharp end’]
for puncture. Application of the keyword method to chunks, though,
is hardly ever at all obvious, and will in any case usually involve only
one word per chunk. For example, one might try to create an associ-
ation for Dutch-speaking students between the more, the merrier and the
Dutch word merrie (‘mare’) by evoking a scene of a group of mares frol-
icking happily in the meadow, but learners may well find associations
of this sort over-contrived. The truth of the matter is that few of the
mnemonics that have been proposed for word-learning (see Schmitt,
1997; Sőkmen, 1997, and Thompson, 1987, for excellent surveys) are at
all well-suited for chunk learning. This is an issue we will pick up again
in Chapter 8.
Words and phrases that can be made more memorable in an effi-
cient way we consider highly teachable. Highly teachable words and
phrases are characterized by their considerable mnemonic potential,
but it requires an intervention on the part of the teacher (or materials
writer) to unlock that potential for the students. Importantly, the inter-
vention required to unlock their mnemonic potential must be relatively
straightforward, not time-consuming, and perceived by the students
as valid. In principle, of course, all lexis is teachable. After all, creative
teachers can, for instance, insert any word or chunk in an appealing
story with a view to helping students remember it, since inserting target
lexis in a narrative is one of the mnemonic techniques that is applicable
to chunks. What we will call teachable chunks, however, are chunks
(1) which have intrinsic mnemonic potential but (2) whose intrin-
sic mnemonic potential is unlikely to be taken full advantage of by
70 Optimizing a Lexical Approach

the students themselves unaided. The common characteristic of such

chunks is that they are motivated in one way or another.
Linguistic motivation – a term we borrow from the paradigm of
Cognitive Linguistics (CL) (Langacker, 1987, 1990; Lakoff, 1987; but see
also Ullman, 1957) – refers to non-arbitrary aspects of linguistic ele-
ments. In mainstream linguistics, language has habitually been treated
as arbitrary; that is, the connection between a linguistic form and its
meaning is considered nothing more than a convention. Only two
exceptions are commonly acknowledged in that view. One is onomato-
poeia (the modelling of linguistic forms on real-world sounds) of which
the following are examples: bang!, splash, oink-oink and cock-a-doodle-doo
and also, by extension, forms such as screech, croak, growl and flip-flop.
Another is the shared but elusive sound symbolism evident in certain
whole sets of lexical items, sometimes called phonaesthemes (Firth,
1930: 45–54; see also Kies, 1990), such as swab, sway, sweep, swing, swipe,
swirl, swish, swivel, and swoop. Still, among mainstream linguists, such
form–meaning connections are considered to be the exception to the
rule that the linguistic sign is essentially arbitrary, and Michael Lewis
is no exception:

The single most fundamental principle of linguistics is the arbitrari-

ness of the sign. The importance of this principle cannot be over-
emphasized. [...] All lexical items are arbitrary – they are simply the
consensus of what has been institutionalized, the agreed language
which a particular group do use, selected from what they could use,
actual language as opposed to theoretically possible language. [...]
Students frequently ask why the language behaves in a certain way,
and are unhappy to be told English is like that, but unfortunately
that is the only accurate answer. The tendency to seek explanations
affects teachers too. Most readers probably suspect they know why
a bus stop is so called – probably because buses stop there. But the
so-called explanation is an illusion – taxis stop not at a *taxi stop,
but on a taxi rank; trains do not stop at *train stops, but at stations.
Bus stop is as arbitrary as taxi rank; by coincidence it looks as if it
can be explained, and no doubt the coincidence makes it easier to
recall, but, and this cannot be stressed too strongly, its construction
is not susceptible to explanation; it is as arbitrary as pen or book,
and attempts to explain will in the end cause confusion rather than
help. Learners need to be taught that some questions are helpful, but
others are not, as ‘answers’ to them simply do not exist. Etymology,
for example, may reveal the sources of some words or patterns, but
 Selecting Chunks for the L2 Classroom 71

the explanations are never more than of details. Language remains

essentially arbitrary. (1997: 17–18)

Put briefly, according to Lewis, there can be no answer to the question

‘Why might word sequence X and not some other word sequence have
become institutionalized to mean Y?’ But if there is no rhyme or rea-
son in the formation and meaning of chunks generally, then there is
very little teachers can reveal about them apart from saying ‘this just
happens to be the way it is said in this language – just try to remember
this phrase as a whole’. And the result of this is that learners are denied
access to many pathways towards kinds of elaborated coding of mean-
ing and form that have been shown to foster durable learning. The only
route for elaboration (though he does not use this term), we have found
Lewis to recommend is the comparison of L2 collocational patterns with
their counterparts in the mother tongue (1997: 63–6).
We concede that many linguistic phenomena do indeed defy explan-
ation, or at least explanation that language learners are likely to profit
from. On the other hand, there are linguistic elements, including many
lexical phrases, whose features, even in the L2 classroom, can be shown
to be motivated. That is, although it is impossible to predict what par-
ticular word sequences will become conventionalized (and thus ele-
vated to chunk status), it is often possible to discern reasons for such
conventionalisation in retrospect. It is in fact one of the principal ambi-
tions of Cognitive Linguists to discern types and instances of motiv-
ation in natural language (Radden and Panther, 2004). This endeavour
follows from the basic tenet of CL that linguistic cognition is supported
by, is pervaded by, and is of the same fundamental character as cogni-
tion generally: there is no modular, autonomous language faculty such
as Universal Grammar or a Language Acquisition Device. Let us, then,
briefly illustrate some of the ways in which answers can be sought to
the question ‘Why has word sequence X in particular been institution-
alized to mean Y?’
Sometimes, the motivation lies in iconicity, that is, the mirroring, in
language, of extra-linguistic phenomena, as already illustrated by refer-
ence to onomatopoeia (Nänny and Fisher, 1999). The role of iconicity in
the formation of chunks is also evident in some conventionalized simi-
les; for example, as cold as ice, as dry as dust and as slippery as an eel, are
direct translations of the extra-linguistic experience of ice as something
which is particularly cold, of dust as something which is especially dry,
of eels as creatures which are paragons of slipperiness. Iconicity is also
evident in another class of chunks – frozen binomials, that is, phrases
72 Optimizing a Lexical Approach

of the form ‘A and B’. For example, the institutionalized word order in
such expressions may reflect the chronology of events in the real world,
as in crash and burn, spit and polish, kiss and tell and bow and scrape.
The iconicity may also be of a scalar nature such that the word order
engenders a ‘crescendo’ effect, as in sixes and sevens, nickel and dime, alive
and kicking, and head and shoulders above [someone]. The word order of
a frozen binomial may also reflect perceptual salience, with the noun
for the larger entity coming first, as in suit and tie, cloak and dagger and
cat and mouse. Chronological and perceptual iconicity can also work in
tandem, as in milk and honey and bread and butter (one tends to first take
a slice of bread and then spread the butter onto it).
Iconicity is also fundamental in conceptual metaphor (Lakoff and
Johnson, 1980; Kövecses, 1990), whereby we conceptualize non-tangible
domains of experience via analogies to physical, often bodily, experi-
ence. For example, we know that adding things to a pile results in the
pile getting higher, while taking things away will reduce its size; that
adding water to a half-full glass results in the water level going up, while
drinking some will make the level go down; and so on. This up–down
experience is easily ‘mapped’ onto the way we quantify non-tangible
things (under age, above average, turn down the music, turn up the heat).
We also associate health and activity with physically being up on our
feet, while illness and inactivity is associated with being down. This is
projected onto our conception of currently functional as opposed to
defective inanimate things (my car broke down, the project is up and run-
ning). Also, having access to something in the physical world usually
coincides with being close to it, while inaccessibility is typically due
to distance. This physical experience is mapped via conceptual meta-
phor onto the way we describe accessibility or inaccessibility in abstract
domains (why she decided to do that is beyond me, the reason is far from
clear, we’re getting close to a solution).
Analogies are often drawn to more specific source domains (rather
than, say, the extremely generic physical experience of moving nearer
to or farther from something in the physical world). For instance, ana-
logies are made with physical fights (such as boxing) in talk about non-
physical competition – for example, the verbal competition in debates
and negotiations (not pull your punches, hit someone below the belt, take it on
the chin, be on the ropes, be down for the count, throw in the towel). Similarly,
our experience of transport and travelling is projected through meta-
phor onto the way we describe non-physical action or inaction (stay the
course, steer clear of, get the green light, take a back seat, in the fast lane, make
the grade). It is at the level of such relatively specific source domains
 Selecting Chunks for the L2 Classroom 73

that the impact of (historical) culture seems to be most noticeable in

phraseology, especially in a language’s stock of idioms (Boers, 2003;
Boers and Stengers, 2008a). Domains of experience that many people in
a language community are familiar with are especially likely to inspire
metaphorical analogies which, over time, give rise to conventionalized
figurative chunks or, in other words, figurative idioms. For example,
the (rather selective) Collins Cobuild Dictionary of Idioms (2002) lists sev-
eral dozen expressions (at least 77) that are retraceable to the domain
of seafaring (on an even keel, clear the decks, go by the board, plain sailing,
in the doldrums, a leading light, show someone the ropes and so on), which
is not surprising given the importance of seafaring in English history.
The English stock of idioms also contains a comparatively high number
of expressions derived from popular ball games such as soccer ([score] an
own goal, be on the ball) and cricket (playing a straight bat, hit someone for
six) and from horse racing (win hands down, neck and neck, in the running,
too close to call, a dark horse). By contrast, the Spanish stock of idioms,
for instance, contains a high number of expressions derived from bull
fighting (a popular sport, at least until recently) and from religion (from
Catholic tradition). This illustrates how the figurative segment of the
chunk repertoire of a language can to some extent be interpreted as a
reflection of the perceived importance (at some stage in history or in
the present) of certain domains of experience in the particular language
community.
Once the source domain of an idiom is identified, it often becomes
feasible to infer its figurative meaning from the context in which it ori-
ginally was (or still is) used in a literal sense. For example, if one learns
that the idiom jump the gun derives from usage in track sports where
the expression refers to any case of an athlete starting to run before
the starting pistol has been fired, then its figurative (that is, idiomatic)
meaning (‘do something before the appropriate time’) becomes quite
transparent. In this important sense, the meaning of many idioms is
not arbitrary but is instead motivated by analogies with concrete source
domains. This fact serves as a key to seeing the motivation for the rela-
tive prominence of certain lexical fields in the idiom repertoire of a
given language – for example, the presence of many terms from sailing
[helm, keel, to tack], ball games [wicket, to tackle, ballpark] and horse rid-
ing/racing [saddle, rein, tether] in the English stock of idioms.
Metaphor is also a motivating factor in chunks which are not often
viewed as figurative at the present time – standardized phrases for mak-
ing a polite request, for example. A CL account of this state of affairs
is as follows. We tend to conceive of time in terms of space: the past
74 Optimizing a Lexical Approach

is behind us, the present is ‘here’ and the future lies ahead of us. This
reflects the conceptual metaphor TIME IS A PATH AND WE MOVE ON IT. (In
CL it is customary to present conceptual metaphors in small capitals
and the linguistic instantiations of the metaphors in italics.) The ‘here
and now’ is close to us and is associated with what we are presently
involved with: LACK OF INVOLVEMENT IS PHYSICAL DISTANCE. Because of the
association of time and space, distance between the speaker and her or
his request can be signalled by the use of past tenses. This is a way of
voicing requests in a polite way, as it suggests a lesser degree of com-
mittal to what is requested on the part of the speaker than the use of a
present tense would (Littlemore and Low, 2006: 167). Compare Can I use
your pocket calculator? and Could I use your pocket calculator? Metaphorical
distance is also created by using a conditional tense rather than the
indicative. Compare Will you look this up for me? and Would you mind
looking this up for me? An additional dimension of iconicity is added
by distancing the thing that is requested from the agent/subject at the
head of the sentence by adding words/chunks in-between, as in I was
wondering if I could possibly use your ruler, which is another instantiation
of the LACK OF INVOLVEMENT IS PHYSICAL DISTANCE metaphor. In short, con-
ceptual metaphor can help motivate the conventionalization of such
sentence-head chunks as Would you mind and I was wondering if.
While metaphor is a process of analogy (for example, talking of time
in terms of space and talking of a debate in terms of a boxing match),
metonymy is a process of association (Barcelona, 2000). For example, our
hands play a prominent role in a very great proportion of the activities
that we perform. The association of hands with certain of these activ-
ities helps motivate the many chunks in which hand or hands are used
metonymically – lend a hand, have a hand in something, a safe pair of
hands, my hands are full and so on. While these metonymic uses are rela-
tively transparent, because the prominent role of hands in all sorts of
activities is universal, other institutionalized metonymies may be much
harder for language learners to interpret. Examples are let’s have tea early
today, where tea stands for a meal in British English, and in what’s on for
pudding? where, in British English, pudding has taken on the meaning
of dessert in general. These cases are motivated by the common custom
in this particular language community of having tea to accompany the
meal and of having pudding as a typical choice of dessert.
Meaning extensions of words and phrases are often the result of
metaphor and metonymy. In a broad sense, metonymy includes the
phenomenon of invited inferences, which has driven the meaning exten-
sions of modal auxiliaries, for example (Sweetser, 1990; Traugott and
 Selecting Chunks for the L2 Classroom 75

Dasher, 2002). The typical direction of semantic change (or, rather,

semantic widening) is from the deontic meaning of the verbs (that
is, the meaning of obligation, permission and ability) to an epistemic
meaning (that is, the meaning of deduction and possibility). In each of
the following pairs, the second meaning was derived from the first (see
the Oxford English Dictionary for the chronology in which various uses
appeared): you must come home before midnight / she must be home because
the lights are on; you may take your leave now / she may leave you if you keep
treating her like that; she can swim / she can get very impatient at times. The
invited inferences which motivate the meaning extensions are as fol-
lows: if someone obliges you to do something, then it is highly probable
that you will indeed do it (leading to the deduction sense of must); if
you are given permission to do something, then it is quite possible that
you will do it (leading to the possibility sense of may); if you are able to
do something, then there is a possibility that sooner or later you will
do it (leading to the more theoretical-possibility sense of can). Similar
processes of meaning extension motivate the future meaning of going to
(when you literally go somewhere, then you will arrive at your destin-
ation at some point in the future) and will (whose original meaning was
want/wish/desire, with the implication that if you strive for something,
chances are that you will get it at some point in the future). A straight-
forward case is the obligation sense of ought to, which was derived from
the meaning ‘owe’.
Apart from the general cognitive processes of analogy through icon-
icity and metaphor, on the one hand, and of association through meton-
ymy and invited inferences, on the other, there are forces of motivation
for chunks that exploit analogies and associations of a language-internal
kind. For example, the existence of a strong word partnership (such
as commit a crime) seems capable of priming additional partnerships
with words from the same lexical field. As already mentioned, the fre-
quent collocation commit a crime, may prime other collocates from the
‘crime’ field, such as offence, murder, manslaughter, arson, a burglary, rape,
fraud, as well as adultery and suicide (which also used to be considered
as crimes in England). The strong collocation wreak havoc may prime
wreak vengeance/revenge. Cause has developed a semantic prosody that
invites collocates that are of a negative kind, such as damage, injury,
harm, trouble, problems and illness (Stubbs, 1995). The same seems to
hold for the noun process (a slow/ painful/ difficult process). Even word
sequences that may not look like semantic units at first sight may have
semantic prosody. For instance, a bit of a is very often used in ironic
utterances (especially understatement), as in the financial markets are in
76 Optimizing a Lexical Approach

a bit of a mess, I’ve lost all my savings, which is a bit of a setback. The same
applies to not exactly a/the, as in Saving Private Ryan is not exactly the kind
of film you wish to see if you cannot stand blood and Saddam was not exactly
[a] mister Nice Guy, was he?
In addition to semantic prosody, it might be possible to discern pat-
terns of stylistic prosody, in the sense that formal words may tend to
prefer formal collocates over non-formal ones (for instance, seeking sol-
ace and be remanded in custody). We will return to this possibility in
Chapter 8.
So far, we have looked at ways of motivating chunks by considering
broadly semantic factors. However, factors to do with phonology also play
a part in the institutionalization of chunks (Boers and Lindstromberg,
2008a; Boers and Stengers, 2008b; Lindstromberg and Boers, 2008 a, b),
especially at the level of lexical selection. It is striking, for example,
that most of the strong noun-collocates of seek are s- words (solace, soli-
tude, asylum). It appears that phonological repetition – especially alliter-
ation – facilitates the standardization of some word strings over others
which could be taken as synonymous: time will tell (rather than time
will show/ *say/ reveal), it takes two to tango (rather than it takes two to
waltz/ jive), from dawn till dusk (rather than from dawn till sunset/ twilight/
evening/ nightfall), feel ten feet tall (which is over 248 times more com-
mon than feel eight / nine / eleven feet tall put together), and so on. Salient
phonological repetition is likely to enhance the memorability of the
word strings and may thus stimulate their institutionalization. This
could be the case also for rhyming chunks, such as high and dry (rather
than, say, up and dry), the name of the game (rather than, say, the name of
the business) and when the cat’s away, the mice will play (rather than ... the
mice will come out).
On the other hand, phonological repetition may also facilitate articu-
lation (by reducing the number of differentiated neuromotor actions).
That may be one of the reasons why less salient kinds of sound repeti-
tion, such as assonance and consonance, are also common in English
phraseology: small talk (rather than little talk), high time (rather than
urgent time, stark naked (rather than stark nude), casual acquaintance
(rather than superficial acquaintance). We mentioned in Chapter 2 that
through frequent use and hence, increasing familiarity, chunks tend
to be pronounced faster and with more phonological reduction than
non-chunks (Bybee, 2002). It is also possible, though, that it is word
strings that are comparatively easy to pronounce that stand a better
chance of becoming conventional in the first place. This possibility fol-
lows the ‘principle of least effort’ (Zipf, 1949, quoted in N. Ellis, 2008a),
 Selecting Chunks for the L2 Classroom 77

according to which speakers tend generally to minimize articulatory

effort (while maintaining communicative effectiveness).
From this brief exploration of linguistic motivation, we hope it has
become clear that not all chunks defy teacher explanation, contrary
to what Michael Lewis has claimed. In fact, some of the examples he
gives to back up the arbitrariness thesis tend rather to undermine it.
Lewis states (1997: 18), for instance, that there is no explanation why we
say bus stop, but train station and taxi rank. However, the use of station
rather than (bus or shelter) may have to do with the size of the place in
question and the fact that a station prototypically includes a building
(which explains why bus station also exists). The use of rank may have
to do with the fact that taxis typically form a line in wait for customers.
Note also the presence of assonance and consonance in train station and
assonance in taxi rank as possible additional motivating factors – per-
haps the word-end ⫹ word-initial consonance in bus stop may also be a
motivating factor.
In this chapter we have made a plea for the adoption of teachability
as one of the criteria for the selection of chunks for pedagogical treat-
ment. We have argued that some chunks are especially inviting tar-
gets for teacher-guided elaboration by virtue of their motivated nature.
However, the validity of this additional criterion depends on two things:
(1) linguistic motivation must have sufficient scope, that is, more than
just a small fraction of the panoply of chunks that make up a language’s
repertoire must be shown to have non-arbitrary features, and (2) the
pedagogical effectiveness of exploiting various kinds of motivation as
pathways for teacher-guided elaboration must be based on the results of
empirical studies. In Chapters 5 and 6 we will present more examples of
how chunks may be motivated, interwoven with summaries of studies
set up with the express purpose of measuring the pedagogical effective-
ness of revealing that motivation to learners.
As we have seen, elaboration may concern (aspects of) meaning or
(aspects of) form. The former is called semantic elaboration, the latter
structural elaboration (Barcroft, 2002). This is not a clear-cut distinc-
tion, however, as meaning and form can hardly be completely divorced
from one another. Still, the types of elaboration we will be assessing
in Chapter 5 exploit semantics more directly than they do form. In
Chapter 6, we will move on to assess types of predominantly structural
elaboration of chunks.
We shall see that a number of elaboration techniques we propose
involve students in analysis of chunks and contemplation of the indi-
vidual words that make up the chunks. This is not uncontroversial.
78 Optimizing a Lexical Approach

Lewis encourages learners to ‘Try to learn whole expressions containing

useful words, rather than just the words, even though this seems much
more difficult’ (1997: 78), and this advice to learn chunks ‘holistically’
would elicit broad approval (but also scepticism about its feasibility)
from Alison Wray. As we mentioned in Chapter 2, in Wray’s (2002)
model, children first acquire formulaic sequences (in their L1) holistic-
ally, as they do not yet discern word boundaries in the flow of speech.
Only at a later stage, when spurred by literacy skills, will native speakers
tend to start analysing chunks (and then typically on a needs-only basis)
at the level of the words which make them up. The great advantage of
unitary storage of multiword chunks in the mental lexicon from an early
age lies in ease of retrieval. When it comes to learning an additional lan-
guage, however, teenagers and adults are at a disadvantage, according to
Wray (2002: 206–10), because they will already have adopted ‘analyt-
ical’, word-by-word processing, which makes it much more difficult for
them to take in multiword sequences as wholes.
Wray (2002: 189–90) acknowledges, however, that holistic storage
may not be the only fluency facilitator: even though teenage and adult
learners tend not to learn L2 chunks holistically, their procedural know-
ledge of these chunks may become profound enough for fast recognition
and production. The position we take in response to these arguments
is as follows. If it is really so difficult for teenage and adult learners to
switch off their analytical mode, then let us explore and assess ways
of harnessing this mode rather than merely noting its robustness. If
students are not likely to process newly encountered chunks truly hol-
istically, and if they must be taught to glue the constituent words of
them together anyhow, then what could be so wrong with elaboration
techniques that help the students appreciate the meaning and lexical
makeup of these chunks in the first place?
So, that is what the next two chapters will do: explore and assess
pathways for semantic and structural elaboration that complement the
noticing of chunks. It must be mentioned from the start, however, that
the power of elaboration lies in the creation of stable memory traces. In
itself, elaboration may quite often not be enough to ensure the develop-
ment of procedural knowledge so far as to guarantee fluency in real-time
production of the learnt chunks. This last step may require automatiza-
tion, an issue we will take up in Chapter 7.
5
Semantic Elaboration

5.1 Exploiting imagery

As learners engage in elaboration of any meaningful element, they

invest cognitive effort. Originally referred to (metaphorically) as deep
processing, elaboration results in better retention than processing which
is superficial, or ‘shallow’ (Cermak and Craik, 1979; for a classic work on
memory, see Baddeley, 1997 [1990]). Although nowadays the adjectives
rich and elaborative are increasingly favoured over the one-dimensional
deep, the basic idea remains the same.
One pathway for rich, or elaborative, processing is dual coding (Paivio,
1986), which involves the association of verbal with non-verbal stimuli.
For instance, the meaning of a word or phrase is associated with a men-
tal image, the recollection of the lexeme thereby being facilitated since
the memory trace of the image constitutes an additional access point
(Paivio and Desrochers, 1979). Associations with mental pictures are
fairly direct in the case of words that denote something concrete. In the
case of abstract words, dual coding is possible, although less straight-
forwardly so, as when the keyword method (see Chapter 4, Section 4.2)
is applied to a target word which is not particularly or not at all con-
crete. For example, the abstract word jeopardy can be associated with
the concrete, easily imageable word leopard by virtue of both seman-
tic association – since leopards happen to be dangerous animals – and
phonological resemblance. Abstract words can also become dually coded
if they are embedded in a narrative that speaks to the imagination. The
adjective precious, for example, may have become associated in many
people’s minds with particular scenes from The Lord of the Rings.
Chunks which lend themselves to the dual coding strategy most read-
ily are, of course, idioms. Idioms have often been characterized as ‘dead’

79
80 Optimizing a Lexical Approach

or ‘frozen’ metaphors, because their figurative meaning has become so

conventionalized that they are no longer usually experienced as figura-
tive, that is, experienced as a meaning extension from an earlier, literal
usage. This is not to say, though, that the original figurative mean-
ing cannot be revived. For example, native speakers are occasionally
reminded of the figurative nature of an idiom when it is used creatively,
as in a pun that evokes a literal reading (for example, I felt I’d been sent
back to square one, but left without any dice) or when a standard lexical
component is replaced by something different, another member of the
same set, for instance (It’s not that big a set-back; you’re back at square
five or so). In fact, because so few idioms are really utterly opaque (Grant
and Bauer, 2004), it is likely that the literal meaning of almost all idi-
oms can be resuscitated to some degree. And luckily for teachers, there
are idiom dictionaries, such as The Collins Cobuild Dictionary of Idioms
(2002 edition) and Speake’s (1999) Oxford Dictionary of Idioms, which
include information about origins in a substantial proportion of the
entries. This is good because awareness of the original, literal usage of
an idiom establishes an association between the idiomatic meaning of
the phrase and a mental image of a concrete scene (which may involve,
for example, motoric as well as visual components). In other words,
informing learners about the original, literal sense of an idiom stimu-
lates dual coding to a degree likely to aid retention.
In the wake of Lakoff and Johnson’s Metaphors We Live By (1980) and
subsequent monographs that were to lay the foundations of Cognitive
Semantics (CS) (Gibbs, 1994; Johnson, 1987; Kövecses, 1990; Lakoff,
1987; Langacker, 1987; Talmy, 1988), a number of writers – among them
applied linguists, language teaching methodologists and lexicogra-
phers – have made the case that materials and activities which raise
awareness of metaphor and metonymy (although the latter is some-
times under-discussed) merit a prominent place in mainstream ISLA
(Baker, 1998; Deignan, Gabrys and Solska, 1997; Holme, 2004; Lazar,
1996, 2003; Littlemore and Low, 2006; MacLennan, 1994; Ponterotto,
1994; Rundell, 2001; Scott, 1994). The question we address in the pre-
sent section is whether and in what ways students’ retention and under-
standing of targeted idioms is indeed enhanced when they are helped
to associate each such chunk with the domain of experience in which it
was once used in a literal sense. Most of the pedagogical techniques we
will describe and assess involve a certain amount of ‘analysis’ or ‘decom-
position’ of idioms, which at first sight may seem to run counter to the
ultimate objective of helping students process idiomatic word strings
more or less holistically.
 Semantic Elaboration 81

It may therefore be useful first to point out that models of L1 idiom

processing do not univocally adhere to the view that idioms are proc-
essed holistically in the first place. These models typically concern com-
prehension during reception in L1: see Cieślicka (2006, 2008), Libben
and Titone (2008) and Sprenger, Levelt and Kempen (2006) for recent
reviews. Some models do consider idiom processing to be holistic, but
several others see a reader’s or hearer’s processing of idiomatic word
strings to be analytical at some times and to some degree. According to
early models (such as Bobrow and Bell, 1973 and Swinney and Cutler,
1979), an idiom is stored as a separate lexical unit in the mental lexi-
con, and it is accessed when a literal interpretation of the word string is
found implausible. According to later models in which idioms are also
considered to be separately stored units, the idiomatic meaning can be
retrieved before a search for any literal interpretation is even begun
(Gibbs, 1980, 1985). Given the fact that the word string making up an
idiom is often more frequently encountered in its idiomatic than in its
literal meaning, there seems no reason (from a psycholinguistic stand-
point that takes frequency effects into account) to posit a processing
sequence in which readers or hearers first try a literal interpretation as a
default option. For example, unless one is in the habit of playing board
games on a daily basis, one has probably encountered the word string
back to square one more often in its idiomatic, figurative sense than in
a literal sense. If the idiomatic meaning is familiar and fits the context
where one encounters it, then there is no reason to expect its processing
to be any slower than the processing of the same word string used with
a literal meaning. Recent studies with the aid of eye-tracking confirm
this (Siyanova, Conklin and Schmitt, 2008): whether or not a common
word string is used literally or figuratively makes no difference to read-
ing time, as long as in either case the string is used according to expec-
tations (and thus used in a way that corresponds to strong memory
traces). However, the same eye-tracking studies reveal that L2 learners
do tend to need more time to access the idiomatic meaning of a word
string (ring a bell in the sense of ‘sound familiar’) in comparison with
the same string used literally in a fitting context (ring a bell in the sense
of ‘making a bell ring’). This may be due to insufficient familiarity with
the word string used as an idiom, while the literal meanings of the con-
stituent words are familiar.
While in the above models L1 idioms are considered to be processed
holistically and stored in the mental lexicon as semantic units separ-
ately from the lexemes of which they are composed, other models have
been proposed that treat idioms as analysable (or decomposable) to
82 Optimizing a Lexical Approach

different degrees. According to one of these (Cacciari and Tabossi, 1988),

the meaning of an idiom is said to be ‘distributed’ among the nodes in
memory which encode the words that make it up. This suggests that the
meaning of an idiom resides to some degree in its component words
and to some degree in the memory connections between them. An
interesting feature of this model is the notion of the recognition point
or ‘idiomatic key’, which is the point at which enough of an idiom has
been read or heard for non-literal interpretation to be triggered (that
is, for the whole network of nodes to be activated). For instance, for a
native-speaking receiver, the idiom key for let the cat out of the bag is
likely to lie (depending, no doubt, on the situation of use) after cat and
before bag. If the key occurs early in the idiom, the idiomatic mean-
ing should be accessed more quickly than if the key comes towards or
at the end. In this model, literal processing seems to cease when the
idiomatic meaning has been accessed (but see Titone and Connine,
1999, below). According to a second model that treats idioms as decom-
posable (Glucksberg, 1993, 2001), repeated encounters with an idiom
cause the meaning of each constituent key word in the idiom to take
on a new meaning that reflects the word’s sense in the idiom. Thus, the
meaning of an idiom can be accessed via the appropriate sense of each
of the (now polysemous) words that make it up. This model also allows
for direct look-up of idiomatic meanings that have become entrenched
over time. Titone and Connine (1999) propose a hybrid model in which
there is both retrieval of stored idiomatic meanings and decomposition
of meaning by analysis, with the latter proceeding even after the idiom-
atic meaning has been accessed. The more conventionalized a figura-
tive word string is, the more likely it is to be processed un-analytically.
At the same time, analytical processing occurs to the extent that an idi-
om’s wording is subject to figurative interpretation. In this model, word
meanings are always activated when an idiom is processed and these
meanings play a greater or lesser role in the interpretation of the idiom
depending on the degree to which the idiom is opaque or transparent.
The idiomatic meaning is retrieved at the ‘key’, but, since the meanings
of component words are accessed automatically, literal and idiomatic
interpretations chronologically overlap. The authors’ eye-tracking stud-
ies indicate that interpretation is facilitated when idiomatic and literal
interpretations are similar, but delayed when the two interpretations
clash. Also, in Cutting and Bock’s (1997) model (modified by Sprenger,
Levelt and Kempen, 2006), a connection between figurative and lit-
eral meaning is maintained. In their model, the meaning of an idiom
is stored as a unit at a node which itself is connected to the nodes at
 Semantic Elaboration 83

which the literal meanings of the individual words are stored. In other
words, the literal senses of the constituent words of the idiom remain
present in the periphery of the network. The thesis that literal meaning
is present during idiom interpretation (and thus that a certain amount
of semantic analysis of idioms takes place) has been corroborated in the
context of L2 idiom interpretation by Cieślicka (2006, 2008). We will
return to the implications of Cieślicka’s experiments further below.
Apart from recent experiment-based psycholinguistic models which
cast doubt on the thesis that idioms are by definition processed non-
compositionally, a number of cognitive scientists and semanticists have
pointed out that virtually all idioms have an internal semantic struc-
ture that is to some degree in line with the overall idiomatic meaning.
It has even been observed of the idiom kick the bucket, which has so
often served as the paragon of non-decomposability, that its meaning
nevertheless shows the influence of the verb kick (Glucksberg, 1991).
That is, because kick denotes a quick, brief action, kick the bucket cannot
naturally be used to refer to a protracted demise.13 Glucksberg illustrates
this by means of the relatively plausible John kicked the bucket in a car
accident and the much odder John lay kicking the bucket due to his chronic
illness. On the assumption that the internal semantic structure of idi-
oms comes into play in L1 processing in languages other than English,
it is reasonable to believe that, when learning L2 idioms, teenagers and
adults bring with them the potential to process these expressions in a
similar fashion, all the more so since, if Wray (2002) is correct, after
childhood there is a decline in one’s propensity to holistically learn
word strings.
Summing up, it looks as though adopting a somewhat non-holistic
mode of processing idioms may not be so unnatural in adult native
speakers after all, and it may even be quite natural among adults learn-
ing an L2. In the remainder of this section we will review evidence
that learners’ inclination to process L2 idioms in terms of their internal
semantic structure and/or literal meaning can be channelled in ways
that are quite fruitful for learning. Let us start by assessing the peda-
gogical use of imagery (and thus the stimulation of dual coding) based
on literal readings of (constituent words of) idioms.
Boers (2001) reports an experiment in which language majors were
first asked to look up in a monolingual dictionary the meaning of ten
English idioms they were not yet familiar with: pass the baton, champ
at the bit, a poisoned chalice, a chink in your armour, haul someone over the
coals, go off at half cock, a steady hand on the tiller, gird your loins, run some-
one ragged and a dummy run. The experimental students were given the
84 Optimizing a Lexical Approach

supplementary task of writing down a hypothesis about the origin of

each idiom while the control students were asked to invent a commu-
nicative situation in which each idiom could be used. It was assumed
that both tasks would involve a comparable investment of cognitive
effort (and comparable richness of processing). It was hoped as well that
hypothesizing about the literal origin of an idiom would summon up
in the learner’s mind an image of a concrete scene and that dual coding
(for example, associating pass the baton with a mental image of a relay
race) would then occur. One week later, the students were presented
with the dictionary definitions of the idioms and asked to try to write
down the corresponding idioms. The students who had hypothesized
about the origin of the idioms (and who – presumably – had associ-
ated the expressions with concrete scenes) outperformed the others
significantly on this post-test, by 21%. Five weeks after the dictionary
look-up task, the students were presented with the list of ten idioms and
asked to write down what they remembered of the (figurative) mean-
ings. Again, the students who had hypothesized about the origins of
the expressions did significantly better, by 34%. The students’ hypoth-
eses about the origins did not always correspond to the etymological
explanations given in idiom dictionaries, but, as students had first of all
familiarized themselves with the figurative, idiomatic meaning of the
expressions, their conjectures about origins were nevertheless plausible.
For example, a number of students associated pass the baton with the
scene of an orchestra conductor passing on his baton to a successor
rather than with the scene of an athlete in a relay race passing on the
baton to a team mate. But both scenes are essentially compatible with
the figurative meaning of the idiom.
Encouraged by these results, Boers, Demecheleer and Eyckmans
(2004a) conducted an experiment with the aid of on-line exercises,
whereby one group of language majors were first prompted by means of
a multiple-choice question to hypothesize about the origin of idioms,
after which their hypotheses were corroborated or amended. Box 5.1
gives some examples.
It is important to note that the feedback given in the ‘identify-the-
source’ task in Box 5.1 does not explicitly mention the figurative, idiom-
atic meaning of the expression. It was going to be part of the challenge
for the students to establish the connection between that meaning and
the origin of the idiom on their own. By contrast, another group of
students were given multiple-choice exercises targeting the figurative,
idiomatic meaning of the same expressions. Box 5.2 gives examples.
The correct response per item was given as feedback.
 Semantic Elaboration 85

Box 5.1 ‘Identify-the-source’ multiple-choice items and feedback

What domain of experience do you think the following idiom comes from?
SHOW SOMEONE THE ROPES
A. Punishment B. Sailing C. Sports

Feedback: Sailing. A novice sailor needs to be taught by an experienced sailor

which ropes he should handle.

What domain of experience do you think the following idiom comes from?
CUT NO ICE WITH SOMEONE
A. Sailing B. Food C. Sports

Feedback: Ice skating. If the blades of your skates are too blunt, they will not
cut into the ice.

What domain of experience do you think the following idiom comes from?
JUMP THE GUN
A. Punishment B. Sports C. War

Feedback: Sports. A contender who jumps the gun sets off before the starting
pistol has been fired.

What domain of experience do you think the following idiom comes from?
RUN THE GAUNTLET
A. Sports B. Commerce C. Punishment

Feedback: Punishment. Running the gauntlet used to be a form of punishment

in the military in which the wrongdoer was forced to run between two lines
of men armed with sticks, who beat him as he passed.

