0

Descriptive and Theoretical Aspects of English Idioms

Eirene C. Katsarou

A Monograph
1

To my Family
 2

Foreword

This monograph consists part of my research undertaken for the completion of my Ph.D thesis at
the University of Essex and focuses mainly on current trends of Theoretical Linguistics with respect
to descriptive aspects of idiom in English. It primarily aims to offer a thorough overview of the key
theoretical work in the field of idiomatology from a syntactic, semantic and socio-cultural point of
view. The lexical multi-word unit of idiom is thus discussed at every level of linguistic analysis and
is presented as an entity of its own lexical integrity that forms the core of the phraseological system
of the English language.

Given the pressing need for a clear and concrete delimitation of idiom as a linguistic term that
will be used profitably as a working definition for further research into a wide range of issues in
relation to the area of second language idiom acquisition in EFL and ESL educational contexts, it is
hoped that the present study will function as a reference point capturing several angles of the topic.
 3

Table of Contents

Foreword ii
Table of Contents iii
List of Tables iv
List of Figures iv

Chapter Page

Chapter 1 Introduction 5
1.1. Research on Idioms in Theoretical Linguistics 5
1.2. Towards a Definition of Idiom 6

8
Chapter 2 The Socio-Cultural and Syntactic Approach
8
2.1. The Socio-Cultural Approach: Murat H. Roberts (1944)
10
2.2. The Syntactic Approach
10
2.2.1. J.J. Katz and P.M. Postal (1963)
13
2.2.2. Wallace Chafe (1968)
17
2.2.3. Uriel Weinreich (1969)
22
2.2.4. Bruce Fraser (1970)
26
2.2.5. Frederick Newmeyer (1972; 1974)

Chapter 3 The Semantic Approach 30

3.1. Adam Makkai (1972) 30
3.2. Chitra Fernando and Roger Flavell (1981) 34

Chapter 4 The Pragmatic Approach 41

4.1. Jurg Strassler (1982) 41
4.2. Chitra Fernando (1996) 45
4.3. Rosamund Moon (1998) 60

Chapter 5 The Psycholinguistic Approach 68

5.1. The Idiom List Hypothesis (Bobrow and Bell, 1973) 68
5.2. The Lexical Representation Hypothesis (Swinney and Cutler, 1979) 69
5.3. The Direct Access Hypothesis (Gibbs, 1980; 1984) 71
5.4. The Cognitivist or Decomposition Approach 72
5.5. A Working Definition of Idiom in English 83

Bibliography
90
 4

List of Tables
Table 2.1. Transformational Defectiveness of Idioms with respect to Governed Rules. 24

Table 2.2. Fraser’s Frozenness Hierarchy (adapted from: Fraser, 1970:39). 24

Table 3.1. Degrees of Semantic Motivation according to Fernando and Flavell (1981:28). 35

Table 3.2. Ranking of Idioms based on their Semantic Properties (adapted from: Flavell 38
et al, 1981: 54).

Table 4.1. Time Deixis of Idioms according to Strassler (1982: 90-1). 44

Table 4.2. Scale of Idiomaticity of Multiword Expressions (adapted from: Fernando, 47

1996b:32).

Table 4.3. Fernando’s (1996b:72-4) Functional Classification of Multiword Expressions.

50
Table 4.4. A Cross-Comparison of Lexicogrammatical and Functional Idiom Classes.
52
Table 4.5. Semantic Classifications of Idioms.
53
Table 4.6. Distribution of Subtypes across all Macrocategories (adapted from: Moon,
1998:62-4). 62

Table 5.1. Overview of Definitions on Idiom in the Literature 90

List of Figures
Figure 2.1. P-marker for ‘John kicked the bucket’ (Katz et al, 1963:280). 11

Figure 2.2. Deep structure for the Passive voice of the Idiom ‘John kicked the bucket’ (Katz 11
et al, 1963:280).

Figure 2.3. Analysis of the idiom ‘shoot the breeze’ according to Weinreich (1969:54). 20

Figure 2.4. Idiom Comparison Rule (Weinreich, 1969:59). 21

Figure 2.5. Derivation of the idiom ‘kick the bucket’ according to Newmeyer (1974:339). 27

Figure 4.1. Wood’s (1986) Continuum of Lexical Phrases 56

Figure 4.2. Cermak’s (1988) Continuum of Phraseological Units 56

 5

Chapter 1

Introduction
‘Idiomaticity is important for this reason, if for no other, that there is so much of it in every language’.
Uriel Weinreich (1969: 23)
1.1. Research on Idioms in Theoretical Linguistics
The issue of conventionalized sequences of words has long been a problem in post-Chomskyan
linguistic theory, though still a relatively marginal one. Despite the pervasiveness of such forms as
idioms, formulas, prefabricated patterns, fixed expressions and phrases in general, and the devoted
efforts of a few interested linguists, conventionalized language has never really found a clear place
in language theories. A brief overview of the literature on the topic suffices to reveal a noticeable
lack of both theoretical and empirical background that result in serious definitional problems. In
fact, this inability to establish a solid theoretical framework within which to define, let alone,
circumscribe and account for their formal as well as functional featural peculiarities appears to be
especially pronounced in the case of idioms in which we are mainly interested here.

From time to time, various European phraseological models have been developed in an attempt
to capture and offer a viable explanation for the lexico-grammatical characteristics and special
discoursal functions of idioms in a concrete and consistent linguistic manner. Obviously, different
models foreground and prioritize different properties and realizations of these properties based on
the theoretical construct employed for the study of idioms. Analyses of idioms that have been
undertaken entirely within the spirit of Chomskyan syntactic theories place much emphasis on
idioms’ surface structure representation at the deep level as well as on their flexibility in terms of
their syntactic transformational potential (Weinreich, 1969; Fraser, 1970). Studies centered upon the
semantics of idiomatic phrases, on the other hand, sought to adequately account for the
interpretation of idioms in the light of their semantic decomposition pertinent to them (Makkai,
1972; Wood, 1986). In addition, corpus-based approaches were also adopted viewing idioms as
frequently fixed co-occurrence sequences (Strassler, 1982). Finally, functional approaches focused
on the study of idiom use in a range of discoursal environments in order to reconcile both sides of
research-the Lexical and the Functional-yielding a unified theoretical model of idioms (Fernando,
1996).

Given the necessity of a theoretical background for a better and deeper understanding of the
idiosyncratic nature of idioms as distinct lexical entities in the vocabulary system of a language, the
 6

focus of this first of the two literature review chapters rests on a thorough and detailed discussion of
idiom research devoted to key descriptive studies aiming to: (i) reveal the controversial status of
idioms in definitional terms stressing the need for an adequately explanatory and theoretically
coherent account of idioms by setting out consistent and clear-cut criteria that circumscribe the
linguistic substance of idioms most unambiguously in the wider area of fixed expressions. (ii)
indicate potential areas of learning and/or processing difficulties that may stem from the use of
idioms in receptive and productive communicative situations by L2 learners of English as a foreign
language. The discussion that follows will be mainly centered on key theoretical language idiom
studies that will be presented and categorised according to the level of analysis adopted in each
treatment in our effort to pinpoint the features that will be incorporated in our working definition for
the type of idioms used for the purposes of the current study.

1.2. Towards a Definition of Idiom

One of the thorniest issues in idiom research has been the question of how to define ‘idioms’.
Throughout time, the study of idioms have run into two main difficulties, i.e. the terminology
involved and the delimitation of the concept resulting in a proliferation of definitions by various
researchers working in the field (for a brief review see, Meier, 1975). It is therefore important to
begin this chapter with an understanding of both the term idiom and the selection criteria adopted
for the idioms chosen for the main study. On the face of it, current definitions of the term idiom as it
is found in widely used English dictionaries and applied by different scholars are presented next in
an effort to adequately represent the confluence of the various interpretations of idioms. Thus, the
sole intention of this chapter is to show to the reader that there is a blurred edge between the
different definitions given by the scholars regarding a coherent interpretation of the ill-defined term
idiom and offer, consequently, a concise working definition of the idiom types considered for the
purposes of this study.

In A First Dictionary of Linguistics and Phonetics (1980) and in The Cambridge Encyclopedia
of the English Language (2003), Crystal defines idioms as:

‘A term used in GRAMMAR and LEXICOLOGY to refer to a SEQUENCE of

WORDS which is SEMANTICALLY and often SYNTACTICALLY restricted,
so that they function as a single UNIT. From a semantic viewpoint, the
MEANING of the individual words cannot be summed to produce the meaning
of the ‘idiomatic’ expression as a whole. From a syntactic viewpoint, the words
often do not permit the usual variability they display in other CONTEXT, e.g.
it’s raining cats and dogs does not permit *it’s raining a cat and a dog/dogs
and cats, etc.’ (Crystal,1980:179 as cited in Awwad, 1990:57)
 7

‘Two central features identify an idiom. The meaning of the idiomatic expression
cannot be deduced by examining the meanings of the constituent lexemes. And
the expression is fixed, both grammatically and lexically. Thus, put a sock in it!
means ‘stop talking’ and it is not possible to replace any of the lexemes and
retain the idiomatic meaning. Put a stocking in it or put a stock on it must be
interpreted literally or not at all.’ (Crystal, 2003:163)

An even more precise entry is given in the Longman Dictionary of English Idioms:

‘An idiom is a fixed group of words with a special different meaning from the
meanings of the separate words. So, to spill the beans is not at all connected
with beans: it means ‘to tell something that is secret.’
(Longman Dictionary of English Idioms, 1990:inside front cover)

According to The New Webster’s Encyclopedic Dictionary of the English Language, ‘idiom’
derives from the Greek lexeme idios which means ‘proper or peculiar to one’s self.’ The following
general entry is given:

‘A mode of expression peculiar to a language or to a person; a phrase or

expression having a special meaning from usage, or a special grammatical
character; the genius or peculiar cast of a language; a peculiar form or variety of
language; a dialect.
(The New Webster’s Encyclopedic Dictionary of the English Language, 1980:420, as cited in Liontas,
2003b:2)

Closely mirroring the definitions given above is the one given in the Oxford English Dictionary
(OED) and is regarded by many scholars as the classic one. Again only the relevant entry, 3a (out of
seven entries in total), is provided below:

‘A form or expression or expression, grammatical construction, phrase, etc.,

peculiar to a language; a peculiarity of phraseology approved by the usage of the
language, and often having a significance other than its grammatical or logical
one.’ (Oxford English Dictionary, 1989:624)

All of these entries contain one definition that emphasizes the difficulty of inferring the meaning
of the idiom as a whole from the meanings of its constituent parts. In addition, some kind of
grammatical peculiarity and a predetermined social usage appears to be attached to the idiom.
Based on these features alone, we can distinguish at least three dimensions to the term ‘idiom: (a)
the feature of meaning, i.e. the semantic opacity dimension, (b) the feature of grammar, i.e. the
structural dimension and (c) the feature of institutionalized usage, i.e. the conventionalized
pragmatic dimension. By way of definition then, one way of looking at an idiom is to regard it at
the very least as a complex tri-dimensional expression which is not explicable in terms of its
individual words. Though idiomatologists each have a different opinion on this matter, the mixing
of the dimensions just mentioned is so pervasive in the research literature that a clarification on
these grounds is warranted.
 8

Chapter 2
The Socio-Cultural and Syntactic Approach

2.1. The Socio-Cultural Approach: Murat H. Roberts (1944)

One of the very first attempts recorded towards a definition of idiom is that advanced by Roberts
as early as 1944. In contrast to later theoretical studies on idiom, Roberts adopts an approach that is
not so specifically concerned with the linguistic character of idioms as complex lexical units with
particular lexico-grammatical attributes. Rather, his view of what constitutes an idiom appears to be
heavily dominated by the notion of institutionalization that idioms undergo through frequent use by
members of a given speech community in their common, everyday communication exchanges.
Irrespective of whether the idiom is a figure of speech, an anomaly of grammar or merely a group
of words whose meaning cannot be gathered from its component parts when separately considered,
it is used in the generic sense of idioma that supplies the combinatorial government of the tongue in
question. In this sense, idiom is ‘the idiosyncrasy of permutation which a given language exhibits in
contradistinction to all other language or a given period exhibits in contrast to all previous periods’
(Roberts, 1944:300).

Following Roberts, idiom is the manifestation of ‘a discourse’ (exprimendum, logos, ratio,

discourse, Rede), i.e. ‘the inner design’ or structure of thought being communicated via a given
language code (exprimens, glotta, lingua, langue, Sprache). Idiom reflects a particular organization
of thought, the mental design of language through different means of expression that permit the
edifice of thought be fully and explicitly represented, not scantily and elliptically indicated. Resting
on the belief that ‘a language is best when it is most idiomatic, that is, most transparent, so that the
thought of the speaker appears distinctly and without linguistic distortion’ (1944:302-3), Roberts
cites Ancient Greek as an example of a highly idiomatic language that can symbolize the contours
of thought with fullness and precision as a result of the harmonic and balanced coexistence of
grammar and rhetoric that capture the atmosphere of the Greek mind. In this respect then, the idiom
of the language ‘…resides in the experiences and predilections of the speakers of the language
…and governs the architecture of all its sentences…..beginning in the substance of ideas and ending
in the form of words’ (1944-291).

Based on the aforementioned observations, Roberts concludes that idiom and its manifestations
constitute an ever-recurrent riddle whose origin will remain inaccessible if approached by the
 9

methods of formal linguistics alone as the causes of idioms are hidden in the speaker’s choice of
perceptions, his structure of concepts and his habitual emotions lying beyond its frozen linguistic
form. Obviously, Roberts is not interested in the identification of distinctive definitional properties
for idioms based on linguistic grounds. That idiom exists and is clearly recognizable as such is
regarded as an established fact. What needs to be examined, according to Roberts, is how the
cognitive design of a given language reveals itself through what He (1989:150) calls creative
idiomaticity referring to the ingenious manipulation of idiomatic expressions by speakers of given
speech community that normally requires cultural or literary awareness on their part.

In this connection, Roberts refers briefly to examples from a number of languages including
English, Russian and Spanish making use of de Saussure’s langue/parole dichotomy to present the
notion of how individual creativity becomes in time part of the common system of elements that
comprise a language. In this view, every idiom is the result of a personal innovation at a particular
point of time (parole) which when it is taken up by a language community becomes part of
common currency (langue). Idioms, in Roberts’ terms, are viewed as a conformity to the cultural
conventions of language use in thinking and talking, as a matter of cultural awareness, with which
the language user can think, talk and live effectively in the culture underlying his creative power in
thinking and generative talent in language use. Seen under this perspective, Roberts’ definition of
idiom is restricted to the institutionalization criterion presenting them as the particular linguistic
means speakers across languages use in order to express their thoughts and communicate their
needs, wishes and beliefs in common interactions. Clearly, such an approach to the definition and
delimitation of idiom as an independent linguistic phenomenon fails its purpose as it embraces a
host of different types of fixed phrases such as proverbs, catchphrases and clichés that are also used
as institutionalized lexical units for the fulfillment of communication needs in a language
community.
 10

2.2. The Syntactic Approach

2.2.1. J. J. Katz and P. M. Postal (1963)

The first Transformational Generative (TG) attempt to deal with idioms was developed by Katz
and Postal in their short article entitled ‘Semantic Interpretation of Idioms and The Sentences
Containing Them’ (1963). In it, idioms are considered to be the ‘exceptions that prove the rule’ of
compositionality since they do not get their meaning from the meanings of their syntactic parts.
Thus, if an idiom is treated as if it were compositional, false predictions are made about its semantic
properties and relations. To emphasize the non-compositional meaning of idioms, the authors (Katz
and Postal, 1963; Katz, 1973) offer the following definition:

‘The essential feature of an idiom is that its full meaning, and more generally the
meaning of any sentence containing an idiomatic stretch, is not a compositional
function of the meanings of the idiom’s elementary grammatical parts.’
(Katz and Postal, 1963:275)

According to Katz (1973) then, idioms are syntactically complex concatenations in a language
that consist of two or more morphemes that the semantic component of the grammar treats as
lexical items and are entered as unanalyzed wholes in the semantic dictionary. The dictionary
should now be thought of as having two parts, a lexical-item part and a phrase-idiom part that
consists of both lexical idioms such as compound words (e.g. telephone, baritone), noun + adjective
(e.g. snow-white), verb + particle (e.g. put up) and similar clusters and phrasal idioms like verb +
object groups (e.g. kick the bucket). Entries in the latter part will have the form: first, a particular
string of morphemes, the idiomatic stretch; next, some associated constituent that must dominate
the idiomatic stretch in the phrase markers that are to be assigned the semantic information
associated with the pair of the idiomatic stretch and its dominating constituent, and, finally, the
semantic information itself.

Access to the semantic information in the two subparts of the dictionary is obtained by different
methods of assigning sets of readings associated with dictionary entries to the minimal semantic
elements in the underlying phrase markers of sentences. In the case of lexical items and possibly for
lexical idioms too, sets of reading are associated with the lowest level of syntactic categories or
terminal elements of underlying phrase markers while in the case of phrase idioms their readings
are assigned to higher level constituents in underlying phrase markers, not to terminal symbols. For
instance, in the case of the idiomatic stretch kicked the bucket, which has as part of its phrase-idiom
dictionary entry the dominating constituent MV, the reading representing ‘die’ is associated not
 11

with any of the morrphemes coomposing this

t stretch, but ratheer with the constituen
nt MV thatt
dominates occurrencees of kick thee bucket in underlying phrase marrkers. Thus,, the underly
ying phrasee
marker forr a sentence such as John kicked thhe bucket iss that presennted schematically in the
t diagram
m
below.
Figure 2.1. P-mark
ker for ‘Joh
hn kicked th
he bucket’ (Katz et all, 1963: 280
0)

Following the phrasee marker, iff a deep sttructure term

minal stringg-in this caase kicked the bucket--
coincides with
w a stringg listed in thhe idiom paart of the dicctionary, annd if it is doominated by
y the correctt
syntactic category,
c theen it may opptionally bee interpreted
d as havingg the idiomaatic meaning
g, i.e. ‘die’..
Alternativeely, if this option
o is noot chosen, itts semantic interpretatiion will be one paraph
hraseable ass
‘strike the pail with one’s
o foot’. In other woords, it cou
uld be said that
t accordiing to this analysis,
a ann
idiomatic expression
e retains the same deepp structure representatiion as its liiteral countterpart and,,
consequently, both litteral and iddiomatic meeanings wh
hose co-exisstence renders idioms ambiguouss
and polyseemous lexical units, aree differentiaated only on
n the basis of
o interpretaation of the sentence inn
different sentential
s contexts. Inn a similar vein, Wein
nreich (19669) also atttempts to explain
e thee
ambiguity of idioms inn semantic terms as a result
r of ‘reeciprocal coontextual sellection of su
ubsenses off
idiom’s constituent paarts’ (see furrther below
w 2.2.2.3.).