After having tackled one or the other of the multiple-choice tasks,

the students were directed to a post-test, which consisted of fill-in-
the-blanks exercises. Box 5.3 gives examples of the kinds of context
that were used to elicit the previously encountered idioms. The correct
responses were presented to the students as feedback.
The idioms were presented in sets of 25 per stage, that is, students first
did one of the two types of multiple-choice exercises on 25 idioms and
subsequently tried the gap-fill items on the same 25 idioms. Altogether,
six such sets of idioms were presented to the students. The students
who had only been made aware of the idioms’ origins outperformed the
students who had learned the idioms’ meanings in the gap-fill exercises
significantly, by 11%. This not only suggests a mnemonic effect of dual
coding, but it also suggests that information about the original, literal
86 Optimizing a Lexical Approach

Box 5.2 ‘Meaning’ multiple-choice items

What is the meaning of the following idiom?

SHOW SOMEONE THE ROPES
A. Disclose the truth to someone
B. Give someone a severe penalty
C. Teach someone how to do a task

What is the meaning of the following idiom?

CUT NO ICE WITH SOMEONE
A. Have a misunderstanding
B. Get on well with someone
C. Make no impression on someone

What is the meaning of the following idiom?

JUMP THE GUN
A. Defend someone at your own peril
B. Do something before the appropriate time
C. Be startled by an unexpected event

What is the meaning of the following idiom?

RUN THE GAUNTLET
A. Run away from your hometown
B. Be in a position of power
C. Go through an unpleasant treatment

Box 5.3 ‘Gap-fill’ post-test

Fill in the missing word.

When I started working here as a novice, nobody bothered to teach me how things
were done around here. I had to find out all by myself how to do my new work
properly. You could say that nobody showed me the ____________.

Fill in the missing word.

Scientists argue that high voltage power lines increase the risk of cancer, but their
arguments cut no ____________ with the big bosses of the electricity industry.
The scientific evidence does not seem to make any impression on them.

Fill in the missing word.

Although we had agreed not to tell anyone about my pregnancy until we were
absolutely certain about it, my husband jumped the ____________ and told his
parents straightaway.

Fill in the missing word.

When her fellow-students found out she had started a relationship with one of their
lecturers, she had to put up with a lot of verbal abuse. Her fellow-students really
made her run the ____________.
 Semantic Elaboration 87

use of the expressions was often sufficient for the students to be able
to infer their idiomatic meaning sufficiently well to match them with
(admittedly very transparent) contexts.
According to the studies by Cieślicka (2006, 2008) referred to above,
language learners often spontaneously try to interpret idioms via a
literal reading of the expressions. However, an obvious prerequisite
for this interpretation strategy is that the literal meaning of the con-
stituent words must be known by the learner. For example, learners
would have to understand the word tether in at the end of my tether
and the word keel in on an even keel to try and interpret the idiom
via a literal reading. It is therefore not surprising that, in the Boers,
Demecheleer and Eyckmans (2004a) experiment, half of the time the
students failed to pick the correct answer to the ‘identify-the-source’
multiple-choice exercises, and so learned about the idioms’ origins
only when reading the feedback. An example in the experiment was
the idiom follow suit, which all students related to the source domain
of clothes instead of card games. It is clear that homonymy of key-
words of idioms renders learner-autonomous identification of the
expressions’ origins more unlikely, especially when it is the less fre-
quent sense of the word that needs to be activated for a learner to have
any chance of adequately interpreting the idiom. The idiom may also
contain an obsolete word (as in run the gauntlet and in the doldrums).
Even when an idiom is made up entirely of familiar words, these are
often multi-interpretable either because of vagueness or polysemy. For
instance, in show someone the ropes, are the ropes on a sailing vessel
or around a boxing ring? In The gloves are off, are these boxing gloves
or the gloves of, say, a bricklayer? In a shot in the arm, is it gunshot or
an injection? In get a leg up, is the leg that of a human being or a dog?
These examples indicate why students need help to guess the origins
of idioms and, on that basis, to infer the correct, conventional figura-
tive meanings so that they can reap the mnemonic benefits of dual
coding in particular and elaborated coding in general. For the strategy
to be fruitful, students need guidance from the teacher (or materials
writer). That is why idioms should be counted as chunks which are
highly teachable. Firstly, teachers can enhance the memorability of
these chunks through the relatively simple intervention of resuscitat-
ing their literal origins; secondly, without teaching, durable learning
is much less likely to occur.
In a follow-up to the above experiment, again with the aid of on-
line exercises, Boers, Demecheleer and Eyckmans (2004a, b) gave a new
cohort of language majors all three exercise types to do, in the fol-
lowing order: (1) ‘meaning’ multiple-choice, (2) ‘identify-the-source’
88 Optimizing a Lexical Approach

multiple-choice, and (3) the gap-fill post-tests. By giving the ‘mean-

ing’ task first, it was possible to exclude from the post-test analysis
the idioms which students already appeared to comprehend prior to
the treatment. The research question in this follow-up experiment
was whether giving information about the origins of the idioms as a
stimulus for dual coding was equally beneficial for all idioms in the
package. This time a total of 275 idioms were targeted. The mean score
on the gap-fill post-test items concerning unknown idioms was 68%,
but there was a small group of idioms which students appeared to find
especially hard to recollect. These were expressions whose (relatively
culture-specific) source domain the students were unfamiliar with –
namely, idioms originating from cricket (such as bat on a sticky wicket
and hit someone for six). It is possible that the students found it hard to
conjure up a clear image to associate with these expressions. To cater
for the likelihood that the students were not familiar with the source
domain of cricket, the explanations given as feedback to the ‘identify-
the-source’ exercises had been made slightly more detailed and thus
longer than average, and it is therefore possible that the students did
not always go to the trouble of reading through all relevant mater-
ial, especially as it was presented on a computer screen. Apart from
these few difficult idioms, the mnemonic effect of giving explana-
tions about the literal origin of an idiom was generally just as strong
for a novel one whose origin students could not initially guess (such as
follow suit – card games) as for a novel one whose origin they did guess
(such as break ranks – warfare).
With a view to estimating the longer-term effects of informing learn-
ers of the original, literal meaning of the expressions, the students who
participated in the experiment just discussed were presented with the
same gap-fill exercises as before but this time after a two-year time lapse.
On completion of the gap-fill exercises, they were also re-presented with
the ‘identify-the-source’ multiple choice exercises concerning the same
idioms. The mean score obtained in the gap-fill exercises was a stunning
85%. It is very possible, of course, that these students (being language
majors) had consolidated their knowledge of the target expressions dur-
ing the intervening two years through incidental encounters. In any
case, it is interesting that if the students managed to recollect an idiom,
they also typically remembered its source domain. The correlation coef-
ficient between correct gap-fill items and correct ‘identify-the-source’
items was extremely high (.80). This strongly suggests that the students
had indeed learned the idiomatic meaning of the expressions in asso-
ciation with the information about their origins, and hence that the
 Semantic Elaboration 89

idioms had to some mnemonically effective degree been dually coded

in long-term memory.
In yet another experiment that made use of the same battery of on-
line idiom exercises, Boers, Eyckmans and Stengers (2007) aimed to
establish which sequence of the tasks was most beneficial for retention.
Some groups of students first did the ‘identify-the-source’ multiple-
choice exercises, then the ‘meaning’ multiple-choice exercises and
finally the gap-fill post-test. For other groups, the first two stages were
reversed so that they did the ‘meaning’ exercises before tackling the
‘identify-the-source’ exercises. The mnemonic effect of informing
students of the origin of the idioms seems to have been optimized
by giving students access to this information before they tackled the
‘meaning’ exercise. The mean scores on the gap-fill post-tests were
around 81% under this condition, as compared to 72% when the
students found out the origins only after learning the figurative mean-
ings. We speculate that the former sequence adds deeper processing
to the learning process because the students can use their knowledge
of the literal, original usage of the expressions in trying to figure out
their idiomatic meaning – which makes it more likely that students
will engage in insightful problem-solving rather than blind guessing
in the meaning exercise. This account is supported by the finding that
knowledge of the origins of the expressions was associated with higher
mean scores on the ‘meaning’ multiple-choice exercises. For students
in the ‘origin first’ condition, the mean score was 83% whereas the
mean score of the students who had not yet been furnished with infor-
mation about origins, was 68%.
A weakness of the battery of on-line exercises we described above is
obviously that, in the two multiple-choice tasks, the idioms are pre-
sented with no context. In authentic texts, students encounter idioms
in context, and contextual clues can direct them towards an appropri-
ate interpretation of the idiomatic meaning of unfamiliar phrases. In
such cases, students might consider information from the teacher about
idiom origins to be unnecessary or even distracting. However, we know
from the following experiment, which was part of the Boers, Eyckmans
and Stengers (2007) study, that context seldom suffices to disambigu-
ate unfamiliar idioms and that information about the origins of the
idioms can effectively point students towards successful interpretation.
In this pen-and-paper experiment, 16 English idioms were presented
to the students in verbal contexts that were borrowed from the Collins
Cobuild Dictionary of Idioms (2002), under the assumption that its com-
pilers were likely to have selected from all the authentic contexts in
90 Optimizing a Lexical Approach

their corpus precisely those judged to illustrate use and meaning most
clearly. The students were asked to write down what they thought the
(contextualized) idioms meant. After the task sheets were collected, the
same idioms in the same contexts were re-presented to the students but
this time accompanied by a brief explanation about the origin of each
idiom. As in the on-line exercises (see above), the origin hints men-
tioned neither the idiomatic meaning nor the connection between that
meaning and the original usage. It was up to the students to make that
connection themselves. Box 5.4 gives some examples of the items and
the hints.
Most of the time (61%), the contexts were insufficient for the students
to correctly interpret the idioms. Moreover, some of the few correct
interpretations may have been of idioms the students were already
vaguely familiar with. The added hints about the origins of the expres-
sions raised the mean figure for correct interpretation by 30%.
In light of the above, it seems safe to say that it is sound practice to
make students aware of the original, literal usage of idioms occurring in
classroom materials. While this can be done after the figurative mean-
ing is clarified, the mnemonic benefits may be greatest when students
are made aware of original, literal usages beforehand, since this raises
the likelihood that their initial inference will be correct. This way of

Box 5.4 Examples of contextualized idioms accompanied by ‘origin’ hints

Task: write down what you think the underlined phrases mean

‘Dozens of tanks are being put through their paces to check that they’re
running correctly before they’re subjected to the rigours of the Saudi
Arabian desert.’
Hint: Originally the different paces a horse is trained to perform.

‘Almost all the oil companies were making money hand over fist.’
Hint: Hand over fist was originally used in nautical contexts with reference
to the movement of a sailor’s hands when rapidly climbing a rope or hauling
it in.

‘Defeat on this embarrassing issue might just tip the PM into throwing in
his hand.’
Hint: Hand refers to the set of cards you are holding in a card or poker
game.

‘As a diplomat he has impressed all sides by his ability to negotiate and his
willingness to roll with the punches.’
Hint: In boxing, a boxer rolls with the punch by moving his body away from
an opponent’s blow so as to lessen the impact.
 Semantic Elaboration 91

ordering the two steps – first consideration of the origin hint and then
guessing – also seems likely to take less time than guessing, considering
the origin hint, and then guessing again if the first guess was wrong.
Either way, this technique is most applicable when the teacher judges
that a text affords insufficient clueing about what an idiom means.
Let us now return to the on-line idiom exercises described above.
With a view to facilitating dual coding, especially in cases where the
students might find the verbal explanations insufficiently clear, pictori-
als were added to the feedback students received after the ‘identify-the-
source’ multiple-choice items. Box 5.5 gives examples.
Boers et al. (2008) report that new cohorts of students who were given
this pictorially enhanced version of the ‘identify-the-source’ exercises
obtained (even) higher scores on the subsequent ‘meaning’ exercises
than did previous cohorts. They also report, however, that the new
cohorts’ scores on the gap-fill post-tests were generally lower than those

Box 5.5 Examples of pictorial elucidation of the origins of idioms

What domain of experience do you think the following idiom comes from?
A CARROT-AND-STICK METHOD
A. Religion B. Animals C. Cooking

Feedback: Animals. Donkeys can be urged on by dangling a carrot before

them and at the same time by hitting them with a stick.

What domain of experience do you think the following idiom comes from?
GO FOR THE JUGULAR
A. Animals B. Entertainment C. Sports

Feedback: Animals. The jugular is a vital vein in your neck. Predators

(e.g. lions and tigers) tend to kill their prey by biting into this jugular.
92 Optimizing a Lexical Approach

obtained in the previous, picture-less version of the battery of exercises.

So, it may be that pictorial elucidation promotes comprehension and
retention of meaning but not form since memory of a picture is very
unlikely to facilitate recall of any particular form of words.14 It is plaus-
ible that when verbal information about the origin of an idiom has
to compete for the learner’s attention with pictorial elucidation, the
latter is likely to be retained better – possibly even at the expense of
the former. This is in line with what experimental psychologists have
termed the ‘picture superiority effect’ whereby, as attested by a con-
siderable body of research, pictures are more memorable than words
(Nelson, Reed and Walling, 1976). However, when words that are to
be recalled are already well-entrenched in the learner’s memory, then
remembering an associated picture may provide access to a mnemonic
pathway leading between the memory trace for the picture and the firm
memory trace for the word whose retrieval is desired. But in attempt-
ing to recollect a word that has left only a weak trace in memory, a
memory search that begins with the trace for an associated picture may
lead no further. This hypothesis is consistent with the results of yet
another study making use of the battery of on-line idiom exercises men-
tioned above (Boers et al., 2009). This time only half of the idioms were
presented with pictorial elucidation of their origins. The other idioms,
which were matched in terms of frequency bands, morphological com-
plexity and mnemonically facilitative features (such as the availability
of L1 cognates or the presence of rhyme and alliteration), came only with
verbal explanations of the origins. Overall, the students’ recollection of
the idioms in the gap-fill post-tests seemed unaffected by the presence
or absence of pictorials in the instructional materials. However, idioms
containing keywords belonging to very low frequency bands (such as
come up trumps, be at the end of your tether, not rest on your laurels, rap
someone on the knuckles, play second fiddle, throw down the gauntlet, ride
roughshod over someone and pass muster), which were thus most likely
to be new to the students, were recollected significantly better when
they had not been presented to the students along with a picture. There
is also evidence that students often remembered the thing or activity
elucidated by a picture but failed to recollect the precise word for it. For
instance, instead of fiddle in play second fiddle to someone, they would
write violin; instead of toss in argue the toss, they would write throw; and
instead of rein in keep a tight rein on someone, they would write rope, and
so on. In short, pictorial elucidation is undoubtedly helpful for receptive
purposes (on condition the pictorials do indeed elucidate rather than,
say, just provide light relief), but there is no reason to assume that a
 Semantic Elaboration 93

picture significantly facilitates retrieval-for-production of the complete

phonological form of an expression whose meaning is weakly known.
(We will say more about this and related issues in Chapter 6, which
begins with a discussion of the difference between forming increas-
ingly elaborated knowledge of the meaning of a chunk and forming
elaborated knowledge of its form.)
Another way of stimulating the dual coding of figurative phrases,
apart from using drawings and pictures, is via mime and physical enact-
ment (see Lindstromberg and Boers, 2008c, for detailed classroom pro-
cedures). An experiment in which pictorial support and miming were
part of one of the conditions is reported by Csábi (2004). Targets for
learning were uses of the verbs hold and keep. One group of the teenage
participants was told that the core meaning of hold involves an agent’s
hand (as in she held the purse in her right hand). Meaning extensions
were then explained via conceptual metonymies and metaphors such
as THE HAND IS FOR CONTROL and POSSESSING IS HOLDING (for example, the
terrorists held them hostage and he did not hold the right certificate). These
participants were told that the core meaning of keep implies a more
durative state of possession (as in you can keep the change). This dura-
tive sense was then invoked to explain other uses of keep as in earn
enough to keep a family and keep that dog out of my study. The teacher
elucidated the core senses by miming or by drawing on the blackboard.
The other group of students was presented with the same examples of
hold and keep, but these were explained by means of translations into
their mother tongue. In an immediate post-test and a delayed post-test
in which students were asked to fill in gaps with either hold or keep, the
former group of students outperformed the latter.
Focusing entirely on the effect of miming, Lindstromberg and Boers
(2005a) conducted an experiment in which the targets for learning
were English manner-of-motion verbs (such as hobble, stagger, teeter, veer,
saunter and slump). Being a ‘satellite-framed’ language in which manner
of motion is often expressed in the verb (for example, stroll) rather than
by an adverbial (walk casually) (Slobin, 2000; Talmy, 1985), English has
a rich repertoire of manner-of-motion verbs that presents learners with
a considerable learning challenge. The challenge is especially formid-
able for learners whose mother tongue (1) is verb-framed (so that manner
of motion is typically expressed not in the verb itself but by means of
an adverbial phrase – as in French Il sortit de la maison en courant), and
(2) affords few L1–L2 cognates (Cadierno, 2004), such as German schwin-
gen and English [to] swing. Moreover, these manner-of-motion verbs are
not only, as a class, frequent in narratives – especially in descriptions
94 Optimizing a Lexical Approach

of action scenes in literature for younger readers (Lindstromberg and

Boers, 2005a) – but have tended to become conventionalized in a wide
range of figurative phrases, as illustrated in Box 5.6.
In the experiment (which was replicated several times, always yield-
ing basically the same results), 24 manner-of-motion verbs were put on
display in the classroom. In one experimental condition, the students
took turns miming the verbs so that their peers could guess which of the
verbs that were on display was being acted out. In the other condition,
students explained the meaning of the verbs verbally so that, again,
their peers could guess which verb was being explained. In both groups,
the activity went on until all the displayed verbs had been successfully
identified twice. An immediate post-test (a gap-filling exercise) targeted
the verbs as used in their literal senses – that is, the senses the students
had just learned. The students who had mimed verbs and guessed which
verbs their peers enacted scored significantly better on this recollec-
tion task than the other group. If a student had enacted a given verb
himself or herself, then the chances of it being recollected were espe-
cially high. This suggests that successful dual coding can also involve
motoric imagery. What we were particularly interested in, however, was
whether the dual coding (be it of a visual or motoric nature) could help
the students interpret phrases in which the manner-of-motion verbs
were used metaphorically. In a delayed post-test, the students were

Box 5.6 Examples of figurative uses of manner-of-motion verbs

• Holmes stumbled on a piece of evidence.

• Sometimes it’s necessary to make a snap decision.
• Hitler was famous for his sudden mood swings.
• The PM dodged all questions about the war.
• The Republicans tried to sway voters with promises of tax cuts.
• She’s trying to wriggle out of her responsibilities.
• The economy is teetering on the brink of a recession.
• Accepting this new job was a leap in the dark, but I’ve learned
by leaps and bounds.
• Occasionally, a patient in pain will lash out at the nurses.
• She’s already shrugged off last week’s disappointment.
• People are referring back to the economic slump of the 1930s.
• I never thought she would stoop so low to get what she wants.
• The negotiations soon began to feel like a tug-of-war.
• Their marriage is on the skids. They’ll end up divorced soon.
 Semantic Elaboration 95

therefore presented with sentences similar to those given in Box 5.6,

each sentence being accompanied by an L1 translation which was delib-
erately kept fairly vague. For example, the Dutch translation of He hurled
insults at his girlfriend did not convey the idea of extreme hostility and
the translation of He stumbled on a crucial piece of evidence did not indi-
cate that the evidence was found incidentally. Students were asked to
evaluate the proposed translations. The results show that students who
had been in the mime group were significantly more likely to spot the
lack of precision in the Dutch translations of these metaphorical usages.
This suggests that deeper comprehension of the literal meanings was
fostered in the miming condition and that this, in turn, aided students’
appreciation of the metaphorical extensions. The experiments reported
by Csábi and by Lindstromberg and Boers also demonstrate how a com-
bination of teaching techniques adapted from Total Physical Response
(TPR) (Asher, 1966, 1988) can serve the semantic elaboration of figura-
tively used words and chunks.
Part of the Boers, Eyckmans and Stengers (2007) study with the aid
of on-line idiom exercises looked beyond the purely mnemonic effects
of dual coding into the issue of in-depth comprehension as well. The
question addressed here was whether information about the origins of
idioms could help students make educated guesses about their usage
restrictions. A new exercise, an ‘identify-the-informal-idiom’ multiple-
choice, was incorporated in the battery. Box 5.7 gives examples of this
new task.
All the students participating in this experiment were presented with
the same ‘identify-the-informal-idiom’ items twice: once at the begin-
ning of the exercise sequence and once again half-way through the
sequence. Half of the student groups met the items the second time
after they had learned about the origins of the expression (through the
feedback on the ‘identify-the-source’ exercise), while the other half met
the items the second time without first having been informed of the
origins. The question was whether the freshly acquired insight into

Box 5.7 Examples of ‘identify-the-informal-idiom’ exercises

In each of the following sets of expressions, one is most typically used in

INFORMAL contexts. Which one do you think it is?

A. lose your shirt B. fall from grace C. take up the gauntlet

A. bury the hatchet B. hit a home run C. touch a nerve
A. keep a finger on the pulse B. run the gauntlet C. jump the gun
96 Optimizing a Lexical Approach

the origins of the expressions would help students rectify the choice of
‘informal’ idioms they had made the first time around. The students who
had been informed of the origins outperformed the others significantly
when they did the exercise again. This suggests that the association of
certain idioms with their source domains can sometimes give a clue as
to the register expressions are appropriately used in. For example, asso-
ciating sticking one’s nose into something with the scene of a dog curiously
poking its nose into things has evident potential to suggest that this is
an informal expression. At a more general level, hand counts of idioms
that are characterized in Speake’s (1999) Oxford Dictionary of Idioms as
‘informal’ reveal that these idioms are not evenly distributed over dif-
ferent source domains. ‘Informal’ idioms were counted in a bank of
1300 English idioms that are retraceable to concrete source domains,
such as gardening (nip something in the bud), clothes (try something on for
size), commerce (wipe the slate clean), fauna (put out feelers), food (know
which side your bread is buttered), handicraft (break the mould), jurisdic-
tion (read the riot act), transport (miss the boat) and the weather (be under
a cloud). The analysis has revealed that the source domains of games
(have an ace up your sleeve, keep your eye on the ball) and entertainment
(a one-man band, play to the gallery) have generated a significantly higher
proportion of informal figurative idioms (29% and 27%, respectively)
than certain other source domains, such as war (break ranks, be in the
front line) and religion (fall from grace) (10% and 5%, respectively). This
leads us to speculate that experiential domains with predominantly ‘ser-
ious’ connotations (such as war and religion) may have been less likely
to generate idioms that are commonly associated with ‘light-hearted’
conversation. Instead, idioms derived from less fraught and formidable
experiential domains (such as games) may generally have been felt to be
more appropriate in informal contexts. To a degree, such usage restric-
tions may have been passed down from one generation to the next.
If there is any truth to this speculation, then re-establishing the link
between idioms and their origins may indeed occasionally serve as a
hint about usage restrictions.

5.2 Organizing lexis

In an instructive survey chapter on vocabulary learning methods,

Sökmen (1997) emphasizes that vocabulary which is somehow organ-
ized (thematically, for instance) is easier to learn than vocabulary in
random lists, although it must be added that grouping words into tight
paradigmatic sets (into sets of near synonyms or sets of antonyms, for
 Semantic Elaboration 97

example) is decidedly not helpful (Erten and Tekin, 2008; Nation, 2000;
Schmitt and Schmitt, 1995). Presenting vocabulary in small sets of
related words that invite the addition of more related items over time
also provides the students with a framework in which to connect newly
encountered words with acquired knowledge, which is known also to
facilitate learning. Michael Lewis also recommends organizing lexis in
order to accelerate learning (1997: 67–85). The organizing principles
he mentions include ‘topic’ (for example, grouping words and phrases
related to the topic of food), ‘situation’ (for example, grouping words
and phrases one tends to use in an administrative office), ‘notion’ (for
example, grouping ways of apologizing) and ‘narration’ (for example,
grouping words and phrases one needs to describe semi-fixed sequences
of events such as developing, manufacturing and marketing a product).
In addition to well-established organizing principles like these (which
have guided the unit-structure of a great many course books), Lewis
more innovatively proposes grouping by kind of metaphor and by col-
location. We will look into the organization principle of collocation
further below. With regard to metaphor, Lewis (1997: 71) refers to work
by Cognitive Semanticists Lakoff and Johnson (1980) and Kövecses
and Szabó (1996), who demonstrate that many institutionalized figura-
tive phrases can be grouped under the headings of more general meta-
phor themes called ‘conceptual metaphors’.15 Let us survey a number
of experiments conducted with a view to measuring the pedagogical
effectiveness of this type of presentation of lexis in the classroom – that
is, presentation in light of metaphor.
In the first of these experiments Kövecses and Szabó (1996) focused
on English phrasal verbs with up and down. One group of students were
asked to study ten phrasal verbs and accompanying explanations of
their underlying conceptual metaphors (such as MORE IS UP and HAPPY
IS UP). Another group of students were asked to study the same phrasal
verbs but were given the L1 translations as an aid. In an immediate post-
test (a gap-filling exercise) the former group outperformed the latter by
almost 9%. Another experiment reported by Boers (2000b) focused on
phrasal verbs instantiating a wider range of conceptual metaphors than
the two addressed by Kövecses and Szabó. One group of students were
asked to study 26 phrasal verbs (each glossed with a synonym) which
expressed a variety of conceptual metaphors (Box 5.8 gives a small sam-
ple). Another group of students were asked to study the same phrasal
verbs listed alphabetically, with each phrasal verb being glossed by
means of multiple synonyms copied from a well-known grammar book.
In an immediate post-test (a text-based gap-filling exercise) that targeted
98 Optimizing a Lexical Approach

Box 5.8 Phrasal verbs grouped per metaphor theme

MORE IS UP; LESS IS DOWN:

turn up the heating; cut down expenses; blow up a story
ACTIVE IS UP; INACTIVE IS DOWN:
set up a business; the machine broke down; the factory was closed down
GOOD IS UP; BAD IS DOWN:
cheer up; feeling down; feel up to a task
VISIBLE IS UP; SEEING IS KNOWING:
look it up in the dictionary; the police turned up late; she never showed up
VISIBLE IS OUT; SEEING IS KNOWING:
find out the truth; it turned out quite difficult; figure out a problem

ten of the phrasal verbs studied, the students in the ‘metaphor’ condi-
tion significantly outperformed those in the ‘synonym’ condition.
The findings by both Kövecses and Szabó (1996) and Boers (2000b)
suggest that metaphor can indeed serve as a helpful, additional organ-
izational principle for the presentation of lexis. Beside the facilitat-
ing effect of simply grouping the target chunks somehow, it is also
likely that this particular grouping principle stimulates dual coding
by raising students’ awareness of the figurative nature of the phrasal
verbs. In pedagogical practice, one could start with a relatively small
number of phrasal verbs grouped under a small number of metaphor
themes and ask students to make additions to the sets as they encoun-
ter more phrasal verbs during the rest of the course. For instance, on
meeting the expressions point something out and things are looking up,
the students could be asked to add these chunks to the appropriate
sets (SEEING IS KNOWING and GOOD IS UP, respectively). First of all, merely
choosing a set would increase the chances of remembering these two
expressions, on account of the cognitive effort involved. Secondly,
making this a routine practice would result in distributed learning
(which is well-known to be beneficial) and also enable students to
connect new items with acquired knowledge, as recommended in
the literature on vocabulary learning generally. The extent to which
students are able (and willing) to autonomously recognize the con-
ceptual metaphors behind phrasal verbs is another matter, however.
Condon (2008) reports a study during which phrasal verbs were
taught in a piecemeal fashion throughout an English course for first-
year university students of Business and Economics. In some groups of
students, half of the phrasal verbs were presented as instances of meta-
phor themes (a exemplified above), while the other half of the phrasal
 Semantic Elaboration 99

verbs, which were instances of the same metaphor themes, were clari-
fied only by means of L1 translations. In other groups of students,
none of the phrasal verbs were taught with reference to metaphor.
Post-test results (scores on gap-fill tests) show that students benefited
from the metaphor-oriented explanations but only concerning phrasal
verbs which had been explicitly taught in that way. There was no sign
of differential uptake between the groups of students when it came to
the phrasal verbs not taught with explicit reference to metaphor. This
suggests that the students did not autonomously relate new items to
the metaphor-based sets established by the teacher even though these
same sets were appropriate. In other words, there was no evidence of
strategy transfer in this study.
Other experiments on the use of metaphor to organize lexis focus on
a range of idiomatic expressions. Boers (2000b) reports an experiment
in which pupils were presented with a list of 18 expressions (borrowed
from Kövecses, 1990) that are used to describe anger and angry behav-
iour. In some classes, these expressions were grouped under headings
referring to conceptual metaphors (see Box 5.9). In other classes, the
same lexis was organized under functional headings referring to whether
the expressions were used to describe sudden anger (for example, blow
up at someone) a more gradual process (simmer down), or angry person-
alities (a ferocious temper). This was done to ensure the same degree of
organization of the input under both conditions, so that any superior
learning effects under the metaphor-oriented condition could not be
dismissed as being merely the result of the lexis being grouped rather
than listed. In an immediate post-test (a text-based gap-filling exercise
targeting ten of the expressions), the experimental group outperformed
the control group significantly.

Box 5.9 Anger-related expressions grouped according to metaphor themes

(based on Boers, 2000b)

Anger as a hot fluid in a container:

anger welled up inside me; I was boiling with anger; she was all steamed up;
she erupted; simmer down; she flipped her lid; I was fuming; she blew up at me.
Anger as fire:
an inflammatory remark; adding fuel to the fire; he kept smouldering for days;
she was breathing fire; she exploded.
Angry people as dangerous animals:
a ferocious temper; don’t snap at me; don’t bite my head off; he unleashed
his anger; that ruffled his feathers.
100 Optimizing a Lexical Approach

Beréndi, Csábi and Kövecses (2008) conducted a similar experi-

ment, also targeting idiomatic expressions related to anger (including
additional phrases such as hot under the collar, fly off the handle and
being beside oneself ). All the targeted expressions were embedded in a
text which students were asked to read. Afterwards they were given
the task of translating the expressions into their mother tongue. One
group of students were shown a display of the expressions grouped
under headings referring to conceptual metaphors. Another group of
students were shown a display of the expressions listed alphabetically.
The former group produced the better translations and also scored
significantly better in a post-test (a gap-fill task) meant to elicit the
target phrases. This suggests not only that metaphor provides a viable
way of organizing figurative lexis, but also that it can be exploited
in order to stimulate cognitive involvement in the form of educated
meaning-guessing.
The effectiveness of using metaphor headings as clues to help students
guess the meaning of idioms is also investigated by Skoufaki (2008). In
her study, students read texts containing expressions that instantiate
MORAL IS UP (such as take the high ground), MORAL IS CLEAN (dish the dirt
on someone), SEEING IS KNOWING (blind someone with science), and HOLDING
IS CONTROLLING (get a handle on something). In one condition, the texts
were accompanied by an explanatory glossary in which the expressions
were grouped under their metaphor themes. In another condition,
the expressions were also grouped under their metaphor themes but
it was the students’ task to write down what they thought the expres-
sions meant. The meaning of unfamiliar words inside the idioms was
clarified to facilitate the task. After this cued meaning-guessing, the
students received the completed glossaries so they could check their
interpretations. In a gap-fill post-test designed to elicit the target idioms,
these students significantly outperformed the group that had received
the complete glossary from the start. This indicates that the extra cog-
nitive effort invested during the meaning-guessing activity left more
durable memory traces as compared to reading clarifications provided
by someone else. It must be mentioned, however, that fewer than 30%
of the cued meaning-guessing attempts were actually successful. It does
not appear, therefore, that corrective feedback can be dispensed with
after a guessing exercise.
Another part of Beréndi, Csábi and Kövecses’ (2008) study investigated
the feasibility of enhancing students’ cognitive involvement during the
contemplation of grouped figurative lexis. A group of language majors
were presented with a list of figurative expressions referring to the
 Semantic Elaboration 101

concept of anger again and they were asked to group these expressions
according to metaphor themes. Two such themes (ANGER IS LIKE A HOT
FLUID IN A CONTAINER and AN ANGRY PERSON IS LIKE A DANGEROUS ANIMAL)
were proposed to the students. Two more themes were expected to be
discerned by the students themselves. This metaphor-recognition task
turned out to be well beyond the students’ abilities and, consequently,
their categorization of the expressions was highly flawed. The students’
recollection of the idioms in a gap-fill post-test was also much poorer
than that shown in a previous experiment (see above) where students
had simply been presented with the metaphor themes and the idioms
grouped accordingly. These findings, like many others we report in this
book, indicate that as a general rule teachers (or materials writers) must
provide some guidance during semi-learner-independent meaning-
guessing activities.
Instead of using conceptual metaphors as a means of grouping figura-
tive idioms, one may also adopt a more historical approach and group
idioms according to their experiential source domains. Examples are
given in Box 5.10.
Again, in practice it is pedagogically most sound to create such sets at
first with a small number of items encountered in classroom materials
and to encourage the students to add items to the sets as the course pro-
ceeds. To heighten the likelihood of students periodically encountering
more idioms that derive from source domains that have already been

Box 5.10 Grouping idioms according to their source domains

Idioms from card games:

call someone’s bluff, force someone’s hand, follow suit, turn up trumps, above
board, get something off your chest, have something up your sleeve, get a raw
deal, not miss a trick
Idioms from warfare:
be up in arms, fight a rearguard action, stand shoulder to shoulder, come under
fire, a last ditch attempt, break ranks, stick to your guns, be at loggerheads, a
Trojan horse
Idioms from boxing:
not pull your punches, a body blow, go the distance, lower your guard, take it on the
chin, the gloves are off, be on the ropes, down for the count, throw in the towel
Idioms from seafaring:
clear the decks, a leading light, a shot across someone’s bows, take something on
board, steer clear of someone, in the doldrums, break the ice, show someone the
ropes, a close call, be on an even keel, in the wake of something, be left high and
dry, the tip of the iceberg, pass muster, all at sea
102 Optimizing a Lexical Approach

broached, one can focus right from the beginning on domains that are
known to have generated considerable numbers of idioms still in cur-
rent use. According to quantitative data reported in Boers and Stengers
(2008a), source domains that fit the bill in English include those listed
in Box 5.10 as well as certain other popular games and sports such as
horse racing (win hands down, neck and neck, off the rails, too close to
call, in the running, a dark horse, keep a tight rein on someone, put some-
one through their paces, ride high, give someone free rein, across the board
and so on). Lists of ‘frequent’ idioms grouped according to more source
domains can be found in Boers and Lindstromberg (2008b: 389–91).
One advantage of grouping idioms by source domain is that it seems
to have potential for making learners more aware of the culture behind
the language being studied. For example, the fact that English has such
a large number of horse-related idioms (here are a few more: from the
horse’s mouth, flog a dead horse, hold your horses, get on your high horse
and change horses in midstream) hints at the past and present economic
and recreational importance of horses and, indeed, at the affinity the
British have traditionally felt towards this animal. Treatment of idioms
in relation to source domain affords plentiful opportunities to interpol-
ate vivid anecdotes that can make discussion of culture more interesting
for more learners (especially if they are teenagers). For instance, against
the background of an accumulated set of English horse idioms, one can
quite relevantly mention that native Anglophones tend to frown on
the idea of eating horse meat (which may help explain why the idiom
I could eat a horse has the force that it does).
Let us now move on to ‘collocation’, which Michael Lewis (1997:
71, 78–81) introduces as an additional principle by which lexis can be
organized. Lewis proposes the use of collocation boxes, such as the exam-
ples given in Figures 5.1 and 5.2.

Dismiss

Express

Meet Objection

Raise

Withdraw

Figure 5.1 Example (1) of a collocation box proposed by Lewis (1997: 78)
 Semantic Elaboration 103

Bleak

Dismal

Daunting Prospect

Exciting

Vague

Figure 5.2 Example (2) of a collocation box proposed by Lewis (1997: 79)

In order to avoid confusion and to reduce the learning load, Lewis rec-
ommends limiting the number of collocates in these boxes to five (of
course some words have even fewer common collocates and so for them
some boxes would remain partly unfilled). He sensibly adds that it is
not necessary to fill all five slots from the start but that it might rather
be a good idea to spread learning over time.
Lewis strongly advocates that learners use personalized vocabulary
notebooks:

Probably the only learning aid which every learner has is a vocabu-
lary notebook. [...] Given the central importance of the learner’s lexi-
con, the role and format of a truly lexical notebook deserves our
close attention. (1997: 75)

At first, it is the teacher’s task to train students to make good use of their
notebooks:

New language may be collected in class during a particular lesson, or

in a less organized way over a period of time, then class time used to
review and organize it to ensure what is recorded is accurate, useful,
organized and retrievable. [...] Using such a book in class over the
period of a course should leave learners fully equipped to continue
using it on their own. (1997: 76)

In other words, eventually the onus is on learners to select lexis for

their notebooks and to record that lexis in a helpful manner. The prin-
cipal advice Lewis has for them is to record whole expressions rather
than single words, and one structured way of doing that is by using
104 Optimizing a Lexical Approach

collocation boxes. With regard to these, Lewis cautions that:

[t]he language should be edited before being recorded, first with the
teacher’s guidance and later by learners themselves. The idea is not to
fill the box with any words which could collocate but to selectively
record only those which collocate strongly or frequently. (1997: 80)

From the above quotations it is clear that Lewis’s objective with respect
to lexical notebooks is learner autonomy. This is in accordance with
his general preference (which we discussed in Chapter 3) for giving pri-
ority in class to the teaching of strategies rather than prolonged focus
on individual items. Leaving aside the assumption that a high pro-
portion of learners will take up the use of personal vocabulary note-
books, a problem with Lewis’s recommendation here is that it must be
hard for learners to decide on their own whether a particular instance
of a word string betokens a word partnership strong enough and fre-
quent enough to merit recording. Even teachers may find it hard to
make this decision. For example, among the five collocates given for
the noun objection in Figure 5.1, only two (raise and withdraw) actually
come up on the Collins Cobuild on-line collocations sampler. Neither
are the missing three mentioned as collocates of objection in the Oxford
Collocations Dictionary for Students of English (2002). Similarly, two of the
five adjective collocates given for the noun prospect (dismal and vague)
in Figure 5.2 are absent from the list generated by the sampler as well as
from the relevant entry in the collocations dictionary. In both colloca-
tion boxes, these relatively weak collocates could easily be substituted
by other words which do occur in the lists generated by the sampler and
which are mentioned in the dictionary.16
It could be argued, of course, that it is up to learners to check appro-
priate sources of information such as a collocations dictionary so as to
ensure that what they record in their notebooks is likely to be useful. To
this one could respond that any learner diligent enough to consult a col-
locations dictionary might as well keep on consulting it and just forget
about the notebook (unless its entries reflect the student’s assumptions
about what is likely to come up on a particular exam), since it is not clear
that the mere act of writing word combinations down in a notebook (or
anywhere else) is likely to trigger as much elaborative processing as is
required for durable memory traces to be created. And yet, it is possible
to pre-design collocation boxes in ways that stimulate more insightful
groupings. For one thing, collocates can be grouped according to their
shared semantic prosody. The collocation box for prospect in Figure 5.2,
 Semantic Elaboration 105

for instance, could be re-designed so that there are two separate group-
ings, one with room for three positively connoted collocates (an attract-
ive prospect, an exciting prospect, an inviting prospect) and three negatively
connoted ones (an alarming prospect, a bleak prospect, a daunting prospect).
With a view to adding cognitive involvement, students could be asked
also to tick the collocate which seems the most positive and the one
that seems the most negative (see De Rycker, 2004 for additional design
possibilities). Further, collocation boxes can be designed with the aim
of exploiting the fact that some words collocate with clusters of words
from the same semantic field. For instance, one strong collocate can be
established for each of two or three base words, for example:

commitbase word → a crimecollocate

conductbase word → an investigationcollocate
performbase word → a playcollocate

Students can then be asked to range other nouns they are likely
already to be familiar with, before or after (or over or under) the appro-
priate base word (commit ⫹ a robbery /a murder /an offence /suicide or
conduct ⫹ an experiment /an inspection /a survey / research; perform ⫹ a
song /a dance /a solo /a ceremony). Thereafter, they can be encouraged
to add more collocates as they are encountered in later coursework (for
instance, commit ⫹ adultery /arson /a burglary /an assault /an atrocity; con-
duct ⫹ an interview /an inquiry; perform ⫹ a miracle). Pre-designed record-
ing formats such as these have the potential to stimulate insightful
processing and cognitive involvement – and, thereby, elaborated men-
tal representations of meaning. They also make it more likely that what
is recorded will be adequate, apt and accurate.
In brief, we accept that keeping a personalized notebook of chunks
may help some learners to develop a sense of responsibility for their
own learning. We believe, though, that the most effective notebook will
be one that includes instructions and templates designed to maximize
the proportion of useful entries and to stimulate cognitive involvement
and elaborative processing.
6
Structural Elaboration

6.1 Phonological motivation in the formation and

standardization of chunks

In Chapter 5, we described and assessed various routes of elaboration,

taking as a starting point the semantic motivation of certain groups
of chunks. Some of those routes were shown to lead learners towards a
deeper appreciation of the meaning of figurative idioms in particular
and were proposed as aids to learning, mostly for receptive purposes.
In these cases, the elaboration is predominantly meaning-oriented.
Other routes we presented, such as the ones suggested for clustering
collocates that share the same semantic prosody (for example, the
collocates of commit), are more form-oriented; they can help learners
appreciate the lexical makeup of chunks and thus foster chunk learn-
ing for productive purposes too. In the present chapter, we will explore
an additional pathway for such form-oriented, or ‘structural’, elabor-
ation. We will show both that the lexical composition of a substantial
number of chunks is phonologically motivated and that this kind of
motivation can easily be exploited by teachers (and materials writers)
in order to trigger mnemonically effective form-focused elaboration in
the students. Further, we will argue that many word sequences have
become standardized owing to the appeal and/or mnemonic effect of
phonological repetition.
Phonological repetition is conspicuous in everyday language. We see
this in conventionalized phrases which show, for instance:

(i) word repetition: on and on, wonder of wonders, fair’s fair

(ii) rhyme: meals on wheels, make-or-break, flower power
(iii) slant rhyme: in actual fact, last gasp, have all the hallmarks of

106
 Structural Elaboration 107

(iv) alliteration: man-made, cash-and-carry, from pillar to post)

(v) assonance: small talk, jump the gun, take a good look
(vi) consonance: casual acquaintance, further afield, wide awake.17

Further below we will present direct evidence that types of repetition

(ii), (iv) and (v) make chunks especially memorable and indirect evi-
dence that type (iii) and, in some cases, type (vi) are unusually mem-
orable as well. We will show too that teachers can play a valuable role
in helping students capitalize on the extra memorability of ‘repetitive
chunks’. First, however, let us turn to the evidence that phonological
repetition (alliteration, for instance) is indeed a motivating factor in
the formation of chunks. Also, in order to be able to argue that phono-
logical repetition merits classroom attention, we need to demonstrate
that it is at play in a relatively large proportion of chunks, at least in
English.
It is easy to argue through examples that – all else being equal – word
strings which show sound repetition are at a comparative advantage in
the competition for chunk status within a language community. We
see, for instance, that the rhyming steer clear of is far more standard-
ized than, say, sail clear of, swerve clear of, navigate clear of, walk clear of.18
Essentially the same thing can be said of It takes two to tango versus It
takes two to waltz / ... jive ...; faint praise versus weak/feeble praise; above
board versus above the table; and so on. On the other hand, the devil’s
advocate may rightly say that this evidence is too anecdotal and that
other factors than phonological repetition, or even sheer coincidence,
may lie behind conventionalization of steer clear of, It takes two to tango
and the other examples just given. After all, if alliterating word strings
(for example) had a competitive advantage over non-alliterating ones,
then one would have to explain why, for instance, a couch potato and
a thorny problem have become standardized rather than the arguably
more euphonious couch cabbage and prickly problem and why fight to the
finish is not as common as fight to the end. To truly make a case for
phonological motivation, one has to present statistical evidence that
the chunk repertoire of a particular language, in this case English, con-
tains a greater proportion of repetitive chunks than would be predicted
by chance.
Boers and Lindstromberg (2008a) made use of collocation samplers
to show how, for example, seek – in contrast with its synonym look
for – has an impressive number of strong collocates which either begin
with s (such as solace, solitude, sanctuary) or which show consonance
after a weak initial vowel (asylum). However, so far as we know, such
108 Optimizing a Lexical Approach

investigations have been too small in scale to strongly suggest that

phonological motivation at all frequently trumps semantic motivation
in collocation formation. (For example, it is possible that seek seeks the
company of other relatively formal words, some of which just coinci-
dentally begin with s.)
Another way of attempting to evaluate the phonological-motivation
hypothesis consists in investigating idioms that have well-attested vari-
ants, that is, which permit a choice between one content word that
contributes to phonological repetition and another that does not. We
screened the Collins Cobuild Dictionary of Idioms (2002 edition) for such
cases, excluding all those that are signalled as having regional differ-
ences (British versus American versus Australian). This yielded a set of
51 pairs, including cut it close / cut it fine, in full flow / in full spate, larger
than life / bigger than life, a rough ride / a bumpy ride, sing the same song /
sing the same tune, tall tales / tall stories, search high and low / hunt high
and low, plain as day / clear as day and take the bait / rise to the bait. To
estimate whether the phonologically repetitive variants might be more
commonly used than their ‘competitors’ we conducted an ‘exact word-
string’ Google search for both variants in each pair and compared the
number of hits this generated. The phonologically repetitive variants
came out on top significantly more often than the non-repetitive ones.
This lends some additional support to the hypothesis that phonologic-
ally repetitive word sequences can have a comparative advantage in the
competition for conventionalization; but a word of caution is again in
order, as it would be very difficult, if not impossible, to determine in
every case which of the two variants in each pair was in circulation first
(diachronically speaking) and may thereby have had a head start in the
competition.
Much more convincing evidence in support of the claim that phono-
logical repetition plays a part in the formation of multiword lexis is
provided by the following larger-scale investigation. We screened the
Macmillan English Dictionary for Advanced Learners (2007, 2nd edition;
hereafter, MED) for all multiword entries (but not sub-entries) starting
with /b/, /d/, /f/, /k/, /p/, /s/ and /t/. These include compounds with or
without hyphenation (baby boom, blue-blooded), binomials (black and
blue), strong collocations (bad breath), short figurative phrases (blind
man’s bluff ) and proper nouns (Bible Belt). We decided to make an inven-
tory of multiword headwords that start with these seven consonants for
two reasons. Firstly, we were focusing on alliteration as type of phono-
logical repetition because – like rhyme – it is easy to recognize and also
because we knew from previous counts (Boers and Stengers, 2008b) that
 Structural Elaboration 109

alliteration is comparatively common – far and away more common

than rhyme, for instance. Thus, since it was alliterative multiword units
we wanted to count, we needed (by definition) headwords starting with
a consonant. Secondly, we chose to screen the dictionary for chunks
whose headwords start with these seven consonants in particular
because the sections for these consonants are relatively large and thus
likely to contain relatively large numbers of multiword entries per con-
sonant. It turned out that the seven sections each yielded over 550 items
of the relevant kinds. As we were interested in repetition of phonemes
rather than graphemes, screening was not entirely straightforward.
For example, in screening for /p/-headwords we skipped ph-entries, in
screening for /t/-headwords we skipped th-entries, and in screening for
/s/-headwords we skipped sh-entries. In order to make the inventories
of multiword units starting with /k/ and with /s/, we also screened the
dictionary section on c-words (giving entries such as cold comfort and
city slicker). In total, 5667 multiword items were identified. We then
counted the number of instances of word-initial consonant repetition
(from the headword, such as cold, to a given collocate, such as comfort),
the ultimate aim being to determine whether /k/-words are especially
likely to follow /k/-headwords, whether /b/-words are especially likely
to follow /b/-headwords, whether /d/-words are especially likely to fol-
low /d/-headwords, and so on. The Fisher Exact Test (Chi-Square statistic)
was applied to the resulting tallies to determine whether any observed
trends were significant. Tables 6.1 and 6.2 exemplify the procedure.
Table 6.3 gives an overview of the results for the seven consonants.
Note that the percentage figures in that table are inevitably skewed by
the differences in number of multiword units collected per consonant.

Table 6.1 /b/_ ⫹ /b/_ vs. /f/_ ⫹ /b/_ multiword lexis in the MED

Followed by /b/-word Not followed by /b/-word

First word /b/_ (n 807) 121 (14.99%) 686

First word /f/_ (n 661) 37 (5.60%) 624

Table 6.2 /k/_ ⫹ /k/_ vs. /s/_ ⫹ /k/_ multiword lexis in the MED

Followed by /k/-word Not followed by /b/-word

First word /k/_ (n 802) 124 (15.46%) 678

First word /s/_ (n 1395) 73 (5.23%) 1322

Table 6.3 Evidence of the role of alliteration in the formation and standardization of multiword units
(sourced from the Macmillan English Dictionary for Advanced Learners, 2007, 2nd edn)

Followed by Followed by Followed by Followed by Followed by Followed by Followed by

/b/-word /d/-word /f/-word /k/-word /p/-word /s/-word /t/-word

1st word /b/_ 121 31 24 53 62 75 47

(n 807) 14.99%1 3.84% 2.97% 6.57% 7.68% 9.29% 5.82%

1st word /d/_ 52 54 17 47 28 46 37

(n 576) 9.03% 9.38%2 2.95% 8.16% 4.86% 7.98% 6.42%

1st word /f/_ 37 40 73 47 55 69 50

(n 661) 5.60% 6.05% 11.04%3 7.11% 8.32% 10.43% 7.56%

1st word /k/_ 72 51 49 124 67 101 56

(n 802) 8.98% 6.36% 6.11% 15.46%4 8.35% 12.59% 6.98%

1st word /p/_ 61 44 31 69 119 103 50

(n 852) 7.16% 5.16% 3.63% 8.10% 13.97%5 12.09% 7.84%

1st word /s/_ 113 116 60 73 118 187 71

(n 1395) 8.10% 8.32% 4.30% 5.23% 8.46% 13.41% 6 5.09%

1st word /t/_ 48 38 34 41 45 55 59

(n 574) 8.36% 6.1% 5.92% 7.14 5.87% 9.58% 10.28%7

See facing page for Notes to Table 6.3

10.1057/9780230245006 - Optimizing A Lexical Approach to Instructed Second Language Acquisition, Frank Boers and Seth Lindstromberg
Notes to table 6.3: Application of the Fisher Exact (Chi-Square) Test yields the following:
1. Multiword entries starting with a /b/-word are more likely (at p .000) to contain another /b/-word than those starting with /d/, /f/, /k/, /p/,
/s/ or /t/.
2. Multiword entries starting with a /d/-word are more likely (at p .000) to contain another /d/-word than multiword entries starting with /b/, more
likely (at p .003) to contain another /d/-word than those starting with /p/, and more likely (at p < .05) to contain another /d/-word than those starting
with /f/, /k/ and /s/.
3. Multiword entries starting with a /f/-word are more likely (at p .000) to contain another /f/-word than multiword entries starting with /b/, /d/,
/k/, /p/ and /s/), and more likely (at p .002) to contain another /f/-word than those starting with /t/.
4. Multiword entries starting with a /k/-word are more likely (at p .000) to contain another /k/-word than those starting with /b/, /d/, /f/, /p/, /s/
and /t/.
5. Multiword entries starting with a /p/-word are more likely (at p .000) to contain another /p/-word than multiword entries starting with /b/, /d/,
/f/, /k/ and /s/, and more likely to contain another /p/-word (at p .01) than those starting with /t/.
6. Multiword entries starting with a /s/-word are more likely (at p .000) to contain another /s/-word than multiword entries starting with /d/, more
likely (at p .004) to contain another /s/-word than those starting with /b/, and more likely (at p < .05) than those starting with /f/, /p/ and /t/.
7. Multiword entries starting with a /t/-word are more likely (at p .000) to contain another /t/-word than multiword entries starting with /s/, more
likely (at p .003) to contain another /t/-word than those starting with /b/ and /p/, and more likely (at p < .05) to contain another /t/-word than those
starting with /d/ and /k/.

10.1057/9780230245006 - Optimizing A Lexical Approach to Instructed Second Language Acquisition, Frank Boers and Seth Lindstromberg
112 Optimizing a Lexical Approach

For example, given the fact that /s/-words are far more frequent gener-
ally than /d/-words, one would expect more /s/-words than /d/-words to
follow /d/-headword multiword entries. And yet, this kind of frequency-
based probability seems systematically overridden by the appeal of
alliteration: in all the dictionary sections examined, the probability of
word-initial consonants being repeated in multiword entries is statistic-
ally much higher than would be predicted by chance.
Just in case the MED were for one reason or the other biased toward
inclusion of alliterating phrases among its entries, we thought it pru-
dent to repeat the above procedure with the aid of the Oxford Advanced
Learners Dictionary (OALD, 6th edition, 2000). Since we were now just
looking for further corroboration of the robust MED-based findings,
we were content to compare only two pairs of consonants: /f/ versus
/m/ and /b/ versus /p/. These pairs were chosen for ease of comparison
because our previous counts led us to expect the corresponding dic-
tionary chapters to include approximately equal numbers of multiword
entries (that is, chapters F ⫽ M and B ⫽ P). The procedure was also dif-
ferent from the MED exercise in the sense that we were now especially
interested in the question of whether alliteration plays a part in the pro-
cess of compounding. We therefore focused on two-word chunks (big
business, blue-blooded) and also on compounds written as single words
(billboard). Compounds which we deemed to consist not of two ‘con-
tent’ words but of a prefix and a word (for example, prepayment) were
not included.
Once again, we observe a statistically significant propensity towards
alliteration in both comparisons. Compound entries starting with
/f/ are more likely to contain a second /f/-word (32 instances) than a
/m/-word (21 instances), whereas compound entries starting with /m/
are more likely to contain a second /m/-word (20 instances) than a
/f/-word (5 instances). Likewise, compound entries starting with /b/ are
more likely to contain a second /b/-word (72 instances) than a /p/-word
(25 instances). In this analysis, however, compound entries starting in
/p/ were found to be just as likely to contain a /b/-word (for example,
pinball) as a second /p/-word (pinpoint): 44 and 40 instances, respect-
ively). Might /b/-words be more prone to alliteration than /p/-words?
According to estimates based on page counts in three dictionaries
(two for native speakers and one for advanced learners), /p/-words are
about 1.4 times as numerous as /b/-words, even after the deduction of
ph-words (actually /f/-words) and ps-words (actually /s/-words). However,
the size advantage for the P chapters is due largely to the presence of
markedly formal, polysyllabic Latinate and Greek-derived (LG) lexis.
 Structural Elaboration 113

It is, after all, the case that the P chapter of any remotely comprehen-
sive English dictionary abounds with words beginning with paleo-, per-,
pre-, pro-, para-, post-, proto- and pur-. But, as we will see below, such
words seem highly unlikely to occur in alliterative chunks. When all
words beginning with these prefixes are discounted, the P and B chap-
ters of the OALD are then of approximately equal size in terms of words
that are at all likely to occur in alliterative chunks (91.7 and 89 pages,
respectively, with page space devoted to illustrations and tables also
having been deducted). Yet, we have estimated on the basis of a bank of
alliterative phrases which we have been compiling from daily reading
and listening, that /b/ alliteratives are about 1.4 times as numerous as
/p/ alliteratives (Lindstromberg and Boers, 2005b). On examining this
collection of alliterative phrases, we find that Latin- and Greek-derived
(LG) words are very much under-represented, particularly LG words of
three syllables or more. Possibly the effect of alliteration (about which
we will say more below) is inversely proportional to the number of pho-
nemes or syllables between points of repetition, so that the effect of
alliteration is stronger in best bet than in either better than best or the
Latinate political posturing. Also, the relative scarcity of LG words in allit-
erative chunks may partly stem from the fact that such words have, on
average, been in the language less long than highly frequent Germanic
words like best. As a matter of fact, we have found no polysyllabic LG
word which comes close to figuring in as many chunks (whether allit-
erative or not) as certain short, Germanic words such as way (way off,
No way!, fall by the wayside, by the way, way to ⫹ VERB, on the way, Where
there’s a will, there’s a way and so on). Additionally, everyday speech is
notably non-LG (Corson, 1985) and it stands to reason that those words
which are most commonly used in speech would be those most likely to
figure in the commonest chunks. (We will return in Chapter 8 to the
question whether LG words are indeed under-represented in English
chunks generally.)
While all this appears to be a plausible explanation for the above
finding that fewer /p/ compounds than /b/ compounds alliterate, we
did exclude from our chunk counts all dictionary headwords with a LG
prefix (for instance, pre⫹suppose), which of course excluded a particu-
larly large number of the GL lexemes from the P chapter page count,
which meant that the chunk lists collected from the B chapter and the
P chapter words were then likely to include roughly equal proportions
of conspicuously LG lexis. Nevertheless, the ratio of /b/ to /p/ allitera-
tives in this sample is 72 b__b__ / 40 p__p__ (⫽ 1.8:1). So, it does after
all appear that /b/-words are more likely to figure in alliterative chunks
114 Optimizing a Lexical Approach

than /p/ words. Why this should be the case is unclear. Still, the bot-
tom line from the whole collection of dictionary-based counts we have
reported here is that there is robust evidence that alliteration is a factor
in the conventionalization of word combinations – even if there may
be unsuspected complications when looking at two consonants, such as
/b/ and /p/, which are articulated almost identically.
Having found confirmation that phonological repetition is, as we sus-
pected, a motivating factor in the formation of everyday-English word
partnerships, we can now turn to estimates of its scope. For a start, of
the total number of 5667 multiword MED entries we looked at in the
above, larger-scale study, no fewer than 737, that is, 13%, alliterate. To
this should be added another 2% of entries which rhyme (brain drain,
fat cat, fun run, fender-bender, pooper scooper, floating voter) and almost
7% which assonate (slow motion, dummy run, crazy paving, face saving,
fat camp, fish and chips, French letter). It seems safe to say, then, that at
least 20% of our bank of 5667 multiword chunks shows some kind of
sound pattern that may help motivate their formation and standard-
ization. It may be that this proportion will eventually be seen to be
even greater. Except for alliteration, the effects of consonance (as in flick
knife) remain unexplored. Rhythm and prosody too may eventually be
shown to be a factor in the appeal and conventionalization of certain
word strings (Naciscione, 2001: 25). For example, the rhythmic pattern
of ‘threesomes’ such as hook, line and sinker, blood, sweat and tears and
tall, dark and handsome is very reminiscent of the – apparently catchy –
rhetorical figure of ‘triples’ (for example, Caesar’s Veni, vidi, vici; the
French motto Liberté, fraternité, égalité; and Jefferson’s life, liberty and the
pursuit of happiness).
Some classes of chunks seem especially prone to the effects of phono-
logical motivation. Our counts in the Collins Cobuild Dictionary of Idioms
(2002, 2nd edition) reveal that almost 18% of English idioms alliterate
(on the back burner, a close call, short shrift, through thick and thin) while,
again, another 2% rhyme (high and dry, eager beaver). If we narrow our
focus to a sample of 508 expressions that are signalled in this diction-
ary as ‘frequently used’, the combined scope of alliteration and rhyme
is nearly 23% of all such idioms. Phonological repetition is even more
common in certain subcategories of idiom, judging by the entries in
this dictionary: 28% of English binomial idioms alliterate and/or rhyme
(chop and change, fair and square, spick and span). This may have to do
with two facts, as follows.
Firstly, binomial chunks tend to be short, which means that the
points of consonant repetition are often close together. Why should
 Structural Elaboration 115

this make a difference? A study by van de Weijer (2003) seems to offer

a clue. He has found, for Swedish, that consonant repetition within
monomorphemic words is rare. His results show that only 11.26%
of Swedish monosyllabic and bisyllabic words contain two or more
identical consonants. He further found that these words are rela-
tively uncommon. Indeed, he estimates that among the more than
28 million word tokens in his sample only 1.57% contain identical
consonants, with the result that a listener or reader in Swedish is likely
to encounter a case of monomorphemic consonant repetition only
once every 65 running words. Spurred by his plausible speculation
that intra-morphemic consonant repetition is likely to be rare in all
Germanic languages, we counted through three texts: the 1402-word
short story ‘A matter of sentiment’ by Saki (available on the internet),
the first 1402 words of an article in Foreign Affairs by U.S. Secretary of
Defense Robert Gates (2009), and the first 1402 words of a story for
girls of about ten (Branford, 1994). Although this data set is small, it
nevertheless appears from our results (Table 6.4) that van de Weijer’s
estimate for Swedish is likely to apply fairly closely to English also.
The major case of divergence (namely, the 2.9% incidence of intra-
morphemic consonant repetition in the Gates’s text) is almost entirely
due to the recurrence in this text of the name the United States. These
results suggest an answer to our question above. First of all, it is gen-
erally accepted that post-early childhood, aural word recognition is
aided by good procedural knowledge of how phonemes tend to be dis-
tributed within morphemes (see van de Weijer, 2003 and 2005, for
references to this literature). Van de Weijer – who does not mention

Table 6.4 Consonant repetition within monomorphemic words in three 1402-

token texts19

Number of Base words Tokens with

base words appearing more consonant
(types) showing than once and repetition
intramorphemic number of as a percent
Text repetition instances Tokens of 1402

Saki 19 interest (3); moment, 25 1.8%

member, mention,
murmur (2 each)
Gates 16 state (19); strategy (5); 40 2.9%
institution (3)
Branford 9 Ma’am (6); did(n’t) (3); 17 1.2%
murmured (2)
116 Optimizing a Lexical Approach

alliteration – therefore observes (2003) that consonant repetition may

aid recognition of words and morphemes, since a second encounter
with the same consonant correlates very highly with the existence of
an intervening morpheme or word boundary. In this regard, we add
that alliteration may well derive perceptual and cognitive salience
from the fact that identical consonants spaced far apart cannot be
as informative about the location of word boundaries as consonants
which are close together. Hence, for a listener, two closely spaced
points of alliteration such as we see in many binomials (such as chop
and change) are likely to be the focus of particular attention, all the
more so since and is often phonetically reduced in speech. It is possible
that speakers, for their part, exploit this fact about alliteration in order
to highlight particular parts of their message. It would appear, with
respect to English, that much the same can be said of the repetition of
a full (that is, unreduced) vowel within a monomorphemic word (we
find no such repetitions in this long paragraph, for example), which
may be part of why assonance, too, is relatively salient.
A second possible reason why conventionalized binomials show sound
repetition to a disproportionate extent is that in a binomial the word
which follows and often serves merely to add emphasis to the message
expressed by the first word (as in part and parcel, signed and sealed and
first and foremost). Because the second word in these cases seems not to
be there on account of its independent meaning, it seems likely instead
that it is there primarily on account of its euphonious effect.
Another category of chunks where phonological considerations often
seem to override semantic considerations is that of standardized simi-
les: no fewer than 41% of standardized similes alliterate and/or rhyme
(as in cool as a cucumber and drunk as a skunk). When we also take the
second-commonest type of sound repetition, namely assonance, into
account (as in airs and graces, cut and run, plain as day and high as a kite)
it looks as though some kind of phonological repetition is at play in
over 50% of binomial idioms and standardized similes (Lindstromberg
and Boers, 2008a).
In any case, the counts through idiom dictionaries indicate that
phonological motivation is sufficiently widespread and frequent for it
to matter both to teachers and to learners. One of the advantages of
exploiting patterns of sound repetition is that doing so can open up
an avenue for structural elaboration that complements the predomin-
antly meaning-focused (and thus primarily reception-oriented) types
of elaboration discussed in the previous chapter. Another advantage is
that focusing on sound repetitions may cater for learners with a more
 Structural Elaboration 117

‘auditory’ learning style or, more generally, a learning style that is

less attuned to the techniques of mental imagery. Perhaps the great-
est advantage of working with the phonological dimension of chunks,
however, is that it also opens up an avenue for the insightful teaching
and learning of vast numbers of chunks that have so far fallen outside
the scope of (CL-inspired) semantic elaboration – that is, ones such as
non-figurative compounds, verb–noun and adjective–noun colloca-
tions, and partitives.
Boxes 6.1 and 6.2 present additional examples of categories of
chunks whose lexical composition learners are more likely to remem-
ber when they become aware of any sound repetition. Note in Box 6.2,
for example, how alliteration may possibly help motivate some col-
locational patterns of do and make which have proven so resistant to
thorough explanation in semantic terms (see Nehls, 1991, for one note-
worthy attempt).
From our counts it appears that in English it is alliteration that is by
far the most common type of phonological repetition in chunks. This
may not be the case in all languages, however. For example, counts in
French and Spanish idiom dictionaries (Boers and Lindstromberg, 2005;
Boers and Stengers, 2008b) have shown lower rates of alliteration than
in English. Nevertheless, the overall ubiquity of phonological repeti-
tion is likely to be similar across the three languages, as more idioms in
Spanish and French than in English display rhyme.
There may be several reasons why alliteration is particularly popu-
lar in English phraseology. Firstly, word stress is typically on the

Box 6.1 Some proverbs, similes and binomials displaying phonological

repetition

throw the baby out with the bathwater; when the cat’s away, the mice will play;
beauty is in the eye of the beholder; the grass is always greener on the other side;
curiosity killed the cat; Better safe than sorry; one swallow doesn’t make a summer;
the proof of the pudding is in the eating; Variety is the spice of life; that’s just the
way the cookie crumbles; birds of a feather flock together.

Blind as a bat; fit as a fiddle; brown as a berry; bold as brass; busy as a bee; clear as
crystal; dead as a dodo; dull as dishwater; dry as dust; good as gold; green as grass;
hot as hell; large as life; pleased as punch; right as rain; thick as thieves.

Bed and board; bed and breakfast; black and blue; cash-and-carry; tried and tested;
forgive and forget; safe and sound; fame and fortune; deaf and dumb; wax and
wane; rack and ruin; rules and regulations; toss and turn; chalk and cheese; moan
and groan; doom and gloom; wear and tear; near and dear; pick and mix.
118 Optimizing a Lexical Approach

Box 6.2 Some compounds and collocations displaying phonological

repetition

Baby boom; bargain basement; beer belly; binge drinking; big bang theory; blood
brothers; bubble bath; bunk bed; civil servant; collect call; collision course; culture-
clash; fair play; far-fetched; force-feeding; fully-fledged; fully functioning; fun run;
health hazard; number crunching; peer pressure; safe sex; screen saver; snail’s pace;
soul-searching; steady state theory; toy boy.

Buckle your belt; commit a crime; crack a case; do a degree; do a doctorate; do

damage; do the dishes; make a mess; make amends; make a mistake; make a move;
make a muddle; make a name; make changes; make mileage; make money; pay the
price; take time; take turns; tell the time; wage war; work wonders.

Bring __ to the boil; hold __ in high esteem; catch __ on camera; put __ into practice;
buy __ in bulk; keep __ in close confinement; give __ cause for concern.

A beast of burden; a code of conduct; contempt of court; the courage of your convic-
tions; a crisis of conscience; the finger of fate; a lady of leisure; the letter of the law;
a man of means; the minutes of the meeting; the name of the game; proof of pur-
chase; the pursuit of perfection; a rate of return; the tricks of the trade.

first syllable in English, which enhances the perceptual salience of

word-initial consonant repetition. In French, by contrast, word stress
is as a rule on the final syllable. It is therefore not surprising to find
that rhyming chunks are more common in the latter language. Mostly
for L1 (and mostly for English), there is by now a great deal of evidence
from a number of research strands in experimental psychology that the
beginnings and ends of words are perceptually and cognitively salient
(see Lindstromberg and Boers, 2008a, b, for brief discussions of the rele-
vant sub-literatures). Aitchison’s (2003) apt and memorable coinage for
this state of affairs is the bathtub effect, whereby a word is likened to a
person lying in a foamy bath with only head (word onset) and the feet
(word ending) sticking out of the water. Within the context of this basic
tendency, word stress can, obviously, increase the perceptual salience
of either the onset or the end of a word, bringing either alliterating or
rhyming patterns more to the fore, in accordance with the word stress
patterns of the language in question.
A second reason why English phraseology is comparatively rich
in alliteration may be its relative poverty with respect to inflection.
English words (those of Germanic origin in particular) tend to be rather
short, and if alliteration occurs across a sequence of short words, the
same consonant is repeated at close intervals and is thus likely to be
extra noticeable.
 Structural Elaboration 119

The third possible reason lies in the comparatively rigid word order
of English. Thus, in English, words that are catchy or euphonious when
strung together have a relatively good chance of occurring over and
over in the same order and proximity, whereas in languages with more
word-order variability they may often be separated and rearranged. Any
given English word string, therefore, seems comparatively likely to leave
more (and therefore more durable) traces in memory as a unit.
Apart from such typological reasons, the popularity of alliteration in
English may also be a cultural heritage, perhaps all the way from Old
English, when poetry was conspicuously alliterative. (Rhyming verse
was only introduced into Middle English via French poetry.) At any
rate, it seems that the use and appeal of alliteration has never wholly
faded from English literature. A discussion of the use of rhyme and allit-
eration in literature would clearly be beyond the scope of this applied
linguistics book. Suffice it to say that examples of literary alliteration
abound in English literary works from all periods, from the likes of Jane
Austen (Pride and Prejudice, Sense and Sensibility) to J. K. Rowling (about
one third of the invented names and chapter titles in the Harry Potter
saga alliterate).