To Katzz and Postaal (1963) suuch a solution seems to

o be a neatt way to account for not
n only thee
semantic innterpretatioon of ambigguous phrasal idioms with
w both ann idiomatic and a literral meaningg
fo their traansformational defectivveness as well.
but also for w This satisfactionn stemmed from theirr
positing thhe deep struucture for a passive senntence like ‘The
‘ buckett was kickeed by Sam’ as given inn
the figure below.
b

Figurre 2.2. Deep

p structure for the passsive voice of the idiom
m ‘John kiicked the bu
ucket’
(Katzz, et al, 1963
3:280)
 12

The idiomatic interpreetation of ‘kick the buucket’ was said

s to be (optionally) possible iff this string,,
and only thhis string, is dominatedd by ‘Mainn Verb’. Butt in this passsive sentennce we find that ‘Mainn
Verb’ dom
minates thee longer strring ‘kick the buckett passive’. The idiom
matic interp
pretation iss
therefore ruled
r out: prrecisely thee result desirred, since ‘The buckett was kickedd by Sam’ does
d not, inn
fact, have the idiomattic interprettation. In thhat way, thee transformaational defiiciencies off idioms aree
accounted for by Katzz and Postall noting thaat (i) transfo
ormations arre triggeredd by formatives alreadyy
present in deep strucctures and (ii)
( these foormatives disturb
d the relationshipp between strings andd
dominatingg nodes in such a wayy as to prevvent idiomaatic interpreetations in jjust those cases
c wheree
they should be prevennted. The question
q obvviously thatt arises is whether
w all such incom
mpatibilitiess
can be sim
milarly accouunted for wiithin this fraamework.

For exaample, we are

a left withh no explannation of why
w (Sam’s)) kicking off the buckeet’ does nott
optionally have the iddiomatic intterpretation while seem
mingly, by their
t theory, it should. Just as thiss
example shows
s thatt Katz andd Postal’s theory preedicts that some deeep structurees will bee
idiomatically interpreeted which are not idiiomatically interpretedd; other exaamples sho
ow them too
predict thaat an idiom
matic interprretation is impossible
i when in faact it is posssible. A seentence likee
‘Sam kickked the buccket gracefuully’ can certainly carrry a meanning paraphhrasable as ‘Sam diedd
gracefully’’. And yet ‘gracefully’
‘ ’ here is a Manner
M Adv
verbial dom
minated by the same ‘M
Main Verb’
which dom
minates ‘kicck the buccket’. If thee passive sentence
s meentioned abbove canno
ot have thee
idiomatic interpretatio
i on for the reason
r given, this senttence shouldd not have such an intterpretationn
either, but this predicction is nott confirmedd. On the faace of this evidence, I believe we
w can onlyy
conclude thhat Katz annd Postal’s explanation
e of the defectiveness off ‘kick the bbucket’ with
h respect too
the passivee transformaation was only
o a luckyy accident at
a best. On the
t other haand, their th
heory seemss
inadequatee to proffer a sufficientt argumentaative line in
n favour of the identityy of the deeep structuree
between ann idiom andd its literal counterpart.
c Whatever the
t details, it seems thaat the deep structure off
the literal sentence
s muust be very different frrom that of the idiomattic one and, this differeence cannott
be relegateed to some optional chhoice withinn the semanttic component. Additioonally, this inability too
provide a theoreticallly well-groounded expplanation for
fo all the transformaational deficciencies off
 13

phrasal idioms with literal counterparts is equally evident in the case of syntactically ill-formed
idioms such as ‘by and large’, ‘kingdom come’ or ‘trip the light fantastic’. As they say, ‘if these
idioms did appear in strings that are generated by the syntactic component, this component would
not be empirically adequate because its output would contain some ungrammatical strings’ (Katz
and Postal, 1963:281) simply relegating them to the rank of semi-sentences.

Despite its theoretical shortcomings, Katz and Postal’s (1963) is considered of crucial historical
importance due to its contribution to the analysis of idioms within the framework of TG grammar.
While their analysis left certain important observations concerning idioms unexplained, it seems
that many of its claims such as the non-compositionality of idioms, the distinction between lexical
and phrasal idioms as well the semantic interpretation of ambiguous phrasal idioms with literal
counterparts obviously demand inclusion in any serious account of idiom undertaken on a cross-
linguistic basis including our own working definition of idiom as used for this study. In view of its
inadequacies, later theoreticians, and among them most notably Weinreich (1969), attempted to
remove these discrepancies inherent in Katz and Postal’s formalization while still working within
the Chomskyan paradigm.

2.2.2. Wallace Chafe (1968)

Wallace Chafe (1968), another transformational-generative grammarian, dealt with the
phenomenon of idioms pointing out the ubiquitous character of idioms and the need for a linguistic
theory to explain them in a natural way since the Chomskyan paradigm has failed to do so.

According to Chafe (1968:111), such a failure is attributed to several peculiarities that an idiom
shows: first, the meaning of an idiom, arrived at through the operation of the semantic component
on a deep structure, which is not some kind of amalgamation of the meanings of the parts of that
structure; instead, the meaning of an idiom is comparable to the meaning of a lexical item. Second,
the transformational deficiencies that most if not all idioms exhibit, for instance the inability of the
idiom ‘kick the bucket’ to undergo passivisation or nominalization and retain its idiomatic meaning.
Third, there are some idioms which are not syntactically well-formed, which could not be generated
by a base component designed to produce well-formed deep structures, such as: by and large,
kingdom come, trip the light fantastic, etc. Fourth, an idiom which is well formed will have a literal
counterpart whose text frequency is usually much lower than that of the corresponding idiom. This
means that the well-formed idiom is less often used in its literal sense, and it is rather frequently
 14

used in its idiomatic one. In this view, the deep structure kick the bucket means something like ‘to
die’ more often than it means something like ‘strike the pail with one’s foot.’

In the second part of his articles, Chafe illustrates his concept of generative semantics with
idioms, using semantic structures into the base component at the beginning of the derivational
process, which, however, are not immediately presented, since two processes have to operate first:
literalization and symbolization. In Chafe’s terms (1968:120-1), idiomaticization is described as
(emphasis mine):
‘….a kind of semantic change which produces a discrepancy between semantic
arrangements(those directly manifested in meaning, concepts, ideas, or whatever the
phenomena of this area may be called) and those ‘post-semantic’ arrangements which
enter directly into symbolization…..[leading to] a special kind of semantic ‘split’. After
such a split has taken place the original semantic arrangement is typically still present in
the language, but in addition a new semantic unit has been formed by a shrinkage of the
composite meaning into a new unitary meaning. Thus, the semantic arrangement ‘kick the
bucket’ shrank into a single unit with a meaning similar to that of ‘die’. The old, ‘literal’
arrangement did not thereby disappear, but remained in the language alongside the new
unit, or idiom. If an idiom has a literal counterpart, then it is symbolized in the same way
as its literal counterpart. That is, before the semantic unit which is an idiom can be
converted by symbolization into pre-phonetic units, it must first be converted into a literal
post-semantic arrangement…[which is] identical with a possible semantic arrangement,
the idiom’s literal counterpart…..A rule which states such a conversion may be called a
‘literalization’ rule [whereby] the historical process of idiomaticization is recapitulated
but in a reverse direction [since] literalization-part of the process of encoding meanings
into sounds- changes the single semantic unit back into the arrangement which is its
historical antecedent’.

Essentially, what Chafe seems to suggest here is that idiomaticization is a historical process by
which certain specific semantic arrangements enter into a special kind of semantic ‘split’. After
such a split has taken place, the original semantic arrangement is typically still present in the
language, but in addition a new semantic unit has been formed by shrinkage of the composite
meaning into a new unitary meaning. Thus, the semantic arrangement of kick the bucket is thought
to have been associated with the methods of killing pigs. In Norfolk England, the bucket was the
beam from which a pig about to be killed used to hang by its hind feet. The nervous reflex actions
of the pig, after its throat was cut to let the blood drain into a receptacle below, caused it to kick the
bucket as it was dying. Then the phrase kick the bucket shrank into a single unit with a meaning
similar to that of ‘die’. The old, ‘literal’ arrangement did not thereby disappear, but remained in the
language alongside the new unit, or idiom. According to this viewpoint then, the process of
idiomaticization lies in diachronic evolution therefore, and cannot be adequately explained by
generative rules as ‘neither a given sense or a given syntactic structure by itself constitutes an
idiom. Rather it is the regular association of one with the other that is the source of idiomaticity.
Such an association is the product of contextual extension in the everyday situation of
 15

communicative use over a period of time’. (Fernando et al, 1981:25). The majority of idioms
exhibit certain stages in their development. They do not randomly acquire new senses, and the
multiple synchronic senses of a given expression will normally be related to each other in a
motivated fashion. No historical shift of meaning can take place without an intervening stage of
polysemy. If an expression once meant A and now means B, it is certain that there was a stage
when the expression meant both A and B, and the earlier meaning of A eventually was lost, as in
the case of idioms here.

However, not all idioms are produced by the shrinkage of well-formed semantic arrangements
into single semantic units. Exceptions are syntactically ill-formed idioms, such as ‘kingdom come’
or ‘trip the light fantastic’ which, according to Chafe, have resulted from semantic shrinkage from
truncated versions of well-known literary quotations. Often the source of an idiom is well-formed at
the time the idiom arises, only to pass out of the language of the idiom’s users while its ghost
remains as the idiom’s literalization. In other words, each idiom has its own special literalization
rule latent within it; and each post-semantic unit has its latent symbolization rule. In fact,
literalization can be looked at as a process rather like symbolization. Thus, the literalization rules
for idioms of the ‘by and large’ type convert them into post-semantic arrangements which are not at
the same time possible semantic arrangements, that is to say they do not have literal counterparts.
This idea has been made clear through Chafe’s (1968:122) claim that:

‘idioms are semantic units like other semantic units, but they require conversion into
arrangements of other semantic units before they are further encoded into sound’.

Chafe also explains the transformational defectiveness of idioms through this generative
semantics approach. Inherent in his theory is the feature of blockage of certain transformations,
however, this is exactly the point at which his argumentation falls short of valid generalizations. For
example, the idiom kick the bucket cannot be passivized since it is regarded as a single semantic
unit much like ‘die’, which cannot be passivized because it is an intransitive verb. Again, we cannot
inflect parts of the literalization of this idiom, for those parts are not present at the semantic stage
where sentences are generated. For instance, if we pluralize bucket we do something that is not
possible with the idiom, for ‘buckets’ is not present semantically when the idiom is present. For the
same reason, we cannot modify bucket with an adjective: ‘John kicked the blue bucket’ carries only
the literal meaning. In contrast, Ernst (1981) underlines the function of adjectives as stylistic
devices that can be used to modify the meaning of an idiom either externally (i.e. the idiom as a
whole) or internally (i.e. only part of the idiom), and thus, specify the domain the idiom is to apply
to (e.g. politics, social life, emotions) although his discussion is equally valid for all types of fixed
 16

expressions irrespective of ‘…….whether their semantics is compositional or not’ (Ernst,

1981:53). As for nominalization, ‘John’s kicking of the bucket’ does not carry the idiomatic
meaning because only the verb kick is nominalized, and this verb is semantically not there while
‘John’s kicking the bucket’ is acceptable as an idiomatic phrase since nominalization here applies to
an entire verb plus object construction.
According to Chafe, the generative semantic model can further explain the fourth peculiarity of
idioms mentioned at the outset of his paper with respect to the typically greater text frequency of an
idiom with that of its literal counterpart. Following his analysis therefore, the semantic units
identifiable as idioms historically arise because their meaning are useful or cogent to the speakers of
the language, so in this sense, it might be expected that they would occur rather frequently. In
addition, if the frequency of occurrence of all semantic units were the same, simple units (like
idioms) would be expected to occur more often than complex combinations of units (like
literalizations), just as ‘m’ occurs more often than the syllable ‘met’. Moon (1994:152) also
demonstrates that ‘rock’ and ‘boat’ in the idiom to rock the boat co-occur 32 times within a 6-word
window in the 18-million-word corpus (OHPC): 3 occurrences are literal while 29 metaphorical
(idiomatic). This observation could be a further indication that language speakers avoid using fixed
expressions literally in free texts: the idiomatic meaning always blocks the literal one. In this
respect, Chafe (1968:123) points out:

‘The usually greater frequency of idioms with respect to their literal counterparts can be
seen as a product of both these factors: their intrinsically high frequency because of their
semantic cogency, and their unitary character in contrast to the complexity of their
literalizations.’

Thus, in Chafe’s analysis, idioms are treated as anomalies in the traditional TG paradigm due to
the four features they exhibit as linguistic entities, i.e. non-compositionality, transformational
defectiveness, ungrammaticality and frequency asymmetry. Since idiomaticity cannot be accounted
for within that part of the Chomskyan paradigm that considered syntax to be central, Chafe adopts
an alternative model based on generative semantics where grammatical structures are generated by
the semantic component and these structures are further subject to conversion into phonetic
structures. However, the most immediate problem with his formulation is that it has to be rephrased
into more standard terminology before it can easily be compared with alternative theories, as we
observed before. Though Chafe’s proposal suggests the mapping of the meaning of the idiom
directly onto the meaning of its literal equivalent via the process of literalization, his approach
seems far too sketchy for adequate evaluation: lexical insertion is not discussed nor is the
mechanism by which proliteralized semantic units affect the operation of the transformational rules.
Despite the fact that his analysis is carried out along semanticist lines, Chafe seems not to realize
 17

the possibility that compositionality in the case of idioms might be a matter of degree, as mentioned
in later studies.

2.2.3. Uriel Weinreich (1969)

The initial attempt to deal with idioms within transformational-generative theory was
supplemented by Uriel Weinreich’s work, originally based on a series of lectures held at the
Linguistic Institute of the Linguistic Society of America in 1966. His definition of the idiom is
largely influenced by the school of Soviet Phraseology and reads as follows:

‘A phraseological unit that involves at least two polysemous constituents, and in

which there is a reciprocal contextual selection of subsenses, will be called an
idiom. Thus some phraseological units are idioms; others are not.’
(Weinreich, 1969:42)

In an effort to specify the features that differentiate idioms from the wider set of phraseological
units, Weinreich draws a distinction between the ‘idiomaticity of expressions’ and the ‘stability of
collocations’ (e.g. assets and liabilities, Latin and Greek). To Weinreich (1969:71), such a
distinction lies in that both collocations and idioms reflect a co-occurrence phenomenon but that the
co-occurrence of words in an idiom results in a special semantic relationship not evident in the
former. Thus, in contrast to collocations that are merely stable and familiar, idioms acquire their
idiomatic and phraseological status due to the potential ambiguity that results from the occurrence
of homonymous literal counterparts of each of the constituent components of the idiom in other
discoursal environments beyond the idiom itself. In this respect, Weinreich defines idioms along a
two-dimensional axis treating ‘pure’ idioms as lexically fixed sequences with non-compositional
meaning that have literal counterparts, much like the class of what Fernando (1996) calls semi-
literal idioms.

Weinreich (1969:40-41) sets three variables which we may find helpful in better understanding
his fairly rigorous definition of idioms. As summarized by Bressan (1979:34), these variables are:
(i) Amount of overlap between subsenses, or ratio of shared to unshared components. A word like
red can have more than one meaning, and if we take the meaning we find in red hair, clearly this is
a subsense of red but is not entirely divorced from it- we are still talking about two colours though
these are not identical. If we proceed in this way, we find that the overlap between subsenses will
get wider or narrower depending on the context and the particular words we use. (ii) When a
polysemous morpheme appears in a construction, the construction is not correspondingly
 18

polysemous (only one of the subsenses is realized). If we take, e.g. row we can imagine two
contexts, one in which we assign to it the meaning of paddle and the other in which we assign to it
the meaning of series, range. Clearly, the two contexts share no semantic components, and are
mutually suppletive. On the other hand, we may have a word like father versus to father. These are
mutually suppletive syntactically in that one is a noun and one is a verb, yet they do not share a
certain amount of the semantic component. (iii) Two-directional selection. In blind date, blind is
used in a subsense other than the main sense (unseeing), and the same applies to date. The fact that
this phrase only occurs in that particular idiomatic meaning must mean that there has been a two-
way subsense selection.

Bearing in mind these three variables, Weinreich claims that the highest degree of idiomaticity is
registered when all three variables have limiting values, i.e. (i) The subsenses of the morphemes are
suppletive, (ii) Selection is determined by a unique contextual morpheme and (iii) Contextual
selection works both ways. According to Weinreich (1969:38), these three properties conspire to
destroy any natural isomorphism between the syntactic organization of an idiomatic expression and
its semantic analysis, its paraphrase. For the sake of the argument, Weinreich gives the example of
red herring = ‘phony issue’, which could also mean ‘blood-colored herring’ or ‘smoked herring’
ascribing a subsense ‘phony’ to the adjective red and a subsense ‘issue’ to the noun herring. The
two senses of red are mutually suppletive, its sense ‘phony’ is selected only in co-occurrence with
herring, and the coupling of the two allows the idiomatic meaning of the phrase. In this sense then,
idiomaticity is defined as uniqueness of subsense and ambiguity is an essential characteristic of true
idioms. Based on this definition, Weinreich excludes the group of idioms with unique constituents
or ‘pseudo-idioms according to Makkai (1972) such as spic and span, kith and kin, hem and haw as
the unique occurrence of, say spic in the phrase spic and span guarantees that spic has only one
subsense found only in this phrase and nowhere else.

With respect to the semantics of idioms, the author asserts that the meaning of an idiom cannot
always be accounted for as a compositional function of the meanings its parts have when they are
not parts of idioms. To clarify this Weinreich (1969:26-31) argues that a phrase consisting of two
expressions (A, B), whose meanings are (a, b) respectively, when they are parts of this phrase,
would not be regarded as idiomatic if it was the result of formula (1) because its meaning would be
the summation of the meanings of its parts (a, b); in contrast, the phrase is idiomatic if it was the
result of formula (2) since its meaning (X) is neither a sense of its constituent (C) nor of the
constituent (D).
 19

In (1), the
t construcction of A/aa with B/b yields
y a pho
onemic sequuence A + B and a sensse that is inn
some way a function of the ingrredient sensses, ‘a’ and ‘b’. In (2), the obtaineed sense ‘x
x’ is not thee
expected function
fu of the
t senses ‘cc’ and ‘d’. The
T cumulaative sense is
i some kinnd of functio
on (f) of thee
componentt senses, buut hardly a sequence, ass shown in (3)
( and (4):

Alternattively, the non-compoositionality of idioms is

i equally explained
e thhrough the devices off
componenttial analysiss in which the
t net diffeerences betw
ween the exxpected α annd the obtain
ned sense y
in factoriall terms can be represennted. Suppoose that α consists
c of components
c s α1, α2, and
d α3 relatedd
by semantiic function f1.
f

α = f1 (α
α1, α2, α3)
y = f1 (α
α1, α2)
y´= f1 (α
α1, α2, α3, α4)
y’’= f1 (α
( 1, α5, α6)

Sense ‘y’ may difffer from seense ‘a’ by just lackin

ng a compoonent, or byy having an
n additionall
componentt, or by diiffering by one or more
m compo
onent. For instance,
i w
when we co
ompare thee
‘absolute’ sense of cooat, ‘a coveering of som
me kind’, with
w its sense in the exppression co
oat of paint,,
we may saay that that in
i the latterr some speciifying comp
ponent is laacking and oonly those components
c s
that renderr the generaal meaning of
o ‘coveringg’ are presen
nt. Again, in comparinng ‘red’ in general
g withh
‘red’ in ‘reed hair’, wee may say thhat a partly different seet of components is reaalised. In su
um then, thee
idiomatic sense of a complex exxpression may
m differ from its litteral sense either in th
he semanticc
function orr in the sem
mantic constituents. Thee differencee between exxpected andd obtained constituents
c s
may amounnt to a supppression of some
s compoonent of meeaning, or thhe addition of some com
mponent orr
 20

a replacem
ment of com
mponents. However,
H moost English
h adjective + noun lexeemic idiomss cannot bee
assigned such subsennses as one may find in the casee of red herring. For instance, th
here are noo
ordinary suubsenses off ‘hot’ andd ‘dog’ as of
o ‘red’ and
d ‘herring’ amountingg to ‘phony
y issue’. Too
Weinreich thus, an iddiom is a poolymorphem
mic unit wh
hereas monoomorphemicc morphem
mes may nott
be called iddioms.