6.2 The memorability of phonologically

motivated chunks

In the worlds of marketing and entertainment, the aesthetically pleasing

effect and memorability of rhyme and alliteration have long been taken
for granted. This can be inferred from the high number of brand names,
advertising slogans, titles of TV series, film titles and names of fictitious
characters that display rhyme and/or alliteration: Coca Cola, Probably
the best beer in the world (now probably in the best bottle), Wherever you go,
go Texaco, The bold and the beautiful, Black Beauty, Sex and the City, Bend
it like Beckham, Desperately Seeking Susan, Mickey Mouse, Donald Duck,
Bugs Bunny, Peter Pan, and so on. It must be said, however, that con-
sumer research itself hardly abounds with experimental studies on the
impact of phonological repetition, and the few studies we have found
in this branch are concerned with rhyme (McQuarrie and Mick, 1999;
Mothersbaugh, Huhmann and Franke, 2002) rather than alliteration,
even though it is the latter which is most frequent in branding and
advertising. For example, browsing through a large collection of adver-
tising slogans at www.ad-mad.co.uk reveals that approximately 30% of
English language slogans use alliteration as compared to approximately
8% that use rhyme.
120 Optimizing a Lexical Approach

Most evidence of the mnemonic effect of phonological repetition –

sometimes including alliteration – comes from the fields of oral litera-
ture and experimental psychology (see Rubin, 1995, for an excellent
survey). A recent study by Lea et al. (2008), for example, confirms the
memorability of alliteration in the context of poetry. Psycholinguistic
experiments confirming the memorability of rhyme have been reported
by Fallon, Groves and Tehan (1999) and Gupta, Lipinski and Aktunc
(2005).
To ascertain whether phonological repetition has mnemonic poten-
tial in the context of L2 learning, and more particularly in the learning
of lexical phrases (rather than single words), we set up the two experi-
ments summarized below. (More detailed reports of these experiments
can be found in Lindstromberg and Boers, 2008a, b.) We decided to focus
on alliteration and assonance, as our dictionary counts (see Section 6.1
above) had shown these two sound patterns to be much more common
in (English) multiword lexis than rhyme.
To assess the mnemonic effect of alliteration, we asked students of
English to sort 26 phrases in an alliterative and a non-alliterative set.
Each phrase was composed of two familiar monosyllabic words. Half
of these phrases alliterated (for example, ring road, sea salt, green grass),
while the other half were ‘un-patterned’ (for example, bath soap, fresh
air, key hole). The phrases, both categories of which were matched as
well as possible with regard to frequency bands, word classes and lex-
ical fields, were presented to the students in random order. The students
were unaware that their retention of the phrases would shortly be tested.
When they had completed the sorting task and handed in their papers,
they were asked to write down as many of the phrases as they could
remember. The students turned out to be significantly more likely to
recollect alliterative than un-patterned target phrases. Two weeks later,
they were given an unannounced delayed test, in which they were pre-
sented with the same 26 phrases jumbled with 13 alliterative and 13 un-
patterned distracters. The students’ task was to tick the phrases they
remembered having sorted the previous week. Students’ recollection of
the alliterative target phrases was again significantly better.
We replicated the above experiment with the participation of a dif-
ferent cohort of students, but this time the students were asked to sort
assonant phrases (for example, home phone, sea breeze, right size) and
similar ones that were un-patterned (for example, storm cloud, good taste,
bad luck). The purpose was to assess whether a type of phonological
repetition evidently less salient than rhyme and alliteration also has
mnemonic potential. The results strongly suggest that the answer is
 Structural Elaboration 121

affirmative: students were again significantly more likely to remember

the assonant phrases than the un-patterned ones, both in the immedi-
ate test and in the delayed test.
Given the empirical evidence of the mnemonic potential of the com-
mon sound patterns of alliteration and assonance in L2 lexical phrases,
the question is whether learners make spontaneous use of this mne-
monic potential or whether prompting from the teacher (or materials
writer) is required. After all, in the above experiments, the students
were made aware of the presence of alliteration or assonance through
the explicit task instruction. They were prompted by the researchers to
engage in structural elaboration regarding the sound patterns of the
stimuli phrases, something they might not have done in the absence of
explicit pointers. In order for students to use a given feature of a chunk
as a trigger for elaboration, they first need to notice that feature. To
estimate whether students are likely to recognize patterns of phono-
logical repetition autonomously, we asked a third group of students to
sort phrases into four categories by considering their sound patterns.
The set of phrases presented to the students contained six rhyming
phrases (such as deep sleep), six alliterative ones (fast food), six assonant
ones (nice try) and six un-patterned ones (brick wall). The students were
not told explicitly what sound patterns they should be looking for, only
that they should try to discern three sound patterns in the set. The out-
come of this less guided, but nevertheless still form-focused, activity
gives grounds for scepticism about learners’ ability to notice phono-
logical motivation on their own: only a very small minority of the
students sorted the phrases even approximately according to expect-
ation, with most failing to discern even a set of alliterative phrases,
let alone assonant ones. After failing to sort the phrases according to
the task instruction, several students admitted they attempted to sort
the phrases by semantic associations instead. The failure of most par-
ticipants to perform the assigned task is perhaps especially surprising
given the fact that the participants in this experiment (as in the two
previously described experiments) were advanced-proficiency language
majors, that is, students whom we may assume have developed a well-
above-average language awareness.
In a way, it should not be so surprising if phonological repetition passes
unnoticed by teenage and adult L2 learners. Beyond childhood, learners
tend to emphasize meaning over form (Skehan, 1998; VanPatten, 1990)
and they do not often attend to any kind of formal aspect of language
spontaneously (Williams, 1999). Information about the exact wording
of messages tends to be discarded from working memory as soon as the
122 Optimizing a Lexical Approach

gist of the message has been worked out, and it is consequently seman-
tic information rather than phonological form that stands the best
chance of being stored in long-term memory (Baddeley, 1997 [1990]).
Exceptions to this pattern do exist, of course. For example, it is well doc-
umented that song lyrics can leave very durable traces in memory, even
if the words are not understood (Bartlett and Snelus, 1980), but then
again the processing may not have been meaning-focused in the first
place. The phonological shapes of the words are mnemonically glued to
the melody and the rhythm (Wallace, 1994).
It is well known, though, that young children take delight in the
sound patterning of words and phrases (Cook, 2000), including alliter-
ation (Jusczyk, Goodman and Baumann, 1999), even if those words and
phrases are not ‘meaningful’ (and in most nursery rhymes and hand-
clap games they tend not to be). However, it seems that this phonological
orientation is gradually replaced by semantically driven processing as
children grow older (Dewhurst and Robinson, 2004). Indeed, although
our everyday language abounds with alliterative phrases, as adults we
are seldom aware of them anymore, especially when we are in the midst
of real-time communication. Rhyme and alliteration are in fact ‘catchy’
only when we momentarily switch our phonology-oriented (or, more
generally, form-oriented) mode of processing back on. It is also only
when we switch this mode back on that we may deliberately coin rhym-
ing or alliterative words or chunks. During real-time communication,
however, constant alertness to phonological repetition would undoubt-
edly distract too much from the encoding and decoding of messages. All
this seems very much in line with the role of literacy in Wray’s (2002)
thesis that children gradually replace their holistic processing of chunks
by a more analytical, word-for-word style of processing. Perhaps one’s
growing familiarity with the written word weakens one’s inclination for
phonologically oriented processing of chunks. This fits in completely
with Oral Formulaic Theory (Lord, 1956, 1964; Parry, 1971) wherein it is
particular preliterate societies that produce memorized oral literature
(poetic epics, for instance) characterized by heavy use of recurring word
strings (known as ‘oral formulae’) which fulfil both aesthetic and mne-
monic functions. Also, according to priming experiments, words are
much better recalled when they are primed by a phonologically similar
word than when they are primed by an orthographically similar word
(Ziegler, Muneaux and Grainger, 2003).
It is quite possible that adult beginner-learners of a foreign language
do initially – like children – engage in a fair amount of phonological
elaboration, for example by sounding out their first L2 words, as these
 Structural Elaboration 123

may still seem pleasantly exotic. It is also quite possible that, due to
cognitive-style differences, some learners are more liable than others
to notice the phonological properties of words and phrases, including
alliteration (Boers and Lindstromberg, 2008a). Still, it seems safe to say
that adult intermediate-to-advanced learners – and these are the learn-
ers the LA has mostly been intended for – are less likely than children
and beginning learners to dwell on the phonological properties of lex-
ical phrases, especially when the constituent words are already known.
If these learners are not very likely to notice patterns such as alliteration
in chunks spontaneously, then it follows that phonologically motivated
chunks merit a place among the sets of chunks to be targeted in the
classroom (and/or in pedagogical materials), and this by virtue of the
fact that they are particularly ‘teachable’. That is, they are not only rela-
tively easy to teach in a way that results in them being remembered, but
they are unlikely even to be noticed, unless they are taught, for although
these chunks have considerable mnemonic potential, relatively few
learners will unlock it without prompting or guidance.
The thesis that teacher intervention can make a difference in this
context is backed up by the results of another experiment reported
in Lindstromberg and Boers (2008b), the essence of which was that
the teacher briefly alerted one group of students to alliteration in
the chunks they encountered in reading and listening texts during a
36-hour EFL course while he refrained from doing so when he taught
a parallel group. In a post-test (an unannounced gap-filling exercise
targeting both alliterative and un-patterned chunks encountered dur-
ing the course), the former group outperformed the latter with respect
to the alliterative targets (where brief structural elaboration had been
triggered). No differential uptake of the other phrases (where classroom
treatment had been the same) was observed. This suggests that simply
drawing students’ attention to the phonological repetition in a given
chunk and perhaps suggesting that its lexical makeup might not be
completely accidental is sufficient to give a measurable return on very
minimal classroom investment.
Further, we considered the assumption that phonologically repeti-
tive word strings have a competitive advantage for conventionalization
because they are aesthetically pleasing and catchy (and hence memor-
able). While this assumption seems uncontroversial with regard to two
evidently salient types of phonological repetition, rhyme and alliter-
ation, it appears less likely to apply to assonant chunks. For one thing,
massive evidence of a bathtub effect suggests, on the face of things, that
assonance must be less noticeable, perhaps far less noticeable, than
124 Optimizing a Lexical Approach

alliteration and rhyme. So, there may be another reason besides extra-
memorability for the formation and standardization of phonologically
repetitive chunks. That potential reason is that repetition may facili-
tate fluency in language production. We mentioned this possibility in
Chapter 4, where we referred to Zipf’s (1949) ‘principle of least effort’,
according to which there is a trade-off between articulatory effort and
semantic distinctiveness. Phonologically repetitive strings such as
assonant chunks may be comparatively economical as far as articula-
tion is concerned, since the same neuromotor actions are reiterated. In
fact, so-called tongue twisters are difficult precisely because the smooth
reiteration of the same phoneme in a given word string is interrupted by
the insertion of a word or morpheme with a variant phoneme, which
then creates an obstacle. Neither is it surprising that young children’s
first attempts at articulation typically involve repetitive word forms
such as dada and pipi (Smith, 1973; Vihman, 1978), which also suggests
that the pronunciation of repetitive forms is comparatively easy.
Boers and Lindstromberg (2008a: 344–5) have suggested that a prin-
ciple of least effort may also help to motivate the word order in some
English binomials. In cases where the word order is neither a matter
of iconicity nor a matter of metre (native speakers of English disfavour
successions of un-stressed syllables and will consequently put mono-
syllabic words before, and bi-syllabic words after, unstressed and, as in
salt and pepper, rough and tumble and pins and needles), the word whose
initial consonant assimilates most readily with the /_d/ of and will
tend to come second, as in hammer and tongs, give and take and supply
and demand. Given the special role of many binomials as ‘semi-fillers’
(when the second half of the phrase adds little meaning), it stands to
reason that – all else being equal – their word order would economize
on articulatory effort.
The notion of fluency facilitation will be taken up again in the next
chapter, but in a different light. In this chapter and the one before, we
have proposed and assessed ways of optimizing the LA by complement-
ing teacher-guided noticing of chunks with teacher-guided semantic
and/or structural elaboration. We have seen that when students engage
in the kinds of elaboration we have described, this increases the chances
that the noticed chunks will leave durable traces in long-term mem-
ory. However, the learning process does not stop there. As we know,
vocabulary knowledge needs to be deepened and consolidated. The
degree of consolidation that is required will obviously depend on the
objectives of the learner. Regular review exercises designed to encour-
age noticing and semantic elaboration might perhaps be enough for
 Structural Elaboration 125

adequate enhancement of fluency not only in listening and reading but

also in production, provided that planning time is available (in the case
of writing or preparing a speech, for example). But for chunk knowledge
to benefit students’ productive fluency in real-time communication, a cer-
tain degree of automatization is bound to be required. This is one of the
issues we discuss in the next chapter.
7
Bearing in Mind

7.1 The need for width and depth of chunk knowledge

The recommendations for optimizing the LA that we have made in

this book so far can be summarized as follows: (1) teach chunks instead
of relying on learner-autonomous, incidental chunk-uptake owing to
awareness-raising alone; (2) select chunks for targeting not just on the
basis of frequency but also on the basis of evidence of collocational
strength and ‘teachability’; (3) in order to improve the chances of reten-
tion, complement noticing by also encouraging elaboration of meaning
and form; and (4) reveal non-arbitrary properties of chunks to make them
more memorable.
In developing this programme we have considered – like Michael
Lewis and his colleagues – encounters with chunks in authentic input
to be the main source for learning: students are helped to notice selected
chunks in this input and, insofar as is possible, the teacher stimulates
elaboration with regard to these chunks in order to encourage uptake.
In addition, graphic templates are provided so that students may record
the chunks they encounter in ways that stimulate additional insight-
ful learning (see Chapter 5, Section 5.2). These steps are arguably not
enough, though, especially when learning aims are ambitious and
include attainment of the ability to produce chunk-rich language under
real-time conditions with fluency. For this, text-independent exercises
and classroom activities with a chunk-orientation may be necessary in
order to spur on and to consolidate the learning process.
One of the arguments in favour of doing classroom exercises on par-
ticular chunks is the need to entrench memory traces. While elabor-
ation with regard to chunks met in texts will help students turn the
noticed chunks into uptake, consolidation of these freshly created

126
 Bearing in Mind 127

memory traces is indispensable for learning to be durable. In fact, even

well-entrenched knowledge needs to be re-activated occasionally to pre-
vent its gradual attrition. Instead of waiting for the noticed chunks to
be re-encountered incidentally in another text, teachers can deliber-
ately re-present them in classroom exercises. Even so, it cannot be taken
for granted that receptive knowledge of chunks acquired from input
will automatically also aid students’ language production.
A second argument in favour of such exercises is that students are
likely to benefit from working on chunks over and above the ones they
are encouraged to pick up from listening and reading. Although most
texts are densely packed with chunks and thus furnish ample opportun-
ity for chunk learning, they may not provide students with a satisfactor-
ily wide array of chunk types, not only because the stock of chunks is so
enormous but also because some types of chunks (such as social routine
formulae and other phrases that are especially useful for conversational
purposes outside the classroom) may be under-represented in the text
materials most commonly explored in contemporary ISLA – song lyrics
being an extreme but pertinent example here. In other words, students
may need exposure to more chunks than those they come across inci-
dentally. This view is shared by Morgan Lewis, one of the contributors
to Michael Lewis’s (2000) edited volume on teaching collocation. She
reports that she seizes on her students’ encounters with collocations in
their course materials as opportunities to inform them about additional
collocations. As a result, she says, the board in her classroom is usually
filled with collocations at the end of a lesson. She acknowledges that it
would be a miracle if her students remembered even 10% of what she
presents (Morgan Lewis, 2000: 25–6). To this acknowledgement we may
add that – all else being equal – it is perhaps only a couple of chunks
presented at the beginning of the lesson and a couple at the end that
stand a particularly robust chance of being retained, owing to primacy
and recency effects in memory uptake (see Sharifian, 2002, for a concise
summary of such effects). It could be argued that this does not matter,
since the chunks that are not retained by students in one lesson are
likely to be presented again in another, so that incremental learning
will take place eventually. In Chapter 3 of this book we referred to a
number of studies in L2 vocabulary learning that give reason for scepti-
cism about that assumption. Presenting a great quantity of potentially
new chunks in one go may have further, more substantial drawbacks.
One is affective in nature: overwhelming students with an abundance
of word partnerships they were not yet aware of may be de-motivating.
The other drawback is that presenting a great deal of information in
128 Optimizing a Lexical Approach

too short a time-span can easily result in mental crowding, with the
result that the majority of the items are processed shallowly at best. We
believe it is generally more effective to err on the side of caution and
target relatively few new chunks per lesson, but to do so in ways likely
to leave deeper memory traces.
Michael Lewis, despite his outspoken preference for using class time
to teach strategies rather than individual lexical items, does concede
that certain benefits can come from classroom activities which focus
on particular chunks:

The emphasis on input runs through the Lexical Approach, but

changes as learners’ level increases. A mature adult L1 lexicon is
simply too large to have been acquired by formal vocabulary teach-
ing. In both L1 and L2 a mature lexicon is acquired in very similar
ways – firstly by large quantities of listening which is largely com-
prehensible, and later by similar quantities of comprehensible read-
ing. I therefore advocate learners listening and reading as much as
possible, confident that this is the best way to develop their lexicons.
This alone is, however, obviously unacceptable. Learners expect to
be taught and we ignore this at our peril. With a certain scepticism
about the value of certain productive practices, I ask what kinds are
most likely to aid acquisition and at the same time satisfy the reason-
able expectation of the learners. (1997: 86)

Lewis (1997: 89–106) proposes a collection of chunk-oriented sample

exercises. By ‘exercises’ he means worksheet-based language-focus work,
usually done on an individual basis. This collection is supplemented by
Hill, Lewis and Lewis (2000: 88–116) mostly with exercises whose aim is
to help students make good use of collocation dictionaries.20 Let us here
briefly review some of the exercise types proposed in Lewis (1997).
Apart from the fundamental LA exercise where students are expected
to identify chunks in texts (so-called pedagogical chunking), the major-
ity of the exercises are matching exercises (under different guises). For
example, students are asked to match verbs in one list with their strong
collocate nouns in another. The matching is often to be made between
a list of words and a series of gapped sentences, so that students choose
the appropriate word to fill in the blank in the sentence. It goes without
saying that the focus of the exercises is systematically on multiword
expressions such as collocations. The exercises serve as eye-openers
(1997: 96) that raise students’ awareness of the importance of chunks
and phraseology in general (so that – it is hoped – they will become
 Bearing in Mind 129

more cautious about combining words randomly and become bet-

ter versed at noticing chunks in the texts they will be using). But the
exercises are definitely also intended as a means of introducing and
practising individual chunks, as can be inferred from the following rec-
ommendation: ‘If such exercises are to teach rather than merely test,
learners must recognise some answers and deduce others by the process
of elimination’ (Lewis, 1997: 88). Next to general awareness-raising, the
intention behind the exercises thus appears to be dual: the consolida-
tion in memory of (partially) acquired chunks, and the introduction of
novel ones, to be added to the student’s repertoire. The design of exer-
cises that live up to this intention is a delicate balancing act, though. As
Lewis himself points out, ‘[i]f there are too many examples, too many
possible answers, or the items are badly ordered, the exercise is simply a
rather perverse activity based mainly on guesswork’ (1997: 88).
A matching exercise, for example, has to include a sufficient number
of familiar or half-familiar items. This raises the success rate and thus
fosters positive affect; also, students stand a good chance of getting the
novel target items right as well if, by a process of elimination, they can
narrow down the number of matching options. But, if too many target
items are unfamiliar, then students will have to resort to blind guess-
work (unless clues to likely collocational patterning – phonological
clues, for example – are available). In the worst case, confusion may set
in if mistaken matches also leave their traces in memory. If, on the other
hand, every item in the matching exercise is familiar, or all except for
one, then the exercise will lack challenge (cases of one unfamiliar item
being resolvable by a process of elimination). Designing a pedagogic-
ally sound matching exercise must therefore proceed from a sufficiently
accurate estimate of students’ chunk knowledge. The risk of creating
confusion is especially high when new lexical items that resemble each
other (formally or semantically) are introduced simultaneously (Schmitt
and Schmitt, 1995, cited in Lewis, 1997: 48; see also Waring, 1997). The
success of matching and/or completion exercises which are intended to
point out the differences in collocational behaviour of near synonyms
such as speak, say, talk and tell (Lewis, 1997: 97) will therefore depend
on students being already reasonably familiar with the meaning and
usage of each verb.21
At any rate, the problem remains that, as long as the makeup of colloca-
tions and lexical phrases in general is presented as completely arbitrary,
the correct matching of novel items must inevitably remain a matter of
either elimination (at best) or blind guessing, neither of which is likely to
involve elaborative processing.22 And yet, it is possible both to reduce wild
130 Optimizing a Lexical Approach

guessing and to increase insightful learning if one exploits the linguis-

tic motivation that is inherent in many chunks. Box 7.1 illustrates how
information about source domains can offer clues for matching idioms
with meanings. Box 7.2 exemplifies how a gap-fill exercise can be made
more do-able through drawing students’ attention to sound repetition.
Box 7.2 also shows how reviewing of the same material can, step-wise,
be made more challenging with a view to deepening entrenchment in
memory. In this example, we simulate the assumption that the students
already partially know some of the targeted chunks while others are new
to them. In this regard, the approach resembles that proposed by Lewis
(1997); it differs, though, in that the exploitation of linguistic motiv-
ation allows for the inclusion of more novel items alongside the items
targeted for consolidation. Blind guessing (which seems unlikely to be
very motivating even when successful) is replaced by ‘educated’ guess-
ing (inferring), involving contemplation of non-arbitrary semantic and
formal features of chunks, which – as was shown in Chapters 5 and 6 –
positively influences the chances of retention.
The approach illustrated in Boxes 7.1 and 7.2 differs from that taken
in many of Lewis’s examples also in the sense that the targeted chunks
are introduced as wholes instead of being initially presented to students
as sets of single words which they have to assemble.

Box 7.1 Reducing blind guessing in matching exercises

Match the figurative expressions with the best paraphrase (write the corre-
sponding number after each expression). The hints about the origins of the
expressions may help.

List of expressions:
From seafaring: take something on board (__); on an even keel (hint: the keel is
the bottom part of a ship) (__); be left high and dry (hint: the boat is stuck on
a beach or sandbank) (__); a leading light (__)
From other means of transport: keep a tight rein on someone (hint: the reins
are the ropes used to direct a horse) (__); take a back seat (hint: in a car) (__);
get into gear (hint: 1st gear, 2nd gear, etc. of a car engine) (__); hit the buffers
(hint: the buffers are the obstacles that stop a train at the end of a railway
line) (__)

List of paraphrases:
(1) keep someone under control; (2) accept something such as a piece of
advice; (3) a role model or example to be followed; (4) start working effect-
ively; (5) stop making progress; (6) let someone else make the decisions;
(7) making steady, calm progress; (8) be in an uncomfortable situation
 Bearing in Mind 131

Box 7.2 Reducing blind guessing in completion exercises

Review 1: Fill in the blanks (one word each). The hints may help.
1. The first couple of months of the school year, she took things easy, but
when the exams approached she got into __________ and started working
quite hard. (hint 1: think of cars; hint 2: the expression alliterates)
2. My uncle is very intelligent and I very much value his opinion. I always
try to take his advice on _______________. (hint: think of a ship)
3. In the ‘war on terror’, the US and the UK took the leading role, while sev-
eral European countries decided to take a back ______________. (hint: think
of a car)
4. Our school has been going through a difficult period, but now there
is a new management and I’m confident we’ll soon be on an even
______________ again. (hint: think of a ship)
5. My girlfriend’s parents keep a tight ______________ on their daughter. For
example, she has to get home before 10pm, even on Saturdays. (hint: think
of horse riding)
6. Most refugees are denied political asylum. They are left high
and ______________ without a home and nowhere to go. (hint 1: think of
boats; hint 2: the expression rhymes)
7. The project had hardly started when it already hit the ______________, sim-
ply because the new board of directors had different priorities. (hint: think
of trains)
8. Ghandi was the leading ______________ in India’s struggle for independ-
ence. (hint 1: think of ships; hint 2: the expression alliterates)

Review 2: Fill in the blanks (one word each). Use this exercise to check your
answers in Review 1.
1. The first couple of months of the school year, she took things easy, but
when the exams approached she ___________ into gear and started work-
ing quite hard. (hint: alliteration)
2. My uncle is very intelligent and I very much value his opinion. I always
try to _____________ his advice on board.
3. In the ‘war on terror’, the US and the UK took the leading role, while sev-
eral European countries decided to take a ______________ seat.
4. Our school has been going through a difficult period, but now there is a
new management and I’m confident we’ll soon be on an ______________
keel again.
5. My girlfriend’s parents keep a ______________ rein on their daughter. For
example, she has to get home before 10pm, even on Saturdays.
6. Most refugees are denied political asylum. They are left ______________ and
dry without a home and nowhere to go. (hint: rhyme)
7. The project had hardly started when it already ______________ the buffers,
simply because the new board of directors had different priorities.
8. Ghandi was the ______________ light in India’s struggle for independence.
(hint: alliteration)

→
132 Optimizing a Lexical Approach

Review 3 (after a time lapse): Fill in the blanks (2 words each)

1. The first couple of months of the school year, she took things easy, but
when the exams approached she got ______________________ and started
working quite hard.
2. My uncle is very intelligent and I very much value his opinion. I always
try to take his advice ______________________.
3. In the ‘war on terror’, the US and the UK took the leading role, while
several European countries decided to take a ______________________.
4. Our school has been going through a difficult period, but now there is a new
management and I’m confident we’ll soon be on an ______________________
again.
5. My girlfriend’s parents keep a ______________________ on their daughter. For
example, she has to get home before 10pm, even on Saturdays.
6. Most refugees are denied political asylum. They are left high
______________________ without a home and nowhere to go.
7. The project had hardly started when it already hit ______________________,
simply because the new board of directors had different priorities.
8. Ghandi was the ______________________ in India’s struggle for independence.

Another exercise type that is recommended in the received version

of the LA is ‘deletion’: students are presented with a set of words all
except one of which are strong collocates of a given keyword. The
task, then, is to identify ‘the odd one out’. The point of this type of
exercise, so the argument goes, is to provide students with negative
evidence – that is, that certain word partnerships which may be com-
mon in the students’ L1 do not exist in L2. In other words, exercise
designers (who may of course be teachers) anticipate that, sooner or
later, the students they are writing materials for will stumble into a
certain pitfall created by L1 interference. The rationale seems to be that
it is better to prevent than to remedy. As far as we know, the jury is
still out about whether presenting learners with ‘negative examples’
(‘don’t say X, but say Y’) is pedagogically effective or detrimental (see
Leeman, 2007, for a thorough discussion). We do know, however, that
one must have a great deal of experience in teaching learners with
a particular L1 in order to be able to anticipate their L2 collocational
errors. So, unless strong evidence of particular recurrent errors made
by the relevant student population is available, for example, evidence
derived from learner corpora (for example, Granger, 1998), we feel the
prudent option is to wait and see what precisely it is in the students L 2
output that has to be remedied in the first place. Naturally, this strat-
egy depends on students having ample opportunity to produce L2 (so
that the teacher can obtain the necessary data). In the received version
 Bearing in Mind 133

of the LA, however, emphasis is on input rather than output (see also
Hill, 2000: 66). Lewis (1997: 108–42) does propose a large collection of
collaborative classroom activities that can be given a chunk-oriented
dimension, but hardly any of these involve extensive language pro-
duction by the students. The activities proposed range from search-
ing out chunks in texts and examining the collocational behaviour of
words to discussing the pragmatics of certain institutionalized expres-
sions, doing game-like matching exercises (using word dominoes and
spaghetti-matching worksheets, for example), and game-like reviewing
(by means of crossword puzzles, for instance). Very few of the activities
require students to produce the chunks they are expected to learn or
review in spoken interaction, let alone in ‘real-time’ authentic speech.
The current version of the LA thus proposes ways of helping students
appreciate the importance of learning a wide range of chunks and it
proposes activities in which students re-produce learnt chunks under
conditions that allow for preparation and monitoring, such as gap-fill
exercises and writing activities (Conzett, 2000); but it is silent about
how students can be trained to put their chunk-knowledge to good
use in more ‘spontaneous’ interaction where chunks may authentic-
ally fulfil one of their major functions – facilitating fluency. This is
another argument in favour of complementing the input-oriented LA
with more output-oriented activities: attaining productive proficiency
requires productive practice. Moreover, chunk knowledge that may be
broad enough to help students produce language under conditions that
allow for planning and auto-monitoring (when writing an essay and
when giving a prepared presentation, for example) may not be deeply
enough entrenched to adequately facilitate fluency when engaged in
real-time language production. As it happens, all varieties of Task-Based
Instruction (TBI) – an extension of Communicative Language Teaching
(CLT) in the direction of learner autonomy – place a premium on struc-
turing class work so that students spend a high proportion of class time
engaged in communication which is as authentic as possible (Prabhu,
1987; Skehan, 1998; and J. Willis, 1996). On the face of things, there-
fore, TBI might be well-suited to development of an ability to rapidly
produce chunks in real, relatively unplanned communication. In gen-
eral, however, advocates of TBI have concentrated much more on acqui-
sition of grammar than on acquisition of vocabulary, with acquisition
of lexical phrases having so far been of very peripheral interest. Also,
while oral fluency is a TBI aim, the crucial issue of automaticity has
received insufficient attention from advocates of TBI or current main-
stream CLT generally (de Bot, 1996; Gatbonton and Segalowitz, 2005;
134 Optimizing a Lexical Approach

Hulstijn, 2001). So, let us now turn to the intersection of the two main
issues at hand: the acquisition of L 2 lexical phrases and automaticity.

7.2 The need for automaticity

In Chapter 2 we surveyed the experimental evidence for the common

assumption that chunk knowledge supports fluency, receptively as well
as productively, including oral fluency. In more recent chapters, we
looked at means whereby teachers and materials writers can do more to
help students learn chunks durably. In the case of figurative idioms, or
‘figuratives’ (Grant and Bauer, 2004), we considered ways of promoting
elaborative processing likely to help students remember meanings. We
concluded that this kind of work is likely to result in enlarged reper-
toires of chunks, of this one broad type, that can be well understood
but not necessarily retrieved on demand. We also argued that many
chunks – relatively high frequency binomials and similes, in particu-
lar – manifest patterns of sound repetition (such as alliteration) which
can be pedagogically exploited in order to increase the number of
chunks that learners can recall and also use in unplanned oral produc-
tion. Later on in this section we will pursue the issue of how teachers
can help learners fluently produce these and other kinds of L2 phrases
through appropriate reviewing activities. First, though, we need to con-
sider a key element of fluency – automaticity – and also the process(es)
of automatization whereby automaticity becomes established.
Near the end of a survey of theories of automaticity and automiza-
tion, Schmidt (1992) remarked that it was ‘unsettling to realize that
the mechanisms made available by psychological theorising for under-
standing L2 fluency derive primarily from the study of skill in such
tasks as typing’ (p. 378). Even today, over 15 years later, there has been
surprisingly little majorly illuminating experimental research into
automatization and automaticity in language, although a few studies
will be mentioned further below. The dearth of insights into linguistic
automaticity certainly stems to a great extent from the methodological
difficulties that are involved – as compared, say, with researching the
automatization of cigar rolling (cf. Anderson, 1982). Indeed, experi-
mental study of the automatization of lexical decision tasks is, so far
as we know, about as close as neuro-psychologists have yet come to the
study of the automatization of language production and reception in
conditions resembling authentic communication. In one experiment
involving a lexical decision task (Logan, 1988: 501–2), participants,
on being presented with a string of four letters, indicated as quickly
 Bearing in Mind 135

as possible whether or not the string was an English word. The same
set of ten words and ten non-words was used 16 times. Then, ten new
words and ten new non-words were added to each set and participants
were presented with the (enlarged) set 16 more times. Over the term of
the experiment, subjects took less and less time to identify words and
non-words. As is virtually always the case in such studies, the decrease
in reaction time conformed to the well-known ‘power law of learning’
whereby in early practice there is a rapid drop in reaction times (in
other words, a rapid acceleration of performance) while in later practice
a curve representing successive reaction times levels off asymptotically.
Prediction of this ‘power law distribution’ of reaction times with prac-
tice is, in fact, one of the main criteria by which competing theories
of automatization have been judged (Schmidt, 1992). Theories which
satisfy this criterion are, very broadly, of two types, those which are
memory-based and those which are process-based.
A well-known example of the former is Gordon Logan’s Instance
Theory (IT) whereby automatic performance results from an accumula-
tion of separate ‘memory traces’, or separate neural representations of
the performance (Logan, 1988). If the accumulation continues, at a cer-
tain point processing will tend to shift from a step-by-step run-through
of the procedure in question – which Logan calls an ‘algorithm’ – to
much faster, all-at-once, holistic retrieval. We know of no similar study
which has addressed the roles of practice (or repeated exposure) in the
automatization of chunk production (or interpretation). Oppenheim
(2000), however, indicates – without a great deal of discussion – that
IT might be applicable. To elaborate considerably on her account, the
storage and automatized retrieval and production of chunks would pro-
ceed in IT as follows (cf. DeKeyser, 2001; Robinson and Ha, 1993: 415;
Schmidt, 1992: 369–71). A word string is neurally stored so that it func-
tions as a motor program for speech production and as a perceptual
unit when language input is being attended to. Holistic storage occurs
when repeated instances of a word string being noticed cause its con-
stituents to become increasingly registered in memory as an ensem-
ble because each noticed instance of the serial co-occurrence of these
constituents establishes a separate trace in memory. (On the assumption
that each trace would involve more than one neuron, a trace is a small
neural network.) Each added memory trace of the word string as an
ensemble makes its retrieval as an ensemble more likely, since the more
traces there are, the easier it becomes to find one to retrieve. An inter-
esting feature of this theory is that algorithmic and holistic operations
on form or meaning run in parallel. For example, if – for production – a
136 Optimizing a Lexical Approach

memory search for a unified chunk of words lasts too long, the algo-
rithm will complete and the chunk will be formulated algorithmically,
that is, procedurally. As memory traces accumulate, it becomes less and
less likely that algorithmic processing will have time to complete and so
holistic retrieval/recognition becomes the norm. Notably, in IT automa-
tization is not only incremental but also probabilistic, since even in later
stages of practice holistic retrieval may, by chance, fail to occur before
the algorithm finishes. One criticism that has been made of this theory
can be summed up by asking: if each memory trace is a highly specific
and separate representation, how is it possible to represent relations of
similarity rather than just ones of identity? With regard to chunks, for
example, a single lexical phrase can come in various guises (with vari-
ous verb inflections, and with various variable slot-fillers, as in it took/
has taken/ is taking [someone] [period of time] to __). To remedy this weak-
ness, Nosofsky and Palmieri (1997) have outfitted IT with a theory of
categorization. This capacity for categorization would explain how sev-
eral variants of a ‘single’ collocation can be perceived as a single chunk
(and possibly as derivations from a ‘canonical’ form).
As already mentioned, few studies relevant to currently favoured
approaches in ISLA have taken account of recent automization/ auto-
maticity theory in neuro-psychology (for reviews, see DeKeyser, 2001;
2007; Segalowitz, 2003; Segalowitz and Hulstijn, 2005). In one study,
Robinson and Ha (1993) aimed to determine whether practice in apply-
ing an explicit grammar rule led to the power law pattern of perform-
ance acceleration predicted by IT. They found that IT was confirmed in
some respects but that the key prediction of a power law acceleration of
reaction times was not borne out. Robinson (1997) reports similar find-
ings from another study (see also DeKeyser, 2001: 142–3 and Segalowitz
and Hulstijn, 2005: 376 for insightful discussion of these two studies).
We can find no suggestion in the later literature that these findings
would be different if viewed in light of Nosofsky and Palmieri’s (1997)
extended version of IT.
A second category of memory-based theories are so-called strength
theories (see DeKeyser, 2001; and Logan, 1988, passim). Oppenheim
(2000) suggests that strength theory generally is a candidate account
of the automatization of chunk retrieval. In strength theory (unlike IT),
N distinct events of noticing a word string do not result in N distinct,
relatively fine-grained memory traces of form and meaning. Instead,
repeated experiences of the string bring about a strengthening of the
connection between a single coarse-grained, generic representation of its
meaning and a single (therefore probably lemmatic) representation of
 Bearing in Mind 137

its form (for remarks on the disadvantages of this, see Logan, 1988: 494
and passim). In other words, recognition of the similarities between sep-
arate noticing events is an integral part of noticing itself. Connectionist
Theory (for example, Rumelhart, McClelland et al., 1986) is also a
strength theory in the sense that repeated practice and noticed expo-
sures strengthen connections between particular elements in a string
of items.
In process-based theories, as opposed to the above memory-based
theories, the speed-up which is so characteristic of automatization
occurs because practice (that is, repeated noticed exposures or episodes
of performance) results in faster and faster run-throughs of a basically
unchanging procedure. Anderson’s well-known Adaptive Control of Thought
(ACT) Theory (1982; 1993) is an example. According to DeKeyser (2001:
135–7), within the memory-based camp there has been some accept-
ance that procedural speed-up may be an additional factor in automa-
tization (or, in process-based theory, ‘proceduralization’) and within
the process-based camp there has been some limited acceptance that
holistic retrieval may play a role. DeKeyser concludes, however, that
disagreement about the mechanisms of automatization is still profound
(see also Segalowitz, 2003; Segalowitz and Hulstijn, 2005).23
So far in this book we have tended to adopt the usual shorthand
explanation of why it is that knowing chunks facilitates fluency. With
respect to oral fluency, this facilitation is assumed to occur because it
is quicker and easier for a speaker – because it requires less attention
and less mental processing – to recall a chunk from memory than to
produce the chunk from scratch by recalling the individual words (and
maybe also various sub-word morphemes such as noun endings) and
combining them according to the speaker’s knowledge of grammar
(Becker, 1975; Pawley and Syder, 1983; Wray and Perkins, 2000). Thus,
in the blink of an eye, a chunk (which is considered, by definition, to
be in some sense holistically stored in long-term memory) is (1) located
in memory when it is needed to formulate a message, (2) retrieved as an
unanalysed whole, and (3) uttered, again without analysis. Although
in this chapter we glimpse only the tip of the iceberg of complex evi-
dence and conflicting interpretation, we have already seen enough to
correctly guess that there might be considerable differences of opinion
among researchers about the accuracy of this shorthand view, both in
general and in detail. This is the first of several reasons why we are
adopting a rather eclectic stance regarding the types of reviewing activ-
ities that are recommendable for fostering fluency (through automatiza-
tion) in chunk output.
138 Optimizing a Lexical Approach

With specific respect to idioms, we have seen (in Chapter 5, Section 5.1)
that it is now widely believed that chunks of this kind are stored and
processed in a hybrid fashion – partly holistically, partly (de)compos-
itionally. Still, regardless of theoretical stance and differences of opin-
ion about type of processing (whether it be memory-based, algorithmic/
procedural or hybrid) and about other matters such as whether auto-
maticity is a unitary or a multiplex phenomenon (it certainly has both
motoric and cognitive aspects),24 researchers do tend to agree that there
is something which needs to be explained and that this ‘something’ may
as well be called either automaticity or automatization, depending on
whether one is concerned with the result or the process that leads up to
it. For Segalowitz (2003) this ‘something’ has the feel of changing from
a car with a manual shift to one with an automatic shift. For Logan
(1988), its essence consists entirely in radically accelerated performance.
However, in addition to acceleration, other characteristics of automati-
city have been advanced in the literature, including increased accuracy of
performance and greatly diminished allocation of effort and attentional
resources (see, for example, DeKeyser, 2001; Schmidt, 1992). Norman
Segalowitz (2007), perhaps the most prominent researcher of automati-
city with a strong interest in ISLA, considers acceleration of performance
to be symptomatic rather than explanatory. In any case, we will fol-
low Segalowitz (2003, 2007) and others (DeKeyser, 2001; Schmidt, 1992;
Segalowitz and Hulstijn, 2005) in assuming that, yes, automaticity is a
vital component of fluency, which itself is the ability to produce or com-
prehend utterances smoothly, rapidly and accurately.
Viewing fluency substantially in terms of automaticity as a cognitive
phenomenon, Segalowitz (2000: 211–14) has claimed that two research
findings are of particular relevance to the promotion of fluency in
ISLA. Firstly, memory traces of learned information include a record
of perceptual and cognitive processes ongoing at the time of learning,
which means that later recall of this information occurs most readily
in circumstances that are similar to those which obtained at the time
of learning (Tulving, 1983). Thus, L2 words and lexical phrases needed
for a given communicative purpose will come most readily to mind if
they were learned in the course of communicative activities bearing a
strong resemblance to the context in which they will be needed after-
wards. This calls to mind classroom activities such as simulations and
role plays, which have been rather popular in CLT for practising so-
called functions (such as apologizing, requesting, asking for directions,
and so on), which CLT half-borrowed from Speech Act Theory (Searle,
1969). These activities can be set up for students to practise multiword
 Bearing in Mind 139

formulae that are common in oral interaction and which thus require a
high degree of fluency (in order to avoid awkward hesitations or substi-
tution of unconventional phrases that could put a strain on the inter-
action), formulae such as How are you doing?, Excuse me, Here you are,
There you go, Can I give you a hand with __?, That’s very kind of you, Much
obliged, Would you mind __?, I was wondering if __, Not at all, Help yourself,
I’m afraid so/not, No kidding; I’m (so) sorry to hear that; Nice meeting you;
Have a nice day and See you. Simulations and role plays25 might even be
designed that elicit certain reference-oriented chunks (such as certain
verb–noun collocations and partitives), for example in Language for
Specific Purposes courses such as Business English (see Powell, 1996,
for a teacher’s resource book for Business English inspired by the LA),
where students can be trained in the fluent use of both topic-related
collocations and interactional chunks (simulating a negotiation, for
instance). However, experienced teachers will bear witness to the fact
that student groups tend to be divided in their like (or dislike) of simu-
lations and role plays. If only for such affective reasons, these activities
need to be complemented by others that also foster fluency. Another
potential drawback of simulations and role plays is that they can be
rather time-consuming.
The second research finding which Segalowitz (ibid.) has claimed is
of particular relevance to the promotion of fluency in ISLA is that auto-
maticity is promoted by consistent mapping between stimuli and cog-
nitive responses (Schneider and Shiffrin, 1977; Shiffrin and Schneider,
1977). It is therefore important to ensure consistent mapping during
classroom practice which is intended to lead towards fluency in L2. In
this connection, Segalowitz (2000: 212) identifies inconsistent mapping
as a major flaw in the Audiolingual Method (ALM) despite the fact that
development of oral fluency is, or was,26 one of its characteristic aims
(Richards and Rodgers, 1986: 52).
As is well known, the ALM relied heavily on oral pattern practice,
or ‘drilling’. One type of drill is the repetition drill in which a model
sentence is spoken by the instructor (or heard on tape) and repeated by
learners, often chorally first and then individually (see Stevick, 1986:
89–105, for a particularly thoughtful consideration of drilling.) During
repetition drills, great emphasis is placed on accuracy and fluency. But,
as one of us clearly recalls from a classically ALM intensive Russian
course taken in 1966–67, a repetition drill was liable swiftly to turn
into a drill requiring substitutions. Suppose, for example, the model sen-
tence is Dale goes to work. Since go to work is very probably a chunk in
the mind of the typical native-speaker, it would seem that repetition
140 Optimizing a Lexical Approach

of this model sentence has some potential to promote automatization.

In the quickly ensuing substitution drill, however, the teacher might
first say ‘They’ (whereupon students would be expected to say They go
to work), then ‘Town’ (students should now say They go to town), then
‘Park’ (for She goes to the park) and so on. In other words, any chance of
automatizing chunks (such as go to work) had secondary priority in the
ALM. Instead, it was automatized27 manipulation of elements within
grammatical paradigms (in other words, training students to apply so-
called grammar rules in an ‘automatic’ fashion) which was intended
by the ALM (but see Robinson, 1997, and Robinson and Ha, 1993, cited
above, on the questionable results). One might of course argue that a
substitution drill such as the invented example above could be modi-
fied so as to serve the purpose of practising multiple chunks, and within
a short period of time too, for if go to work is a chunk, so also, probably,
is go to town. These could be drilled well-separated from another set of
go collocations without to, such as go home and go downtown, and yet
another set that requires the definite article, such as go to the movies and
go to the theatre. But ALM drills – and there are at least ten further types
(Paulston, 1971; Richards and Rodgers, 1986: 54–5) – are difficult for
teachers to manage unless they read the prompts out of a book, which
merely introduces new disadvantages (dwindling teacher–student rap-
port, for instance). Given how hard it can be for teachers to lead com-
plex drills, it is not surprising that this component of the ALM was
often partly, or largely, consigned to language laboratories. In any case,
it has often been noted that grammar-focused, ALM-style drilling ses-
sions are susceptible to severe problems if not expertly managed and if
learners are not highly motivated. For the ALM to be fruitful – which it
occasionally has been (see further below) for the lower-level learners it
was originally intended for (Howatt, 2004) – a good aptitude for learn-
ing languages may matter even more than usual. Specifically, from the
learner’s point of view, ‘classic’ ALM sessions are too liable to break
down into a welter of formal options whose meanings and functional-
ities are lost sight of and thus come across as divorced from the commu-
nicative needs outside the classroom (or language laboratory).
Drawing on early findings in experimental cognitive psychology
about mental imagery (Bugelski 1982), dual coding (Paivio, Yuille and
Madigan, 1968) and spreading activation (Anderson, 1984), Stevick
(1986) made a number of detailed proposals for remedying the shallow
treatment of meaning in ALM-style practice of grammar. Unfortunately,
by his time of writing the ALM had fallen into general disrepute (at least
in EFL), dragging intensive oral repetition practice a good part of the
 Bearing in Mind 141

way down with it; in any case, Stevick (1986) was relatively uncon-
cerned with vocabulary, phrasal or otherwise. It is also unfortunate that
a significant, mainstream revival of interest in L 2 vocabulary did not
begin until the 1980s (Meara, 1980), by which time the ALM was no
longer popular and thus an unlikely source of inspiration for investiga-
tions into vocabulary teaching, or chunk teaching in particular. Now
though, more than a generation after the onset of the ALM’s swift fall
from grace,28 things may be changing, at least with regard to oral repe-
tition. According to Gatbonton and Segalowitz (2005), the lesson to be
drawn from the demise of the ALM and similar methods is not that
fluent, accurate repetition is inadvisable but that teachers should focus
on utterances for themselves and not as means to teach grammar. To this
end, Gatbonton and Segalowitz (ibid.) have proposed an utterance-based
approach dubbed ACCESS – short for Automatization in Communicative
Contexts of Essential Speech Segments.
According to Gatbonton and Segalowitz, ACCESS differs from CLT
partly as follows. In early CLT (see Byrne, 1976; Harmer, 2001: 80–2;
Paulston, 1971; and Rivers and Temperley, 1978), lessons begin with
form-focused exercises and move towards more or less communicative
activities. Thus, students are drilled before they engage in communi-
cation. In later versions of CLT, known as Task Based Instruction (TBI)
(for example, J. Willis, 1996) a lesson is likely to progress in the oppos-
ite direction so that communication precedes focus on form. In either
case, repetitive practice – if there even is any – is temporally separated
from communication. This is also the case in Day and Shapson’s (2001)
proposal advocating the use of form-focused language games both
before and after communicative activity. In ACCESS, however, lessons
are designed so that automatization of a ‘critical mass’ of target utter-
ances can occur within a communicative activity structured so that
tokens of each utterance are repeatedly elicited, and in this way learn-
ers get the benefits of CLT (as enumerated, for example, by Harmer,
2001) as well as the benefit of plenty of repetition. The task described
in Gatbonton and Segalowitz (ibid.) is for near beginners and is called
‘Family Relationship’. In it, students, working in groups of eight to ten
members, are supposed to: pretend to be one family, decide how they
are all related, draw a family tree and, later, explain their family tree
first to members of other groups and then to the whole class.
What is particular about ACCESS, is that it has been developed to
address massive experimental evidence – seldom adequately recog-
nized in any modern form of CLT – that automatization is promoted
by repetition (Schneider and Chein, 2003).29 Importantly, Gatbonton
142 Optimizing a Lexical Approach

and Segalowitz (ibid.) counter the contention that focus on meaning

impedes focus on form, and vice versa (Skehan, 1998; VanPatten, 1990),
by arguing that this is the rule only when students are given too few
opportunities to hear and use a particular utterance.
Still, it is reasonable to view ACCESS as a variant of TBI, given the
fact that a complete unit of ACCESS instruction is built around a multi-
stage communicative task in accord with Swain’s (1993, 2000) output
hypothesis that production of (well-structured) utterances in real con-
versation drives development of students’ interlanguage. However, the
divergence of ACCESS from other variants of TBI is of relevance to any
attempt to optimize the LA precisely because ACCESS adds to TBI, and
to modern CLT generally, a largely missing provision for the devel-
opment of fluency (seen in terms of automaticity), and it does so by
using tasks designed to encourage a great deal of repetition of ‘essential
speech segments’ (ESSs). Gatbonton and Segalowitz (ibid) consider these
to be formulaic expressions which (1) have clear pragmatic functions;
(2) occur with minimal structural variation; (3) have high potential for
re-use; and (4) can serve as complete utterances. Among the relatively
few examples that they give are: ___ is my father; I am his ___; I am mar-
ried to ___; I stayed home; I went out.30 In light of their definition and
their examples, it should be evident that the class of ESSs must include
many chunks such as the interaction formulae mentioned above in
connection with ‘functions’ or speech acts, which are worth relatively
prolonged attention even at beginner and elementary levels – if, that is,
the students wish to be able to easily and freely converse with native or
other expert speakers in the near future.
A striking element in the rationale for ACCESS, is the claim – which
to us has the ring of truth – that genuine (as opposed to sham) CLT has
not, in fact, taken firm root around the world. With particular reference
to China and Japan, Gatbonton and Segalowitz (ibid.) suggest that one
reason for this dearth of uptake by teachers (as opposed, say, to minis-
tries of education), is that:

[...] many teachers have difficulty seeing the learning value of

communication activities. Teachers in many parts of the world
are used to highly structured activities such as teaching gram-
mar rules, conducting drills, and teaching vocabulary lists, which
makes it hard for them to accept that activities such as games,
role-plays, and problem solving with little obvious language teach-
ing purpose can actually count as ‘real teaching’. (Gatbonton and
Segalowitz, 2005: 327)
 Bearing in Mind 143

‘Nowhere’, they add, ‘is this more evident than in CLT’s approach to
fluency’ (ibid. p. 327).
This raises two important questions. Firstly, while we are inclined to
accept that ACCESS, or some form of TBI very like it, is ideal for teach-
ing and learning ESSs (real communication is, after all, the immediate
context), and while we expect ACCESS could be adapted so as to cover a
wider variety of chunks than envisaged by its originators, is it really the
case that chunks can never effectively be practised in un-TBI-like ways?
And if several options for practising a given set of chunks are available,
would it not be sensible to open-mindedly investigate which of these
are comparatively efficient, even though some of the options are some-
what reminiscent of the rejected ALM? The dearth of empirical research
on automatization in language learning (especially chunk learning) is a
second reason for us to adopt an eclectic stance concerning the choice
of classroom activities designed to foster chunk-fuelled fluency in ISLA.
Even with respect to TBI as a means of teaching grammar (which has
received much more attention than its merits for vocabulary acquisi-
tion, let alone chunk acquisition), Swan (2006) has pointed out that key
claims made for TBI (for example, that a great deal of grammar learn-
ing can happen during student-to-student negotiation of meaning) and
against explicit teaching of grammatical forms by non-TBI methods
(that is, that these methods have been conclusively shown to be fail-
ures) are far from well-substantiated (see also Ranta and Lyster, 2007).
It is interesting in this connection that Richards and Rodgers (1986)
observe that a pre-cursor of the ALM (one deriving from the so-called
‘informant method’ used by linguists such as Leonard Bloomfield) was
adopted by the U.S. Army where, ‘in small classes of mature and highly
motivated students, excellent results were often achieved’ (p. 45). From
what we know of the ALM and its immediate predecessors (see Richards
and Rodgers, 1986: 44–63), we may infer that oral fluency was a key
point of evaluation (especially given that in the ALM vocabulary load
is typically kept to a minimum). In other words, something like the
ALM, with all its evident faults, has apparently been judged to have
succeeded in fostering oral fluency. More generally, Hulstijn (2001: 86)
has remarked:

‘Intentional learning’, ‘rehearsal’, ‘practice’, ‘drill’, and ‘automaticity’

are terms which often elicit negative connotations among L2 special-
ists, being associated with the superficial parroting of meaningless
stimuli ... However, several decades of psycholinguistic research have
made it clear that lexical information simply must be reactivated
144 Optimizing a Lexical Approach

regularly for it to remain quickly accessible. Therefore, these terms

deserve to be updated in the jargon of the L2 specialist, albeit with
the note that the nature of the processing during a rehearsal event
[e.g., shallow or elaborative processing] will determine the likelihood
of the information being rescued from the fate of oblivion. With
that proviso in mind, it is legitimate to conclude that ‘intentional
vocabulary learning’ as well as ‘drill and practice’ must have a [com-
plementary] place in the L2 curriculum ...

We support this view, especially with respect to developing students’

ability to remember and to fluently produce and comprehend chunks
of all sorts. We support it all the more since evidence from research
into instructed L2 vocabulary learning (see especially Nation, 2001) has
discovered value in a substantial number of techniques and procedures,
such as translation (Laufer and Girsai, 2008) and the memorizing of
longish texts (Wray and Fitzpatrick, 2008) which have little or noth-
ing in common with the spirit and principles of TBI. As for teacher-
led oral drills, we believe they can be useful, especially in beginner
and post-beginner classes, provided that each drill always refers to a
well-clarified (kind of) real-world communicative setting and that map-
ping between form and meaning/function is always consistent. For one
thing, use of drills and other non-TBI techniques and exercises may
fit better than TBI itself with educational traditions prevalent in, say,
China. Also, without at all disputing that TBI offers important benefits
(such as the furthering of learner autonomy), it is relevant to note that
most versions of TBI (Skehan, 1998; J. Willis, 1996) tend to proceed
through prolonged, markedly codified, multi-stage units of instruction.
However, it seems to us that it frequently is unnecessary, undesirable
or impractical to deploy TBI in full form in order to focus briefly on a
small set of chunks with the aim of promoting swift recall and fluent
delivery. Thus, while ACCESS does appear highly promising as a means
of fostering fluent control of some types of chunks in the context of a
TBI lesson, we do not believe (and Gatbonton and Segalowitz do not
seem to claim) that acknowledging the merits of ACCESS entails accept-
ing it as the only means whereby fluent control may be established.
In an eclectic mindset, but still drawing lessons from what have been
shown to be the weaknesses of the ALM (see, for example, Prabhu,
1987), we recommend the following two easy-to-apply principles for
the design of oral review practice when the aim is to equip students
with the ability to quickly retrieve and fluently produce certain
chunks of language. Principle 1 is that both fluency in oral interaction
 Bearing in Mind 145

(which is partly a matter of motor skill) and memory formation are

promoted by repeating target language out loud (for example, N. Ellis,
1996a: 103–4). Principle 2 is that it is the establishment of consistent
associative connections in long-term memory that provides an essen-
tial basis for automaticity (Ellis, 1996a: 107). Oral repetition practice of
short, well-contextualized, and thoroughly understood situational dia-
logues (repeated at growing intervals) is in accord with these principles.
Substitution drills, transformation drills and other drills that endanger
consistent mapping of form and meaning are not. Also in accord is con-
solidation practice through the use of prompts that help students to
recall and say the whole of a chunk that they have previously learned.
There are many possible variations of this practice (see Lindstromberg
and Boers, 2008c). For example, on being reminded of a particular situ-
ation, text or topic, students can be shown relevant prompts. On seeing
the prompt, students should call out the whole chunk. In the case of
situational formulae such as fingers crossed, knock on wood (or touch wood)
and small world, prompts can be in the form of mime (for fingers crossed
and knock on wood, for instance) or a simple drawing (for small world, for
instance). Or they can be in the form of hints such as ‘In a distant place
you meet someone from your home town and you say ...’. The prompt
can also be one word of the chunk or the chunk written in abbreviated
form. For example, either ‘move’ or ‘L. m. on’ could be used as prompts
for Let’s move on (to Chapter 8, for instance).31
8
Directions

8.1 Broadening the scope

In this section we will consider a number of possibilities for broadening

the scope of the pedagogical approach to chunks we have outlined in the
book thus far. We will briefly explore the following avenues: (1) apply-
ing the elaboration techniques to single words in addition to multiword
lexis, (2) expanding the armoury of chunk-teaching techniques, and
(3) applying the LA in teaching languages other than English.
In order to keep our book project manageable, we have focused almost
entirely on multiword lexical units, but one should not forget that the
LA originally had to do with foregrounding lexis in general. It is because
the LA emphasizes the point that language should be recognized as
grammaticalized lexis instead of lexicalized grammar (Lewis, 1993: iv)
that the LA gives to the syntagmatic behaviour of words (i.e., how
words combine) the attention it merits, but – although comparatively
little mention has been made of this – helping students acquire single
words has certainly also been part of the overall approach (see especially
D. Willis, 1990). For one thing, it is no easy matter to neatly separate
multiword from single-word lexis. In this book we have tended, impli-
citly, to go by spelling conventions in making the distinction but have
done so out of practical motives rather than firm principle. For example,
we have counted as chunks compounds such as night shift which are
spelled as separate words, but usually not ones like nightcap that happen
to be spelled as single words (we made an exception in one of the allit-
eration studies reported in Chapter 6). Anyone wishing to question the
correctness of distinguishing chunks from non-chunks on the basis of
spelling conventions might point out that hosts of compounds spelled
as two or more separate words in English are spelled as single words in

146
 Directions 147

other Germanic languages – for example, English Christmas tree, and

and so on / and so forth versus Dutch kerstboom and enzovoort. Moreover,
some of the compounds that are spelled as single words in contemporary
English were written as multiword items in earlier stages of the language.
For example, Bunyan (The Pilgrims Progress, 17th century) spelled wayside
as way side (Freeborn, 2006: 366). Other examples are discourse connec-
tors such as nevertheless, notwithstanding, however and furthermore, which
also used to be written as multiword items. When a compound comes to
be spelled as a single word, this may be taken as a reflection of just how
closely bound the constituent parts of the word combination (that is,
of the chunk) eventually became in speakers’ minds. (And it cannot be
doubted by anyone who has seen much writing by native Anglophone
teenagers that new single-word compounds would be appearing with
each passing decade – were it not for the conservatism of modern educa-
tors, lexicographers and publishers.)
If it is useful to have students contemplate the lexical makeup of
chunks spelled out as multiword items (as we have claimed in this
book), then it might be useful to do so also with regard to ones spelled
as a single word. Firstly, this may reveal patterns of sound repetition
with mnemonic potential (see Chapter 7). The composition of chunks
written as single words may be influenced by phonological motives as
much as that of multiword chunks. A quick count in the MED tells us,
for instance, that no fewer than 17 of the 58 compounds starting with
back- which are spelled as single words, show alliteration and/or rhyme
or assonance (for instance, backbencher, backbiting, backpack, backtrack,
backlash and backslapping).
Secondly, if the meaning of multiword chunks is often (partly)
inferable from their constituent parts, then this may also be the case
for compound items that are spelled as a single word. The semantic
decomposability of single words has been the object of a recent study
by Sanchez (forthcoming), who analysed the 2500 highest-frequency
lexemes of English and German. Sanchez adopts the vantage point of
an idealized second language learner in estimating what proportion
of this subsection of vocabulary is intrinsically ‘learnable’ because the
meaning of the words can be worked out through lexical decompos-
ition. For example, the meaning of rarely can in principle be worked out
by a learner who understands both rare and the function of -ly. Sanchez
estimates that over 13% of the English words on the list and over 17%
of the German words are fully decomposable. Given the likelihood that
monosyllabic words are over-represented in the top frequency bands,
it is very possible that the percentage of semantically decomposable
148 Optimizing a Lexical Approach

words in the vocabulary at large is considerable higher than these fig-

ures suggest. In addition, Sanchez estimates that over 45% of the words
on her English and her German word lists are made up of words the
meaning of which the learner will not be able to arrive at through lex-
ical decomposition alone but which are nevertheless partially analys-
able. For example, inferring the meaning of yesterday by breaking it up
as yester ⫹ day will be blocked by the absence of yester as an independ-
ent lexeme in contemporary English. An attempt to infer the meaning
of discover may start well if it is analysed as dis ⫹ cover; but a figurative
interpretation is required on top of this (namely, that taking a cover off
something stands, metonymically, for finding it). We conjecture that
also here teachers can play the role we have made a plea for throughout
our book: they can help students appreciate non-arbitrariness where the
students themselves are unlikely to recognize it.
One of the pathways for semantic elaboration we advocated in
Chapter 5 was dual coding, which teachers can encourage by, for
example, associating the figurative meaning of idioms with the source
domains in which they were originally used literally. A similar tech-
nique can readily be applied to polysemous single words which happen
to have both literal and figurative senses. Polysemy has been one of
the main areas of interest of CL (see, for example, Radden and Panther,
2004, a collective volume devoted to linguistic motivation). Brugman
(1981) set the tone for a series of detailed lexical analyses of highly poly-
semous prepositions – over, in Brugman’s case – where the diverse senses
of a particular word (including figurative senses as in talk it over and
preside over a meeting) are all linked directly or indirectly to a central,
spatial sense (as in jump over the fence). Let us use a simpler word than
over to illustrate this kind of motivated polysemy: the figurative sense
of beyond in Why she decided to tell him is beyond me is most probably a
metaphorical extension of the meaning of beyond we see, for example,
in The ball is beyond the neighbours’ hedge, where beyond means ‘on the
other side of and some distance from’ [the hedge] and thus not at all
near and graspable. The case made by CL for motivated polysemy has
inspired proposals for the insightful teaching and learning of preposi-
tions (see, for example, Lindstromberg, 1997, a book-size treatment of
English prepositions generally) and experiments have been conducted
to measure their pedagogical merit. Boers and Demecheleer (1998), for
instance, found that students were better able to interpret the figurative
senses of beyond (as in This theory is beyond me) if they were first given a
definition of the prototypical spatial sense from which each metaphor-
ical sense extends. (The definition in this case is one that emphasizes
 Directions 149

that beyond implies that there is some distance between the two ref-
erents, a semantic feature which makes beyond a good word to use in
order to express the metaphor ABSTRACT INACCESSIBILITY IS DISTANCE.) This
finding was corroborated in a follow-up study reported in Boers et al.
(2008).
However, it is not only high-frequency words that are polysemous:
the vast majority of medium-frequency words also have several senses or
usages, very often one literal and one figurative. The basic CL strategy
in explaining a derived figurative usage with a view to making it more
memorable is to make learners aware of its connection with the central,
literal sense. In Chapter 5 we exemplified this with reference to experi-
ments (Lindstromberg and Boers, 2005a) where students benefited from
having been taught the literal usage of certain manner-of-movement
verbs (such as leap and stumble) when they were later asked to interpret
the same verbs in figurative chunks. Csábi (2004) showed the efficacy
of essentially the same technique when it was applied to the figurative
uses of the verbs keep and hold.
In Chapter 5 we also made recommendations for the insightful
organization of lexis. For example, we recommended the grouping of
idioms according to the source domains (such as seafaring) in which
the phrases were originally used in a literal sense. The usefulness of
this kind of grouping of single words that are used figuratively was put
to the test in one of the experiments in Boers (2000b). Students taking
a course in Business English were given a rather elaborate vocabulary
list containing words such as soar, plunge and slide with a view to help-
ing them describe upward and downward trends with greater precision
(when commenting on a graph their interlocutor could not see, for
example). For one group of students the words were categorized under
the source domains of ‘aircraft’, ‘diving’ and ‘mountaineering’; that
is, the domains where the target words are used in their literal senses.
For another group of students the words were categorized according to
whether they described comparatively slow or fast change. As a post-
test, the students were asked to write a commentary about two graphs.
Counts of the number of different ‘up-down’ lexemes used in these
commentaries revealed that students in the first group used a signifi-
cantly wider range than did the students in the second group. In sum, it
looks as though the various strategies for elaboration we have outlined
in connection with chunks are applicable also to at least some sets of
single words.
Some descriptive CL investigations of polysemy have taken
a diachronic perspective, to analyse, for instance, processes of
150 Optimizing a Lexical Approach

grammaticalization and semantic shift in (modal) auxiliaries (Hopper

and Traugott, 1993; Traugott and Dasher, 2002). A historical linguis-
tics perspective has not so far been adopted much in CL approaches to
language teaching, nor in fact in language teaching generally (excep-
tions include Ilson, 1983; Pierson, 1989, and more recently, Schmitt
and Marsden, 2006). Yet, historical linguistics could be a welcome
additional source of insights about pedagogically exploitable linguistic
motivation – not just so that more (types of) lexemes (single words as
well as chunks) can be revealed to be highly teachable but also so that a
wider range of learning styles, or cognitive styles, can be catered for. For
instance, we know from correlations of students’ learning gains with
their responses to learning-style and cognitive-style questionnaires
that although particular imagery mnemonics are beneficial to most
learners (see Chapter 5), some seem not to benefit from them at all
(Boers, Eyckmans and Stengers, 2006; Boers, Lindstromberg et al. 2008;
Boers, Piquer P íriz et al. 2009). But as, hopefully, more and more empir-
ically validated teaching techniques become available, the greater the
number of learner profiles that teachers will be able to accommodate.
Additionally, knowledge of the semantic shifts of words over time can
sometimes shed light on the meaning and makeup of word strings that
have been established as chunks for a very long time. For instance, the
semantic symmetry of neither fish nor fowl is perceived more clearly if
fowl is interpreted in its obsolete sense of ‘birds in general’. Similarly, one
man’s meat is another man’s poison becomes more symmetrical if meat is
understood in its obsolete sense of (solid) food in general. The phrase
hazard a guess may perhaps make more sense to learners if they see how
hazard is reconnected to its French source le hasard (⫽ ‘chance’) rather
than its current meaning of ‘danger, risk’ (as in health hazard). Perhaps
also the use of assume in the collocation assume responsibility can also
be better understood if it is reconnected to its French source assumer,
one of whose meanings is ‘accept, come to terms with’. To give just one
further example, knowing that the oldest meaning of draw is some-
thing like ‘pull or drag’ may help learners better understand, and better
remember, chunks such as draw someone’s attention to, draw blood, draw
someone out (of their shell) and draw on (resources).
Diachronic information may even afford insights into why whole
groups of words are more represented in chunks than others. For
example, one may hypothesize that – at least in non-technical dis-
course – it generally takes some time for words to bond into strong
collocations and compounds, for the simple reason that for any com-
bination of individual words to become strongly neurally associated
 Directions 151

with each other in the minds of many speakers, the combination must
occur sufficiently often in the discourse of the language community
in question. In English, the oldest core of vocabulary is Germanic. To
that core was first added Norman French vocabulary in the aftermath
of the Conquest. But English continued to receive words from French
long after the link with Normandy was broken. Another language that
English has extensively drawn vocabulary from – mostly for ‘academic’
and religious purposes), and often via French – is Latin. Learners whose
mother tongue is descended from Latin might well find it interesting
to learn what the chances are that a randomly selected English chunk
will include a Latinate (that is, French- or Latin-derived) word which
is cognate with a word in their own mother tongue. If there were any
plausibility to our hypothesis that ‘time in the language’ is an import-
ant variable in chunk formation, then – at least in non-technical gen-
res – we might expect a difference in type-frequency in chunks between
the comparatively recent Latinate words and older English-core words.
In a preliminary and somewhat informal attempt to see if this expect-
ation might be borne out, we turned to the Oxford Dictionary of English
(ODE, 2005) because it includes many short multiword entries (mostly
compound nouns plus a few adjective ⫹ noun collocations) and because
it gives etymological information. Because it is a big (albeit single-
volume) dictionary for native-speakers, with 2054 large pages, the ODE
includes many entries which are uncontroversially technical, regional
and/or obsolete. These we excluded from analysis because it was more or
less everyday standard international English that we were interested in.
We looked at all the headwords beginning with pa because this stretch
of the dictionary seemed certain to include a goodly number of Latinate
words (viz. the prefixes paedo-, paleo-, para-) as well as ones Germanic
in origin. To summarize somewhat, we decided to call ‘English’ any
headword which is attested from Middle English or before, even if it was
borrowed from French. We decided to call ‘Latinate’ any headword bor-
rowed from Latin and/or from French after the Middle English period.
We did not count words borrowed in the modern period from languages
other than Latin or French (English has also borrowed considerably from
Dutch and Spanish, for example). In counting headwords, we focused on
‘word families’ (a.k.a. ‘base words’), so pace, pacer, pacy were counted as
one item not three. Chunks, though, were counted separately; so, for
the pace family, pacemaker and pace-setter, for example, each got one tally
mark. This allowed us to compare the contribution made by Latinate
word families and ‘English’ word families to the slice of English chunk
repertoire represented in this section of the dictionary. Our tallies in this
152 Optimizing a Lexical Approach

section of the ODE yielded 73 Latinate word families. All together, these
participate in only 28 multiword entries. By contrast, the 90 ‘English’
word families we counted participate in no fewer than 174 multiword
entries. It thus appears that Latinate words participate in (short) English
chunks significantly less frequently than ‘English’ words do.
Conceding the possibility that we chose an unrepresentative section
of the dictionary to look at, we decided to carry out another pilot inves-
tigation (again using the ODE), this time based on counts of data on
every 50th double-page spread, starting with pages 4 and 5 (in other
words, we examined pages 4–5, 54–5, and so on); this amounts to over
80 pages. On these pages we counted 117 non-technical multiword
items. We then checked the etymology of every individual word they
are composed of. We found that over 70% of the words making up
these multiword items turned out to be of Germanic origin (chiefly Old
English and Norse). The finding that not even 30% of the words are of
French or Latin origin is all the more striking considering Finkenstädt
and Wolff’s (1973) estimate that only 25% of the wordstock of English
is Germanic, as compared to about 62% deriving from French, Latin
or Greek (often borrowed via Latin). It cannot therefore be the case
that we found more Germanic chunks simply because there are more
Germanic words from which chunks can be composed. In short,
although the ODE defines a huge number of conspicuously Latinate
words, they figure very minimally in those non-jargon chunks which
the compilers of the dictionary deemed important enough to profile
as headwords. So, while learners of English whose mother tongue is a
Romance language may find solace in the fact that English contains a
great deal of Latinate lexis which can serve them as lexical toeholds
in vocabulary acquisition, it seems that these cognates will not, on the
whole, greatly help them to master English chunks. In fact, our study
suggests that even relatively short French- and Latin-derived words
tend much less than similarly short Germanic words to figure in the
kinds of short chunks accorded headword status in the ODE. In fact,
it seems possible that half a millennium or more of time in the lan-
guage has not sufficed for large numbers of by no means recondite
Franco-Latinate loan words to become completely English, at least with
respect to chunk formation. Certainly this merits further investigation.
Meanwhile, we must of course not lose sight of the fact that Latinate
and Greek-derived words do figure prominently in many thousands
of chunks in the professional discourses of sciences and modern tech-
nologies. However, from what we know, the rather codified, stipulative
processes whereby technical chunks become accepted in this or that
 Directions 153

field are very different from the grassroots processes which operate on
everyday, non-technical vocabulary.
Diachronic information, such as information about borrowings, may
possibly also shed light on patterns of ‘stylistic prosody’ in chunk for-
mation. For example, we know that French lexis was introduced into
English mostly by the aristocracy and the upper classes, and that Latin
was introduced mostly by the clergy and the learned (Baugh and Cable,
1993). It is not surprising, then, that many words from French and
Latin origin tend to be generally more ‘formal’ than their synomyms
of Germanic origin. Could the comparatively formal nature of Latinate
words influence their collocational behaviour – that is, on the com-
paratively rare occasions when they do actually collaborate in unstipu-
lated, ‘grassroots’ chunk formation (see above) – in the sense that they
may seek the company of other comparatively formal, Latinate words?
Is it a coincidence, for example, that severe fatigue (both loans from
French) sounds more idiomatic than severe tiredness and that the med-
ical world has opted for chronic fatigue syndrome as a technical term
rather than chronic tiredness syndrome? Statistical evidence of such a
pattern of diachronically motivated stylistic prosody may be difficult
to find, because many common Latinate words simply do not have
a Germanic synonym to compete with in their collocational court-
ship, and many Germanic words have entered frequency bands where
words tend to have a wide range of collocates.32 When synonyms do
exist, however, the pattern may become discernable. For example, The
Collins Cobuild on-line collocation sampler generates, for the French-
origin word rapid, no fewer than 14 noun collocates ending in -ion (an
ending which is a conspicuous sign of French origin), for example, suc-
cession, deterioration, expansion and induction. The collocation list gener-
ated for its synonym fast contains no -ion noun at all. In addition, the
list for rapid contains many other nouns that are clearly French bor-
rowings, such as development, descent and reform, which are absent from
the fast list (which does, though, include the French-origin word furi-
ous, one of a number of other alliterative combinations such as flowing,
freeze and fun). Again, more and larger-scale research would be needed
to gain a clear picture of these matters, but in the event that more than
‘anecdotal’ evidence for a diachronic motivation in chunk formation
did become available, such motivation could possibly be used as an
additional pathway for pedagogical semantic elaboration, one with a
touch of historical linguistics.
In setting out our proposals for chunk teaching in the previous chap-
ters and also in the discussions in the present section of this chapter,
154 Optimizing a Lexical Approach

we have been guided mostly by the CL premise that language (includ-

ing its phraseology) is much less arbitrary than used to be widely
assumed. Still, we would not deny that many lexical phrases continue
to betray little or no linguistic motivation. Since we have relatively little
faith in incidental uptake of chunks in contexts of ISLA and since we
believe in the superiority of insightful learning over blind memoriza-
tion, we must wonder what teaching and learning strategies – beyond
those which exploit linguistic motivation – could usefully be recruited
to help students in settings of ISLA to commit chunks to memory in
comparatively efficient ways. One potential candidate is an adapted
version of the well-known Paired-Associates (PA) method, which was ori-
ginally developed around 1900 by Mary Whiton Calkins (Furumoto,
1991) and which can be a means of memorizing a list of target items by
pairing each one with some other item, usually one of a different kind
or class. This technique of working with lists of items which are paired
on the basis of some kind of association between them reflects the view,
popular in the early 20th century, that learning involved associations
of stimulus and response. A list of paired associates is considered to be
memorized when the learner can respond to any stimulus by saying
the associated response (also, we would say, vice versa). Unlike most of
the various mnemonics that have been proposed for learning single L2
words (Hatch and Brown, 1995; Nattinger, 1988; Schmitt, 1997; Sökmen,
1997; Thompson, 1987; Boers and Lindstromberg, 2008c: 14–17), the
PA method is at least somewhat adaptable to the project of learning L2
chunks.
For example, the PA technique may be a way of learning conversa-
tional formulae, which we have not given much attention to in this
book (except in Chapter 7, Section 7.2), but which do figure promin-
ently among Nattinger and DeCarrico’s (1992) learning targets: Hi, how
are you doing? → Fine, thanks; and you?; Could I have some more X → Sure,
help yourself; Sorry I’m late → That’s alright; My goldfish has died → I’m so
sorry to hear that; Fancy X? → I’d love to, but ...; and so on. It may benefit
learners’ conversational interaction skill to use such formulae appropri-
ately and fluently (that is, without awkward hesitations), if such formu-
lae are closely paired in memory with phrases or kinds of messages that
typically trigger them. In that way PA learning can quite evidently be a
refinement and extension of the ancient practice of memorizing lists of
L2/L1 meaning equivalents, especially when no one-to-one equivalents
are available.
Memorizing PAs, however, can also be used as a complementary way
of learning types of chunks for which applied CL does have certain
 Directions 155

proposals, such as phrasal verbs. As is well known, one reason why

learners find phrasal verbs difficult to understand and remember is
that it is often difficult, if not impossible, to find accurate synonyms
or translations. Another problem is that the same few verbs and prepo-
sitions tend to occur over and over again in different combinations,
which can lead to confusion about which phrasal verb has what mean-
ing. Fortunately, a number of recently published learners’ dictionaries
deal very well with meaning and usage of common phrasal verbs, and
materials for home study of phrasal verbs have improved in the same
respect. But the problem of how learners can remember phrasal verbs
remains. So, the first step in a PA approach would be for the teacher
and/or students to choose a small set of phrasal verbs to work on. It is
important, however to avoid repetition of main verbs within any tar-
get set and of prepositions, so as to lessen the risk that learners will
forget which meaning goes with which of two or more similar forms.
(Of course, a dictionary of phrasal verbs can be a valuable resource here,
particularly if it gives information about relative frequency.) Next, each
targeted phrasal verb is paired with a short sentence illustrating a com-
mon usage and including, if at all possible, an especially frequent col-
locate; this collocate will serve as the prompt (i.e., the paired associate).
For example, the Oxford Phrasal Verbs Dictionary for Learners of English
(2006) gives the following examples for pull off – She pulled off her gloves
and The goalie pulled off a terrific save. The former, which is distinctly
literal, is a suitable target at pre-intermediate level; the latter could be a
target for intermediate level or above. In the materials preparation stage
a list of PAs is produced which is based on the list of example sentences
as shown in Box 8.1 (based on Lindstromberg and Boers, 2008c: 38–9).
The students’ task now is to remember the (short) list of paired associ-
ates and also, although not necessarily verbatim, the matching example