Weinreiich also adddressed the issue of syyntactic defeectiveness of

o idioms w
within TG trradition andd
his proposaal, essentiallly consists an extensioon of Katz and
a Postal’ss (1963) anaalysis of idiioms. Thus,,
in a similaar line, Weiinreich (19669:57) propposes the in
nclusion witthin a dictionary of an
n idiom listt
where eachh entry is a string of morphemes,
m which may range from
m two morphhemes to a sentence inn
length, witth its associated phrasee marker annd a sense descriptionn. The entryy contains, in
i addition,,
contextual features annd instructiions for obligatory or prohibited transformaations. In th
he examplee
below, for the idiom shoot
s the breeze, the contextual
c features
f speecify an anim
mate plural noun for a
subject andd prohibit thhe presencee of the passive morph
heme; the enntry is also marked as prohibitingg
of-nominallization, sinnce there is no *the booys’ shootin
ng of the brreeze in thee idiomatic sense. Thee
idiomatic sense
s is parraphrased as
a ‘chat idlyy’. By inclu
uding for eaach idiom inn the idiom
m list of thee
dictionary a specificaation of its transformaational pecu
uliarities ‘soo that prohhibited transsformationss
would be blocked
b andd obligatoryy ones woulld be applieed’, Weinreeich attemptts to fill in the gaps off
Katz and Postal’s
P moddel and accoount for thee transformaational behaaviour of idiioms in mucch the samee
way that Fraser (19700) did with his
h frozenneess hierarchy
y.

Figu
ure 2.3. Anaalysis of the idiom ‘sh
hoot the breeeze’ accorrding to Weeinreich (19
969: 54)

As an addition
a to Katz and Postal’s
P forrmalization,, Weinreichh (1969:59)) introducess an ‘idiom
m
comparisonn rule’ whiich operatees on a term
minal string
g before it has enteredd the transfformationall
componentt or the sem
mantic proceess. Such ruule matchess a terminall string agaiinst the idio
om list. If itt
finds an enntry in the idiom list which
w is ideentical with
h all or partt of the term
minal string
g, the idiom
m
comparisonn rule delettes the sem
mantic featurres of the matched
m fraagment of thhe terminall string andd
substitutes the semanttic features and transfoormational instructionss specified ffor the matcching entryy
 21

in the idioom list. Thee idiom com

mparison ruule is option
nal and this optionalityy accounts for the factt
that each idiomatic
i e
expression h a homoophonous liiteral counterpart. Thee diagram th
has hat followss
presents thhe function of
o the idiom
m comparisoon rule scheematically:

Figure 2.4. Idiom Com

mparison Rule (Weinrreich, 19699:59)

Weinreiich also reaalized the im

mportance of
o the probleem created by the fact that Katz and
a Postal’ss
scheme didd not accouunt for idiom
ms which arre syntactically ill-form
med. His solution in th
his case wass
to store suuch items inn the dictionnary as whooles, in the same way that any noormal lexicaal items aree
stored. Thuus, ‘by and large’ or ‘kkingdom com
me’ would be unitary dictionary
d eentries just like
l ‘hit’ orr
‘boy’. Likke Katz andd Postal, he
h considerred and rejected this explanationn for idiom
ms that aree
syntacticallly well-form
med. One reason
r is thhat such a trreatment foor idioms w
with literal counterparts
c s
produces lexical
l entrries which are phonoologically anomalous
a and whichh may com
mplicate thee
operation of
o the phonoological com
mponent coonsiderably. Phonologiccally these iidioms musst be treatedd
as syntactic phrases and
a it seemss unsatisfacttory to inclu
ude them inn the lexicon as if they
y were unitss
without thhe syntacticc structure which, from
m the phon
nological perspective, they obvio
ously have..
Furthermorre, idioms evidently contain
c partts which are subject too the same inflectionaal processess
(including irregularitiies) as corrresponding literal item
ms. In this respect, thhe past tensse of ‘Sam
m
kicked the bucket’ is just
j that, annd not ‘Sam
m kick the bucketed’.
b F all thesee reasons, Weinreich’s
For W s
final propoosal went soo far as to inntroduce ‘unnanalyzablee complexess’ along witth normal leexical itemss
into deep structure
s in the usual way
w and theen to postpo
one the operration of maatching with
h the idiom
m
list until affter transforrmations haave been appplied-that is,
i at the levvel of surfaace structuree. From thee
above disccussion, it seems
s that not only does
d this su
uggestion exxaggerate thhe differencce betweenn
‘kick the bucket’
b andd ‘trip the light
l fantasstic’, makin
ng them totaally different phenomeena, it alsoo
splits the dictionary
d innto two unreelated parts at the samee time that it
i attributes to surface structure
s ann
unprecedennted semanntic importaance. Thouggh Weinreicch’s analysiis is certainnly an impro
ovement off
Katz and Postal’s theory, still it becomes obvious th
hat the prooblem of syyntactically
y ill-formedd
idioms cannnot be unifformly accoommodated within the TG paradiggm as they aare never geenerated byy
 22

the base component. On these grounds, Weinreich defines the idiom in such a way that syntactically
ill-formed concatenations are excluded which are therefore just listed in the lexical-item part of the
dictionary as any other lexeme is.

In sum, for Weinreich (1969) idioms are represented as structured strings of lexical items, with
their own unique syntactic, semantic and phonological properties. His version of the lexical
integrity analysis has idioms stored in an ‘idiom list’ and each entry in it is a string of morphemes
which may be from two morphemes to a sentence in length, with its associated phrase marker and a
sense description. An idiom is viewed as a polysemous lexical unit whose ambiguousness is
dissolved by the reciprocal contextual selection of the subsenses of its constituent parts that
determines their idiomatic or otherwise usage in a specific phrase. The meaning of an idiom cannot
be accounted for as a compositional function of the meanings of its component parts while
syntactically, idioms are entries in an idiom list marked with their transformational properties in the
same way that Fraser (1970) attempted to show that idioms behave like a single unit by allowing
idiomatic lexical entries to consist of strings of complex symbols, with some properties common to
the entire string. Weinreich’s treatment of phrasal idioms as fixed polymorphemic, polysemous
units that abide by the criterion semantic decompositionality is also incorporated in our working
definition of idiom as it is used for the purposes of the present study.

2.2.4. Bruce Fraser (1970)

In a most insightful treatment that enriched the Transformational-Generative (TG) view of

idioms, Fraser’s 1970 paper attempts to defend and restore the status of the theoretical construct of
TG theory in terms of its descriptive and explanatory power in the analysis of idioms (cf. Chafe,
1968). In parallel to Weinreich’s (1969) analysis and adopting Katz and Postal’s (1963) subdivision
into lexical and phrasal idioms, most of the discussion constitutes an effort to offer a theoretically
viable explanation within the framework of TG theory revolving around the issues of the semantic
representation of phrasal idioms in the deep structure of a sentence as well as of their prominent
recalcitrance in terms of particular syntactic transformations.

His definition of the idiom was:

‘a constituent or a series of constituents for which the semantic interpretation is not a

compositional function of the formatives of which it is composed.’ (Fraser, 1970:22)

Based on this definition, idiom is treated and represented in the deep structure as a complex symbol
consisting of three component parts, i.e.(i) a set of insertion restrictions, (ii) a set of syntactic
 23

features which dominates a phonological representation and (iii) a set of semantic markers. For
instance, the lexical entry for the idiom hit the sack will consist of (1) insertion restrictions
specifying that the entry must occur after an adult human subject noun phrase and possibly before
certain types of adverbials (e.g. at 5 p.m.); (2) three complex symbols, the first containing the
syntactic feature [+V] and the phonemic string for hit, the second containing the syntactic feature
[+DET] and the phonemic string for the, and the third containing the syntactic feature [+N] and the
phonemic string for sack; and (3) a set of semantic markers which has the semantic reading ‘go to
bed’. According to this analysis, the representation of the semantic interpretation of an idiom takes
place once the insertion restrictions have been satisfied. Then, each successive complex symbol of
the lexical entry must be determined to be compatible with the corresponding complex symbol of
the base P-marker until the entire idiom is inserted in a single operation and its semantic reading
associated with the lowest constituent dominating the idiom, in the example above, it is the verb
phrase, VP, assuming it dominates the verb and direct object noun phrase of a sentence. While the
above analysis applies to idioms having a literal counterpart, Fraser notes (1970:31) that the same
process cannot be applied in cases where idioms do not have literal counterparts such as noun-noun
compounds (e.g. cheese soup), discontinuous idioms (e.g. bring something to light), verb-noun
phrase idioms (e.g. kill the goose that lays the golden egg) as well as grammatically ill-formed
idioms (e.g. trip the light fantastic). In these cases, idioms are analyzed as having a deep structure
analogous to an expression which resembles the idiom in its surface structure representation.

However, the most innovative feature in Fraser’s discussion of idioms lies in the fact that he
dismisses the indices that mark transformation blockages as insufficient to allow any clear
distinction between governed rules, i.e. rules such as topicalization, question formation, right
dislocation, adverb insertion, pronominalization (see also Alford, 1971) and particle movement that
idioms do not undergo uniformly and ungoverned rules in favour of a hierarchy of transformational
frozenness. According to Schenk (1995:253) this deficient syntactic behaviour of idioms is mostly
attributed to the reluctance of the some idiom parts to undergo certain syntactic operations follows
from the fact that a compound idiomatic expression corresponds to one primitive meaning
expression and therefore, proper parts of idioms do not carry meaning. Thus, not all idioms undergo
one particular operation normally applied to meaningful expressions. For instance, kick the bucket
does not passivize; and blow off some steam does not undergo the particle movement rule, the
extraction operation or passivization (cf. Ross, 1970).
 24

Table 2.1. Transformational Defectiveness of Idioms with respect to governed rules

e.g. *the beans, Margaret spilled. [Topicalization]

*What did Henry shoot? ANSWER: the bull. [Question Formation]
*You are building them, castles in the air. [Right dislocation]
*He kicked today the bucket. [Adverb insertion]
*Mary pulled my leg and Henry pulled it. [Pronominalization]
*Having lost, you should throw the sponge in. [Particle movement]

Therefore, Fraser (1970:39-41) proposed that idioms could be organized into a ‘frozenness
hierarchy’ with five types of transformations ranging from expressions which undergo nearly all
traditional transformations without losing their idiomatic meanings (e.g. throw in the towel) to those
expressions which will not undergo even the most simple transformations and still retain their
idiomatic interpretations (e.g. to face the music). According to Fraser’s perspective, idioms are
classified on seven levels of frozenness indicating transformational deficiency of the following sort:

Table 2.2. Fraser’s Frozeness Hierarchy (adapted from Fraser, 1970: 39)
L6- Unrestricted The group here admits all sorts of transformation.
L5- Reconstitution Idioms in this group do not allow a change in their syntactic function, e.g. action
nominalization transformation. [He spilled the beans*His spilling of the
beans].
to lay down the law, to let the cat out of the bag, to pop the question, to spill the
beans, to throw in the sponge.
L4- Extraction Idioms in this group do not allow one of their constituents to be shifted, e.g.
particle
movement transformation. [He paid attention to the discussion*He paid
attention the discussion to].
to break the ice, to make use of, to add up to, to close up, to ask for, to pay
attention to,
to wait on.
L3- Permutation Idioms in this group do not allow two of their consecutive constituents to be
permuted, e.g.
the yes-no question transformation. [The cat has your tongue*Has the cat your
tongue?]
to keep up one’s end, to put on some weight, to teach new tricks to an old dog ,
the cat has someone’s tongue, to turn back the clock.
L2- Insertion Idioms in this group do not allow insertion of a constituent not belonging to the
idiom, e.g. indirect object movement transformation. [Barbara gave Nigel
hell*Barbara gave hell to Nigel].
to bear witness to, to give chase to, to give the back of one’s hand to, to lend
hand to, to pay homage to, to fish for, to harp on, to run into, to stick to.
L1- Adjunction Idioms in this group do not allow adjunction of a constituent not belonging to
the idiom, e.g. gerundive nominalization transformation. [Suzanne burned the
candle at both ends*Suzanne’s burning the candle at both ends].
to kick the bucket, to care for (children), to stand for, to burn the candle at both
ends, to catch fire, to dance up a storm, to give birth to, to give ear to, to stir up
trouble, to turn over a new leaf.
L0- Completely Frozen This group of idioms admits no transformation at all, such as: to bleed one white,
to build castles in the air, to face the music, to get up one’s energy, to let off
some steam, to be on pins and needles, to turn a deaf ear to.
 25

Followed from bottom to top, this hierarchy reflects an increasing degree of distortion permitted
to the basic idiom shape of an untransformed idiom. In this respect, literally uninterpretable idioms
such as ‘trip the light fantastic’ belong to level L0 as they cannot undergo any operation and are
therefore characterized as the most frozen type of idiom. Interestingly enough, Cutler (1982)
tentatively claims that the syntactic frozenness of such idioms can be attributed to their use in their
idiomatic form in language for the longest time. On the other hand, those idioms that are least
frozen and belong to the level L6 permit considerable alteration. However, any idiom belonging to
one level is automatically marked as belonging to any lower level. For instance, the idiom to pass
the buck to is analyzed as belonging to level L5 so that any reconstitution operation can apply (e.g.,
nominalization) but also any other operations lower in the hierarchy are also acceptable in this
idiom. That way, the extraction operation (passive transformation) and insertion operation (indirect
object movement) can equally apply.

It becomes clear by now that in the context of generative grammar, idioms are viewed as an
anomaly of the system with TG grammarians’ efforts being focused on an analysis of the
transformational behaviour of idioms. However, this insight cannot function as a definitive criterion
of idiomaticity since all idioms exhibit transformational deficiencies and, yet, transformational
deficiencies are also a feature of non-idioms (Mittwoch, 1971). Seen under this perspective,
Fraser’s frozenness hierarchy was often subjected to heavy criticism, as it is entirely based on pure
intuition that often yields inconclusive evidence with respect to the transformational potential of
idioms that seriously need further empirical validation. To this extent, Newmeyer (1972) rejects the
hierarchy as a putative explanatory device for the syntactic behaviour of idioms offering a number
of counterexamples that exemplify its shallowness on theoretical grounds. To him, although to cast
pearls before swine does not undergo Adverb Preposing and lay down the law does not undergo
Particle Movement, nevertheless, both idioms can passivize-an operation that should be
incompatible with not undergoing the two previously mentioned rules. In the same spirit, Machonis
(1970) challenges the predictive power of the frozenness hierarchy in a linear fashion at least with
respect to the transformations of passivization, particle movement and dative shift adopting
information structures of earlier analyses for marking the syntactic behaviour of individual idiom
entries, precluding any type of hierarchy. In this perspective, though transformational deficiency of
idioms in syntactic terms cannot be used as the single criterion for the linguistic description of
idioms, nevertheless, it will be included in our working definition of idioms as used for the
purposes of the present study.
 26

2.2.5. Frederick Newmeyer (1972; 1974)

In two closely related articles, Newmeyer (1972;1974) attempted to demonstrate that there is far
some kind of regularity to the behaviour of idioms in terms of their syntactic behaviour abandoning
the widely held belief that idioms are entered as unit entries in the lexicon (cf. Katz and Postal,
1963; Weinreich, 1969; Fraser, 1970). Instead, he suggested that idioms be treated as non-units and
that their transformational behaviour is determined by a transderivational constraint that relates their
actual meaning, their literal meaning and the lexical items involved. Newmeyer (1974:327) defines
idioms as:
‘…a constituent or series of constituents for which the semantic interpretation
is not a compositional function of the formatives of which it is composed.’
(Newmeyer 1974:327)

Newmeyer (1974:336) states that the grammar of every language contains an idiom inventory,
which is a list of ordered pairs of semantic representations (M1, M2). Each pair on this list is called
an idiom source and any item in the lexicon may be marked with the designation of one or more
idiom sources. The idiom inventory relates to a transderivational constraint (1) and to a condition on
lexical insertion (2) in the following ways:

‘(1) For each idiom source, an idiom with meaning M1, is derived from its paired
M2, a subpart of some initial P-marker. Rules applying to structures derived from
M2 must obey all the government conditions of both M1 and M2.

(2) In each idiomatic derivation, there may be inserted at shallow structure for
those nodes corresponding to subparts of M2, only those lexical items marked with
the designation of the idiom source whose M1 and M2 have governed the
application of the cyclic rules in that derivation.’ (Newmeyer, 1974:336-7)

The initial P-marker in an idiomatic derivation is the semantic representation of the idiom’s
literal equivalent (M2). However, the actual meaning of the idiom is a different semantic marker P-
marker (M1). Both M1 and M2 affect the derivation throughout the application of the cyclic
governed rules of Passivization, Unspecified Object Deletion, Conjunct Movement, Subject
Raising, Tough Movement and There Insertion. In the lexicon, lexical items which appear in idioms
are marked with the designation of a particular idiom source in addition to their usual features.
Since kick and bucket, for example, appear in the idiom kick the bucket, as part of their entries in the
lexicon we would find a designation of the appropriate idiom source. Such designation would
contain two semantic representations: M1 would be the semantic representation [DIE]; M2 would be
the semantic representation [KICK THE BUCKET]. Because of their idiom source designations, the
lexical items kick and bucket at shallow structure, i.e. the level defined by the completion of the
application of the cyclic transformational rules, would replace the semantic formatives [KICK] and
[BUCKET] in the idiomatic as well as in the literal derivation. In extending Newmeyer’s model in
 27

more comppositional teerms, Greim

m (1983:70)) also suggeests that synntactic as w
well as semaantic factorss
in combinaation with a transparenncy feature in
i terms of iconicity (i.e. a pictoriial representtation of ann
idiom’s figgurative meeaning), facttual congruuity (i.e. thee extent to which
w idiom
m elements point to itss
figurative meaning) and
a imputedd congruity (i.e. the ad
ddition meaaning of an idiom not based
b uponn
the semanttic content of
o its constittuents) deteermine the successful
s appplication oof a transforrmation.