Box 8.1 Creating paired associates for phrasal verbs

Associate 2: The
Associate 1: Targeted ‘prompt’ towards
Example sentence phrasal verb the model sentence

She pulled off her gloves. pull off gloves

Look new words up in your look up new words
dictionary.
Put away your toys! put away your toys
While cleaning the attic, came across old photo
she came across an old photo.
156 Optimizing a Lexical Approach

sentences. Entrenchment in memory can be fostered by means of infor-

mal, more or less graded testing (teacher-to-students quizzing, student-
to-student quizzing or individual self-testing (see the end of Chapter 7).
For instance, in pairs, students can take turns calling out the prompt
(the associate in column three of Box 8.1) and invite their partners to
call out the matching phrasal verb or the example sentence, or they can
take turns calling out the verbs and invite their partners to call out the
matched prompt (for more options and examples, see Lindstromberg
and Boers, 2008c).
Apart from the PA method, other old mnemonics might be consid-
ered for their potential usefulness for chunk learning. One of the cri-
teria would probably have to be whether a given mnemonic is likely to
permit the storage of many chunks (for adequate volume of learning)
and to facilitate fast retrieval (for the sake of fluency). For example, the
ancient mnemonic ‘loci’ (Latin for ‘locations’), which consists in men-
tally depositing each individual target datum in a different, visualized
location, is also somewhat serviceable for trying to remember chunks.
For example, if you want to remember a set of items (including, say, fly
by night), you can – in your mind’s eye – place them each, one by one,
at a different location along a route with which you are very famil-
iar. It is doubtless helpful to incorporate additional mnemonics such
as matching target items to locations by first letter (for example, fly by
night: florist’s shop) or adding details to the mental image (for instance,
to visualize the florist’s closed at night or with a big fly buzzing around
inside). The question is whether this type of mnemonic can accommo-
date a lot of chunks (for one thing, you would need a lot of ‘loci’ to asso-
ciate them with), and also whether the mental scanning of the fictitious
itinerary might not slow down retrieval. Again, empirical studies would
be welcome to assess the effectiveness (and, crucially, the efficiency – in
terms of return on investment of time and effort) of mnemonics like
these when they are applied to the learning of lexical phrases.
Let us now turn to the question as to how relevant and useful the LA
is when it is applied to the teaching of languages other than English.
Nattinger and DeCarrico (1992: 66) assert that ‘Lexical phrases [...] exist
in the same abundance and perform the same conversational functions
in other languages as they do in English, and they can therefore be
exploited no matter what the language being taught’; and they give
plenty of examples of lexical phrases in Chinese, Spanish and Russian
to back up this assertion. Myles, Hooper and Mitchel (1998) and
Forsberg (2006) demonstrate the importance of formulaic sequences
also in the acquisition of L2 French. Given their fundamental role, it
 Directions 157

would be odd indeed if chunks were found to be more prevalent in one

language than another. And yet, popular belief persists that English
is an exceptionally idiomatic language.33 If this popular perception
were accurate, then the relevance of the LA would be exceptional for
the teaching of EFL. In preparing the ground for a teaching experi-
ment around the LA with Spanish as L2 (which we will review below),
Stengers (2007) therefore made a two-tier quantitative comparison of
the presence of chunks in Spanish and English. She first looked up the
frequencies of occurrence in comparable corpora (The Collins Cobuild
English and Spanish Wordbanks) of 500 English and 500 Spanish idi-
oms which she had randomly picked from same-sized idiom dictionar-
ies of both languages. The mean frequency of occurrence of the English
and Spanish idioms in the respective corpora turned out identical (24.5
occurrences per 56 million words). In the second investigation, she con-
sidered the presence of chunks at large (not just ‘idioms’) by having ten
native speakers of English and ten native speakers of Spanish listen to
a one-hour compilation of radio interviews in their language and note
down every formulaic sequence they heard. The English and Spanish
audio compilations were matched in terms of length, subject matter,
number of speakers and gender of speakers. On average, the respond-
ents recognized 5.8 formulaic sequences per minute of discourse in the
English sample, as compared to a mean of 5.4 in the Spanish sample.
This is but a slight difference and statistically not at all significant. In
brief, Spanish appears to be as idiomatic as English, and it follows that
students of Spanish face the same challenge as learners of English in
learning to string words together idiomatically.
The next question is whether learners’ use of chunks in Spanish as L2
also correlates with their level of proficiency as assessed by blind judges,
as we found it did in L2 English (see Chapter 2, Section 2.2). Stengers
(2009) set up an experiment where one group of upper-intermediate
students of Spanish was given training in pedagogical chunking in a
26-hour course (spread over eight months), while for a parallel group
of students that course was geared more towards single-word learning
and grammar revision. Apart from this course, in which the two groups
were given differential treatment, the students took several other
Spanish language courses (including translation) in the same period.
The content and methods used in these other courses were the same
for all. At the end of the treatment the students were given a re-tell
task: they were given a text written in their mother tongue and were
asked to reproduce its contents in Spanish (a number of single key-
words were laid out on a sheet to aid memory). The students’ recorded
158 Optimizing a Lexical Approach

performance was subsequently listened to by six experienced Spanish

teachers. Three of these were asked to give each performance a score for
oral proficiency (focusing on the parameters of fluency, accuracy and
range of expression), while the other three were asked to make a note of
all the chunks they heard in each performance. The students’ ranking
in terms of proficiency scores and their ranking for number of chunks
produced was then subjected to correlation analysis. This yielded coef-
ficients of .36 for the parameters fluency and accuracy correlated with
the chunk-counts and .39 for the parameter range of expression corre-
lated with the chunk-counts. While these correlations are statistically
significant, they are considerably weaker than the highly significant
ones obtained in a parallel experiment with the participation of same-
level students of English, where coefficients were as high as .65 (see
Chapter 3, Section 3.2).
According to Stengers (2009), the weaker relationship between
students’ chunk use and their perceived oral proficiency in Spanish
may be due to the fact that Spanish is much more inflectional than
English. Accurate use of verb ⫹ noun collocations, for instance, requires
correct choices among many more verb inflections in Spanish than it
does in English.34 In other words, the risk of ‘erring’ at the level of con-
jugation is greater in Spanish; for example, the volunteers who noted
down the chunks used by the students of Spanish in the above re-tell
experiment commented that the students on occasion got the word
combination basically right but erred at the level of morphology. It is
therefore possible that considerations at the level of morphology over-
ride the assessors’ view that students handle phraseology more easily in
Spanish than in English as L2. We therefore speculate that it may gener-
ally take longer for LA-inspired ISLA to make a profound impact on per-
ceived (oral) proficiency when the L 2 is richly inflectional. Depending
on the models of chunk storage and automatization one adheres to
(see Chapter 5, Section 5.1 and Chapter 8, Section 8.2), the learner will
either have to holistically store more morphological variants of a given
collocation or be able to derive those variants from a single ‘canonical’
chunk via automatized mastery of ‘grammar’. For exemplar-based (and
thus chunk-based) learning of grammar to be successful, the learner
will need to encounter (and notice) a great number of exemplars of dif-
ferent kinds. 35 When it comes to grammar patterns beyond the highly
frequent ones, the usual contexts of ISLA (that is, learning that is usually
non-intensive and predominantly classroom-based) may not provide
sufficient exposure for exemplar-based learning to progress at a satis-
factory pace. In other words, a certain amount of ‘top-down’ grammar
 Directions 159

instruction might usefully complement ‘from-chunks-up’, exemplar-

based learning. Whether both approaches should meet halfway, or one
third or some other fraction of the way, will probably depend on the
typology of the L2 and the degree of L1-L2 similarity.
However, there is a second reason why a certain form-focus in ISLA is
required even in connection with chunk learning in English, and that
is that learners attend more to content words than to small function
words (such as articles and prepositions) in the chunks they pick up
(Ellis, 2008b). For one thing, in spoken discourse it is the small func-
tional elements of formulaic sequences, especially sequences of high
frequency, that are prone to phonological reduction, which must make
them hard for learners to discern in the first place. While Nick Ellis
(2008b) shows this is clearly problematic for learners in contexts of nat-
uralistic L2 acquisition, as it explains persisting ‘small’ inaccuracies in
their output, in ISLA at least there is an opportunity for teacher-guided
remediation in cases where students produce, for example, with other
words (instead of in other words), play in someone’s hands (instead of play
into someone’s hands), on the driving/driver’s seat (instead of in the driving/
driver’s seat), and so on. We need to acknowledge that also in our own
proposals for insightful learning of chunks, the focus has mostly been
on content words (especially, where we have recommended resuscita-
tion of the literal meanings of idioms, exploitation of phonologically
motivated lexical selection, and drawing attention to semantic prosody
in sets of collocations). However, to meet the needs of ambitious learn-
ers, teachers and materials writers should not forget work on the ‘little’
words as well. That is why, if phrasal lexis is recorded in the form of
verb–noun or adjective–noun collocation boxes or grids (as is recom-
mended in Lewis, 1993, 1997, 2000), these should not just mention the
verbs, nouns and adjectives, but possibly also articles and prepositions.
It is also here that repetition drilling for fluency and accuracy may yield
particular benefits.
Before closing this section, let us briefly return to Stengers’ (2009)
experiment set up to assess the merits of a lexical approach to Spanish
as L2. As mentioned above, this was a controlled experiment with one
group of students receiving training in pedagogical chunking which
was withheld from a parallel group. In accordance with the received
version of the LA, awareness-raising and the strategy training were
expected to help the former group of students add more lexical phrases
to their repertoires than would be added by the other group, phrases
both from the texts used in the treatment course and from the samples
of Spanish they encountered in the other courses they were taking, as
160 Optimizing a Lexical Approach

well as outside the college. This expectation was not borne out by the
data. The number of chunks used by both groups of students in their
re-tell task at the end of the experiment was the same. There was no
evidence of differential uptake of chunks either from the course materi-
als used in the treatment course or from elsewhere. This confirms the
findings we have reported (in Chapter 3, Section 3.2) with regard to
similar experiments set up with the participation of English majors:
pedagogical chunking activities alone do not seem to make much of a
difference to chunk uptake. It could be argued, of course, that the re-tell
task was too challenging to elicit chunks the students may have picked
up but not yet mastered to the point that they could retrieve them for
an oral production task. That is why Stengers (2009) also administered
a receptive chunk-knowledge test to both of these groups of students of
Spanish. This Discriminating Collocations Test, which we will describe in
the next section, did not reveal any differential learning either. We feel
that Stengers’ combined corroborative findings that L2 proficiency is
associated with chunk mastery, but that this mastery is not obtained via
chunk-noticing activities alone, adds to our plea, made throughout our
book, for optimizing the received version of the LA.

8.2 Testing chunk knowledge

Knowledge of the syntagmatic behaviour of a word is an integral, cruc-

ial part of word knowledge, or, to repeat J. R. Firth’s famous assertion,
‘You shall know a word by the company it keeps’ (1957: 11). It follows
that for L2 vocabulary testing to give an accurate estimate of students’
vocabulary knowledge, tests should include a collocational dimension.
Furthermore, given the strong correlations we have found between
students’ chunk knowledge and their level of proficiency, we may con-
jecture that tests which accurately measure chunk knowledge might
even serve as a shortcut for gauging students’ proficiency in general, for
good knowledge of collocations appears to be indicative of the amount
of exposure to L2 a student has had and the amount of intake that has
occurred.
However, apart from one contribution to Lewis (2000), the received
version of the LA has unfortunately made few recommendations for test
formats. That one contribution, by Peter Hargreaves, mentions multiple-
choice items, such as the examples in Box 8.2 (the first example con-
cerns colligation [‘word grammar’] rather than collocation).
Some of the exercises proposed by Lewis (1997, 2000) might also
serve as test formats. For example, the learner could be presented with a
 Directions 161

Box 8.2 Examples of colligation and collocation test items from Hargreaves
(2000)

Only one of the following words can occur in both of the blanks in the two
sentences below. Please circle the appropriate letter:
A. believes B. tends C. boasts D. claims
She _____ that she is more accurate than her sister in her work.
She _____ to be more accurate than her sister in her work.

Which of the options below fits the blank in the following sentence best?
She obviously didn’t want to discuss the matter so I didn’t _____ the point.
A. maintain B. follow C. pursue D. chase

word’s (corpus-verified) strong collocates (revise for, re-sit, pass, fail, take)
and be asked to infer what word these items are collocates of (see also
Chaper 1, Box 1.1). As far as we know, however, none of these poten-
tial test formats have yet been validated in empirical experiments. (By
‘validity’ we mean a test’s ability to measure what it is meant to meas-
ure.) For example, a test consisting of only two or three test items may
not have sufficient coverage to give an accurate estimate of a student’s
entire range of relevant abilities (since the fewer the items, the greater
the risk of skewing due to the role of coincidence). Also, a test of recep-
tive knowledge may not always give an accurate estimate of productive
skills, and so, unless strong correlations with performances on product-
ive language tests are found, no claims beyond the receptive skills actu-
ally measured should be made. To advance one ‘bridge’ further (and
possibly too far), a collocations test could be intended to serve as an
indirect gauge of proficiency in general. If so, however, its results should
definitely show very strong correlations with those obtained for those
same students on tests that measure proficiency directly.
In fact, a test format that is both quick to administer and easy to
score and yet nevertheless allows one to extrapolate overall proficiency
has always been something of a holy grail in language testing. This is
because proficiency is a complex, multifaceted concept and tests which
try to directly measure a student’s proficiency (for example, interview
tests and essay-writing tests) are not only very time-consuming (to
administer and to mark) but also pose problems in terms of inter-rater
reliability (that is, the score awarded by one assessor may differ consid-
erably from that awarded by another) and even intra-rater reliability (for
instance, the same performance may not be awarded the same score by
an individual assessor at different points in time).
162 Optimizing a Lexical Approach

In this, the last section of the book, we will present two chunk-oriented
test formats which we have grown accustomed to using ourselves
and which demonstrate the points made just above. The first format
is intended to measure collocational knowledge per se (that is, chunk
knowledge as a distinct component of proficiency). The second format
integrates the testing of chunk knowledge as part of a test intended to
estimate general proficiency. Let it be clear that both example formats
are intended merely to illustrate some of the avenues for chunk test-
ing that are presently being explored. We do not claim to be experts
in language testing and instead refer the reader to a recent book edited
by Barfield and Gyllstad (in press), which contains a collection of state-
of-the-art proposals for measuring learners’ phraseological competence.
For a recent collection of proposals for assessing vocabulary know-
ledge (mostly at the level of single words) we refer to Daller, Milton and
Treffers-Daller (2007). Some ‘classics’ about language testing generally
are Bachman (1990), Bachman and Palmer (1996) and Weir (2005).
The first format we would like to present was developed by Eyckmans
(in press) and is called the Discriminating Collocations Test (Disco, for
short). It is intended to measure students’ receptive knowledge of verb–
noun collocations, or, more accurately, students’ recognition of such col-
locations. This aim has the advantage that a large number of items can
be tested in a short time-span because, all else being equal, a recogni-
tion task is faster than a production task. As a result, with Disco it is
feasible to cover a comparatively wide array of collocations. Box 8.3 lists
examples of items used in the Disco test.
The test items are developed as follows. Strong verb–noun colloca-
tions are extracted from a corpus (for English, typically the BNC and/
or Collins’ Wordbanks), using verbs as query items. To ensure only strong
collocates are selected, the statistical probability of co-occurrence has
to be above a certain threshold (in the versions of the Disco used so far,
only collocates with a z-score or t-score higher than 3.0 were included).
The resulting set of collocations is divided into three frequency bands.
This is done with a view to enhancing the tests’ discriminative power:
students who have had a lot of L2 exposure are more likely to have met
(and noticed) the comparatively low-frequency collocations in addition
to the more common ones. Each test item is made up of two strong
verb–noun collocations of the same frequency band and one distracter.
Students therefore have to recognize as true collocates two stimuli per
item. This reduces the impact of guessing in comparison with test items
consisting of just two stimuli. To verify that the distracters are indeed
highly unlikely collocations, the corpus is consulted again for their
 Directions 163

Box 8.3 Example items from the Discriminating Collocations Test, from
Eyckmans (in press)

Each of the following test items is made up of two idiomatic and one non-
idiomatic verb–noun combinations in English. Tick both idiomatic verb–
noun combinations.

1. Mean harm Draw a distinction Tie an issue

2. Set a decision Adopt an approach Spread rumours
3. Take a cold Do harm Tell the truth
4. Break the law Keep a record Supply a reason
5. Form a hazard Bite the dust Settle a dispute
6. Find fault Gain worry Take comfort
7. Take turns Raise doubts Undertake life
8. Place discipline Adopt a policy Take an exam

non-occurrence. In addition, native-speaker colleagues are consulted to

verify that there can be no doubt about the correct response to each
test item.
Eyckmans (in press) ascertained the validity of the format by admin-
istering a 50-item Disco test to groups of students of English at different
levels of proficiency (at different stages in their language training, and
as gauged by means of other test formats such as cloze tests). Thanks to
the inclusion of test items corresponding to different frequency bands,
the Disco test yielded significantly different scores for the different
student groups, as predicted on the basis of their overall proficiency
levels. While this validates the Disco test format at least as a receptive
measure of collocational competence, Eyckmans has refrained from
making any claims about the potential power of the test indirectly to
gauge students’ productive knowledge of collocations as well. Stengers
(2009) administered a Spanish version of the Disco test (also 50 items)
to groups of L2 Spanish students, who were also given a re-tell task (see
above). The number of chunks used by each student in the re-tell task
was counted by blind judges. Correlation analyses of the rankings of the
students’ scores obtained in the Disco test with the number of chunks
they produced in the oral task showed only weak correlations. This sug-
gests that having good receptive knowledge of chunks (measured in this
case by means of a recognition task) does not necessarily mean that one
will also be able to use many chunks in a (real-time) production task.
The second test format we would like to present integrates the testing of
chunk knowledge in a general proficiency test, referred to as the Deleted
Essentials Test (DET) by Eyckmans, Boers and Demecheleer (2004).
164 Optimizing a Lexical Approach

In this test, based on a format proposed by Weir (1990) and originally

intended as a test of reading comprehension, an authentic text is laid out
in numbered lines. Per numbered line (but not all lines need be num-
bered), one word is deleted. The word selected for deletion by the test
developer has to be both indispensable and predictable. The place where a
word is missing is not signalled; it is the student’s task to find out and to
infer from the context and the co-text what word it is. Initially and for
several years, we initially used the DET as a general proficiency test and,
more specifically, as a replacement for the so-called cloze test and the
so-called C-test (for a discussion, see Klein-Braley, 1997), but we came
to realize that, given the probabilistic nature of the task in the DET
(the co-text leads one to anticipate the presence of a particular word,
which then turns out to be missing), collocational knowledge must play
a major role in good task-performance. Box 8.4 gives an example of a
DET for advanced learners (the text was compiled and adapted from
several short newspaper articles).
Box 8.5 gives the key to the test in Box 8.4. It is clear from the example,
that the DET does not exclusively test chunk knowledge. Some items
test knowledge of ‘grammar’ (numbered lines 6, 8 and 20, for example),
others test knowledge of single-word vocabulary (lines 2, 7 and 19, for
instance), and still others test reading comprehension and familiarity
with ‘rhetorical structure’ (lines 9, 16 and 25). Still, almost half of the
items require knowledge of the chunk at hand (lines 1, 4, 7, 12, 13, 17,
18, 21, 22, 24 and 25), and this ratio seems to follow from the require-
ment that the deleted words have to be indispensable and predictable.
One advantage of the DET in comparison with the Disco format illus-
trated above is that it goes beyond mere recognition of collocations in
that it requires (semi-) productive knowledge. Another potential advan-
tage in a broader testing context is that Disco could permit the slim-
ming down of general-proficiency test-batteries consisting of several
complementary tests (if the DET were able to capture several skills in
one go). The obvious disadvantage of DET in comparison with the Disco
format is that it takes longer to administer and thus cannot have the
same wide coverage of collocations. In addition, because the starting
point is one or more existing authentic texts, sampling of the test items
is not as controlled as in the Disco test.
The validity of the DET as a general-proficiency test was verified
through the following correlation analyses. Eyckmans, Boers and
Demecheleer (2004) took the English-proficiency exam scores obtained
by four successive cohorts of first-year English majors at a university in
Brussels. Student numbers ranged from 59 to 69. The exam consisted
 Directions 165

Box 8.4 Example of a Deleted Essentials Test

Read the following text. In every numbered line one word is missing. Indicate
by means of a slash (/) where a word is missing and then write down the missing
word in the margin.

ROAD RAGE
The British may have a reputation as a mild-mannered race – always
willing to form a queue and quick to apologise when someone treads
on our toes, but it sometimes seems as a collective red mist of madness 1. __
descends whenever we climb our cars. A recent survey has found that 2. __
no than 80% of motorists have been victims of road rage in the past year. 3. __
At most minor, road rage constitutes a rude hand gesture, a honk of 4. __
the horn or a flash of the. At the other end of the spectrum it involves 5. __
full-blown confrontation, often with tragic consequences. Newspaper
reports over the last two years reveal that dozens of people have stabbed 6. __
– sometimes fatally – following altercations at the side of the road.
One poll found that 30% of men carried some form of ‘security’ item
with them in the car, including knives, repellent spray and crowbars.
Last October a pizza driver in Edinburgh was attacked with a hammer. 7. __
In Watford a month later, a motorist ammonia thrown in his face. 8. __
These, of course, are the more sensational incidents, but there appear 9. __
to be a swelling undertone of aggression on the country’s crowded roads.
Worldwide, only South Africa is to be worse than the UK. According to 10. __
an international of aggressive behaviour in motorists, South African 11. __
drivers are the most prone road rage. Two-thirds reported having been 12. __
on the receiving of roadside aggression in the past 12 months. 13. __
However, the likelihood of drivers resorting to aggression is only 16. __
influenced by the country that they happen to live. For example, Finnish 17. __
scientists have found that daughters are more than sons to inherit their 18. __
fathers’ road rage behaviour: women to have a more aggressive 19. __
driving style if it was their fathers who them to drive. Even the lunar 20. __
cycle could a part: statistics published by the Sussex police have shown 21. __
a correlation between violent incidents and moons. The Oxford English 22. __
Dictionary ‘lunatic’ as ‘affected with the kind of insanity that was 23. __
supposed to have recurring periods, depending changes of the moon’. 24. __
Any link between aggression and full moons has yet be confirmed by 25. __
science, though.
 Box 8.5 Key to the example Deleted Essentials Test

ROAD RAGE
The British may have a reputation as a mild-mannered race – always
willing to form a queue and quick to apologise when someone treads
on our toes, but it sometimes seems as / a collective red mist of madness 1. IF
descends whenever we climb / our cars. A recent survey has found that 2. INTO
no / than 80% of motorists have been victims of road rage in the past year. 3. FEWER
At / most minor, road rage constitutes a rude hand gesture, a honk of 4. ITS
the horn or a flash of the /. At the other end of the spectrum it involves 5. LIGHTS
full-blown confrontation, often with tragic consequences. Newspaper
reports over the last two years reveal that dozens of people have / stabbed 6. BEEN
– sometimes fatally – following altercations at the side of the road.
One poll found that 30% of men carried some form of ‘security’ item
with them in the car, including knives, repellent spray and crowbars.
Last October a pizza / driver in Edinburgh was attacked with a hammer. 7. DELIVERY
In Watford a month later, a motorist / ammonia thrown in his face. 8. HAD
These, of course, are the more sensational incidents, but there / appear 9. DOES

10.1057/9780230245006 - Optimizing A Lexical Approach to Instructed Second Language Acquisition, Frank Boers and Seth Lindstromberg
 to be a swelling undertone of aggression on the country’s crowded roads.
Worldwide, only South Africa is / to be worse than the UK. According to 10. SAID
an international / of aggressive behaviour in motorists, South African 11. SURVEY
drivers are the most prone / road rage. Two-thirds reported having been 12. TO
on the receiving / of roadside aggression in the past 12 months. 13. END
However, the likelihood of drivers resorting to aggression is / only 16. NOT
influenced by the country that they happen to live /. For example, Finnish 17. IN
scientists have found that daughters are more / than sons to inherit their 18. LIKELY
fathers’ road rage behaviour: women / to have a more aggressive 19. TEND
driving style if it was their fathers who / them to drive. Even the lunar 20. TAUGHT
cycle could / a part: statistics published by the Sussex police have shown 21. PLAY
a correlation between violent incidents and / moons. The Oxford English 22. FULL
Dictionary / ‘lunatic’ as ‘affected with the kind of insanity that was 23. DEFINES
supposed to have recurring periods, depending / changes of the moon’. 24. ON
Any link between aggression and full moons has yet / be confirmed by 25. TO
science, though.

10.1057/9780230245006 - Optimizing A Lexical Approach to Instructed Second Language Acquisition, Frank Boers and Seth Lindstromberg
168 Optimizing a Lexical Approach

of several tests, each of which included a DET (a different one each ses-
sion), and an interview (that is, a semi-structured conversation between
the student and the examiner). The interview is one of the methods
employed to get a direct impression of a student’s level of proficiency,
but, as mentioned above, a useful interview is time-consuming and
it is hard to control for inter-rater and intra-rater reliability. Finding a
test format that could reduce the weight awarded to interview scores
(or that could even render interviews dispensable in some student pop-
ulations and under some conditions) would therefore be welcome. The
DET could be a viable candidate for complementary or even substitutive
use if the ranking of students’ scores obtained by this format corre-
lated strongly and systematically with the ranking of scores obtained in
the interviews. That is indeed what Eyckmans, Boers and Demecheleer
(2004) found: systematic, highly significant correlation coefficients of
around .62 for each of the four exam sessions.
As mentioned in the introduction to this section, we merely offer the
two test formats we have illustrated for consideration. We acknowledge
that the domain of chunk testing is still very much uncharted territory.
Exploring it is one of several important courses of action we hope will
be undertaken by educational linguistics researchers of multiword lexis
in coming years.
More generally, we hope we have managed in this final chapter, and
in the book as a whole, to paint a picture of a number of attractive
trailheads, hardly broached as yet, which may lead the venturesome
enquirer to vistas and insights that may show how a lexical approach to
ISLA can continue to be optimized.
Notes

1. Exact word searches on Google in October 2008 turned up 6,130,000 for fully
functional vs. 64,100 for totally functional (96:1), 51,000 for completely functional
(120:1) and 3900 for wholly functional (ca. 157:1). As single words there were 296
million hits on fully (with some relating to a Swiss town of that name), 251 million
hits on completely (1.2:1), 175 million on totally (1.7:1) and 27.1 million on wholly
(11:1). As can be seen, the ratios in the first set are at least one order of magni-
tude higher than those in the second. Unless otherwise noted, in this book we
base claims about relative frequencies on such searches.
2. The ratio of Google hits in October 2008 was 402:1.
3. The ratio of Google hits in October 2008 for large majority:big majority was
157:1.
4. On consultation, we found no entries in the Oxford Collocations Dictionary
for the not-so-uncommon words vacation and zero, for instance.
5. The modern spelling of chunks like anyway and anyhow as single words
underscores the likelihood that each of these word combinations is stored
and mentally processed as a single unit. Much the same can be said for mod-
ern gonna and wanna (instead of going to and want to) and ‘misspellings’ such
as alot (instead of a lot).
6. Following Tucker (2007: 957) and Sag et al. (2002), Wray (2008: 74–6) argues
that generative grammars hardly permit this kind of analysis. (Again, spill
the beans serves as the example).
7. For example, native-speakers of German are notably likely to use the unnat-
ural expression have the possibility to (as in *We had the possibility to see both
the castle and the tunnels.) They may well notice other, natural uses of possi-
bility but, logically, these cannot constitute evidence for them that have the
possibility is not an additional option in natural English.
8. Hulstijn (2001: 260), however, points out this metaphor does not feature in
Connectionism, which sees lexical knowledge as being highly distributed
throughout the mind.
9. Whenever we use the term ‘significant’ in this book, we do mean statistically
significant, that is, at p < .05, and in most of the numerous experiments we
discuss throughout the book in fact at p < .001. We do not expect our read-
ership to be familiar with statistics and have for readability’s sake forgone
repeatedly mentioning such mathematical data, which can be obtained dir-
ectly from us and from the articles we cite.
10. Distinguishing figurative idioms from other idioms is not at all straightfor-
ward. A speaker may only become aware of the figurative nature of a given
idiom on finding out about its original, literal usage.
11. O’Keeffe, McCarthy and Carter (2007: 84) report that 20% of all idioms
they identified in a vast composite corpus (CANCODE plus the Cambridge
International Corpus) occurred only once.
12. What expressions merit the label ‘idiom’ is a matter of some debate. Grant
and Bauer (2004), for example, argue that only a couple of hundred phrases

169
170 Notes

are true idioms under strict application of, in particular, the criteria of
non-compositionality and non-figurativeness. For Grant and Bauer, the
vast majority of phrases that are included in ‘idiom dictionaries’ are what
they call ‘figuratives’. As we are adopting the vantage point of the language
teacher, we will avoid this debate, and simply treat as idioms those phrases
which can be found in well-known idiom dictionaries.
13. Notice the consonance in he kicked the bucket, which may have helped con-
ventionalization of the phrase (see Chapter 6), and the derived assonant
phrase he kicked it. Another variant, he popped his clogs, also assonates.
14. See Gombrich (1960/2004; 1972/1985) and Fodor (1981) for strong state-
ments of the view that ‘a picture can tell a thousand words’. See Nőth (1995)
for a review of the literature on the vagueness of pictures.
15. We find it somewhat remarkable, though, that Lewis – who is elsewhere
adamant that lexis is essentially arbitrary – seems in this case to endorse the
basic Cognitive-Semantic idea that conventionalized figurative expressions
tend to be motivated (by conceptual metaphor).
16. This kind of constructive criticism could be levelled at many of the exam-
ples of collocation boxes that Lewis proposes – for example, neither con-
structive nor level __ at are mentioned in the collocation boxes he proposes
for criticism, while mention is made of much weaker collocates such as help-
ful. In fairness, sources such as the Oxford Collocations Dictionary were not
yet available at his time of writing; nor were his examples intended to be
examples of rigorous corpus-based analysis; still, they suggest that there
may not be strong grounds for optimism about the ability of learners to eas-
ily identify the best chunks to record.
17. All these patterns can be described in terms of either consonance or asson-
ance or in terms of both. Thus, alliteration is one type of consonance.
Rhyme and slant rhyme are types of assonance plus consonance. Another
type of consonance plus assonance is rather uncommon in English – so-
called ‘lead’ repetition as in best bet (where there is repetition of the CV
onset).
18. There were 2,681,000 Google exact-word hits for steer /steers ... clear of in
December 2008. All of the alternatives produce extremely low numbers of
Google hits. The examples which follow in this paragraph have also been
checked by Googling.
19. One complication in gathering this data has to do with deciding what con-
stitutes a monomorphemic word in the mind of a typical speaker of English.
For instance, is the final /d/ in did a separate morpheme semantically akin
to -ed? Also, someone who knows a lot of Latin might conceivably process
words such as interest, mention and expense /ekspens/ (all of which occur in
our small corpus) not solely as wholes but also (sometimes) analytically, per-
haps as follows, inter/est, ment/ion, and ex/pense. For such a speaker, there
would be even fewer instances of intra-morphemic consonant repetition
than our figures show. For the vast majority of English speakers, however,
many erstwhile polymorphemic Latin-derived words such as mention, as well
as certain strings of formerly separate words (for instance, Ma’am; cf. French
ma dame) are doubtless now understood as being completely fused. Such
cases probably constitute an important class of exception from the tendency
towards no intra-morphemic consonant repetition. Van de Weijer (2005)
 Notes 171

notes that other classes of exception include words associated with child
language (such as mama, dada; see van de Weijer 2003, 2005, for pointers to
relevant studies on this kind of consonant harmony), onomatopoeic words
(pop, plop) and colloquial pejoratives (dud, tat). One might add personal
names and nicknames (Lillian, Bobby) and (semi-)taboo sex-related terms
(tit, dildo).
20. The Oxford Collocations Dictionary for Students of English (2002) was not avail-
able yet at the time, but the LTP Dictionary of Selected Collocations (Lewis and
Hill, eds., 1997) was.
21. Commenting on a matching/completion exercise on the collocational
behaviour of speak, tell, talk and say that he proposes, Lewis (1997: 98) says:
‘It may be argued that this type of exercise creates confusion, but the con-
fusion is inevitable, and it is better addressed in the classroom rather than
leaving learners to develop understanding unaided.’ We are not sure how to
interpret this in the light of (a) Lewis’s statements in favour of independent
learning, and (b) his cautionary remarks about the risk of confusion associ-
ated with the simultaneous presentation of similar items.
22. In all fairness, a number of matching and completion exercises proposed
in the received version of the LA do attempt to engage the students in elab-
oration: students are asked to sort the phrases according to whether they
are positively or negatively connoted; they are asked for their personal
reaction to the content of example sentences; and they are asked to make
comparisons with their mother tongue, for example. It is hard to assess the
pedagogical effectiveness of any of the exercises proposed in, for example,
Lewis (1997), for as far as we know they have not yet been put to the test
in any controlled experiments. This lack of empirical validation, of course,
is – unfortunately – not at all unusual, but holds for the vast majority of
exercises presented in teachers’ resource books, course books and books for
independent study (including highly acclaimed, best-selling ones).
23. A further theoretical option is that the initial procedure undergoes one or
more episodes of restructuring (Cheng, 1985). Plainly, two types of restruc-
turing are partial chunking and complete chunking.
24. Segalowitz (2007) discusses two aspects of cognitive fluency (he is careful
not to imply that these two are all that there are): ‘access fluidity’, which
depends on automatic processing and has to do with how links are made
between forms and meanings (for example, how fast and how unstoppably);
and ‘attention control’, which has to do with an ability to construct the
meanings of complex messages as they unfold in real time.
25. The difference between simulations and role plays is that in the former the
students imagine being in a given situation themselves while in the latter
they take on the identity of another (fictitious) character. Role plays may
have the advantage that they cater for the more inhibited individuals in a
class, who may feel less shy if they can put on an act. On the other hand,
teachers who have tried them will bear witness that role plays can drift into
farce if students (particularly teens) do not take them seriously enough.
26. It seems customary now to speak of the ALM in the past tense although,
for all we know, it is still in use somewhere, perhaps under one of its aliases
such as the Structural Approach or the Michigan Oral Method. For a short but
informative history of the ALM, see Howatt (2004: 305–8). See Castagnaro
172 Notes