Howeveer, kick the pail is derived with a literal readiing only, siince in the llexicon paill would nott
carry the designation
d t it is parrt of an idioom. The derrivation of literal
that l kick tthe bucket differs
d from
m
that of kick the pail only
o in the choice of lexical item
m inserted at
a shallow sstructure. [K
KICK THE
E
BUCKET]] will thereffore be the initial
i P-maarker of threee verb phraases: kick thhe bucket (lliteral), kickk
the pail (liiteral), and kick
k the buccket (idiom
matic with meaning
m [DIE]). In otheer words, eaach M2 willl
be the initiial P-markeer of at leastt two derivaations: the M2 which iss provided by conditio
ons on whatt
may consttitute a welll-formed semantic reppresentation
n governs the
t ‘normaal’ (i.e. non
n-idiomatic))
derivation and the identical M2 which
w is provvided by the idiom invventory whicch governs, along withh
its paired M1, the idioomatic derivvation. Gram
mmatically ill-formed
i i
idioms suchh as happy go
g lucky, byy
m come unddergo no traansformational rules ass their M2’ss violate conditions onn
and large, to kingdom
well-formeed semanticc representattions despitte the well-fformedness of their M1’s.

The folllowing diaagram schem

matizes thee derivation
n of idiomaatic kick thhe bucket (N
Newmeyer,,
1974:339)
Figurre 2.5. Derivvation of th
he idiom ‘k
kick the buccket’ accorrding to New
wmeyer (19
974: 339)

Contrarry to Weinreeich (1969) and Fraserr (1970) wh

ho treat idiom
ms as structtured string
gs of lexicall
items withh their own unique synntactic, sem
mantic, and phonologiccal propertiies, Newmeeyer (1974))
claims thaat idioms haave an inteegrity only at the leveel of semanntic represeentation thatt are neverr
inserted ass a lexical unit at anyy one pointt in the derrivation. Hiis argumentts against such
s a unitt
 28

treatment are based on the following grounds: (1) Lexical items should replace only constituents.
Nevertheless, in idioms such as pull one’s leg and lead one on a merry chase non-idiomatic and
idiomatic material appear in the same expression. Idioms like it’s all over town that replaces a verb
and its subject (everyone knows), to beat around the bush that replaces a verb and its object (avoid
discussion) and to sit on pins and needles that replaces a verb and adverbial material (wait
anxiously) are even more problematic as they do not replace constituents at all. (2) A unit treatment
does not allow a clear explanation of why well-formed idioms have literal equivalents. (3) Idioms
which range over two S nodes (e.g. Don’t count your chickens before they hatch), where cyclic
rules can apply in the lower S (e.g. Don’t count your chickens before they are hatched) demand to
be treated as two different idioms if we regard idioms as units. (4) A unit treatment does not
account for the morphology of strong verbs occurring in idioms (e.g. *He shooted the breeze, *He
breaked the ice). Since irregular verbal morphology is not semantically determined in English,
verbs in idioms have the same morphological properties as their homophonous literal counterparts.

Consequently, Newmeyer’s concept is that unit treatment should be abandoned introducing

instead, an idiom inventory that consists of ordered pairs of semantic representations (M1, M2). M1
is the meaning of the idiom, whereas M2 is the semantic representation of its literal equivalent. In
the lexicon, all items that appear in idioms are marked. In order to derive an idiom, all lexical items
with the same designation must be inserted. This treatment has several advantages: All idioms are
inserted at shallow structure, and therefore idiomatic causative and idiomatic non-causative verbs
retain their close derivational relationship. If only lexical items are to be inserted, neither
constituency problems nor any morphological problems arise. Literal equivalents are guaranteed by
requiring that all M2 are well-formed while transformational deficiencies can be explained to a large
extent. Even though in his analysis Newmeyer does not consider all the properties of idioms (e.g.
ill-formed idioms, certain transformational deficiencies), he has made it clear that:

‘in general there seems to be much more regularity in each language’s idiom
stock than has been previously believed.’ (Newmeyer, 1972:301)

In sum, we notice that the useful contribution made by Transformational-Generative

grammarians to the study of idioms lies in their focus on the transformational behaviour of idioms.
In the context of generative grammar, idioms are defined as that phenomenon according to which a
string of several words is used to express a notion that is not analysable and distributable over the
different words of the string and is seen as an anomaly of the system, since some of the slots have to
be filled in by several words instead of one: for example, kick the bucket as opposed to ‘die’. Still
the insight of idioms’ transformational behaviour provided by the theory cannot function as a
definitive criterion of idiomaticity. All idioms have transformational deficiencies and, on the other
 29

hand, transformational deficiencies are also a feature of non-idioms as well. The deficiency of
Transformational Generative (TG) Grammar is also acknowledged by Coulmas (1979a; 1979b) as
well as by Makkai (1987) in some of his later writings on idiomaticity, who advocate instead a
pragmatic treatment of idioms as pervading features of everyday discourse with their own
functional and structural significance in a given speech community. Coulmas (1979a:240)
characteristically points out two reasons behind the conceptual weakness of the TG theory in
relation to its inability to explain idiomaticity:

(A)…..idiomaticity destroys the neat conceptual complementarity of competence

and performance. It is not clear whether the filters admitting some expressions as
idiomatic and rejecting others should be assigned to competence or performance.
To look at idiomaticity as a mere performance phenomenon and thereby
dismissing it as from ‘linguistics proper’ is certainly no satisfactory solution.

(B)…with idioms, the Fregean principle loses its validity. According to the
Fregean principle, the meaning of every complex sequence follows from the
lexical meanings of the individual parts. This basic semantic truth, however, has
no bearing on idioms, and hence it is not clear how we should provide for a
semantic description of idioms unless we decide to rid ourselves of this problem
by putting all of them as unanalyzed units into the lexicon.

In this light, idioms are not considered a nuisance but rather a challenge in pragmatic terms as they
stand where the freedom, as well as the regularity, of syntax ends, and where the semantic principle
of compositional meaning fails. Idioms are viewed as an important part of ordinary everyday
conversation and their status as fixed expressions is part of the common knowledge governing
verbal behaviour (see also section 2.2.4. for pragmatic approaches to idioms).
 30

Chapter 3
The Semantic Approach
3.1. Adam Makkai (1972)
On this level of theoretical study, idioms have often been associated with irregularity and
frozenness leading linguists to consider them as obstacles in the way of constructing an optimal
systematic linguistic model. Actually, their definitions of the idiom cannot be said to capture the
phenomenon of idioms in its totality as they define it mono-dimensionally using syntactic properties
as their only critical criterion. Working within the model of stratificational grammar, Makkai’s
(1972) study on idiom is still considered from a descriptive point of view the most comprehensive
treatment of the subject in English. Following the principles of Soviet phraseology as well as the
work of pioneer British English lexicographers such as Healey (1958) among others, Makkai
advocates that the idiom structure of every language is divided into two distinct idiomaticity area,
the ‘lexemic’ and the ‘sememic’, the identification of which constitutes the core theme of his thesis
(see also Bolinger, 1976 for a short review of the book).

In identifying idioms, Makkai (1972:58) argues that it is not only more economical but also more
insightful to use the term idiom only for units realized by at least two morphemes and defines
idioms as ‘polylexonic lexemes made up of more than one minimal free form or word’ (1972:122).
Requiring an idiom to have at least two independent lexical items excludes expressions consisting
one free form and one or more bound forms added by affixation as the grammar provides adequate
decoding rules for such types. Thus, as Makkai argues, the suffix –wards has the meaning ‘in the
direction of’ in forwards and backwards. Noun + wards has a predictable meaning, hence all new
forms created on this pattern are similarly interpretable, though many compounds and derivatives
(e.g. pineapple, hospitality) cannot be equally interpreted from a compositional point of view. By
contrast, the meaning of an idiom is not predictable from its component parts, which are empty of
their usual senses. However, Makkai fails to use his definition of idiom consistently as he includes
instance of one-word compound idioms in his classification (e.g. eavesdrop, blackmail) while at the
same time his definition as it stands is unclear with respect to the features he considers important
for the circumscription of the term idiom. For the reason, Makkai (1972:58) proceeds to the
advancement of four specific criteria for the identification of idioms, namely: (i) the presence of at
least two free morphemes in a given expression, (ii) the ability of these morphemes to function with
different meanings in more than one environment, (iii) the potential ambiguity of all idioms of
decoding arising from the possibility of literal interpretation and (iv) the semantic unpredictability
 31

of idioms arising from the fact that an idiom has a meaning which cannot be deduced from its
component parts.

In this respect, lexemic idioms, that belong to the first area of idiomaticity, are, therefore,
syntactic units whose meanings are not predictable from their syntactic constituents (i.e. free
morphemes) and capable of functioning with a literal meaning in a different environment. Due to
this distribution in other environments of each lexon in other environments, lexemic idioms are
subject to ‘disinformational potential’ that arises from a possible lack of understanding despite
familiarity with the meanings of the components that can even lead to erroneous decoding which
can potentially mislead the uninformed listener. To Makkai, this disinformation is a different form
of ‘misunderstanding’ than the kind of encountered in the case of homonyms that involve
misinformation. Disinformation occurs when the structural composition of the utterance in which
the idiom was heard-or read, in our case-allows the listener/reader to decode the idiom in a logical,
yet sememically erroneous way. By contrast, misinformation does not occur as a result of an act of
logical-but literal, and therefore wrong- decoding of complex polylexonic lexemes on the part of the
listener/reader, but as a result of accidentally homophonous forms occurring in similar
environments with equally meaningful decodings. For example, she bears children, ‘she carries
them’ versus she bears children, ‘she gives them birth’ involves a potential ‘misunderstanding’
based on misinformation, but not on disinformation. Hence it is not idiomatic. In Makkai’s
(1969:51) words then, the definition of the lexemic idiom reads as follows:

‘Any polylexonic lexeme which is made up entirely of lexons all of which occur in other
environments as realizations of lexemes which are simultaneously morphotactically
separate words, is a LEXEMIC IDIOM. E.g. put up with ‘tolerate’, undergo ‘suffer,
experience’, hammer and tongs ‘violently’. Consequently, polylexonic lexemes one or
more of whose constituent lexons, in spite of being morphotactically separate words, are
not simultaneously realizations of independent lexemes in other environments must be
termed PSEUDO-IDIOMS. Examples include the famous cranberry, kith and kin, to and fro,
spic and span, and so on. These are not true idioms because they cannot really mislead
you; all they can do is lead you up a blind alley: There are no sememes in the language
today behind the lexons cran, kith, fro and spic as viewed outside of the familiar
binomials. Form such as these were called ‘cranberry morphs’ by the late Bernard Bloch’.

Seen under this light then, Makkai in fact separates homonymity and idiomaticity in his
misinformation/disinformation distinction as he neglects to explore the implications of
disinformation for idiomaticity in any real depth simply because he fails to relate it to homonymity
(cf. Fernando et al, 1981). To our view, disinformational potential is considered to be the key
feature of idiom that is attributed to the existence of a literal homonymous counterpart which
complements the fact that its syntax is non-correlative, i.e. a true idiom has both a literal and a non-
 32

literal ‘face’ and as such comprises a ‘double exposure’. Thus, the fact that idioms such as kick the
bucket meaning ‘die’ has a homonymous literal counterpart ‘to strike the pail with one’s foot’
reinforces the fact that their meaning is not the result of the compositional function of their
individual syntactic constituents and can subsequently be misleading in certain texts. The relation
between homonymity and disinformation in the case of idioms has also been touched on by
Weinreich (1969: 43-44) who argues that in terms of normal colloquial language use we would not
use the phrase bacon and eggs for packaged bacon and a carton of eggs although this is not an
impossibility in terms of the referents denoted by the phrase. Though the legitimacy of such
possibility constitutes for Weinreich, as it does for Makkai, one of the essential features of the pure
idiom, Weinreich fails to carry this criterion far enough, as in the case of by heart, red herring and
rain cats and dogs while Makkai could have explained more fully the usefulness of this feature.

Accordingly, Makkai (1969) identifies six types of lexemic idioms. These are: (i) Phrasal verbs
(make up, turn out, bring up), (ii) Tournure idioms, i.e. polylexonic lexemes of a larger size-level
than a phrasal verb insofar as it consists of at least three lexons that are lexemes elsewhere (fly off
the handle, rain cats and dogs, kick the bucket, have it out with somebody, be well-off). Tournures
are further sub-classified into nine groups according to the following principles: (a) the compulsory
nonrepresentative IT in the idiom (e.g. to have it out (with)) or at the end of it (e.g. to come off it),
or the compulsory nonrepresentative definite article (e.g. to fly off THE handle) and indefinite
article (e.g. to pull A fast one) or neither of those (e.g. to cash in one’s chips), (b) the verb of the
idiom is either followed by an irreversible binomial (e.g. to rain cats and dogs) or the idiom
containing an irreversible binomial starts with a preposition and not a verb (to all intents and
purposes) or , (c) a direct object is not involved (e.g. to dance on air) and (d) the idiom is headed by
the verb be (e.g. to be up a creek).(iii) Irreversible binomials (e.g. pepper and salt, coffee and
cream). Based on Malkiel (1959), the class of irreversible binomials, or ‘co-ordinated idioms’
according to Austin et al (1988), is further categorized into three major groups: (1)
Morphotactically irreversible idiomatic binomials (e.g. rough and tumble, nip and tuck), (2)
Morphotactically reversible idiomatic binomials that become literal constitutes after the reversal
(e.g. aches and pains, Adam and Eve) and (3) Morphotactically reversible polylexonic binomials
with resulting loss of institutionality (e.g. judge and jury, King and Queen) (iv) Phrasal compounds
(e.g. hot dog, blackmail, high-handed), (v) Incorporating verb (e.g. eavesdrop, manhandle), i.e.
complex lexemes whose first lexon is a noun or an adjective in other environments and literal re-
encoding of them reveals a related structure where the verb leads the construction which is either
followed by a direct object and/or an appropriate choice of prepositional phrase. To eavesdrop
‘surreptitiously to overhear’ is an idiom, but the corresponding literal structure to drop eaves or to
 33

drop (something) from the eaves has no semantic connection with ‘to overhear’. (iv) Pseudo idioms
(e.g. kith and kin, spick and span), i.e. lexemic idioms one of whose constitutes is a cranberry
morph that cannot be simultaneously a realization of an independent lexeme in other environment
and therefore, lacking in the disinformation potential of pure idioms.

Sememic idioms constitute the second major idiomaticity area identified by Makkai (1969:54)
and defined as ‘polylexonic constructions whose aggregate literal meaning functions additionally as
the realization of another combination of meaning units’. Such idioms correlate with the basic
sememes of institutionalized cultural-pragmemic awareness with institutionalization being the fifth
criterion of idiomaticity advanced by Makkai. Into the area of sememic idiomaticity fall (i)
Proverbs (e.g. don’t count your chickens before they’re hatched), (ii) Familiar quotations (e.g.
brevity is the soul of wit), (iii) ‘First base’ idioms associated with a national game like American
baseball (e.g. have two strikes against one), (iv) Idioms of institutionalized politeness (e.g. May
I…X?), (v) Idioms of institutionalized greeting (e.g. How do you do?), (vi) Institutionalized
understatement (e.g. I wasn’t too crazy about him) and (vii) Institutionalized hyperbole (e.g. he
won’t even lift a finger). To Makkai, this classification is well-motivated as it is grounded on the
functional different that exists between the two types of idiom, the lexemic and the sememic:
sememic idioms, in contradistinction to the lexemic type, correlate mainly with institutionalized
culturally pragmatic meanings claiming a more culture-bound and language-specific character as
they are essentially instantiations of interpersonal communication among speakers of a speech
community and could even claim membership to a third idiomaticity area of the hypersememic or
cultural idiom (see also section 2.2.1.above).

Nevertheless, Makkai draws a parallel between lexemic and sememic idioms in respect of the
requirement that constituents of idioms have freedom of occurrence: just as the lexons of lexemic
idioms occur in other syntactic environments so do the constituent lexons of sememic idioms
having thus as common denominator the feature of disinformation potential or erroneous decoding
on the part of the listener/reader, hence their characterization as idioms of decoding. Idioms of
encoding, on the other hand, are characterized as phraseological idioms and involve mainly
collocational restrictions such as for instance the use of at in he drove at 70 m.p.h. that do not obey
either to the semantic decomposition principle or to the disinformation potential of decoding
idioms. In this respect, Makkai’s distinction between idioms of decoding and idioms of encoding,
although quite dichotomous allowing for no fine gradations, seems to be in parallel with
Weinreich’s discussion of idiomatic expressions and stable collocations who also highlights
potential ambiguity as the defining characteristic of the former but not the latter.
 34

In sum, the most valuable part of Makkai’s work concerns the identification of idioms on the
basis of the various criteria commented on in the foregoing discussion. The use of the stratification
grammar model makes his study an early example of a highly formal approach to idioms resulting
in useful idiom categorizations and sub-categorizations on either structural or functional differences
setting the basis for thesaurus-like Ecological lexicography (Makkai, 1977). Idiomaticity is viewed
as a language universal (Makkai,1978;1993), a representation of phoricity, whereby signs in human
languages point not only at their references but at other signs as well and through these at a series of
networks of interrelated references as is the case with the phenomena of metaphor, anaphora and
cataphora. With respect to our definition of idiom, we tend to accept and adopt the criteria put forth
by Makkai with the exception of the institutionalization criterion and concentrate more on the
lexemic idiom and especially on the sub-type of tournure, an expanded phrasal (verbal) idiom of
limited morphology and syntactic freedom. These include idioms such kick the bucket, fly off the
handle and bury the hatchet akin to those selected and incorporated in the present study.

3.2. Roger Flavell and Chitra Fernando (1981)

A more developed investigation of idioms was carried out by Fernando and Flavell (1981:48)
who define the pure idiom as:

‘a non-literal set expression whose meaning is not a compositional function of

its syntactic constraints but which always has a homonymous literal
counterpart.’ (Fernando and Flavell, 1981: 48)

In their monograph ‘On Idiom’, Fernando and Flavell (1981:19) acknowledge the complexity of
idiomaticity as a phenomenon and emphasize the necessity of multiple criteria for inclusion in its
definition with each criterion representing a single property. In their view, if idiomaticity is so
defined, certain types of idiom are seen to possess more distinguishing properties than others
establishing thus varying degrees of idiomaticity along a continuum that correlate with different
types or categories of idiom. Based on the belief that the adoption of multiple criteria would enable
the filtering-out of the non-idiomatic while retaining those forms which show one or more of the
properties of idiom, the authors primarily base their definition of idiom in essentially morpho-
syntactic criteria suggesting two scales of idiomaticity by emphasizing that ‘…the key property of
idiom that manifests itself most unequivocally…is the asymmetry between sense and syntax.’
(Fernando et al, 1981:20). Regarding idiomaticity as largely a semantic matter, they consider only
properties that establish the semantic unity of idioms as more important than those that establish
points of grammatical contrast (e.g. transformational constraints) and, so focus on features of three
sorts: semantic, syntactic and sociolinguistic.
 35

From a semantic point of view then, pure idioms are characterized as (i) non-literal in terms of
the referents denoted by these parts since their meaning is not the result of the compositional
function of its constituent parts, (ii) ‘double exposure’ due to an existence of a literal homonymous
counterpart which complements the fact that its syntax is non-correlative and (iii) they usually have
a compulsory literal semantic counterpart which may substitute with it synonymously in a given
context. In this respect, the authors’ efforts concentrate on a synchronic description of idioms from
a semantic point of view in terms of the strength of the link between the idiomatic meaning of a
phrase and the cumulative literal meanings of the correlative elements of the expression. Based on
native speakers’ judgments as part of a project to assess the degree of figurative motivation implicit
in idioms, they suggest a continuum of semantic transparency ranging from totally transparent to
the totally opaque (pure) idiom. Although they admit that the continuum proposed is divided in a
rather arbitrary way into the four categories, as presented in the table below, nevertheless, we think
it is a convenient scheme that allows us to assess the degree of semantic motivation of a given
expression.

Table 3.1. Degrees of semantic motivation according to Fernando and Flavell (1981: 28)

Literal Metaphors

Idioms
(i) (ii) (iii) (iv)
Transpar Semi-transparent Semi-opaque Opaque
ent
cut wood, skate on thin ice, kill two birds burn one’s boats, tarred with the same pull someone’s leg, foxglove,
break with one stone, add fuel to the brush, jump down someone’s throat pass the buck, trip the light
eggs, fire fantastic
a pink
shirt,
bring in,
rely on

In an attempt to define the semantic opacity of idioms as a result of the asymmetry between
syntax and sense, Flavell et al (1981:24) resort to a diachronic explanation of such phrases as kick
the bucket. What happens in this and other expressions of this sort is that while initially the
independent lexemes of the phrase (kick, the, bucket) are used in a straightforward, literal way such
that the resultant meaning of an expression formed with them is simply the sum of its constituent
parts, diachronically these expressions take on new, idiomatic meanings through various semantic
processes and at the same time calcify syntactically. As a result, it is no longer possible to talk of
lexeme + lexeme giving the full meaning, i.e. a polylexonic construction, but only of one single
lexeme, with a sense not obviously connected with that of the original constituent lexemes, i.e. a
monolexemic construction. The weaker the linkage between literal and figurative meaning in an
 36

expression the more opaque the phrase it is representing the class of pure idioms. Conversely, in the
case of phrases situated at the transparent end of the spectrum, the metaphor is clearly perceived
and still alive and kicking. Thus, a distinction is made among four categories of idioms, i.e. (1)
Transparent expressions, which are not idioms, but merely free collocations with a literal meaning
derived from the meanings of the constituent words, (2) Semi-transparent phrases, which can be
regarded as metaphors having a counterpart with a literal meaning, (3) Semi-opaque phrases,
metaphor idioms which are not completely intelligible and (4) Opaque phrases, which are full
idioms whose meaning cannot be derived from the meanings of the component words. This scale of
idiomaticity is not very useful and rather vague in that metaphors can hardly be separated from-
especially in classes 2 and 3-considering that metaphorization is a general property of idiomaticity.

Complementary to the criterion of semantic decompositionality comes the syntactico-semantic

property of homonymity. In parallel to Makkai’s (1972) discussion of idioms’ disinformation
potential and Weinreich’s (1969) criterion of potential ambiguity associated with true idioms,
Flavell et al (1981:33) also claim that homonymity consists of an essential feature of all pure idioms
that signals a point of differentiation between true idioms, set phrases and metaphorical expressions.
On these grounds, two sets of phrases are excluded from the class of pure idioms: (i) constructions
that violate truth conditions and, therefore, have no literal counterparts (e.g. to rain cats and dogs,
to move heaven and earth, to jump down one’s throat) as well as constructions that are
phraseologically peculiar (e.g. one’s cool, to be beside oneself, to put one’s heart and soul into
something) which are all forms of hyperbole or some other figure. (ii) Set phrases such as by heart,
happy-go-lucky, to kingdom come, by and large. This class of items is also deficient in respect of
homonymous literal counterparts and consists of a class of expressions which in addition to having
some figurative force also shows a phraseological peculiarity. (iii) Similes, both explicit and
implicit (e.g. like the pot calling the kettle black, like a bolt from the blue, like the tail wagging the
dog) and metaphors of the form X is Y (e.g. a big fish in a small pond, a fish out of water, a wet
blanket).

In this respect then, the homonymity of pure idioms serve not only to separate them from various
types of semi-idioms (e.g. red herring), which although non-literal and phraseologically peculiar,
they, nevertheless, lack a homonymous literal counterpart, but, also renders the idiom the most
opaque manifestation of figurative language in terms of its individual constituents as well as of the
loss of its origin and original figurative connection to the average user. The fact that smell a rat,
scratch other’s backs have homonymous literal counterparts complements the fact that their
meaning is not the result of the compositional function of their individual syntactic constituents. In
 37

this sense, the non-correlative syntax of the idiom is further emphasised by the fact that it functions
contrastively with its homonymous literal counterpart leading to potential misidentification and
subsequent miscomprehension when encountered in a text. True idioms are also separated from
other polysemous items (e.g. surgical/military operation) and clichés as they tend to have an
institutionalized compulsory literal counterpart which may substitute with it in a given context as
argued also by Makkai (1972:171).

Structurally speaking, an idiom is viewed as a syntactic unit which manifests lexical integrity
and represents a peculiarity of phraseology associated with only one institutionalized idiomatic
encoding out of many possible other encodings (Fernando, 1978:328). In this sense, idioms are
simultaneously idioms of both encoding and decoding that exhibit a certain degree of both lexical
and syntactic variation in their structure. Thus, parts of idioms can be deleted or can also be
modified by items not part of the idiom and sometimes even the order of the idiom reversed in the
service of Wit. In such instances an element of the idiom may be changed, without its original
import being changed at the same time, by substitution, inversion, or deletion.

Substitution Reversal of the order of constituents

Give the green light For the cat to swallow the canary
The ordeal of negotiation is on, the Opera House in Sydney For a while in Hobart this week it looked faintly
has been booked and all that is needed now is the Red light as if the canary had swallowed the cat.
from Peking to go ahead.

Deletion of part of the idiom

Swallow the bitter pill
Jimmy Connors, tennis king of Australia…..has lost his crowns to John Newcombe, Arthur Ashe and yesterday
to Manuel Orantes. A bitter pill for someone who also defeated Australians Rod Laver and John Newcombe in
challenge matches at Las Vengas earlier this year.

Stressing the unproductivity of idioms in syntactic terms, the researchers offer evidence in
favour of the limited flexibility in the idiom syntax associating the application of specific
transformations with the achievement of specific stylistic effects. Substitution, topicalization,
conjunction reduction and pronominalization are normally inadmissible while idioms composed of
adjective + noun (e.g. a hot dog) do not allow predictive usages, nominalizations, the formation of
comparatives or superlatives, or modification. Although the researchers accept Fraser’s (1970)
frozenness hierarchy as a useful means for classifying idioms syntactically, they nevertheless
suggest that it would have greater value for idiomaticity per se if it were possible to establish a
syntax-semantics correlation as a complementary basis for the hierarchy where idioms can be
assigned to different classes throughout the idiomaticity continuum defined along bi-directional
 38

axes, i.e. semantic opacity and syntactic frozenness. Table 2.4.below shows their idiom
classification based on these two structural properties (adapted from Flavell, et al, 1981:54).

Table 3.2. Ranking of Idioms based on their Syntactic Properties (adapted from Flavell et al,
1981: 54)

Set 1 smell a rat Set 4 white lies

twist somebody’s arm
in the dark Set 5 thin excuses
scratch each other’s backs thank goodness
sharing drinks
Set 2 stand down
off one’s hands Set 6 as different as chalk from cheese
grown up
cut down on Set 7 all very poshy
count oneself lucky
Set 3 storm in a tea cup
to be born with a silver spoon in one’s mouth
sitting on a time bomb
dead duck

According to the researchers, the four idioms at the top of the scale are distinguished from the
others in that in addition to showing non-correlative syntax, lexical integrity and institutionalization,
they also have homonymous counterparts that carry a literal signification. Such idioms, therefore,
manifest the highest degree of idiomaticity since they are the most opaque. However, even here we
can make slight distinctions between smell a rat and twist somebody’s arm, in the dark and scratch
each other’s backs, all three of which are less opaque in terms of their constituents than smell a rat.
The second set of idioms can bear a literal interpretation, still they do not have homonymous literal
counterparts in the way the items at the top of the scale do. The items included in the classes 3, 4
and 5 are more obviously figurative since they either violate truth conditions (a storm in a tea cup)
or constitute situational improbabilities (sitting on a time bomb). Such figurative connections can be
worked out on an analogical basis. White lies constitutes an intermediate point between metaphor
and items which have, like white lies, one non-literal component, but not the same degree of lexical
integrity. This excuses, thank goodness and sharing drinks are predictable collocations and while
they share the key feature of non-literalness, they lack fixity. Similes in set 6 lack two of the key
properties of pure idioms, i.e. a non-correlative syntax resulting in non-literalness and the
complementary property of a homonymous literal counterpart and for this reason they are placed
towards the bottom end of the idiomaticity scale. All very poshy and count oneself lucky are not only
literal but also, like thin excuses and thank goodness, predictable collocations rather than fixed
expressions. To Flavell et al (1981:44) then, the essence of idiomaticity does not lie in the syntax of
 39

a language and idioms are at best only a sub-class of all transformationally deficient structures.
Therefore, any attempt to set up syntactic deviance as the primary norm to establish idiomaticity
fails, since it cannot distinguish between idioms and these other transformationally deficient
structures. For as far as our working definition of idiom is concerned, we tend to accept and adopt
this view and focus on strictly semantic criteria in the delimitation of the term ‘idiom’ while limited
lexical and syntactic freedom are viewed as supplementary features that delineate the linguistic
entity of idiom, as used in this study.

Idiom is also defined in sociolinguistic terms as ‘an institutionalized expression, i.e. it is

approved by the usage of the language.’ (Flavell et al, 1981:44). The criterion of institutionalization
is considered to be useful as it separates items which show a peculiarity of encoding and are in
addition non-literal such allusions and clichés and idioms among them from those that are nonce-
idioms or predictable collocations that can be coined and used between two or three individuals for
a very limited period of time. Institutionalization, in terms of currency an idiom has among the
members of a specific speech community or a sub-group of such a community, is believed to be part
of the idiomaticity that reinforces the linguistically arbitrary character of idioms as reflected in the
asymmetry between sense and syntax as discussed above. In contrast to Flavell et al (1981), we
reject institutionalization as a defining characteristic of idiom due mainly to its all-embracing and
non-linguistic character. Since institutionalization can best be regarded in terms of variable degrees
rather than as a fixed phenomenon that means that if included in its definition idiomaticity can
range from non-institutionalization as in the case of a private idiom such as It’s a nice shade of
blue, isn’t it, used by a wife to indicate her husband’s insensitivity to colour through slang terms
(e.g. blood box, thumb merchant) and jargon (e.g. black box, Occam’s Razor) to idioms (e.g. smell a
rat, cross one’s fingers).

In summary then, Flavell and Fernando (1981) acknowledge and stress the fact that idiomaticity
should not be treated as a monolithic homogeneity but, rather as categories that shade off into one
another in actual usage, much like a scale. In this respect, they classify idioms into four categories
of transparent, semi-transparent, semi-opaque and opaque on the grounds of the criteria of non-
correlative syntax, lexical integrity, literal homonymous counterpart and institutionalization. Pure
idioms are the most opaque as they manifest the highest degree of idiomaticity on each one of these
points that represent different properties of the universal phenomenon of idiomaticity. With respect
to our own working definition of idiom, we adopt three out of the four criteria set forth by Flavell
and Fernando (1981) and therefore agree that idioms (a) vary in their degree of semantic
compositionality ranging from semi-transparent to totally opaque, (b) have a literal homonymous
 40

counterpart which, in fact, complements the non-correlative syntax of the idiom, i.e. a true idiom
has both a literal and a non-literal ‘face’ and as such comprises a ‘double exposure’ and (c) is a
syntactic unit that manifests relative lexical integrity as well as transformational potential in its
structure. Institutionalization is not adopted in our definition as a criterion since it does not clearly
distinguish idiom from other phraseological peculiarities.
 41

Chapter 4

The Pragmatic Approach

4.1. Jurg Strassler (1982)
Strassler’s (1982) pragmatic analysis of idioms consists one of the first serious attempts to
investigate the discoursal functions of idioms drawing heavily on the speech-act theory of Searle,
Grice, Sadock and other post-Chomskyan theoreticians. The aims of his study are twofold, i.e. (i)
the identification of idiom functions based on current pragmatic theories as well as (ii) the
identification of the special features of idioms which set them apart from the rest of the English
vocabulary and, which also constitute their raison d’etre. Strassler himself (1982:85) summarizes
the core theme of his thesis as follows:

‘………every idiom has a non-idiomatic synonym on the semantic level. The

question now remains as to why idioms exist and why they can only be used under
certain circumstances….. . I trust I shall find elements within idioms which they do
not share with their literal counterparts.’ (Strassler, 1982:85)

Like Makkai (1972), Strassler (1982:13) cites sense 3a of the Oxford English Dictionary (OED)

of idiom (quoted in section 2.2 above) claiming that it provides the best framework for a

categorization of the different notions of idiomaticity. Nevertheless, he does give his own working

definition of an idiom as:

‘….a concatenation of more than one lexeme whose meaning is not derived from the
meanings of its constituents and which does not consist of a verb plus an adverbial
particle or preposition. The concatenation as such then constitutes a lexeme in its
own right and should be entered as such in the lexicon.’ (Strassler, 1982:79)

As it becomes obvious, Strassler’s definition partly agrees with those offered by Weinreich (1969)
and Makkai (1972) as a complex structures that consists of more than one lexical units whose
meanings do not add up to give the meaning of the idiom as a whole. However, what makes
Strassler’s definition of idiom noteworthy is the exclusion of ‘a verb plus an adverbial particle or
preposition’ he acknowledges that an idiom dictionary such as the ODCIE Vol.1 (1975) is devoted
primarily to expressions of this sort. Such an arbitrary exclusion is made on no other reasons but
practical ones which are, nevertheless not further sufficiently elaborated on. Additionally,
metaphors, similes and collocations are also excluded and treated as individual phenomena quite
apart from idioms themselves based on the fact they have only one deviant element which affects
 42

the semantics of the phrase without, nevertheless, presupposing a non-correlative syntax-sense

structure. Thus, a metaphor does not express a different meaning of a lexeme, but rather represents
the use of a lexeme with some features violated. In the case of for instance parts of the human body
(e.g. leg, foot, hand, tongue) we can use the relevant lexemes in combination with other nouns such
as chairs, tables, shoes. Splitting the literal meanings into their binary features and comparing them
with the transferred (i.e. figurative) meanings, the main components are still there. Equally, similes
are differentiated from idioms as the integration of an utterance such as X is as cool as a cucumber
requires only that the hearer should decide whether it is used ironically or not. Semantically
speaking the meaning is straightforward and from a discourse point of view there are only two
possibilities: either X is really cool or X is not cool at all. The same semantic transparency applied
to the case of collocations that are often thought to be idioms but, whose meaning, upon closer
scrutiny, is unequivocally transparent appearing at the bottom of the idiomaticity scale as their
meaning is a function of the compositional principle (see Flavell et al, 1981 above).

In his pragmatic analysis of the functions of idioms, Strassler (1982:78) used a corpus of
approximately 106,000 words so that a number of socially significant variables such as social status,
age, education, profession as well as the gap between the partners with respect to these parameters
could be taken into consideration as part of the empirical justification in the subsequent study of
idiom functions in a variety of discoursal environments. The material included in the corpus is only
of the conversational sort and gathered from transcripts of trials, recordings of therapeutic sessions
and excerpts from the White House transcripts yielding 92 idioms in total. In fact, such low
occurrence of idiom is mainly attributed to corpus limitations whose size is not representative of a
language and, therefore, cannot guarantee any reliable conclusions. Parts of the private recordings
and of the transcripts were telephone conversations, some of which recorded from radio and
television while the larger part of the corpus consisted of conversations between two partners only,
as these were easier to evaluate, especially with regard to the hearer’s reaction.

Based on the belief that when using an idiom the speaker conveys more information than its
semantic content establishing either a social hierarchy or testing the hearer’s opinion in this matter,
Strassler (1982:116) demonstrates the workings of such a social hierarchy by analyzing the deictic
use of idioms in conversation between participants of equal or higher/lower social status in a variety
of situations. In this respect, he notes that the deictic use of idioms covers cases of (i) personal
reference (first person idioms), (ii) reference to the communicative partner (second person idioms)
and (iii) reference to a third person or object. All four possibilities appear in his corpus, as
 43

illustrated below, although, in cases where an idiom is used in order to refer to an object, is not
linked with person deixis.

(i) First Person Idiom: Speaker-Speaker

Roger: An’ yer al – yer dancing?
Ken: Sure. I can’t dance worth shit, I just move around hehh’s all you gotta do.

(ii) Second Person Idiom: Speaker-Hearer

B: Well a seven one four is where a person is unemployed an’ ‘e an’ ‘e submits
his own subcontractor’s sheet now
A: how come you’ve looked so so deeply into this, I mean, why why have you
got such a sort of a chip on your shoulder about it?

(iii) Third Person Idiom: Speaker-Third Party

Dean: ….I told this fellow O’Brien, ‘If you want money, you came to the wrong
man, fellow. I am not involved in the money. I don’t know a thing about. I
can’t help you. You better scramble about elsewhere. O’ Brien is a ball
player. He carried tremendous water for us.

Third person deixis is reported by Strassler to be the commonest identifiable function of idioms
in his conversational data and concludes on this evidence that such usage is unmarked and neutral
while, on the other hand, idioms used for first and second person reference are marked. The reason
offered for this markedness lies in the social relations between conversational partners, what
Strassler calls deixis. Examination of a variety of participant exchanges in different situations
(court-accused, therapist-patient, US President-communicative partners in the White House
transcripts, etc) led Strassler to conclude that the deictic use of idioms is determined by the social
status of the users in relation to their conversational partners:

‘The 2nd person idiom is restricted to the communicative partner of higher status,
the 1st person idiom to the lower status partner. This pattern is so strong that
there are hardly any exceptions to be found.’

Third person idioms being neutral are unrestricted and so may be used by anybody unless the status
difference between participants is too great as in the case of a patient and his therapist. Similarly,
the use of first person idioms is open anybody but is avoided by dominating speakers as, according
to Strassler (1982:103), they have a ‘self-abasement effect’. Second person idioms are the most
restricted as their use is socially acceptable only among peers. Strassler, therefore, concludes that
idioms function as status markers and, accordingly their use or non-use among conversational
partners is a form of social membership. Consequently, the additional deictic information idioms
convey constitute their functional raison d’être while at the same time accounting for their presence
or their absence in different situations.
 44

On the other hand, idioms cannot exhibit neither time nor place deixis as they normally do not
occur with adverbs of time and place. The time deixis of idioms lies exclusively within tense and
the dominant pattern in 85% of all cases in the corpus show a strict concordance of tense with
verbal idioms being used in the same form as their context. Even when grammatical rules require a
different tense (conditional, direct speech), this pattern is maintained by adding an additional clause.