(2006) for a fierce and persuasive refutation of the routine claim (though
not made by Howatt) that the ALM is a product of Behaviourism.
27. Of course, in the heyday of the ALM language teaching methodologists gen-
erally spoke of ‘habit formation’ rather than ‘automatization’, but Howatt
(2004) offers some evidence that, for some audiolingualists at least, ‘habit
formation’ did mean ‘automatization’.
28. While the ALM seems rarely to have been rigorously followed outside the
United States and areas influenced by American fashions in ISLA (e.g.,
Japan), contemporary methods followed elsewhere included a good deal of
ALM-style controlled oral practice; see, for example Prabhu’s (1987) discus-
sion of what he calls the Situational-Oral-Structural Approach.
29. In this connection, Gatbonton and Segalowitz (2005) also make the very
sensible point that ample provision of opportunities to repeat the target
language means that students who have produced an incorrect version may
well have an opportunity to produce an accurate version later on.
30. Two additional examples are a fire broke out and arson is suspected. It is debat-
able whether all examples of ‘essential speech segments’ given by Gatbonton
and Segalowitz (2005) would be recognized as chunks by, for example, LA
practitioners. However, they (p. 243) make the point that what is or is not a
chunk depends on context and that pedagogically relevant chunks may be
overlooked if they are sought in large, mixed corpora rather than in ones
clearly marked for pragmatic context (Coulmas, 1981) and communicative
situation (Kecskes, 2002; Read and Nation, 2004).
31. This reviewing can be done in a somewhat student-centred way by agree-
ing with them on a set of chunks to review. Individually, students produce
a prompt for each chunk. Then, in pairs or threes, they ‘test’ each other by
taking turns displaying or calling out their prompts.
32. For example, with reference to etymological information in The Concise
Oxford Dictionary (1990, 8th edn) we have calculated that over 90% of Averil
Coxhead’s (2000) Academic Word List (a collection of 570 highly useful –
and thus frequent – word families for university students regardless of spe-
cialization) is derived from Latin or Greek. We have sampled the strongest
collocates of 57 of these relatively frequent Latinate words (for example,
acquire, brief, concentrate, detect, evaluate, function, investigate, obtain and pre-
dict), but have found no evidence that these relatively frequent words from
Latin or Greek origin collocate significantly more with other Latinate words
than with Germanic ones. If such a trend exists, we suspect it will pertain
mostly to Latinate words of lower frequency bands.
33. Exact-word Google searches of ‘English is a very idiomatic language’, ‘English
is an idiomatic language’, ‘English is very idiomatic’ and ‘English is idiom-
atic’ produced almost 600 hits (in December 2008), typically from ESL web-
sites and language learner forums. The same search for Spanish produced
only seven. It is possible, of course, that English is simply a more frequent
topic on the World Wide Web than Spanish is. All the same, stating that a
given language is (very) idiomatic carries the implicature that some lan-
guages might not be idiomatic to the same extent. We also know from tes-
timonies of fellow teachers and from classroom observations of teachers
and teacher trainees that EFL teachers are wont to warn students that they
would be mistaken to think of English as an easy language to learn because
 Notes 173

the comparative simplicity of English grammar is counterbalanced by its

high degree of idiomaticity.
34. To illustrate that Spanish is more inflectional than English: Spanish verbs
are marked for 1st, 2nd, and 3rd person, singular and plural, and this in all
tenses; further, Spanish (still) has subjunctive tenses, again signalled in verb
morphology; on top of that, articles and adjectives agree with nouns for
gender and number.
35. Even with respect to English, one may well doubt whether chunk teach-
ing of the kind we have discussed in this book can provide sufficient
input for exemplar-based grammar acquisition in ISLA contexts. Very few
of the chunks we have dealt with in this book are more than three words
long – the exceptions being a number of idioms and proverbs, at least when
they are not used elliptically. This holds for the received version of the LA
as well – although Michael Lewis (personal communication) often advises
students to try to discern chunks that are much longer. While collocations
may provide exemplars for grammar at the level of morphology (plural for-
mation and subject–verb agreement, for example), it is hard to see how,
by focusing on collocations alone – let alone idioms, some of which flout
grammatical regularities – knowledge of ‘higher level’ syntactical patterns
of usage can de induced. Consider, for example, the choice of a tense to
fit the speaker’s message and the need to achieve the standard agreement
between, say, a conditional verb form with had ⫹ past participle in the sub-
ordinate clause and the complex but usually different verb form in the main
clause (would ⫹ have ⫹ past participle). For exemplar-based learning of syn-
tax to be possible, students would need to take in as exemplars much larger
stretches of discourse than relatively short chunks alone. For them to be
able to do this often enough to make a difference, memorization of repre-
sentative (and well-understood) samples of text may not at all be a waste
of time (Wray and Fitzpatrick, 2008). Again, L1 acquisition has an unmis-
takable advantage over ISLA because of the much greater amount of input.
But even in L1 it seems that there is at least a perceived need to fuel chil-
dren’s acquisition of (the more challenging) grammar patterns by inviting
children to sing and to memorize exemplary song lyrics (see, for example,
Cook, 2000). As an example, here is the chorus line of a well-known French
children’s song, which exemplifies the distinction between conditional and
future tenses:
Promenons-nous dans les bois
Tant que le loup n’y est pas
Si le loup y était, il nous mangerait
Puisqu’il y est pas, il nous mangera pas.
References

Dictionaries and corpora

British National corpus, http://www.natcorp.ox.ac.uk.
Cambridge International corpus, http://www.cambridge.org/elt/corpus/interna
tional_corpus.htm.
CANCODE is part of the Cambridge International Corpus.
Collins Cobuild Advanced Learner’s English Dictionary (2006, 5th edition) (Glasgow:
HarperCollins Publishers).
Collins Cobuild Dictionary of Idioms (2002, 2nd edition) (Glasgow: HarperCollins
Publishers).
Collins Cobuild Wordbanks, http://www.collins.co.uk/corpus/CorpusSearch.
aspx.
LTP Dictionary of Selected Collocations (1997) (Hove, UK: LTP / Boston: Thomson
Heinle).
Macmillan English Dictionary for Advanced Learners (2007, 2nd edition) (Oxford:
Macmillan Publishers Ltd.).
Oxford Advanced Learner’s Dictionary of English (2000, 6th edition) (Oxford:
Oxford University Press).
Oxford Collocations Dictionary for Students of English (2002) (Oxford: Oxford
University Press).
Oxford Dictionary of English (2005, 2nd revised edition) (Oxford: Oxford
University Press).
Oxford Idioms Dictionary for Learners of English (2001) (Oxford: Oxford University
Press).
Oxford Phrasal Verbs Dictionary for Learners of English (2006) (Oxford: Oxford
University Press).
Oxford Dictionary of Idioms (1999), ed. Jennifer Speake (Oxford: Oxford University
Press).
The Concise Oxford Dictionary (1990, 8th edition) (Oxford: Oxford University
Press).

Articles and books

Aitchison, J. (2003) Words in the Mind (3rd edn) (Oxford and New York: Basil
Blackwell).
Anderson, J. R. (1982) ‘Acquisition of cognitive skill’, Psychological Review, 89:
369–406.
Anderson, J. R. (1984) ‘Spreading activation’ in J. R. Anderson and S. M. Kosslyn
(eds.) Tutorials in Learning and Memory, pp. 61–90 (San Francisco:
W. H. Freeman).
Anderson, J. R. (1993) Rules of the Mind (Hillsdale, NJ: Lawrence Erlbaum
Associates).

174
 References 175

Arnaud, P. J. L. and S. J. Savignon (1997) ‘Rare words, complex lexical units

and the advanced learner’ in J. Coady and T. Huckin (eds.) Second Language
Vocabulary Acquisition, pp. 157–73 (Cambridge: Cambridge University Press).
Asher, J. J. (1966) ‘The learning strategy of the total physical response: a review’,
Modern Language Journal, 50: 79–84.
Asher, J. J. (1988) Learning Another Language Through Actions: A Teacher’s Guidebook
(3rd edn) (Los Gatos, CA: Sky Oaks).
Atkinson, R. C. (1975) ‘Mnemontechnics in second-language learning’, American
Psychologist, 30: 821–8.
Avila, E. and M. Sadoski (1996) ‘Exploring new applications of the keyword
method to acquiring English vocabulary’, Language Learning, 46: 379–95.
Bachman, L. F. (1990) Fundamental Considerations in Language Testing (Oxford:
Oxford University Press).
Bachman, L. F. and A. S. Palmer (1996) Language Testing in Practice (Oxford:
Oxford University Press).
Baddeley, A. (1997) Human Memory: Theory and Practice (Hove: Psychology Press
Ltd.) [Revised from Baddeley, A. (1990) Human Memory: Theory and Practice
(Needham Heights, MA: Allyn and Bacon)].
Baker, J. (1998) ‘Metaphor’, English Teaching Professional, 7: 13–14.
Barcelona, A. (2000) Metaphor and Metonymy at the Crossroads (Berlin and New
York: Mouton de Gruyter).
Barcroft, J. (2002) ‘Semantic and structural elaboration in L2 lexical acquisition’,
Language Learning, 52: 323–63.
Barfield A. and H. Gyllstad (eds.) (in press) Researching Second Language Collocation
Knowledge (Basingstoke: Palgrave Macmillan).
Barlow, M. and S. Kemmer (eds.) (2000) Usage Based Models of Language (Palo
Alto: CSLI Publications).
Bartlett, J. C. and P. Snelus (1980) ‘Lifespan memory for popular songs’, American
Journal of Psychology, 93: 551–60.
Baugh, A. C. and T. Cable (1993) A History of the English Language (4th edn)
(London: Routledge).
Becker, J. D. (1975). ‘The phrasal lexicon’ in Proceedings of the 1975 Workshop
on Theoretical Issues in Natural Language Processing, Cambridge, MA, pp. 70–7
(Morristown, NJ: Association for Computational Linguistics).
Bensoussan, M. and B. Laufer (1984) ‘Lexical guessing in context in EFL reading
comprehension’, Journal of Research in Learning, 7: 15–32.
Beréndi, M., S. Csábi and Z. Kövecses (2008) ‘Using conceptual metaphors and
metonymies in vocabulary teaching’ in F. Boers and S. Lindstromberg (eds.)
Cognitive Linguistic Approaches to Teaching Vocabulary and Phraseology, pp. 65–99
(Berlin and New York: Mouton de Gruyter).
Bishop, H. (2004) ‘Noticing formulaic sequences: a problem of measuring the
subjective’, LSO Working Papers in Linguistics, 4: 15–19. (Madison: University of
Wisconsin-Madison).
Bobrow, S. A. and S. M. Bell (1973) ‘On catching on to idiomatic expressions’,
Memory and Cognition, 1: 343–6.
Boers, F. (2000a) ‘Enhancing metaphoric awareness in specialised reading’,
English for Specific Purposes, 19: 137–47.
Boers, F. (2000b) ‘Metaphor awareness and vocabulary retention’, Applied
Linguistics, 21: 553–71.
176 References

Boers, F. (2001) ‘Remembering figurative idioms by hypothesising about their

origin’, Prospect, 16: 35–43.
Boers, F. (2003) ‘Applied linguistics perspectives on cross-cultural variation in
conceptual metaphor’, Metaphor and Symbol, 18: 231–8.
Boers, F. and M. Demecheleer (1998) ‘A cognitive semantic approach to teaching
prepositions’, English Language Teaching Journal, 53: 197–204.
Boers, F. and S. Lindstromberg (2005) ‘Finding ways to make phrase-learning
feasible: The mnemonic effect of alliteration’, System, 33: 225–38.
Boers, F. and S. Lindstromberg (2008a) ‘Structural elaboration by the sound (and
feel) of it’ in F. Boers and S. Lindstromberg (eds.) Cognitive Linguistic Approaches
to Teaching Vocabulary and Phraseology, pp. 330–53 (Berlin and New York:
Mouton de Gruyter).
Boers, F. and S. Lindstromberg (2008b) ‘From empirical findings to pedagogical
practice’ in F. Boers and S. Lindstromberg (eds.) Cognitive Linguistic Approaches
to Teaching Vocabulary and Phraseology, pp. 375–93 (Berlin and New York:
Mouton de Gruyter).
Boers, F. and S. Lindstromberg (2008c) ‘How cognitive linguistics can foster
effective vocabulary teaching’ in F. Boers and S. Lindstromberg (eds.) Cognitive
Linguistic Approaches to Teaching Vocabulary and Phraseology, pp. 1–61 (Berlin
and New York: Mouton de Gruyter).
Boers, F. and H. Stengers (2008a) ‘A quantitative comparison of the English and
Spanish repertoires of figurative idioms’ in F. Boers and S. Lindstromberg
(eds.) Cognitive Linguistic Approaches to Teaching Vocabulary and Phraseology,
pp. 355–74 (Berlin and New York: Mouton de Gruyter).
Boers, F. and H. Stengers (2008b) ‘Adding sound to the picture: An exercise
in motivating the lexical composition of metaphorical idioms in English,
Spanish and Dutch’ in M. S. Zanotto, L. Cameron and M. C. Cavalcanti
(eds.) Confronting Metaphor in Use: An Applied Linguistic Approach, pp. 63–78
(Amsterdam and Philadelphia: John Benjamins).
Boers, F., M. Demecheleer and J. Eyckmans (2004a) ‘Etymological elaboration
as a strategy for learning figurative idioms’ in P. Bogaards and B. Laufer (eds.)
Vocabulary in a Second Language: Selection, Acquisition and Testing, pp. 53–78
(Amsterdam and Philadelphia: John Benjamins).
Boers, F., M. Demecheleer and J. Eyckmans (2004b) ‘Cultural variation as a vari-
able in comprehending and remembering figurative idioms’, European Journal
of English Studies, 8: 375–88.
Boers, F., J. Eyckmans and H. Stengers (2007) ‘Presenting figurative idioms
with a touch of etymology: More than mere mnemonics?’ Language Teaching
Research, 11: 43–62.
Boers, F., J. Eyckmans, J. Kappel, H. Stengers and M. Demecheleer (2006)
‘Formulaic sequences and perceived oral proficiency: Putting a lexical
approach to the test’, Language Teaching Research, 10: 245–61.
Boers, F., S. Lindstromberg, J. Littlemore, H. Stengers and J. Eyckmans (2008)
‘Variables in the mnemonic effectiveness of pictorial elucidation’ in F. Boers
and S. Lindstromberg (eds.) Cognitive Linguistic Approaches to Teaching
Vocabulary and Phraseology, pp. 189–216 (Berlin and New York: Mouton
de Gruyter).
Boers, F., A. M. Piquer Píriz, H. Stengers and J. Eyckmans (2009) ‘Does pictorial
elucidation foster recollection of idioms?’ Language Teaching Research, 12.
 References 177

Branford, H. (1994) Dimanche Diller in Danger (London: Collins Children’s

Books).
Brugman, C. (1981) Story of Over. MA thesis (University of California, Berkeley).
Bugelski, B. R. (1982) ‘Learning and imagery’, Journal of Mental Imagery, 6(1):
1–92.
Butler, C. (2005) ‘Forumulaic language: an overview with particular reference to
the cross-linguistic perspective’ in C. Butler, M. Gómez-González, S. Doval-
Suarez (eds.) The Dynamics of Language Use, pp. 221–42 (Amsterdam: John
Benjamins).
Bybee, J. (2002) ‘Phonological evidence for exemplar storage of multiword
sequences‘, Studies in Second Language Acquisition, 24: 215–21.
Bybee, J. and S. Thompson (2000) ‘Three frequency effects in syntax’, Berkeley
Linguistic Society, 23: 65–85.
Byrne, D. (1976) Teaching Oral English (Harlow, UK: Longman).
Cacciari, C. and P. Tabossi (1988) ‘The comprehension of idioms’, Journal of
Memory and Language, 2: 668–83.
Cadierno, T. (2004) ‘Expressing motion events in a second language: A cogni-
tive typological perspective’ in M. Achard and S. Niemeier (eds.) Cognitive
Linguistics, Second Language Acquisition and Foreign Language Teaching (Studies
on Language Acquisition, 18), pp. 13–49 (Berlin and New York: Mouton
de Gruyter).
Castagnaro, P. J. (2006) ‘Audiolingual Method and Behaviorism, from misunder-
standing to myth’, Applied Linguistics, 27: 519–26.
Cermak, L. S. and F. I. M. Craik (1979) Levels of Processing in Human Memory.
(Hillsdale, NJ: Lawrence Erlbaum).
Cheng, P. W. (1985) ‘Restructuring versus automaticity: Alternative accounts of
skill acquisition’, Psychological Review, 92: 414–23.
Chomsky, N. (1980) Rules and Representations (Oxford: Basil Blackwell).
Cieślicka, A. (2006) ‘Literal salience in on-line processing of idiomatic expres-
sions by second language learners’, Second Language Research, 22: 115–44.
Cieślicka, A. (2008) ‘Formulaic language in L 2: Storage, retrieval, and produc-
tion of idioms by second language learners’, Paper presented at the 33rd
International LAUD Symposium, March 10–13, 2008. University of Duisburg-
Essen.
Colson, J. P. (2008) ‘Cross-linguistic phraseological studies: an overview’ in
S. Granger and F. Meunier (eds.) Phraseology: An Interdisciplinary Perspective,
pp. 191–206 (Amsterdam and Philadelphia: John Benjamins).
Condon, N. (2008) ‘How Cognitive Linguistic motivations influence the learn-
ing of phrasal verbs’ in F. Boers and S. Lindstromberg (eds.) Cognitive Linguistic
Approaches to Teaching Vocabulary and Phraseology, pp. 133–58 (Berlin and New
York: Mouton de Gruyter).
Conklin, K. and N. Schmitt (2008) ‘The processing advantage of formulaic
sequences’, Applied Linguistics, 29: 72–89.
Conzett, J. (2000) ‘Integrating collocation into a reading and writing course’ in
M. Lewis (ed.) Teaching Collocation: Further Developments in the Lexical Approach,
pp. 70–87 (Hove: LTP / Boston: Thomson Heinle).
Cook, G. (2000) Language Play, Language Learning (Oxford: Oxford University
Press).
Corson, D. (1985) The Lexical Bar (Oxford: Pergamon Press).
178 References

Coulmas, F. (1981) ‘Introduction: conversational routine’ in F. Coulmas (ed.)

Conversational Routine: Explorations in Standardized Communication Situations
and Prepatterned Speech, pp. 1–18 (The Hague: Mouton).
Coxhead, A. (2000) ‘A new academic word list’, TESOL Quarterly, 34: 213–38.
Croft, W. (2001) Radical Construction Grammar: Syntactic Theory in Typological
Perspective (Oxford: Oxford University Press).
Cruttenden, A. (1981) ‘Item-learning and system-learning’, Journal of
Psycholinguistic Research, 10: 79–88.
Csábi, S. (2004) ‘A cognitive linguistic view of polysemy in English and its impli-
cations for teaching’ in M. Achard and S. Niemeier (eds.) Cognitive Linguistics,
Second Language Acquisition, and Foreign Language Teaching, pp. 233–56 (Berlin
and New York: Mouton de Gruyter).
Cutting, J. C. and Bock, K. (1997) ‘That’s the way the cookie bounces: Syntactic
and semantic components of experimentally elicited idiom blends’, Memory
and Cognition, 25: 57–91.
Daller, H., J. Milton and J. Treffers-Daller (2007) Modelling and Assessing Vocabulary
Knowledge (Cambridge: Cambridge University Press).
Day, E. and S. Shapson. (2001) ‘Integrating formal and functional approaches
to language teaching in French immersion: An experimental study’, Language
Learning, 51: 7–80.
de Bot, K. (1992) ‘A bilingual production model: Levelt’s speaking model
adapted’, Applied Linguistics, 13: 1–24.
de Bot, K. (1996) ‘The pyscholinguistics of the output hypothesis’, Language
Learning, 46: 529–55.
Dechert, H. W. (1983) ‘How a story is done in a second language’ in C. Faerch
and G. Kasper (eds.) Strategies for Interlanguage Communication, pp. 175–95
(London: Longman).
De Cock, S. (2004) ‘Preferred sequences of words in NS and NNS speech. Belgian
Journal of English Language and Literatures, New Series, 2: 225–46.
Deignan, A., D. Gabrys and A. Solska (1997) ‘Teaching English metaphors using
cross-linguistic awareness-raising activities’, ELT Journal, 51: 352–60.
DeKeyser, R. (2001) ‘Automaticity and automatization’ in P. Robinson (ed.)
Cognition and Second Language Instruction, pp. 125–51 (Cambridge: Cambridge
University Press).
DeKeyser, R. (2007) ‘Situating the concept of practice’ in R. DeKeyser (ed.)
Practice in a Second Language: Perspectives from Applied Linguistics and Cognitive
Psychology, pp. 1–18 (Cambridge: Cambridge University Press).
De Ridder, I. (2002) ‘Visible or invisible links: Does the highlighting of hyper-
links affect incidental vocabulary learning, text comprehension and the read-
ing process?’ Language Learning and Technology, 6: 123–46.
De Rycker, T. (2004) ‘Lists, boxes or maps? Recording and retrieving Collocations
in an ELT/ESP context’, Belgian Journal of English Language and Literatures, New
Series, 2: 315–28.
Dewhurst, S. and C. Robinson (2004) ‘False memories in children: Evidence for a
shift from phonological to semantic associations’, Psychological Science, 15: 782–6.
Doughty, C. and J. Williams (eds.) 1998. Focus on Form in Classroom Second
Language Acquisition (Cambridge: Cambridge University Press).
Draaisma, D. (2000) Metaphors of Memory: A History of Ideas about the Mind (Cambridge:
Cambridge University Press). [Originally published in Dutch in 1995.]
 References 179

Ellis, N. (1996a) ‘Sequencing in SLA: Phonological memory, chunking, and

points of order’, Studies in Second Language Acquisition, 18: 91–126.
Ellis, N. (1996b) ‘Analyzing language sequence in the sequence of language
acquisition: some comments on Major and Ioup’, Studies in Second Language
Acquisition, 18: 361–8.
Ellis, N. (2002) ‘Frequency effects in language processing’, Studies in Second
Language Acquisition, 24: 143–88.
Ellis, N. (2008a) ‘Phraseology: The periphery and the heart of language’ in
F. Meunier and S. Granger (eds.) Phraseology in Foreign Language Learning and
Teaching, pp. 1–13 (Amsterdam and Philadelphia: John Benjamins).
Ellis, N. (2008b) ‘The dynamics of second language emergence: Cycles of lan-
guage use, language change, and language acquisition’, The Modern Language
Journal, 92: 232–49.
Ellis, N. and D. Larsen-Freeman (2006) ‘Language emergence: Implications for
applied linguistics’ (introduction to the Special Issue), Applied Linguistics, 27:
558–89.
Ellis, N., R. Simpson-Vlach and C. Maynard (2008) ‘Formulaic language in
native and second language speakers: Psycholinguistics, corpus linguistics,
and TESOL’, TESOL Quarterly, 42: 375–96.
Ellis, R. (1990) ‘Individual styles in classroom second language development’
in J. de Jong and G. Stevenson (eds.) Individualizing the Assessment of Language
Abilities (Clevedon, UK: Multilingual Matters).
Ellis, R. (2003) Task-based Language Learning and Teaching (Oxford: Oxford
University Press).
Ellis, R. and G. Barkhuizen (2005) Analysing Learning Language (Oxford: Oxford
University Press).
Erman, B. and B. Warren (2000) ‘The idiom principle and the open choice prin-
ciple’,
. Text, 20: 87–120.
Erten, I. H. and M. Tekin (2008) ‘Effects on vocabulary acquisition of presenting
new words in semantic sets versus semantically unrelated sets’, System, 36:
407–22.
Eyckmans, J. (2007) ‘Taking SLA research to interpreter-training: Does knowledge
of phrases foster fluency?’ in F. Boers, J. Darquennes, R. Temmerman (eds.)
Multilingualism and Applied Comparative Linguistics: Pedagogical Perspectives,
pp. 89–104 (Newcastle: Cambridge Scholars Publishing).
Eyckmans, J. (in press) ‘Towards an assessment of learners’ receptive and product-
ive syntagmatic knowledge’ in A. Barfield and H. Gyllstad (eds.) Researching
Second Language Collocation Knowledge (Basingstoke: Palgrave Macmillan).
Eyckmans, J., F. Boers and M. Demecheleer (2004) ‘The Deleted-Essentials Test:
an effective and affective compromise’, Humanising Language Teaching 6. www.
hltmag.co.uk (Pilgrims).
Eyckmans, J., F. Boers and H. Stengers (2007) ‘Identifying chunks: Who can see
the wood for the trees?’ Language Forum, 33: 85–100.
Fallon, A., K. Groves and G. Tehan (1999) ‘Phonological similarity and trace
degradation in the serial recall task: When CAT helps RAT, but not MAN’,
International Journal of Psychology, 34: 301–7.
Fillmore, C. (2000 [1979]) ‘On fluency’ in H. Riggenbach (ed.) Perspectives
on Fluency, pp. 43–60 (Ann Arbor, MI: The University of Michigan Press).
Reprinted from C. Fillmore, D. Kempler and W. Wang (eds.) (1979) Individual
180 References

Differences in Language Ability and Language Behavior, pp. 85–101 (New York:
Academic Press).
Fillmore, C., P. Kay and M. O’Connor (1988) ‘Regularity and idiomaticity in
grammatical constructions’, Language, 64(3): 501–38.
Finkenstädt, T. and D. Wolff (1973) Ordered Profusion: Studies in Dictionaries and
the English Lexicon (Heidelberg: Carl Winter).
Firth, J. R. (1930) Speech (London: Ernest Benn).
Firth, J. R. (1957) ‘General linguistics and descriptive grammar’ in Papers in
Linguistics 1934–1965, pp. 216–28 (Oxford: Oxford University Press).
Fitzpatrick, T. and A. Wray (2006) ‘Breaking up is not so hard to do: Individual
differences in L 2 memorization’, Canadian Modern Language Review, 63:
35–57.
Fodor, J. A. (1981) ‘Imagistic representation’ in N. Block (ed.) Imagery, pp. 63–86
(Cambridge, MA: MIT Press).
Forsberg, F. (2006) Le Langage Préfabriqué en Français L2 : Etude Acquisitionelle et
Comparative (Stockholm : Stockholm University).
Freeborn, D. (2006) From Old English to Standard English: A Course Book in
Language Variation across Time (3rd edn) (Basingstoke and New York: Palgrave
Macmillan).
Fries, C., R. Lado and the Staff of the Michigan English Language Institute
(1958) English Pattern Practices: Establishing the Patterns as Habits (Ann Arbor,
MI: University of Michigan Press).
Furumoto, L. (1991) ‘From “Paired Associates” to a psychology of self: The intel-
lectual odyssey of Mary Whiton Calkins’ in G. A. Kimble, M. Wertheimer
and C. White (eds.) Portraits of Pioneers in Psychology, pp. 57–72 (Hillsdale, NJ:
Lawrence Erlbaum Associates).
Gatbonton, E., and N. Segalowitz (2005). ‘Rethinking communicative language
teaching: a focus on access to fluency’, Canadian Modern Language Journal, 61:
325–53.
Gates, R. (2009) ‘A balanced strategy: reprogramming the Pentagon for a new
age’, Foreign Affairs, January/February. (Accessed on the internet via Slate
Magazine in Nov. 2008).
Gibbs, R. W. Jr (1980) ‘Spilling the beans on understanding and memory for
idioms in conversation’, Memory and Cognition, 8: 449–56.
Gibbs, R. W. Jr (1985) ‘On the process of understanding idioms’ Journal of
Psycholinguistic Research, 14: 465–72.
Gibbs, R. W. Jr (1994) The Poetics of Mind: Figurative Thought, Language and
Understanding (Cambridge: Cambridge University Press).
Glucksberg, S. (1991) ‘Beyond literal meanings: the psychology of allusion’,
Psychological Science, 2: 146–52.
Glucksberg, S. (1993) ‘Idiom meanings and allusional context’ in C. Cacciari
and P. Tabossi (eds.) Idioms, Proccessing, Structure, and Interpretation, pp. 3–27
(Hillsdale, NJ: Lawrence Erlbaum Associates).
Glucksberg, S. (2001) Understanding Figurative Language: From Metaphors to Idioms
(Oxford: Oxford University Press).
Godfroid, A., A. Housen and F. Boers (2009) ‘A procedure for testing the Noticing
Hypothesis in the context of vocabulary acquisition’ in M. Pütz and L. Sicola
(eds.) Inside the Learner’s Mind: Cognitive Processing and Second Language
Acquisition (Amsterdam and Philadelphia: John Benjamins).
 References 181

Goldberg, A. E. (2006) Constructions at Work: The Nature of Generalization in

Language (Oxford: Oxford University Press).
Gombrich, E. H. (2004 [1960]) Art and Illusion: A Study in the Psychology of Pictorial
Representation (London: Phaidon).
Gombrich, E. H. (1985 [1972]) Symbolic Images (Edinburgh: Phaidon).
Granger S. (1998) ‘Prefabricated patterns in advanced EFL writing: collocations
and formulae’ in A. P. Cowie (ed.) Phraseology, Theory, Analysis and Applications,
pp. 145–60 (Oxford: Oxford University Press).
Grant, L. and L. Bauer. (2004) ‘Criteria for re-defining idioms: Are we barking up
the wrong tree?’ Applied Linguistics, 25: 38–61.
Gupta, P., J. Lipinski and E. Aktunc (2005) ‘Re-examining the phonological
similarity effect in serial recall: The roles of type similarity, category cueing,
and item recall’, Memory and Cognition, 33: 1001–16.
Han, Z., E. S. Park and C. Combs (2008) ‘Textual enhancement of input: Issues
and Possibilities’, Applied linguistics, 29: 597–618.
Hargreaves, P. (2000) ‘Collocating and testing’ in M. Lewis (ed.) Teaching
Collocation, pp. 205–23 (Hove: LTP).
Harmer, J. (2001) The Practice of English Language Teaching (3rd edn) (Harlow, UK:
Longman).
Hatch, E. V. and C. Brown (1995) Vocabulary, Semantics, and Language Education
(Cambridge: Cambridge University Press).
Hill, J. (2000) ‘Revising priorities: From grammatical failure to collocational suc-
cess’ in M. Lewis (ed.) Teaching Collocation: Further Developments in the Lexical
Approach, pp. 47–68 (Boston, MA: Thomson Heinle).
Hill, J., M. Lewis and M. Lewis (2000) ‘Classroom strategies, activities and exer-
cises’ in M. Lewis (ed.) Teaching Collocation: Further Developments in the Lexical
Approach, pp. 88–116 (Boston, MA: Thomson Heinle).
Hoey, M. (2005) Lexical Priming: A New Theory of Words and Language (London:
Routledge).
Holme, R. (2004) Mind, Metaphor and Language Teaching (Basingstoke and New
York: Palgrave Macmillan).
Hopper, P. and E. Traugott (1993) Grammaticalization (Cambridge: Cambridge
University Press).
Howarth, P. (1998) ‘Phraseology and second language proficiency’, Applied
Linguistics, 19: 24–44.
Howatt, A. P. R. (2004) A History of English Language Teaching (revised edn)
(Oxford: Oxford University Press).
Hu, M. and P. Nation (2000) ‘Unknown vocabulary density and reading compre-
hension’, Reading in a Foreign Language, 13: 403–30.
Huckin, T. and J. Coady (1999) ‘Incidental vocabulary acquisition in a second
language: A review’, Studies in Second Language Acquisition, 21: 181–93.
Hulstijn, J. (1992) ‘Retention of inferred and given word meanings: Experiments
in incidental vocabulary learning’ in P. J. L. Arnaud and H. Béjoint (eds.)
Vocabulary and Applied Linguistics, pp. 112–25 (London: Macmillan).
Hulstijn, J. (1993) ‘When do foreign-language learners look up the meaning of
unfamiliar words? The influence of task and learner variables’, The Modern
Language Journal, 77: 139–47.
Hulstijn, J. (2001) ‘Intentional and incidental second language vocabulary learn-
ing: a reappraisal of elaboration, rehearsal and automaticity’ in P. Robinson
182 References

(ed.) Cognition and Second Language Instruction, pp. 258–86 (Cambridge:

Cambridge University Press).
Hulstijn, J. and B. Laufer (2001) ‘Some empirical evidence for the Involvement
Load Hypothesis in vocabulary acquisition’, Language Learning, 51: 539–58.
Hulstijn J., M. Hollander and T. Greidanus (1996) ‘Incidental vocabulary learn-
ing by advanced foreign language students: The influence of marginal glosses,
dictionary use, and reoccurrence of unknown words’, The Modern Language
Journal, 80: 327–39.
Ilson, R. (1983) ‘Etymological information: Can it help our students?’ English
Teaching Journal, 37: 76–82.
Ioup, G. (1996) ‘Grammatical knowledge and memorized chunks’, Studies in
Second Language Acquisition, 18: 355–60.
Irujo, S. (1993) ‘Steering clear: Avoidance in the production of idioms’,
International Review of Applied Linguistics in Language Teaching, 31: 205–19.
Jackendoff, R. (2003) Foundations of Language: Brain, Meaning, Grammar, Evolution
(Oxford: Oxford University Press).
Johnson, M. (1987) The Body in the Mind: The Bodily Basis of Reason and Imagination
(Chicago: University of Chicago Press).
Jusczyk, P., M. Goodman and A. Baumann (1999) ‘Nine-month-olds’ attention
to sound similarities in syllables’, Journal of Memory and Learning, 40: 62–82.
Kecskes, I. (2002) Situation-bound Utterances in L1 and L2 (Berlin: Mouton de
Gruyter).
Kelly, P. (1990) ‘Guessing: No substitute for the systematic learning of lexis’,
System, 18: 199–208.
Kennedy, G. (2008) ‘Phraseology and language pedagogy: Semantic preference
associated with English verbs in the British National Corpus’ in F. Meunier
and S. Granger (eds.) Phraseology in Foreign Language Learning and Teaching,
pp. 21–41 (Amsterdam and Philadelphia: John Benjamins).
Kies, D. (1990) ‘Three principles underlying iconicity in literature: the poet-
ics of nonsense in children’s and general literature’. A paper prepared for the
Seventeenth International Systemic Congress, July 3–7, 1990, at the University
of Stirling, Scotland. http,//papyr.com/hypertextbooks/grammar/iconicity.
htm. Accessed, 26 Nov. 2008.
Klein-Braley, C. (1997) ‘C-Tests in the context of reduced redundancy testing:
An appraisal’, Language Testing, 14: 47–84.
Knight, S. (1994) ‘Dictionary use while reading: The effects on comprehen-
sion and vocabulary acquisition for students of different verbal abilities’,
The Modern Language Journal, 78: 285–99.
Kövecses, Z. (1990) Emotion Concepts (New York: Springer).
Kövecses, Z. and P. Szabó (1996) ‘Idioms: A view from Cognitive Semantics’,
Applied Linguistics, 17: 326–55.
Krashen, S. D. (1982) Principles and Practice in Second Language Acquisition (Oxford:
Pergamon Press).
Krashen, S. D. (1985) The Input Hypothesis: Issues and Implications (London:
Longman).
Krashen, S. D. (1989) ‘We acquire vocabulary and spelling by reading: additional
evidence for the input hypothesis’, Modern Language Journal, 73: 440–64.
Krashen, S. D. and T. D. Terrell (1983) The Natural Approach: Language Acquisition
in the Classroom (Hayward, CA: Alemany Press).
 References 183