Table 4.1. Time deixis of idioms according to Strassler (1982: 90-1)

1. Present-Present 2. Past-Past
D:….I don’t say that the uh the uh the
A: Yeah, it’s been a rough week, I- everybuddy is- yihknow. opinionated polemic journalist hasn’t an
B: Mm- enormous function to play in the editorial but
A: -talking about it, `n everbuddy, course I don’t know whether it’s more things have been changed, more
that, er just thet we’re just – completely bogging down et work, influence has taken place by journalists
hhhm. disclosing facts and enquiring into things say
the appalling scandal of the Crimean War by
Delaine in the Times, that did more than
editorials. How many boots did they have?

In (1) the conversation starts off in the present perfect tense and, logically, the idiom should be in
the same tense, as it refers to the state A was in during that past week. However, the immediate
context is not in the present tense. The central point is therefore not, as usual, the time of encoding,
but the tense of the context. In (2) again we notice that the tense of the idiomatic expression is the
same as the literal sentence immediately preceding it. Similar to time deixis, idioms are kept as
neutral as possible also do not allow the insertion of locative deictic elements such as
demonstratives, locative prepositions, adverbs as well as motion verbs. Whereas in the case of time
deixis the speaker is forced to neutralize the external temporal feature by strict concordance as there
is no way of avoiding tense, locative features, however, are not compulsory and therefore the easiest
way to avoid them wherever possible. In addition, idioms can also context deictic elements and be
used mainly as anaphoric expressions pointing back to something already present in the context.

With respect to Grice’s co-operative principle and his four conversational maxims of quantity,
quality, relation and manner, idioms do now show any intrinsic peculiarities. Idioms cannot be
considered to be elements which always result in flouting one of the maxims and thus
conversationally implicate specific information. As idioms are used and understood by the
participants of a conversation as lexical units, they can also be used like lexemes with respect to the
Gricean maxims. A speaker can obey or flout these maxims with idiomatic utterances and thus
implicate additional meaning. However, idioms, according to Strassler (1982:135), are to be
distinguished by other lexemes by virtue of the conventional implicatures, which relate to the social
relationship among the conversational partners and third persons to which reference is made. These
conventional implicatures class idioms as a sub-category of lexis as they are not only lexemes that
 45

capture complex everyday situations semantically, but also linguistic units that reduce the
complexity of social interactions. According to Lattey (1986:224), ‘idioms function as a kind of
euphemism, the use of which makes it more socially acceptable to make a personal-opinion
comment about an event or relationship than the use of a non-idiomatic expression of the same
massage would be.’

Obviously, Strassler’s (1982) treatment of idioms from a pragmatic point of view markedly
differs from previous discussions on the topic already reviewed in this section. Idioms are not only
viewed as semantico-syntactic irregularities with a stipulated, non-decomposable meaning and a
defective transformational potential in syntactic terms but they can also be assigned a specific
function in everyday discourse. To Strassler (1982), thus, the most relevant feature of idioms that
can be used to classify idiomatic expressions and, further establish them as a closed set is the social
constellation of the conversational participants as they tend to occur at specific points within
conversations with the key pragmatic function of establishing a social hierarchy between
participants of equal, higher or lower social status in diverse discoursal environments. In this
respect, idioms behave like other lexemes as they contain semantic information on the one hand, but
they also provide a method of handling special situations.

4.2. Chitra Fernando (1996b)

In a parallel fashion to Strassler (1982), Fernando (1996b) in her more recent study on idioms
investigates the phenomenon of idiomaticity within the tradition of corpus linguistics with the
following aims in mind: (a) to provide a neat classification of the different types of idioms by
establishing the criteria that most clearly distinguish them from the wider set of English multiwords
and (b) to account for the particularities of idiom use in a variety of specific situational contexts
along the lines of ideational, interpersonal and relational Hallidayan functional higher level macro-
functions (see also Moon, 1997;1998 below). The main data for this study comes from two sources,
i.e. (i) Fernando’s own corpus, predominantly a product of intensive collecting from newspapers,
general reading, literary and academic, personal correspondence, conversation, the electronic
media, and seminars and (ii) the Birmingham Collection of English Text (BCET) approximated 20
million words at the time of consultation.

Originally, in her analysis idioms and idiomaticity are two phenomena which are closely related
but not identical. According to Fernando (1996b:30), the basis of both is the habitual and, therefore,
predictable co-occurrence of specific words, but with idioms signifying a narrower range of word
 46

combinations than idiomaticity. All idioms show idiomaticity but not all word combinations
showing idiomaticity, for instance, habitual collocations (e.g. rosy cheeks, sallow complexion, black
coffee) are idioms for they are relatively unrestricted in their adjectival and nominal variants:
rosy/plump cheeks, rosy dawn, and a sallow skin are all possible. In this respect then, the term
idiom is largely used as a cover term for the description of conventionalized multiword expressions
and habitual collocations, restricted or unrestricted in their variability as manifestations of
idiomaticity and is defined accordingly as:

‘….indivisible units whose components cannot be varied or varied only within definable
limits. No other words can be substituted for those comprising for example, smell a rat, or
seize/grasp the nettle, which take either of these two verbs but no others: thus grab is
unacceptable. Nor are the words of an idiom usually recombinable.’(Fernando, 1996b:30).

‘.....idiomatic expressions, both canonical and non-canonical [grammatically], are only those
expressions which become conventionally fixed in a specific order and lexical form, or have
only a restricted set of variants….’ (Fernando, 1996b:31).

In her effort to demonstrate the shading off of the sub-classes of idioms into another as well as
the overlap between idioms and their lexical kin, collocations, Fernando (1996b:32) employs a
combination of scalar and columnar format in her proposed classification of English multiword
expressions. The table below has been adopted from Fernando (1996b) attempts an illustration of
idioms and habitual collocations as idiomatically related but different lexical types primarily based
on the degree of lexical variability each of the two classes displays. The semantics of idioms and
collocations is also taken into account as a second dimension in their classification distinguishing
between literal and non-literal expressions in both classes. Dotted and shaded areas in the diagram
indicate the overlapping areas between idioms and habitual collocations that include: (i) Semi-
literal idioms with semi-literal collocations of restricted variance, (ii) Literal idioms with literal
collocations of restricted variance and (iii) Literal idioms whose optional elements area usually
retained with literal collocations whose optional elements are usually dropped both displaying
restricted variance.
 47

Table 4.2. Scale of Idiomaticity of Multiword Expressions

(adapted from: Fernando, 1996b:32)

Idioms Habitual Collocations

1. Pure Idioms
(a) Invariant, Non-literal
(e.g. devil-may-care, backlash, chin wag, red herring, make
off with, spick and span, smell a rat, the coast is clear, etc)
(b) Restricted Variance, Non-literal 1. Restricted Variance, Semi-literal
(e.g. pitter-patter/pit-a-pat, take/have forty winks, (e.g. explode a
seize/grasp the nettle, get/have cold feet, etc) myth/theory/notion/idea/belief, catch the
post/mail, thin/flimsy excuse, etc)
2. Semi-literal Idioms
(a) Invariant
(e.g. drop names, catch fire, kith and kin, foot the bill, fat
chance, you’ve got, etc)
(b) Restricted Variance 2. Restricted Variance, Literal
(e.g. chequered career/history, blue film (e.g. addled brains.eggs, in-the-not-too-
/story/gag/comedian, good morning/day, etc) distance
past/future, for certain/sure,
potato/corn/wood, etc chips, etc)
3. Literal Idioms 3. Unrestricted variance, Semi-literal
(a) Invariant (e.g. catch a bus/plane/ ferry, etc, train,
(e.g. on foot, one day, in sum, in the meantime, on the run a business/company, etc, theatre, by
contrary, arm in arm, very important person (VIP), potato dint of hard work/patience/repetition, etc)
crisps, tall, dark and handsome, waste not, want not, happy
New Year, etc.
(b) Restricted Variance 4. Unrestricted Variance, Literal
(e.g. opt in favour of/for, for example/instance, in order (e.g. beautiful/lovely, etc, sweet woman,
that/to, happy/merry Christmas, etc) smooth/plump, etc, glowing/rosy cheeks,
etc)
4. Literal Idioms 5. Restricted Variance, literal, optional
(a) Restricted variance, optional elements usually elements usually dropped
retained (e.g. shrug (one’s shoulders), nod (one’s
head), clap (one’s hands), etc)

Considering Fernando’s (1996b) taxonomy presented here, a couple of points are further worthy
of mentioning at this point: First of all, idioms and habitual collocations are treated together as
instances of the larger set of English multiword expressions that are grammatically well-formed,
lexically invariant and semantically non-literal conventionalized phrases associated with culturally
salient encoding likely to recognized ‘by members of the language community as a standard way of
referring to a familiar concept or situation’ (Pawley, 1986:101). According to the table, lexical
fixity and semantic literalness are a matter degree, consequently, multiword expressions can range
from the completely fixed, semantically non-literal (e.g. pins and needles ‘the tingling sensation
following numbness’) through the possibility of some grammatical changes like those for tense (e.g.
spill/spilled the beans ‘commit a indiscretion’) to lexical variation from the restricted and semi-
literal (e.g. blue film/joke etc ‘obscene’) to the unrestricted semi-literal (e.g. catch a bus/train, etc
 48

‘be in time for’). Invariance of the words making up the expression combined with non-literalness
also contribute to the conventionalized multiword status. Clearly, Fernando’s (1996b) distinction
between idioms and collocations is based on formal criteria in the sense of amount of possible
lexical substitution that each one of these classes allow as well as semantic literalness presented in a
cline of gradient lexical restrictedness and transparency in meaning as a result of it. In this sense,
her distinction does not agree with Weinreich’s (1969) and Makkai’s (1972) earlier discussions with
respect to the ‘idiom vs. collocation’ distinction where ambiguity due to homonymous literal
counterparts leading to disinformation potential of decoding idioms is what most clearly
differentiates them from the set of restricted collocations.

With respect to the classification of idioms, Fernando offers three different sub-classes, i.e. pure
idioms, semi-idioms, and literal idioms. Pure idioms are defined as ‘a type of conventionalized,
non-literal multiword expressions. Spill the beans, for example, has nothing to do with beans. In
contrast to its literal meaning counterpart meaning ‘letting fall leguminous seeds’, a non-literal
meaning is imposed on the idiom as a whole: ‘commit an indiscretion’. This class of idioms is the
one we are mostly interested in as their linguistic character most commonly interferes with reading
comprehension in L2. Semi-idioms, on the other hand, has one or more literal constituents and at
least one with a non-literal subsense, usually special to that co-occurrence relation and no other:
drop has the meaning ‘overuse’ only when it occurs with names. Equally, Cowie (1981: 229) talks
about figurative idioms such as do a U-turn, change gear and open the bowling which have
figurative meanings but which also preserve a current literal interpretation as distinguished from
idioms proper that seem completely semantically unmotivated and petrified much like pure idioms
discussed by Fernando (1996b) here. Lastly, the class of literal idioms consists of largely restricted
collocations and is less semantically complex than either pure and semi-idioms. In fact, the term
appears counter-intuitive in itself and literal idioms seem to constitute something like an
intermediary class between idioms, restricted and free collocations (see more Aisenstadt, 1979,
1981; Rose, 1978) although as Bolinger (1977:158) has suggested ‘…in reality there is no clear
borderline between the two’.

As far as size goes, Fernando (1996b:41) sets the compound as the lower limit for inclusion in
the class of idioms as they represent habitual co-occurrences between two or more words (e.g.
foxglove, overtake, pitter-patter) and can be easily assigned to the classes of literal idioms (e.g.
mother-in-law), semi-literal idioms (e.g. baby-sitter) or non-literal idioms (e.g. foxglove) while the
complex clause (e.g. when the cat away, the mice will play) is the recognized as the upper limit.
From a syntactic point of view, the transformational potential of idioms is a function of their
 49

semantic transparency and lexico-grammatical make-up in a differential manner. For instance,

while some idioms (e.g. say no more, fat chance you’ve got, pins and needles) allow no additions or
deletions even in the form of inflectional variations for tense or number as the case may be, others
inflect readily in order that the idiom appears in its correct form in terms of number concord (e.g. I
twisted Richie’s arm/I can twist the arms of a few friends) or in terms of the correspondence
between time and tense (e.g. This bloke got such a fright he kicked the bucket!). Such alterations of
tense and number are regarded as forms of replacement of the original grammatical category, of
change, that is general enough to subsume the processes of addition, deletion and permutation
performed in an idiom (see more Fernando, 1996b:42-52). In this sense then, pure idioms are placed
at the top of the hierarchy as lexically invariant expressions made up of words that are semantically
void in that they do not carry an independent meaning of their own, resulting thus to non-literalness.
Semi-idioms and restricted collocations placed in the middle as they are characterized by minimal
lexical variation and partial semantic transparency while unrestricted collocations belong to the
bottom of the hierarchy due to unlimited lexical substitution possibilities and meaning transparency.

Based on the inclusion of the category of sememic idioms and discoursal expression in earlier
idiom taxonomies by Makkai (1972) and Alexander (1984) respectively, Fernando (1996) extends
her study of idiomatology by examining the specific uses of idioms in discourse identifying three
separate idiom classes based on Halliday’s functional theory. Thus, ideational idioms either signify
message content, experiential phenomena including the sensory, the affective, and the evaluative, or
they characterize the nature of the message, for example, as being specific or non-specific.
Interpersonal idioms fulfil either an international function or they characterize the nature of the
message. In their interpersonal function they initiate, maintain, and close an exchange and are
closely associated with politeness routines. In such roles these idioms exemplify the mores of social
interactions, exemplifying as well the operation of some of the maxims of Grice’s Co-operative
Principle: sincerity, newsworthiness, and brevity, which characterize the nature of the message.
Finally, relational or textual idioms ensure the cohesion, and can therefore aid the coherence of
discourse. They can accordingly be grouped along with conjunctions, for example, and, but, or, and
so, because, if….then, as having a textual function. They may be categorized into those which
integrate information and those which sequence information. The table in the following page
summarizes and offers examples of the adapted Fernando’s (1996:72-4) functional categorization.
 50

Table 4.3. Fernando’s (1996b:72-4) Functional Classification of Multiword Expressions

1. Ideational or ‘the state and way of the world’s 2. Interpersonal Idioms 3. Relational Idioms
idioms’ a. Interaction Strategies a. Integrative
a. Message Content 1. Greetings and Farewells 1. Adversative
1. Actions (e.g. good morning, how are (e.g. on the contrary,
(e.g. tear down, mess about with, do a U-turn, twist sb’s you?, see you later, bye for far from, etc)
arm, spill the beans, wear different hats, give sb an inch now, etc) 2. Comparison
(and he’ll take a mile etc), wave/offer/hold the olive 2. Directives (e.g. on the one
branch etc. (e.g. let’s face it, tell you what, hand…on the other,
2. Events say no more, believe you me, etc)
(e.g. turning point, the straw that breaks the camel’s back, not to worry/don’t worry, never 3. Causal
out of the mouth of babes, have blood on one’s hands, etc) mind, do you mind, etc). (e.g. so that when, the
3. Situations 3. Agreement more X….the more Y,
(e.g. be in Queer Street, be in a pickle, be up a gum tree, (e.g. that’s true, you’re telling no wonder, etc)
etc) me, say no more, (that’s) a 4. Concessive
4. People and Things good question, etc). (e.g. at the same time,
(e..g a back-seat driver, a man about town, a scarlet 4.‘Feelers’,eliciting opinions etc)
woman, a fat cat, a red herring, a lounge lizard, etc) (e.g. what do you think?, how 5. Addition
5. Attributes do you feel?, etc) (e.g. in addition to,
(e.g. cut-and-dried, matter-of-fact, lily-white, as green as 5. Rejections what is more, etc)
grass, from A to Z, etc) (e.g. you’re kidding/joking, b. Sequencing of
6. Evaluations come off it, tell it to the Chaining Information
(e.g. turn back the clock, it is a pity, matter-of-fact, cut- Marines, I wasn’t born 1.Sequencing
and-dried, lily-livered, beauty is/lies in the eye of the yesterday, etc) Metadiscoursal
beholder, waste not, want not, a watched pot never boils, if b.Characterizing the Message Information
you can’t stand the heat get out of the kitchen, a Trojan 1. Newsworthiness (e.g. in the first place,
horse, I’m not my brother’s keeper, etc). (e.g. guess what!, what do you last but not least, etc)
7. Emotions know?, what, you ask?, etc) 2.Sequencing
(e.g. green with envy, have one’s heart in one’s mouth, a 2. Sincerity Temporal
lump in one’s throat, lose one’s heart, for one’s blood to (e.g. quite seriously, believe Information
boil, walk on air, down in the dumps, tear one’s hair, etc) you me, as a matter of fact, etc). (e.g. one day, a long
b. Characterizing the Message 3. Calls for Brevity time ago, up to now,
1. Specific Information (e.g. cut the cackle, get to the etc)
(e.g. to be exact/precise, for example, that is, the question point, etc).
is, what I am saying is, my guess is, I felt like saying, etc). 4. Uncertainty
2. Non-specific Information (e.g. I daresay, mind you, etc)
(e.g. kind of/sort of, or something, such and such, and so
on, etcetera, etcetera, blah, blah, blah, etc)

Adhering to the notion of idiom in the sense of locutions that reflect the special cut -as
exemplified in the above taxonomy- the different structural design of various languages (English,
German, and Sinhala) and varieties of the English language which have developed differently either
to the influence of indigenous language (e.g. Sri Lankan English) or in isolation from the parent
language as in the case of Australian English (see further, Fernando, 1996a), Fernando (1996:75-
93) further discusses on a cross-linguistic basis the different means used by each community to
express its ‘specific character, property or genius’ of its language. According to the researcher,
usage in different languages exemplifies how various communities of speakers have drawn in
different ways on the universal property of lexicogrammar, so that their use of language shows a
distinctive, though conventionalized, structural organizations of elements: different kinds of
 51

complex verb formations and differences in the utilization of syntactic processes such as bracketing,
deletion, and reduplication (see also Roberts, 1944 above).

Fernando’s (1996) analysis of idioms and idiomatology is heavily influenced by earlier scalar
categorizations, each reflecting the theoretical basis of their creators. Fraser’s scalar categorization
reflects a transformational orientation while Makkai’s (1972), though non-scalar, a stratificational
one. Although Fernando’s lexicogrammatical and functional idiom categorizations discussed in this
section show no specific model-orientation, they seem to be based partly on the kinds of descriptive
categories found in general linguistic theory as exemplified most notably in the work of Alexander
(1984) and McCarthy (1992). Both categorizations display areas of overlap with Makkai’s idiom
classification but including an even greater range of multiword or ‘fixed expressions’ often
categorized along an idiomatic cline in terms of either the degree of their lexical/syntactic fixedness
and/or semantic opacity. A cross-comparison of the four idiom classifications presented in table
(2.1) below summarizes the overlapping idiom classes in both lexicogrammatical and functional
terms yielding a combined taxonomy that includes the following classes: 1. Tournures (Full-
idioms/Semi-idioms/Literal Idioms), 2. Irreversible Binomials/Trinomials, 3. Phrasal Verbs, 4.
Phrasal Compound Idioms, 5. Habitual Collocations (Restricted/Unrestricted), 6. Incorporating
verbs, 7. Frozen Similes, 8. Pseudo-idioms, 9. Possessive ’s phrases, 10. Proverbs/Cultural
Allusions & Quotations/Catchphrases/Idiomatic Similes, 11a. Discoursal Expressions (Social
formulae, gambits, stereotypes), 11.b Discoursal Expressions (Connectives/Structuring devices). Of
all categories, we are particularly interested in full or ‘pure’ idioms, a sub-type of tournure
constructions with verb+object (e.g. pull sb’s leg) or verb+ (adjunct)+object (e.g. go off one’s head)
exhibiting the highest degree of syntactic fixity and semantic opacity.
 Table 4.4. A Cross-Comparison of Lexico-grammatical and Functional Idiom Classes 52
A. Lexico-grammatical
52 Idiom Classes
Makkai (1972:132-179) Alexander (1984) McCarthy (1992:56-7) Fernando (1996b)
1. Tounures (‘turn of phrase’) 1. Full-idioms 1. Tournures 1. Pure Idioms
(fly off the handle, rain cats and dogs) (run up (a bill), tear off, to smell a rat) (pull somebody’s leg) (smell a rat, make off with, seize/grasp the nettle)
2. Semi-idioms 2. Semi-Literal Idioms
(beefy-looking, dead drunk, a fat salary) (drop names, catch fire, blue film/story/joke)
3. Literal Idioms (opt in favour of, abstain from)
2. Irreversible Binomials 3. Irreversible Binomials/Compound Idioms 2. Binomials and Trinomials
(pepper and salt, coffee and cream) (spick and span, dead drunk) (give or take, ready, willing and able)
3. Phrasal Verbs 3. Phrasal Verbs
(make up, turn out, bring up) (take somebody off)
4. Phrasal Compounds 4. Compound Idioms 4. Opaque Compounds
(hot dog, blackmail, high-handed) (dead drunk) (a mish-mash)
5. Restricted Collocations 4. Habitual Collocations (Restricted/Unrestrictd)
(breakneck speed)
5. Incorporating Verbs 6. Frozen similes
(eavesdrop, manhandle, boot-lick) (as cold as charity)
6. Pseudo-Idioms 7. Possessive ’s phrases
(kith and kin, spic and span, to and fro) (a king’s ransom)
B. Functional Idiom Classes
1. Proverbs 1. Proverbs 4. Allusions/Quotations 1. Cultural Allusions 1. Ideational Idioms
(don’t count your chickens before they’re hatched) (a watched pot never (you’ve never had it so good) (to be or not to be) Message Content
2. Familiar Quotations boils) 5. Idiomatic Similes
(brevity is the soul of wit) 2. Stock Phrases (so sober as a judge)
3. ‘First-base’ idioms associated with a (a recipe for disaster)
national game like baseball 3. Catchphrases
(have two strikes against one) (that’s another fine
mess you got me into)
4. Idioms of ‘institutionalised’ politeness 6. Discoursal Expression 2. Conversational routines 2. Interpersonal Idioms
(May I….X?) 6/1. Social Formulae/Cliches (by the way, how’d you do) Interactional Strategies
5. Idioms of ‘institutionalised’ greeting’ (How do you do?)
(How do you do?) 6/2. Conversational Gambits
6. Idioms of ‘institutionalised (Guess What!)
understatement’ 6/3. Stylistic Formulae
(I wasn’t too crazy about him) (My lords, ladies and gentlemen)
7. Idioms of ‘institutionalised hyperbole’ 6/4. Stereotypes
(he won’t even lift a finger) (We’re just good friends)
6. Discoursal Expressions 3. Relational Idioms
6/5. Connectives; Structuring Devices Integrative/Sequencing/Chaining Information
(finally, to conclude, once upon a time)
 53

The class of pure or opaque categories is even narrowly defined by Alexander (1987:110) in
terms of the degree of semantic transparency and lexical fixity they display as end-points on different
spectra along which many fixed expressions can also be distributed in line with the idiom
taxonomies offered by Alexander (1987), McCarthy (1992), Fernando and Flavell (1981) and
Fernando (1996b) above. Similarly, Wallace (1979:67) claims that there is no hard and fast divisions
in the classification of the vast amount of phraseological data categorizing stereotyped expressions
along a vertical axis in terms of a structural hierarchy starting at compound-word level and ending at
sentence level and a horizontal axis in terms of decodability, ‘going from expressions which are
easily decoded or understood from their constituent parts (‘transparent’) to those which are liable to
erroneous decoding or not decodable’ due to the meaning opacity they display.

It is exactly due to this semantic interpretation or decoding as well as encoding difficulties that
idioms often cause to L2 learners that is used as a primary criterion in Alexander’s (1987) definition
of pure or opaque idioms. As can be seen from the table below, pure idioms are always placed at the
bottom of the syntactico-semantic restrictedness spectrum as a result of the maximum degree of
fixity they display in lexical, grammatical and primarily semantic terms. While some idioms (spill
the beans) are considered to be opaque and at the extreme end of ‘pure idioms’, others can be listed
in different categories resulting in several subtypes (e.g. Semi-idioms, Metaphorical idioms,
Figurative idioms) and a relative lack of agreement about transparency/opacity.

Table 4.5. Semantic Classifications of Idioms

Alexander (1987) Fernando and Flavell (1981) Fernando (1996b)
1. Literal Phrase 1. Literal and/or Transparent 1. Literal idiom
(hit the ball) (cut wood, break eggs) (tall, dark and handsome, on foot
2. Semi-idioms 2. Metaphor/Semi-transparent 2. Semi-literal idiom
(hit a six) (skate on thin ice, kill two birds (kith and kin, drop names)
with one stone)
3. Semi-idiom
(catch your breath)
3. Metaphorical Idiom
(hit the jackpot)
4. Figurative Idiom
(hit list)
5. Opaque/Pure Idiom 3. Full-idiom/Opaque 4. Pure Idiom,
(hit the sack) (pull someone’s leg (spill the beans)

Glaser (1998, 1988) also considers idioms as the prototypes of a set expression or phrase
assigning them to the centre of the phraseological system of modern English language. These
phraseological embrace idioms as well as non-idioms, i.e. non-idiomatic restricted collocations, in
nominative function, namely, word-like units which designate phenomena, objects, events,
 54

processes, actions, states, qualities, relations, etc. in the outside world, and a few word groups which
only function as operators in that they designate relations between phenomena or objects also known
as function words. Idiomatized prepositions (by virtue of, by dint of) and conjunctions (in order to,
on condition that) may serve as examples. Idioms form the majority, and are thus regarded the
prototypical class due to some unique feature of form in the sense of fixed wording that is unaltered
and deficient in possible lexical substitution that might affect their idiomatic interpretation or
meaning in the sense of semantic noncompositionality. Non-idioms, on the other hand, are
characterized only by having fixed content and order determined by custom which dictates that they
be expressed by a particular selection and order or words, they have transparent meanings and
include terminological word groups (e.g. unconditional surrender), onymic entities, i.e. phrases
which are proper names (e.g. the Black Sea, the Golden Twenties), clichés (e.g. an eloquent silence),
paraphrasal verbs, and other set expressions (Gates, 1973:70).

Sentence-like phraseological units whose logical structure consists of a nomination and a

predication and their finite verb may be absent in the case of reduction or ellipsis belong to the
transition area. This area comprises the following types: 1. Fragments or reductions of proverbs (e.g.
a new broom (sweeps clean)), 2. Proverbial sayings (e.g. to throw out the baby with the bath-water),
3. Irreversible binomials (e.g. bread and butter, alive and kicking), 4. Stereotyped comparisons or
similes (e.g. as proud as a peacock) and 5. Literary allusions and fragments of quotations (e.g. a
Jekyll and Hyde). Finally, the periphery, or outer circle of the phraseological system, covers
phraseological units which are chiefly propositions, though their idiomatic character varies greatly.
The majority of them are non-idioms, which depend on the given subtypes of phrases, i.e. proverbs
(e.g. make hay while the sun shines), commonplaces or truisms (e.g. boys will be boys), routine
formulas (e.g. mind the step), slogans (e.g. value for money), commandments and maxims (e.g. thou
shalt not kill) and quotations (e.g. speak softly and carry a big stick). Again, idiomaticity of a given
fixed phrase is viewed as a matter of different degrees of idiomaticity dealing with more idiomatic to
less idiomatic phrases as one moves from the center to the periphery of the phraseological wordstock
of the English language.

This tendency to describe language on a continuum instead of assuming a qualitative, either-or

distinction between idiomatic language and regularly generated language, seems also pronounced in
the collocationist approach with completely invariant clusters at one end of the continuum, freely
combining morphemes at the other, and all degrees of combinatorial flexibility in between. Wood
(1986) perhaps offers the best model of this approach by making a case that there is no neat
 55

categorization between prefabricated and creative syntax and that these exist only at opposite ends of
a continuum, separated by infinite shadings of syntactic and semantic variation. In contrast to most
collocationist accounts that use only a semantic criterion of whether a combination is fully
‘compositional’ or not, that is, whether the meaning of the collocation is fully predictable from the
individual meanings of the words that compose it, (see for example, Nuccorini, 1990), Wood’s
(1986) work goes further in employing a syntactic criterion as well of whether the combination is
fully productive or not, i.e. whether a form is structurally unique, whether it is fully canonical, or
whether it is somewhere in between these two poles. Following Wood (1986:2), idioms are defined
as complex expressions which are wholly non-compositional in meaning and wholly non-productive
in form that can properly be described in terms of two continua, i.e. (a) a continuum of semantic
transparency, shading by degrees from the utterly opaque to the fully predictable and (b) a
continuum of form productivity ranging from expressions which permit no variation to those with
freely variable constituents. Contrary to previous accounts such as Weinreich’s (1969) and Makkai’s
(1972), Wood (1986) also includes as idioms, expressions containing an opaque cranberry-morph
(come a cropper), phrases of irregular syntactic structure (by and large) as well as any compound
words that meet the criteria of the definition. It also rejects any necessary connection between
idiomaticity and ambiguity and denies the viability of attempts to distribute the meaning of a true
idiom over the meaning of its parts in a lexicon.

Using these parameters, Wood is able to define the continuum more precisely. A fully non-
compositional, non-productive collocation is a true idiom, a truly frozen piece of language. Phrases
such as hell for leather, by and large qualify, for they make no compositional ‘sense’, nor do they
have syntactic patterns that generate other such structures. At the other end of the continuum, a
phrase like see the river is fully compositional and productive; its meaning is a combined meaning of
the individual units, and its form is the basis for an unlimited number of phrases. In between these
two extremes are phrases that are limited in compositionality and productivity in various degrees.
For example, the oft-cited kick the bucket is not completely frozen, since the word kick in the sense
of die exists in a few other phrases, such as kick off and kick out, and is therefore less than an idiom
than hell for leather (Ruhl,1977:462). This meaning of kick is certainly restricted, however, and so
the phrase is not as compositional as drink milk/tea/coffee, which demonstrates full compositionality.
Likewise, the phrase off with his head permits only limited lexical variation, (down with the king,
away with all X), always carrying the meaning of ‘Directional Particle + with + NP’, and therefore is
more restricted in the combinations it can form than is a fully productive phrase like see the river.
 56

Using these criteria, Wood suggests that language patterns itself along the continuum represented in
Figure 4.1. below.
Figure 4.1. Wood’s (1986) Continuum of Lexical Phrases

idioms – collocations – colligations - free combinations

Thus, to Wood (1986) idioms are one of a kind, compositionally and productively, and are
completely unpredictable in their meaning and form. True idioms (by and large, hell for leather,
happy go lucky) are thus completely frozen and relatively few in number, since most phrases permit
some degree of compositionality or productivity. Collocations are roughly predictable yet are
restricted to certain specified items and thus are nameable by words. Take umbrage, for example, is a
highly restricted collocation that invites a compositional interpretation, but the unique noun and the
empty verb baffle it. Substitution of synonyms is possible in many collocations, and where possible,
will produce what are called ‘idiom families’ (pay heed/attention, open-and-shut
case/issue/problem). When substitution is bounded only by syntactic category and semantic features
– and predictability of meaning is greater than ever – one arrives at colligations, a term coined by
J.R. Firth. Colligations are generalizable classes of collocations, for which at least one construct is
specified by category rather than as a distinct lexical item. As phrases are less restricted to specific
lexical items and can be described in terms of more word classes, the more they approach the freely
compositional and productive end of the continuum.

Within the tradition of early Soviet phraseological research, Cermak (1988) also explores issues
relating to idioms and the ways in which they are to be distinguished from other collocations, free
and restricted, through a detailed study of Czech and English. In his study, Cermak (2001:2) claims
that degree of collocational restriction exhibited by idioms, restricted and free collocations is the
decisive factor that differentiates them suggesting a cline of collocational restrictedness since all
word combinations are thought to be restricted in one way or another. In this respect then, an idiom
is defined as:

‘…such a unique and fixed combination of at least two elements for which it holds that
at least two elements for which it holds that at least some of these do not function in the
same way, in any other combination of the kind, or do function only in a highly restricted
number of them only.’ (Cermak, 1991:262)
 57

Following this operational definition, idiom is viewed as a result of the intricate interplay of the
anomalous syntagmatic and paradigmatic positions of an item in the language system. Generality and
regularity are completely or almost completely lost and replaced by combinatorial anomaly. Thus,
for example, idioms such as hold good, play truant, red tape, kith and kin and as to which are fixed
combinations with a familiar nominative function and meaning, have for the most part, no analogous
counterparts in which their components recur with the same function. This can be tested by their
commutation into the following tentative forms *hold bad, *hold nice/fine, *act/behave truant,
*green/blue tape, *kith and family/relatives/clan, *as from/through. As shown in figure 4.2. below,
an idiom is viewed as a fixed syntagm at least one item of which is, a member of a severely
restricted and closed paradigm always in syntagmatic function or paradigmatic membership, though
often in meaning, too.

Figure 4.2. Cermak’s (1988) Continuum of Phraseological Units

__________________________________________________________

idioms collocations colligations free

[by and large] [kick the bucket] [off with his head] [see the river]

According to Cermak (1991:263), these combinatorial or syntagmatic and class or paradigmatic

aspects are regarded as primary and universal elements in any language system that constitute the
substance of the idiom and consequently become the universal criterion for the judgement of
idiomaticity. As a basic property, combinatorics is thought to give rise to all of the other idiomatic
qualities, such as the lack of compositional function between constituents’ meanings, their
transformational deficiency or formation based on no model which allows for no exact repetition or
replica in another idiom. In this light, Cermak (1994:187) suggests that idioms can be characterized
as:

‘inescapable resolutions of the combinatorial blind alleys of any language, once the
language deviates in its combinatorial structures from its regular possibilities and once its
numerous anomalous combinations (of many kinds and types) become frozen and
established as stable parts of the system of its nominations’. (Cermak, 1994:187)

Idiom formation as a process is thus viewed as universal based on the interplay of several factors, i.e.
(i) of too large combinatorial capacity unused, (ii) of language creativity drawing on analogy and
similarity and (iii) of the human capacity to view a functionally recurrent item as a sign. Idioms are
usually the result of the language creativity of a hit-and-miss kind, formed in a haphazard manner
and following no plan which explains why idioms cannot be made, coined and tailored to suit the
 58

needs of a moment. They are seen as peculiar signs, amalgams with arbitrarily stipulated meaning,
different from the meaning of their component words, regarded as marks of language economy. In so
far as our working definition of idiom is concerned, collocational restrictedness in terms of lexical
and syntactic fixedness is one of the features adopted in the current study together with semantic
properties that constitute the essence of the phenomenon of idiomaticity as has been outlined so far.

However, the notion of scalarity in idiomaticity has been recently adopted and applied in multi-
dimensional descriptive models by various theoreticians in their effort to draw clear demarcation
lines between idioms and the broader set of fixed expressions. Carter (1987:70-2) briefly discusses
three different scales, namely, collocational restriction, syntactic structure and semantic opacity in
the form of sets of continua with fixed points but several intermediate categories to determine how
fixed particular lexical patterns are. The cline in his collocational restriction runs from less fixed to
more fixed based on the extent to which words in a fixed expression can be substituted in different
syntactic slots, distinguishing between unrestricted collocations (e.g. take a look/a holiday/a rest/a
letter/notice/a walk) or what Wood (1986) calls free constructions, semi-restricted collocations (e.g.
harbour doubt/grudges/uncertainty) or ‘colligations’ (Wood, 1986) and restricted collocations (e.g.
stark naked, pitch black) or simply collocations in Wood’s (1986) scale to distinguish them from
idioms with the latter displaying the highest degree of lexical fixity. This classification of lexicalized
expressions is in line with Barkema’s (1996a:145-6) categorization of lexical phrases in lexical
restriction collocability terms into (a) free constructions, (b) collocationally open and (c)
collocationally closed phrases. Carter (1987) also includes the intermediary class of familiar
collocations (e.g. innocent bystander, unrequited love, lukewarm reception) with fixed expressions
whose ‘words keep regular company with each other’, though its description is left unspecified and
unclear in relation to the rest of the categories in the cline. Furthermore, the class of idioms is not
included in Carter’s scale as in Wood’s (1986) where it becomes evident that idioms are the most
restricted type of fixed expressions in the phraseological system of the language with the smallest
degree of internal lexical substitutability.

On his ‘syntactic structure’ scale, fixed expressions are assigned to three categories based on the
degree of syntactic flexibility they exhibit, i.e. flexible (e.g. break someone’s heart, nice to see you),
regular with certain constraints (e.g. to drop a brick, to smell a rat) and irregular (e.g. to go one
better, to be good friends with somebody). However, while it is true that constructions with
idiosyncratic syntax are generally inflexible, the reverse is not (e.g. kick the bucket), and, for that
reason, Barkema (1992;1993;1994) distinguishes between fully flexible, semi-flexible and inflexible
 59

fixed expressions in terms of the internal grammatical variation they exhibit with respect to the
extent they permit (a) additional syntactic functions to the base pattern of an expression (e.g. an
appalingly tough nut to crack), (b) the substitution of one of the lexical items in the base form from a
closed class by an alternative from the same class (e.g. harder nuts to crack), (c) the rearrangement
(‘permutation’) of the syntactic functions in the base pattern (e.g. a postmodifying instead of a
premodifying adjective in a nut too tough to crack the first time out) and (d) the introduction of
‘foreign’ elements (‘interruption’) in the base pattern, i.e. ones that do not fit in with the syntactic
structure of the base pattern (e.g. interruption by an adverbial: a hard nut, as always, to crack). The
existence of these three degrees of flexibility is expressed as a scale with two extremes: fully
flexible, semi-flexible and inflexible. Idioms belong to the far end of the scale exhibiting the
minimum degree of syntactic flexibility as already has been discussed (Fraser, 1970; Fernando et al
1981).