Kuiper, K. (1996) Smooth Talkers: The Linguistic Performance of Auctioneers and

Sportscasters (Englewood Cliffs, NJ: Lawrence Erlbaum).
Lakoff, G. (1987) Women, Fire, and Dangerous Things: What Categories Reveal about
the Mind (Chicago: University of Chicago Press).
Lakoff, G. and M. Johnson (1980) Metaphors We Live By (Chicago: University of
Chicago Press).
Langacker, R. W. (1987) Foundations of Cognitive Grammar, Volume 1: Theoretical
Prerequisites (Stanford: Stanford University Press).
Langacker, R. W. (1990) Foundations of Cognitive Grammar, Volume 2: Descriptive
Applications (Stanford: Stanford University Press).
Langacker, R. W. (2000) ‘A dynamic usage-based model’ in M. Barlow and
S. Kemmer (eds.) Usage-Based Models of Language, pp. 1–63 (Stanford: Center
for the Study of Language and Information Publications).
Laufer, B. (1989) ‘What percentage of text lexis is necessary for comprehension?’
in C. Lauren and M. Nordman (eds.) Special Language: From Humans Thinking
to Thinking Machines, pp. 316–23 (Clevedon: Multilingual Matters).
Laufer, B. (1997) ‘The lexical plight in second language reading’ in J. Coady
and T. Huckin (eds.) Second Language Vocabulary Acquisition: A Rationale for
Pedagogy, pp. 20–34 (Cambridge: Cambridge University Press).
Laufer, B. (2005) ‘Focus on form in second language vocabulary acquisition’ in
Susan Foster-Cohen (ed.) EUROSLA Yearbook 5, pp. 223–50 (Amsterdam and
Philadelphia: John Benjamins).
Laufer, B. and N. Girsai (2008) ‘Form-focussed instruction in second language
vocabulary learning: A case for contrastive analysis and translation’, Applied
Linguistics, 29: 694–716.
Laufer, B. and M. Hill (2000) ‘What lexical information do L2 learners select in
a CALL dictionary and how does it affect word retention?’ Language Learning
and Technology, 3: 58–76.
Laufer, B. and J. Hulstijn (2001) ‘Incidental vocabulary acquisition in a second
language: The construct of task-induced involvement’, Applied Linguistics, 22:
1–26.
Laufer, B. and P. Nation (2001) ‘Passive vocabulary size and the speed of mean-
ing recognition: Are they related?’ in S. H. Foster-Cohen and A. Nizegorodcew
(eds.) EUROSLA Yearbook 1, pp. 7–28 (Amsterdam and Philadelphia: John
Benjamins).
Laufer, B. and D. Sim (1985) ‘An attempt to measure the threshold of compe-
tence for reading comprehension’, Foreign Language Annals, 18(5): 405–11.
Lazar, G. (1996) ‘Using figurative language to expand students’ vocabulary’, ELT
Journal, 50: 43–51.
Lazar, G. (2003) Meanings and Metaphors (Cambridge: Cambridge University Press).
Lea, R. B., D. N. Rapp, A. Elfenbein, A. D. Mitchel and R. S. Romine (2008)
‘Sweet silent thought: Alliteration and resonance in poetry comprehension’,
Psychological Science, 19: 709–16.
Leeman, J. (2007) ‘Feedback in L2 learning: responding to errors during prac-
tice’ in R. M. DeKeysar (ed.) Practice in a Second Language: Perspectives from
Applied Linguistics and Cognitive Psychology, pp. 111–37 (Cambridge: Cambridge
University Press).
Lennon, P. (1990) ‘Investigating fluency in EFL: A quantitative approach’,
Language Learning, 40(3): 387–417.
184 References

Leow, R. P. (2000) ‘A study of the role of awareness in foreign language behav-

iour: Aware versus unaware learners’, Studies in Second Language Acquisition,
22: 557–84.
Levelt, W. (1989) Speaking: From Intention to Articulation (Cambridge, MA: MIT
Press).
Lewis, Michael (1993) The Lexical Approach: The State of ELT and a Way Forward
(Hove, UK: LTP / Boston: Thomson Heinle).
Lewis, Michael (1997) Implementing the Lexical Approach: Putting Theory into
Practice (Hove, UK: LTP / Boston: Thomson Heinle).
Lewis, Michael (ed.) (2000) Teaching Collocation: Further Developments in the
Lexical Approach (Hove, UK: LTP / Boston: Thomson Heinle).
Lewis, Morgan (2000) ‘There is nothing as practical as a good theory’ in
M. Lewis (ed.) Teaching Collocation: Further Developments in the Lexical Approach,
pp. 10–27 (Hove, UK: LTP / Boston: Thomson Heinle).
Libben, M. R. and D. A. Titone (2008) ‘The multidetermined nature of idiom
processing’, Memory and Cognition, 36: 1103–21.
Lindstromberg, S. (1997) English Prepositions Explained (Amsterdam and
Philadelphia: John Benjamins).
Lindstromberg, S. and F. Boers (2005a) ‘From movement to metaphor with
manner- of-movement verbs’, Applied Linguistics, 26: 241–61.
Lindstromberg, S. and F. Boers (2005b) ‘Means of mass-memorisation of mul-
tiword expressions, part one: The power of sounds’, Humanising Language
Teaching, 7; www.hltmag.co.uk (Pilgrims).
Lindstromberg, S. and F. Boers (2008a) ‘Phonemic repetition and the learning of
lexical chunks: the mnemonic power of assonance’, System, 36: 423–36.
Lindstromberg, S., and F. Boers (2008b) ‘The mnemonic effect of noticing
alliteration in lexical chunks’, Applied Linguistics, 29: 200–22.
Lindstromberg, S. and F. Boers (2008c) Teaching Chunks of Language (Rum,
Austria: Helbling Languages).
Littlemore, J. (2001) ‘Metaphor as a source of misunderstanding for overseas
students in academic lectures’, Teaching in Higher Education, 6: 333–51.
Littlemore, J. and G. Low (2006) Figurative Thinking and Foreign Language Learning
(Basingstoke and New York: Palgrave Macmillan).
Logan, G. (1988) ‘Toward an Instance Theory of automatization’, Psychological
Review, 95: 492–527.
Logan, G. (1997) ‘Automaticity and reading: Perspectives from the Instance
Theory of automatization’, Reading and Writing Quarterly, 13: 123–46.
Long, M. H. (1991) ‘Focus on form: A design feature in language teaching meth-
odology’ in K. de Bot, R. B. Ginsberg and C. Kramsch (eds.) Foreign Language
Research in Cross-Cultural Perspective, pp. 39–52 (Amsterdam and Philadelphia:
John Benjamins).
Lord, A. (1956) ‘The role of sound patterns in Serbo-Croatian epic’ in M. Halle,
H. Lunt, H. McLean and C. van Schooneveld (eds) For Roman Jakobson. Essays
on the Occasion of his Sixtieth Birthday, pp. 301–5 (The Hague: Mouton).
Lord, A. (1964) The Singer of Tales (Cambridge, MA: Harvard University Press).
MacArthur, F., and J. Littlemore (2008) ‘A discovery approach to figurative
language learning with the use of corpora’ in F. Boers and S. Lindstromberg
(eds.) Cognitive Linguistic Approaches to Teaching Vocabulary and Phraseology,
pp. 159–88 (Berlin and New York: Mouton de Gruyter).
 References 185

Mackey, A., J. Philp, T. Egi, A. Fujii and T. Tatsumi (2002) ‘Individual differences
in working memory, noticing of interactional feedback and L2 development’
in P. Robinson (ed.) Individual Differences and Instructed Language Learning,
pp. 181–209 (Amsterdam and Philadelphia: John Benjamins).
MacLennan, C. H. G. (1994) ‘Metaphors and prototypes in the teaching and
learning of grammar and vocabulary’, International Review of Applied Linguistics,
32: 97–110.
Major, R. (1996) ‘Chunking and memory: a response to Ellis’, Studies in Second
Language Acquisition, 18: 351–54.
McCarthy, M. (1998) Spoken Language and Applied Linguistics (Cambridge:
Cambridge University Press).
McCrone, J. (1999) ‘States of mind’, New Scientist 2178: 30–33.
McDermid, V. (2007) Beneath the Bleeding (London: HarperCollins).
McQuarrie, E. and D. Mick (1999) ‘Visual rhetoric in advertising: text-
interpretative, experimental and reader-response analyses’, Journal of Consumer
Research, 26: 37–54.
Meara, P. (1980) ‘Vocabulary acquisition: a neglected aspect of language learn-
ing’, Language Teaching and Linguistics: Abstracts, 13: 221–46.
Mondria, J. A. (2003) ‘The effects of inferring, verifying, and memorizing on the
retention of L2 word meanings: An experimental comparison of the “mean-
ing-inferred method” and the “meaning-given method” ’, Studies in Second
Language Acquisition, 25: 473–99.
Moon, R. (1998a) ‘Frequencies and forms of phrasal lexemes in English’ in
A. P. Cowie (ed.) Phraseology: Theory, Analysis and Applications, pp. 79–100
(Oxford: Clarendon Press).
Moon, R. (1998b) Fixed Expressions and Idioms in English (Oxford: Clarendon
Press).
Mothersbaugh, D., B. Huhmann and G. Franke (2002) ‘Combinatory and separa-
tive effects of rhetorical figures on consumers’ effort and focus in ad process-
ing’, Journal of Consumer Research, 28: 589–602.
Myles, F., J. Hooper and R. Mitchell (1998) ‘Rote or rule: Exploring the role of for-
mulaic language in classroom foreign language learning’, Language Learning,
48: 323–64.
Naciscione, A. (2001) Phraseological Units in Discourse: Towards Applied Stylistics
(Riga: Latvian Academy of Culture).
Nagy, W. (1997) ‘On the role of context in first- and second language vocabu-
lary learning’ in N. Schmitt and M. McCarthy (eds.) Vocabulary: Description,
Acquistion and Pedagogy, pp. 64–83 (Cambridge: Cambridge University Press).
Nänny, M. and O. Fisher (eds.) (1999) Form Miming Meaning: Iconicity in Language
and Literature (Amsterdam: John Benjamins).
Nation, P. (2000) ‘Learning vocabulary in lexical sets: Dangers and guidelines’,
TESOL Journal, 9: 6–10.
Nation, P. (2001) Learning Vocabulary in Another Language (Cambridge: Cambridge
University Press).
Nation, P. and P. Meara (2002) ‘Vocabulary’ in N. Schmitt (ed.) An Introduction to
Applied Linguistics, pp. 35–54 (London: Arnold).
Nattinger, J. R. (1988) ‘Some current trends in vocabulary teaching’ in R. Carter
and M. McCarthy (eds.) Vocabulary and Language Teaching, pp. 62–82 (Harlow:
Longman).
186 References

Nattinger, J. R. and J. S. DeCarrico (1992) Lexical Phrases and Language Teaching

(Oxford: Oxford University Press).
Nehls, D. (1991) ‘Do/Make compared with German tun/machen and Dutch doen/
maken: a synchronic-diachronic approach’, International Review of Applied
Linguistics, 29: 303–16.
Nelson, D. L., V. S. Reed and J. R. Walling (1976) ‘Picture superiority effect’,
Journal of Experimental Psychology: Human Learning and Memory, 2: 523–8.
Nesselhauf, N. (2003) ‘The use of collocations by advanced learners of English
and some implications for teaching’, Applied Linguistics, 24: 223–42.
Nosofsky, R. M. and T. J. Palmieri (1997) ‘An exemplar-based random walk model
of speeded classification’, Psychological Review, 104: 266–300.
Nőth, W. (1995) ‘Can pictures lie?’ The Semiotic Review of Books, 6: 10–12.
O’Keeffe, A., M. McCarthy and R. Carter (2007) From Corpus to Classroom: Language
Use and Language Teaching (Cambridge: Cambridge University Press).
Oppenheim, N. (2000) ‘The importance of recurrent sequences for non-native
speaker fluency and cognition’ in H. Riggenbach (ed.) Perspectives on Fluency,
pp. 220–40 (Ann Arbor: University of Michigan Press).
Paivio, A. (1986) Mental Representations: A Dual-coding Approach (Oxford and New
York: Oxford University Press).
Paivio, A. and A. Desrochers (1979) ‘Effects of an imagery mnemonic on second
language recall and comprehension’, Canadian Journal of Psychology, 33:
17–28.
Paivio, A., J. C. Yuille and S. A. Madigan (1968) ‘Concreteness, imagery and
meaningfulness values for 925 nouns’, Journal of Experimental Psychology
Monograph 76 (1), pt. 2: 1–25.
Palmer, H. E. (1999 [1925]) ‘Conversation’. Reprinted in R. C. Smith (ed.) (1999)
The Writings of Harold E. Palmer, pp. 185–91 (Hon-no-Tomosha).
Paribakth, T. S. and M. Wesche (1997) ‘Vocabulary enhancement activities and
reading for meaning in second language vocabulary acquisition’ in J. Coady
and T. Huckin (eds.) Second Language Vocabulary Acquisition: A Rationale for
Pedagogy, pp. 174–200 (Cambridge: Cambridge University Press).
Parry, M. (1971) ‘The making of Homeric verse’ in A. Parry (ed. and trans.) The
Collected Papers of Milman Parry, pp. 1–239 (Oxford, Clarendon Press).
Paulston, C. B. (1971) ‘Structural pattern drills: A classification’, Foreign Language
Annals, 4: 187–93.
Pawley, A. and F. Syder (1983) ‘Two puzzles for linguistic theory: Nativelike
selection and nativelike fluency’ in J. Richards and R. Schmidt (eds.) Language
and Communication, pp. 191–226 (London: Longman).
Pawley, A. and F. Syder (2000 [1976]) ‘The one-clause-at-a-time hypothesis’ in
H. Riggenbach (ed.) Perspectives on Fluency, pp. 163–99 (Ann Arbor, MI: The
University of Michigan Press). A revised version of a paper read at the First
National Conference of the Linguistic Society of New Zealand, Auckland,
August 1976.
Peters, A. M. (1983) Units of Language Acquisition (Cambridge: Cambridge
University Press).
Peters, E. (2007a) ‘Manipulating L2 learners’ online dictionary use and its effect
on L2 word recognition’, Language Learning and Technology, 11 (June).
Peters, E. (2007b) ‘The relationship between L2 learners’ online dictionary use
and word retention’, Language Forum, 33(2): 45–64.
 References 187

Petitjean, J. (2008) El papel de la traducción en la adquisición semi-incidental de

vocabulario. MA Thesis (Erasmus University College Brussels, Belgium).
Pierson, H. D. (1989) ‘Using etymology in the classroom’, English Teaching Journal,
43: 57–63.
Pimsleur, P. (1967) ‘A memory schedule’, Modern Language Journal, 51: 73–5.
Ponterotto, D. (1994) ‘Metaphors we can learn by: How insights from cognitive
linguistic research can improve the teaching/learning of figurative language’,
Forum, 32: 2–7.
Powell, M. (1996) Business Matters (Hove, UK: LTP / Boston: Thomson Heinle).
Prabhu, N. S. (1987) Second Language Pedagogy (Oxford: Oxford University
Press).
Pressley, M., J. Levin and M. McDaniel, (1987) ‘Remembering versus inferring
what a word means: mnemonic and contextual approaches’ in M. McKeown
and M. Curtis (eds.), The Nature of Vocabulary Acquisition, pp. 107–27 (Hillsdale,
NJ: Lawrence Erlbaum Associates).
Pullum, G. K. and B. C. Scholz (2002) ‘Empirical assessment of stimulus poverty
arguments’, The Linguistic Review, 19: 9–50.
Radden, G. and K. U. Panther (eds.) (2004) Studies in Linguistic Motivation (Berlin
and New York: Mouton de Gruyter).
Ranta, L. and R. Lyster (2007) ‘A cognitive approach to improving immersion
students’ oral language abilities: the Awareness-Practice-Feedback sequence’
in R. M. DeKeysar (ed.) Practice in a Second Language: Perspectives from Applied
Linguistics and Cognitive Psychology, pp. 141–60 (Cambridge: Cambridge
University Press).
Read, J. and P. Nation (2004) ‘Measurement of formulaic sequences’ in N. Schmitt
(ed.) Formulaic Sequences: Acquisition, Processing, and Use, pp. 23–35 (Amsterdam
and Philadelphia: John Benjamins).
Richards, J. C. and T. S. Rodgers (1986) Approaches and Methods in Language
Teaching: A Description and Analysis (Cambridge: Cambridge University
Press).
Rivers, W. M. and M.S Temperley (1978) A Practical Guide to the Teaching of English
as a Second Language (New York and Oxford: Oxford University Press).
Robinson, P. (1995) ‘Task complexity and second language narrative discourse’,
Language Learning, 45(1): 99–140.
Robinson, P. (1997) ‘Generalizability and automaticity of second language learn-
ing under implicit, incidental, enhanced, and instructed conditions’, Studies
in Second Language Acquisition,19: 223–47.
Robinson, P. (2003) ‘Attention and memory during SLA’ in C. J. Doughty and
M. H. Long (eds.) The Handbook of Second Language Acquisition, pp. 631–78
(New York: Blackwell Publishing).
Robinson, P. and M. Ha (1993) ‘Instance theory and second language rule
learning under explicit conditions’, Studies in Second Language Acquisition, 13:
413–38.
Robinson, P., S. Ting and J. Unwin (1995) ‘Investigating second language task
complexity’, RELC Journal, 26(2): 62–79.
Roediger, H. (1980) ‘Memory metaphors in cognitive psychology’ Memory and
Cognition, 8: 231–46.
Rosa, E. and R. P. Leow (2004) ‘Awareness, different learning conditions, and L2
development’, Applied Psycholinguistics, 25: 269–92.
188 References

Rosa, E. and M. D. O’Neill (1999) ‘Explicitness, intake, and the issue of awareness:
Another piece to the puzzle’, Studies in Second Language Acquisition, 21: 511–56.
Rott, S. and J. Williams (2003) ‘Making form–meaning connections while read-
ing: a qualitative analysis of word processing’, Reading in a Foreign Language,
15: 45–75.
Rowling, J. K. (2007) Harry Potter and the Deathly Hallows (London: Bloomsbury
Publishing Plc.).
Rubin, D. (1995) Memory in Oral Traditions: The Cognitive Psychology of Epic,
Ballads, and Counting-out Rhymes (London: Oxford University Press).
Rumelhart, D., J. McClelland and the PDP Research Group (1986) Parallel
Distributed Processing, vol. I (Cambridge, MA: MIT Press).
Rundell, M. (2001) ‘Cats, conversations, and metaphor’, Humanising Language
Teaching, 3 (www.hltmag.co.uk. Pilgrims).
Sag, I., T. Baldwin, F. Bond, A. Copestake and D. Flickinger (2002) ‘Multiword
expressions: a pain in the neck for NLP’ in A. Gelbukh (ed.) Computational
Linguistics and Intelligent Text Processing: Third International Conference, Mexico
City, pp. 1–15 (Heidelberg/Berlin: Springer-Verlag).
Sanchez, C. S. (forthcoming) ‘Improving word-learnability with lexical decom-
position strategies’ in S. De Knop, F. Boers and T. De Rycker (eds.) Fostering
Language Teaching Efficiency Through Cognitive Linguistics (Berlin and New
York: Mouton de Gruyter).
Schmidt, R. W. (1990) ‘The role of consciousness in second language learning’,
Applied linguistics, 11: 129–58.
Schmidt, R. W. (1992) ‘Psychological mechanisms underlying second language
fluency’, Studies in Second Language Acquisition, 14: 357–85.
Schmidt, R. W. (1993) ‘Awareness and second language acquisition’, Annual
Review of Applied Linguistics, 13: 206–26.
Schmidt, R. W. (2001) ‘Attention’ in P. Robinson (ed.) Cognition and Second
Language Instruction, pp. 3–32 (Cambridge: Cambridge University Press).
Schmitt, N. (1997) ‘Vocabulary learning strategies’ in N. Schmitt, and
M. McCarthy (eds.) Vocabulary, Description, Acquisition and Pedagogy, pp. 199–
236 (Cambridge: Cambridge University Press).
Schmitt, N. (2008) ‘Instructed second language vocabulary learning’, Language
Teaching Research, 12: 329–63.
Schmitt, N. and R. Marsden (2006) Why is English Like That? Historical Answers to
Hard ELT Questions (Michigan: The University of Michigan Press).
Schmitt, N. and D. Schmitt (1995) ‘Vocabulary notebooks: Theoretical under-
pinnings and practical suggestions’, ELT Journal, 49: 133–43.
Schneider, W. and J. M. Chein (2003) ‘Controlled and automatic processing:
From mechanisms to biology’, Cognitive Science, 27: 525–59.
Schneider, W. and R. Shiffrin (1977) ‘Controlled and automatic human information
processing. Part 1, Detection, search and attention’, Psychological Review, 84: 1–66.
Scott, M. (1994) ‘Metaphors and language awareness’ in L. Barbara and M. Scott
(eds.) Reflections on Language Learning, pp. 89–104 (Clevedon: Multilingual
Matters).
Searle, J. (1969) Speech Acts (Cambridge: Cambridge University Press).
Segalowitz, N. (2000) ‘Automaticity and attentional skill in fluent performance’
in H. Riggenbach (ed.) Perspectives on Fluency, pp. 200–19 (Ann Arbor: The
University of Michigan Press).
 References 189

Segalowitz, N. S. (2003) ‘Automaticity and second languages’ in C. Doughty and

M. H. Long (eds.) The Handbook of Second Language Acquisition, pp. 282–308
(Oxford: Blackwell Publishing).
Segalowitz, N. (2007) ‘Access fluidity, attention control, and the acquisition of
fluency in a second language’, TESOL Quarterly, 41: 181–6.
Segalowitz, N. S. and J. Hulstijn (2005) ‘Automaticity in bilingualism and second
language learning’ in J. F. Kroll and A. M. B. De Groot (eds.) Handbbook Of
Bilingualism, Psycholinguistic Approaches, pp. 371–88 (Oxford: Oxford University
Press).
Shapiro, A. M. and D. L. Waters (2005) ‘An investigation of the cognitive
processes underlying the keyword method of foreign vocabulary learning’,
Language Teaching Research, 9: 129–46.
Sharifian, F. (2002) ‘Memory enhancement in language pedagogy: implications
from cognitive research’, TESL-EJ, 6(2), A-2; http://tesl-ej.org/ej22/a2.html;
accessed 21 Dec. 2008.
Shiffrin, R. and W. Schneider (1977) ‘Controlled and automatic human infor-
mation processing. Part 2, Perceptual learning, automatic attending, and a
general theory’, Psychological Review, 84: 127–90.
Shin, D. and P. Nation (2008) ‘Beyond single words: The most frequent colloca-
tions in spoken English’, ELTj, 62: 339–48.
Sinclair, J. (1991) Corpus, Concordance, Collocation (Oxford: Oxford University
Press).
Sinclair, J. and A. Mauranen (2006) Linear Unit Grammar: Integrating Speech and
Writing (Amsterdam and Philadelphia: John Benjamins).
Siyanova, A., K. Conklin and N. Schmitt (2008) ‘Are idioms processed differ-
ently by native speakers and second language learners?’ Paper presented at the
SLRF 2008 conference, 19 Oct. 2008, University of Hawaii.
Skehan, P. (1998) A Cognitive Approach to Language Teaching (Oxford: Oxford
University Press).
Skehan, P. and P. Foster (1999) ‘The influence of task structure and processing
conditions on narrative retellings’, Language Learning, 49: 93–120.
Skoufaki, S. (2008) ‘Conceptual metaphoric meaning clues in two idiom pres-
entation methods’ in F. Boers and S. Lindstromberg (eds.) Cognitive Linguistic
Approaches to Teaching Vocabulary and Phraseology, pp. 101–32 (Berlin and New
York: Mouton de Gruyter).
Slobin, D. (2000) ‘Verbalized events: A dynamic approach to linguistic relativity
and determinism’ in S. Niemeier and R. Dirven (eds.) Evidence for Linguistic
Relativity, pp. 107–38 (Amsterdam and Philadelphia: John Benjamins).
Smith, N. (1973) The Acquisition of Phonology (Cambridge: Cambridge University
Press).
Sőkmen, A. J. (1997) ‘Current trends in teaching second language vocabulary’
in N. Schmitt and M. McCarthy (eds.) Vocabulary: Description, Acquisition and
Pedagogy, pp. 237–57 (Cambridge: Cambridge University Press).
Sprenger, A. S., W. J. M. Levelt, and G. Kempen (2006) ‘Lexical access during the
production of idiomatic phrases’, Journal of Memory and Language, 54: 161–84.
Stengers, H. (2007) ‘Is English exceptionally idiomatic? Testing the waters for
a lexical approach to Spanish’ in F. Boers, J. Darquennes and R. Temmerman
(eds.) Multilingualism and Applied Comparative Linguistics: Pedagogical Perspectives,
pp. 107–25 (Newcastle: Cambridge Scholars Publishing).
190 References

Stengers, H. (2009) Putting the Idiom Principle to the Test: An Exercise in Applied
Comparative Linguistics. Unpublished PhD dissertation (Free University of
Brussels).
Stevick, E. (1986) Images and Options in the Language Classroom (Cambridge:
Cambridge University Press).
Strässler, J. (1982) Idioms in English: A Pragmatic Analysis (Tűbingen: Gunter Narr
Verlag).
Stubbs, M. (1995) ‘Collocations and semantic profiles: on the cause of the trou-
ble with quantitative studies’, Functions of Language, 2(1): 23–55.
Swain, M. (1985) ‘Communicative competence: Some roles of comprehen-
sible input and comprehensible output in its development’ in S. Gass and
C. Madden (eds.) Input in Second Language Acquisition, pp. 235–56 (New York
and Rowley, MA: Newbury House).
Swain, M. (1993) ‘The output hypothesis: Just speaking and writing aren’t
enough’, The Canadian Modern Language Review, 50: 158–64.
Swain, M. (2000) ‘The output hypothesis and beyond: Mediating acquisition
through collaborative dialogue’ in J. P. Lantolf (ed.), Sociocultural Theory and
Second Language Learning, pp. 97–114 (Oxford: Oxford University Press).
Swan, M. (2006) ‘Legislation by hypothesis: the case of task-based instruction’,
Applied Linguistics, 26(3): 376–401.
Sweetser, E. (1990) From Etymology to Pragmatics (Cambridge: Cambridge
University Press).
Swinney, D. and A. Cutler. (1979) ‘The access and processing of idiomatic expres-
sions’, Journal of Verbal Learning and Behavior, 18: 523–34.
Talmy, L. (1985) ‘Lexicalization patterns: Semantic structure in lexical forms’ in
T. Shopen (ed.) Language Typology and Syntactic Description. Vol. 3: Grammatical
Categories and the Lexicon, pp. 57–149 (Cambridge and New York: Cambridge
University press).
Talmy, L. (1988) ‘Force dynamics in language and cognition’, Cognitive Science,
12: 49–100.
Taylor, J. (2002) Cognitive Grammar (Oxford: Oxford University Press).
Taylor J. (2006) ‘Polysemy and the lexicon’ in G. Kristiansen, M. Achard,
R. Dirven and J. Ruiz de Mendoza Ibáñez (eds.) Cognitive Linguistics: Current
Applications and Future Perspectives, pp. 51–80 (Berlin and New York: Mouton
de Gruyter).
Thompson, I. (1987) ‘Memory in language learning’ in A. Wenden and J. Rubin
(eds.) Learner Strategies in Language Learning, pp. 43–56 (Englewood Cliffs, NJ:
Prentice-Hall).
Titone, D. A. and C. M. Connine (1999) ‘On the compositional and noncomposi-
tional nature of idiomatic expressions’, Journal of Pragmatics, 31: 1655–74.
Tomasello, M. (2003) Constructing a Language: A Usage-Based Theory of Language
Acquisition (Cambridge, MA: Harvard University Press).
Traugott, E. and R. Dasher (2002) Regularity in Semantic Change (Cambridge:
Cambridge University Press).
Tucker, G. (2007) ‘Between lexis and grammar: towards a systemic functional
account of phraseology’ in R. Hasan and C. Matthiessen (eds.) Continuing
Discourse in Language: A Functional Perspective, pp. 1951–75 (London: Equinox).
Tulving, E. (1983) Elements of Episodic Memory (New York: Oxford University
Press).
 References 191

Ullmann, S. (1957) The Principles of Semantics (Oxford: Blackwell).

Underwood, G., N. Schmitt and A. Galpin (2004) ‘The eyes have it: an eye-move-
ment study into the processing of formulaic sequences’ in N. Schmitt (ed.)
Formulaic Sequences: Acquisition, Processing and Use, pp. 153–72 (Amsterdam
and Philadelphia: John Benjamins).
van de Weijer, J. (2003). ‘Consonant variation within words’ in D. Archer,
P. Rayson, A. Wilson and T. McEnery (eds.), Proceedings of the Corpus Linguistics
2003 Conference. University Centre for Computer Corpus Research on Language
Technical Papers, Vol. 16, pp. 184–90 (Lancaster University, UK).
van de Weijer, J. (2005). ‘Listeners sensitivity to consonant variation within
words’, Lund Working Papers 51, pp. 225–39 (Lund: University of Lund).
VanPatten, B. (1990) ‘Attending to form and content in the input: an experi-
ment in consciousness’, Studies in Second language Acquisition, 12: 287–301.
Vihman, M. (1978) ‘Consonant harmony: its scope and function in child
language’ in J. Greenberg (ed.) Universals of Language, vol. 2, Phonology, pp. 281–
334 (Stanford, CA: Stanford University Press).
Wallace, W. (1994) ‘Memory for music: Effect of melody on recall of text’ Journal
of Experimental Psychology: Learning, Memory and Cognition, 20: 1471–85.
Waring, R. (1997) ‘The negative effects of learning words in semantic sets:
a replication’, System, 25: 261–74.
Waring, R. and P. Nation (2004) ‘Second language reading and incidental learn-
ing’, Angles on the English Speaking World, 4: 11–23.
Watanabe, Y. (1997) ‘Input, intake, and retention: Effects of increased process-
ing on incidental learning of foreign language vocabulary’, Studies in Second
Language Acquisition, 19: 287–307.
Weinert, R. (1995) ‘The role of formulaic language in second language acquisi-
tion: a review’, Applied Linguistics, 16: 185–205.
Weir, C. (1990) Communicative Language Testing (Hemel Hemptstead: Prentice
Hall).
Weir, C. (2005) Language Testing and Validation: An Evidence-based Approach
(Basingstoke and New York: Palgrave Macmillan).
Wiese, R. (1984) ‘Language production in foreign and native languages: same
or different?’ in H. Dechert, D. Möhle and M. Raupach (eds.) Second Language
Productions, pp. 11–25 (Tűbingen: Gunter Narr).
Williams, J. (1999) ‘Memory, attention and inductive learning’, Studies in Second
Language Acquisition, 21: 1–48.
Willis, D. (1990) The Lexical Syllabus: A New Approach to Language Teaching
(London: Collins ELT).
Willis, J. (1996) A Framework for Task-based Learning (Harlow, UK: Longman).
Wooland, G. (2000) ‘Collocation – encouraging learner independence’, in
M. Lewis (ed.) Teaching Collocation, pp. 28–46 (Hove: LTP).
Wray, A. (2000) ‘Formulaic sequences in second language teaching: Principles
and practice’, Applied Linguistics, 21: 463–89.
Wray, A. (2002) Formulaic Language and the Lexicon (Cambridge: Cambridge
University Press).
Wray, A. (2008) Formulaic Language: Pushing the Boundaries (Oxford: Oxford
University Press).
Wray, A. and T. Fitzpatrick (2008) ‘Why can’t you just leave it alone? Deviations
from memorized language as a gauge of native-like competence’ in F. Meunier
192 References

and S. Granger (eds.), Phraseology in Foreign Language Learning and Teaching,

pp. 123–47 (Amsterdam: John Benjamins).
Wray, A. and M. R. Perkins (2000) ‘The function of formulaic language: an inte-
grated model’, Language and Communication, 20: 1–28.
Ziegler, J., M. Muneaux and G. Grainger (2003) ‘Neighborhood effects in audi-
tory word recognition: Phonological competition and orthographic facilita-
tion’, Journal of Memory and Language, 48(4): 779–93.
Zipf, G. K. (1949) Human Behaviour and the Principle of Least Effort: An Introduction
to Human Ecology (Cambridge, MA: Addison-Wesley).
Index

alliteration, 15–16, 76, 107–34, 147 dictation, 31

articulatory ease, 76–7, 124 discriminating collocations test
assonance, 76–7, 106–9, 114–24, 147 (DISCO), 162–3, 164
audio-lingual method (ALM), 16–17, distributed learning, 98
139–41, 143, 144, 171–2 drilling, drills: see audio-lingual
auditory learning style, 117 method; rehearsal, repetition
automaticity, 25–7, 32, 34, 36, 61–2, drills, oral, 139–45, 159
133–45, 158 dual coding, 79–98, 140, 148
autonomy, 18–21, 39–48, 59, 87, 98–9,
104, 121, 144 elaboration, 22–3, 41, 68, 71, 77–8,
awareness-raising, 15, 19–21, 39–54, 79–105, 106–24, 126, 134, 146, 148,
80, 85, 90, 98, 102, 117, 121, 128, 149, 153
149, 159 emergentism, 27–9
etymology: see diachronic angle
bathtub effect, 118, 123–4 exemplars, 27–31, 158–9
binomials, 71–2, 108, 114–17, 124, 134 eye-tracking, 37, 45–6, 81–2
breadth of knowledge, 10–11, 36, 126
fluency, 10–11, 23, 25–38, 61–3, 78,
chunking, 19–20, 35, 47–8, 52–3, 157, 124–5, 126, 133–45
159–60 focus on form, 44, 141–2; see also
colligation, 160–1 grammar instruction
collocation boxes, 20, 21, 102–5, 159 frequency, 4–5, 7–8, 13–14, 20, 28,
collocations sampler, 5–7, 57, 59, 60, 29, 34, 41–4, 48, 53, 56–67, 147,
104, 107, 153 157, 162–3
compositional vs. non-compositional, frequency effects, 28, 34, 48, 81
11–13, 46–7, 83, 138 function words, 35, 60, 159
compounding, 112–14, 117–18, 146–7
conceptual metaphor: see metaphor glossaries, glosses, 49, 50–1, 100
conceptual metonymy: see metonymy grammar acquisition (L1), 27–30, 31
consonance, 76, 77, 107, 114, 170 grammar instruction, 31, 133,
constructions, 27 140–3, 158
conversational fillers, 9, 13, 37, 38,
53, 61, 124 holistic storage / processing, 4, 10,
corpus, corpora, 4–5, 7–8, 56–7, 24, 26–9, 31–4, 78, 81–3, 122,
59–60, 132, 157, 161–2 135–8, 158

deep (or rich) processing: see iconicity, 71–2, 74–5, 124

elaboration idioms, 8, 11–13, 29, 30, 47, 49, 56,
deleted essentials test (DET), 163–8 63, 64–73, 79–92, 95, 96, 100–2,
deletion exercises, 132 108, 114, 117, 130, 138, 148–9, 157,
depth of knowledge, 10–11, 36, 126 170–1
diachronic perspective, 4, 108, imagery (mental): see dual
149–53 coding

193
194 Index

incidental uptake, 13, 20, 39–54, 56, paired associate learning, 154–6
61, 68, 154 pedagogical chunking, 19–20, 35, 48;
inflection, consequences of, 23, 26, see also chunking
33, 43, 57, 118, 136, 158, 173 phrasal verbs, 3, 8, 97–9, 155–6
phonetic or phonological reduction,
keyword method, 69, 79 34–5, 43, 76, 116, 159
kinaesthetic imagery: see motoric physical enactment: see mime
imagery pictorial elucidation, 91–3
picture superiority effect, 92
L1 transfer, L1 interference, 1–2, 46, pre-literacy: see literacy
62, 132 prioritization: see selection
Latinate lexis (including borrowings proverbs, 117
from Greek via Latin), 112–13, procedural knowledge, 26–7, 33, 62,
151–3, 170, 172 78, 115, 137; see also automaticity
learner autonomy: see autonomy probabilistic, 10, 136, 164
least effort, principle of, 76, 124 productive vs. receptive skills, 10–11,
levelling, phonological, 34, 35, 43 25–6, 61–2, 125, 133, 161–3
literacy, 24, 25, 31, 41, 78, 122
lyrics, 122, 127, 173 rehearsal, repetition, 20–1, 141–4
rhyme: see assonance
matching exercises, 20, 23,
128–33, 171 selection, 13–14, 55–77, 103–4
memorability, 14–16, 68–9, 76, 87–92, semantic prosody, 59, 75–6, 104,
107, 119–24, 147–9, 154–6 153, 159
metaphor, 15, 72–5, 79–80, 93, 94, semi-incidental uptake, 49–54
97–101, 148–9 similes, 71, 116, 117
metonymy, 74–5, 80, 93, 148 slant rhyme, 106, 170
MI: see mutual information song lyrics: see lyrics
mime and physical enactment, speech acts, 28, 138, 142
93–5, 145 stylistic prosody, 76, 153
mnemonic potential: see
memorability task-based instruction (TBI), 133,
mnemonics, 69, 150, 154, 156; see also 141–4
keyword method, paired associate task-induced involvement load, 50;
learning see also elaboration
motivation (linguistic), 15–19, 70–7, teachability, 67–78, 126
106–8, 114–19, 124, 130, 148, 153 testing, 156, 160–8
motoric imagery, 68, 80, 94 translation, 2, 36, 52, 144
mutual information (MI) score, 5, 7 T-score, 5, 162

naturalistic context, 9, 18, 30, 159 usage-based models of language, 28

negative evidence, 31, 46, 132 usage restrictions, 65, 95–6,
notebooks: see vocabulary
notebooks vocabulary notebooks, 19, 20, 103–5
noticing, 12, 19, 20, 22, 28, 43–54,
68, 121, 123, 137 width of knowledge: see breadth of
knowledge
odd-one-out exercises, 132 word repetition, 106
oral formulae, Oral Formulaic
Theory, 122 Z-score, 5, 162