Additionally, Carter’s third scale concerns the dimension of semantic opacity and, in fact partly
coincides with Fernando et al’s (1981) semantic cline already discussed. In this connection, Carter
classifies fixed expressions into four categories in terms of compositionality and semantic motivation
based on speakers’ judgements about the relations between the literal and the institutionalized
metaphorical meanings of idiomatic expressions. These are: (a) transparent (e.g. long time, no see;
when all is said and done), (b) ‘semi-idioms’/metaphor/idiomatic similes (e.g. we are all in the same
boat, an open-door policy), (c) semi-transparent (e.g. the business really took off, to get round
somebody) and (d) opaque with two sub-categories (i) overt (uninterpretable without
contextual/cultural knowledge) or in Fernando et al’s, 1981 terminology ‘semi-opaque’ phrases such
as O.K., right on, yuk, bottoms up and (ii) covert where according to Fernando et al (1981) as well as
Barkema (1996a;1996b) ‘pure’ idioms such as to be on the wagon, to be on the ball belong due to
their strictly non-compositional lexical meaning. Semantic opacity is one of the key idiom-type
variables included in our study that is determined as a function of the degree of compositionality that
an idiom exhibits. In this respect, our idioms are classified as transparent, semi-transparent and
opaque with respect to the contribution of the meanings of each of their component words to the
meaning of the idiom as a whole unit. In parallel to Fernando et al (1981) and Moon (1998),
Barkema (1996a:132) also considers institutionalization as another type of characteristic for the
identification of fixed expressions treated in a separate cline according to which lexicalized phrases
are assigned to different classes in relation to their frequency and currency of usage among speakers
of a speech community.
 60

In sum, it seems that the borderline between literal and idiomatic expressions, on the one hand,
and metaphorical and idiomatic expressions, on the other, is controversial. Idiomatization is a
process (Michiels, 1977) and a given structure acquires its idiomaticity gradually. Several authors
(Fernando, 1978; Fernando et al, 1981; Makkai, 1972; Wood, 1986) have taken the view that
idiomaticity is more a matter of degree rather than an ‘all-or-nothing’ property of a string in a given
language. Therefore, it might be suggested that literal and figurative might grossly be the endpoints
of a continuum along which one can situate different type of expression, according to their levels of
literalness, idiomaticity, or metaphoricity. In this respect, Carter (1987:72) emphasizes the necessity
of the notion of scalarity being applied in the study of the multiple lexico-grammatical and semantic
dimensions of phraseological phenomena as their points of intersection and overlap would result in
the definition of the ‘most fixed expression as those which are closed in more than one dimension’.
We cannot but agree with this statement as we believe that a definition of idiom that includes and
accounts for its many different linguistic aspects consistently and adequately enough to demarcate
idioms as an phraseological class will serve as a solid framework to undertake empirical work with
respect to the research questions concerning idiom identification and comprehension while reading
texts in a foreign language.

4.3. Rosamund Moon (1998)

Following earlier attempts (Strassler, 1982; Fernando, 1996) that classified idioms in terms of
their functions in both spoken and written genres, Moon’s (1998) study is the latest most well-known
corpus-based work to describe the structural characteristics and discoursal behavior and usage of a
vast amount of multiword units against an 18 million-word corpus of Contemporary English, the
Oxford Hector Pilot Corpus (OHPC). In line with previous approaches (Fernando, 1978; Fernando et
al, 1981; Makkai, 1972; Wood, 1986), Moon (1998:19) discusses phraseological phenomena on a
cline with ‘collocations’ and ‘idioms’ forming two large and amorphous subgroups of fixed
expressions that are non-compositional and transformationally defective to some extent as well as
discoursally or situationally constrained units. As a way of classifying the wide range of fixed
expressions so that particular types could be selected or excluded from the investigation, Moon
(1998:20) focuses on three macrocategories, i.e. anomalous collocations, formulae and metaphors
that are ‘problematic and anomalous on grounds of lexicogrammar, pragmatics, or semantics. Idiom
is not used as a formal category and is subsumed within the broader category of fixed expressions
referring loosely to semi-transparent and opaque metaphorical expressions such as spill the beans
and burn one’s candle at both ends as opposed to other kinds of expression (Moon, 1998:5).
 61

Thus, the fixed expressions which are treated as anomalous collocations are syntagmatically or
paradigmatically aberrant in the sense that they cannot be decoded purely compositionally nor
encoded freely. Further subclassification of anomalous collocations based on the nature of the
anomaly includes such categories as: (i) ill-formed collocations (e.g. at all, by and large, of course)
that go against the conventional grammatical rules of English, (ii) cranberry collocations (e.g. kith
and kin, on behalf of, short shrift) or Makkai’s (1972:43) cranberry morphs that include items that
are unique to the string and not found in other collocations. (iii) defective collocations (e.g. at least,
a foregone conclusion, in effect) that cannot be decoded purely compositionally either because a
component item has a meaning not found in other collocations or contexts or because one or more of
the component items is semantically depleted and (iv) phraseological collocations (e.g. in action,
into action, out of action) that consist of cases where there is a limited paradigm in operation and
other analogous strings may be found, but where the structure is not fully productive. The category
of formulae present problems of discoursal function as they are specialized pragmatically. The
subtypes included in this class generally conform to lexicogrammatical conventions of English,
although a few are effectively truncated utterances. They are generally compositional semantically,
although some similes and proverbs are obscure or metaphorical. These are: (i) simple formulae that
have some special discoursal function or are iterative or emphatic and syntagmatically fixed such as
I’m sorry to say, alive and well, pick and choose, (ii) sayings such as quotations (e.g. an eye for an
eye), catchphrases (e.g. that’s the way the cookie crumbles) and truisms (e.g. home, James, and don’t
spare the horses), (iii) proverbs (e.g. every cloud has a silver lining) and (iv) similes (e.g. as good
gold). Finally, metaphors are semantically non-compositional and their subclassification reflects
degrees of transparency representing a continuum rather than discrete categories. In this sense, there
is the class of transparent metaphors (e.g. alarm bells ring), of semi-transparent metaphors (e.g.
throw in the towel, under one’s belt) and of opaque metaphors where pure idioms are included (e.g.
bite the bullet, kick the bucket).

In parallel to Fernando and Flavell (1981) and Barkema (1996) above, Moon (1998) also believes
that idiomaticity is a complex phenomenon that can best be defined by multiple criteria, each
criterion representing a single property. For that reason, she also suggests that there are three primary
criteria that come into play when assessing the holism of a string as a fixed expression, i.e.
institutionalization, fixedness and noncompositionality. These principles are not present to an equal
extent in all items but in differing degrees that allow the formation of continua along which fixed
expressions can vary. In this sense, there are degrees of institutionalization, from the extremely
frequent of course to the fairly rare cannot cut the mustard; of fixedness, from the completely frozen
 62

kith and kin to the relatively flexible and variable take stick from someone, get a lot of stick from
someone and of non-compositionality, from the opaque bite the bullet to the transparent enough is
enough.

According to the researcher, institutionalization is the process by which a string or formulation is

recognized to be a standard expression for the meaning in question and is accepted as a lexical item
of the language since it denotes a meaning which is culturally authorized as a standard concept in a
speech community (Pawley et al, 1986; Gotz, 1976). Although institutionalization is treated as a
necessary condition for a string to be classifiable as a fixed expression, it is not sufficient by itself.
However, to Moon (1998) institutionalization is an aspect of fixed expressions that differentiate them
from other sets of lexical phrases which can be quantified if assessed by the frequency with which a
given string recurs in the 18-million word corpus she uses in her investigation. Generally, corpus
evidence with respect to frequencies of fixed expressions indicated that just 40% of them occur with
a frequency that must be considered no more significant than random chance, and over 70% have
frequencies of less that 1 per million tokens. More specifically, the majority of metaphors – where
idioms are included - have frequencies of less than 1 per million while anomalous collocations of
some kind appear to be the commonest fixed expressions in the corpus. The distributions of the
subtypes of each of the macrocategories under consideration is presented below. Notably, the
distribution of opaque metaphors, of which pure (or classical) idioms is a part, are rather infrequent
with a frequency occurrence in OHPC of greater than 1 per million including cases such as bite the
bullet, over the moon and red herring (Moon, 1998b; 1988). In the far larger Bank of English corpus,
idioms are slightly less common with frequencies of over 2 tokens per 3 million words while no
metaphor occurs more frequently than 50 per million. Strassler (1982:77) also reports on the limited
distribution of idioms in his corpus of spoken English interaction of various types where he found
only 92 token of idioms, i.e. an average of 1 idiom per 1, 152 words. Gottlieb (1994) also mentions
that the most frequent of the 28 idioms in his research occurs 24 times in the four COBUILD corpora
totaling 44.3 million running words, thus reaching a score of 0.6 parts per million only. The
frequencies of occurrence for each subtype in each of the macrocategories studied in Moon’s (1998)
study are briefly presented in the table below.
Table 4.6. Distribution of subtypes across all Macrocategories
(adapted from Moon, 1998:62-4)
Anomalous Collocations Formulae Metaphors
Ill-formed 10% Simple Formulae 70% Transparent 37%
Cranberries 4% Sayings 2% Semi-transparent 51%
Defective 62% Proverbs 19% Opaque 12%
Phraseological 24% Similes 9%
 63

On the other hand, lexicogrammatical fixedness – or formal rigidity – implies some degree of
lexicogrammatical defectiveness in units, for examples with preferred lexical realizations and other
restrictions on aspect, mood, or voice. Based on corpus evidence, Moon (1998) suggests that the
variation among lexical phrases is particularly noticeable along the fixedness continuum that we
should think in terms of ‘preference of form’ rather than ‘fixedness of form’. Investigating a wide
range of fixed expressions against the OHPC, Moon (1998) found that 40% of the items she studied
regularly varied in the form they took indicating that all multiword units are fixed to various degrees
with idioms tending to have the most restricted formal structure of all fixed expressions. In this light,
the commonest types of lexical variability among fixed expressions included: (i) verb variation
where no real change in meaning of a fixed expression occurs although there may be register
dictinctions (e.g. blow up in ones’s face/explode in one’s face, bend/stretch the rules, hang in the
air/be left hanging in the air, be/come within a ace of something), (ii) noun variation (e.g. a
piece/slice of the action, at all events/at any event, jobs for the boys/girls, a ballpark
figure/estimate), (iii) adjective and modifier variation (e.g. a bad/rotten apple, bleed someone
dry/white, a different/another kettle of fish, all/more power to one’s elbow), (iv) particle variation
that involves no apparent shift in meaning (e.g. by/in leaps, in/at full throttle), (v) conjunction
variation (e.g. hit and/or miss), (vi) cases of specificity and amplification where the variations
consists of some inserted or suppressed material that is often either adjectival (e.g. cut the (umbilical)
cord) or adverbial (e.g. (down) on one’s uppers), (vii) cases of truncation whereby traditional
proverbs and sayings are downgraded from their canonical or earliest forms to lower-level
grammatical units or a compound sentence to a single clause or a clause to a group (e.g. a bird in the
hand (is worth two in the bush)), (viii) reversals (e.g. day and night, night and day), (iv) register
variation where the first variation is usually more formal than the second (e.g. beat one’s
breast/chest), (x)British/American variations (e.g. cut a long story short (BrE), make a long story
short (AmE)), (xi) spelling variations that sometimes reflect historical or etymological developments
(e.g. rack and ruin, wrack and ruin).

Additionally, fixed expressions were also found to differ in terms of the grammatical operations
they undergo that include the following based on OHPC corpus search of fixed expressions: (i)
variation in the expression of the notion of ‘possession’ or indication of attributes (e.g. get/keep/have
one’s eye in), (ii) causative variation that denotes a state, process or action (e.g. come to a head,
bring something to a head) and resultative variation that explicitly mentions the result of the state,
process or action (e.g. beat someone black and blue), (iii) variation in aspect where one variation is
effectively a continuative, generally signaled by the verb keep (e.g. cross one’s fingers/ keep one’s
 64

fingers), (iv) reciprocity in which cases variations reflect the ways in which different participants are
mentioned (e.g. show one’s true colours, reveal one’s true colours), (v) case relationships involving
fixed expressions with beneficiaries so that both ditransitive and prepositional structures are found
(e.g. drop someone a line/ drop a line to someone) and (vi) delexical structures where a verb in one
variation corresponds to a cognate noun or adjective and (often) delexicalised or support verb in the
other variation (e.g. bloody someone’s nose, give someone a bloody nose). Again, idioms allow the
minimum degree of lexical and grammatical flexibility and are therefore transformationally defective
(see also Fraser, 1970).

Moon’s (1998) study also focused on the discoursal functions of fixed expressions following
related classifications developed in earlier work by Strassler (1982) and Fernando (1996). Broadly
speaking, Moon (1998:219) categorizes fixed expressions according to their primary text functions
into five categories, i.e. (i) Informational fixed expressions occurring in 41% of the cases in the
corpus that convey new information and contribute to a discourse propositionally (e.g. rub shoulders
with, in the running). Conveying information is a very basic language function, and informational
fixed expressions are the most basic of all consisting predominantly of anomalous collocations
(under 60%) whereas few of them were formulae (under 10%). According to Moon (1998:223),
anomalous collocations are largely classified as informational as they are lexicogrammatically non-
compositional: they are often unmarked and uniteresting ‘ways of saying things’, adding little to the
text stylistically. The conveying of information is a creative, text-forming process, and is less likely
to be marked and stereotyped a fact that precludes the use of formulae that are by definition
stereotypes. (ii) Evaluative fixed phrases occurring in 38% of the cases in the corpus that convey a
speaker’s evaluation and attitude (e.g. kid’s stuff, near the knuckle). Evaluative fixed expressions are
often metaphors in 47% of the cases found in the corpus. (iii) Situational fixed phrases occurring in
5% of the cases in the corpus that relate extralinguistic context and respond to a situation (e.g. excuse
me!, long time no see). These mostly consists of conventionalized utterances, routines, and gambits.
They are typically formulaic, compositional strings of words with pragmatic functions of, in
particular, rituals of thanking and apologizing. (iv) Modalizing fixed phrases occurring in 10% of the
cases in the corpus that convey truth values, advice, requests (e.g. I kid you not, you know what I
mean). Over half of modalizing fixed phrases were classified as anomalous collocations, and around
a third are formulaic. Last, (v) organizational fixed phrases that occur in 6% of the cases in the
corpus and whose function is to organize a text and signal discourse structure (e.g. by the way, for
instance, be that as it may).
 65

Besides their primary text functions, it was also found that in 47% of the cases, fixed expressions
had two or more text functions contributing to their texts in two or more ways. For instance, many
informational and evaluative fixed expressions also had modalizing secondary functions, typically
epistemic. On stream, take place and come to light further the narrative in the examples below and
make statements about the way things are or seem. Similarly, the tip of the iceberg and go/run deep
evaluate and modalize by suggesting the extent of the phenomenon:

(i) The fully fledged system of departmental reports will come on stream in 1991.
(OHPC: journalism)
(ii) But had the market boom continued a few more weeks, the whole thing might never
have come to light. (OHPC: journalism)
(iii) The usually veiled criticism contained in the report material, and the open comment
coming to light in denunciations and court prosecutions, have necessarily to be regarded as
the tip of the iceberg. (OHPC: non-fiction)

Additionally, a few informational and evaluative fixed expressions also have some organizational
function. For example, the pros and cons gives information and also signals contrasted aspects of the
matter under discussion. Similarly, last but not least evaluates and signals the end of a sequence, and
the conventional wisdom both evaluates and signals an appeal to shared beliefs:

(i) To see why, start with a look at the pros and cons of Mr Codd’s creation. His relational model is
a theory of what data is and how it can be retrieved. (OHPC: journalism)
(ii) They did, however, tend to draw upon academic illusionism, early modernism, or last
but not least, on mass-media imagery. (OHPC: non-fiction)
(iii) It was conventional wisdom in the 1960s that a trade gap could be closed, without
sacrificing growth, by a lower pound. (OHPC: journalism)

In addition to the primary functions described above, Moon (1998:241; 1992) also refers to the
phenomenon of cross-functioning of fixed expressions when used with a function other than an
additional to its primary one. In this view, cross-functioning operates instantially and relates to the
behaviour of individual fixed expressions in individual contexts whereby a speaker/writer uses them
in functions other than their canonical ones, thereby foregrounding or thematizing the selection. A
typical example is that of informational or evaluative fixed phrases used as signaling devices: a text
organizer or speech gambit. Consider, for example, a newspaper column which begins:

I must nail my colours to the mast. I’m a very keen advocate of all sorts of sport for all sorts of people
at all ages, but intensive sport or intensive training for sport could surprisingly, have side effects.
(Daily News, Birmingham, 4 June 1987:10)

Nail your colours to the mast is here used as a preface and as an emphasizer of frankness, rather than
simply to convey information. Like lay your cards on the table and many other expressions, it
signals a report and will normally be accompanied in text by a statement of whatever has been
 66

transmitted so frankly or clearly. Such uses can also be observed in spoken interaction. According to
Moon, (1992:22), in a television interview during the campaign for the 1987 general election in
Britain, the interviewer made three consecutive attempts to interrupt the politician he was
interviewing and to grab the turn. He used fixed expressions in order to so in addition to evaluating
the interviewee’s arguments: ‘You want the best of both worlds’, ‘You’re just beating around the
bush’, and ‘You’re trying to have it both ways’. In a comparable way, Moon (1994) also notes that
proverbs and sayings are not only used didactically but also to evaluate or preface while idioms can
also be used to clarify, summarize and evaluate in line with Drew and Holt’s (1995) discussion in
reference to the use of idioms as signals of topic transition in naturally occurring conversations.

Then only last week, the Director of Public Prosecutions for Northern Ireland and Sir Patrick
Mayhew, the Attorney General (both, incidentally, with spotless criminal records) agreed that no one
should be prosecuted for attempting to pervert the course of justice – not because these things hadn’t
happened, but because putting them in the spotlight of British Justice would ‘not be in the public
interest’. That is to say: it might open up a can of worms.

Exploring 300 idioms and other fixed phrases in a corpus of telephone conversations, Drew et al
(1995) found strong evidence of idioms being used to signal topic transition, summarizing or
assessing a segment of the conversation since ‘the production of an idiom is mutually oriented to or
by participants as bring one topic to a close and providing an opportunity for introducing a next
topic’ (Drew et al, 1995:124). They report and analyse many cases where fixed phrases are
produced, associating them with disagreements between speakers in which case one or both speakers
seems to employ idiomatic formulations of their position in their attempt to close the matter by
getting the other participant to agree with that position.

Clearly, Moon’s (1998) research is not conclusive nor could it be, given the comparatively small
and unbalanced corpus she was working with. Idioms are not studied separately as a distinct category
of lexical units but, rather are subsumed under the broad class of fixed expressions under
investigation in this study. According to Moon (1999), corpus-based research can provide evidence
about idiom and idiom-like expressions in terms of the frequencies, distributions, and currencies of
idioms; of their genre preferences; their conventions of syntactic and collocational behaviour; their
formal instability and variability; and their conventions of discoursal behaviour and connotations.
Acknowledging her study limitations, the researcher supports the need for huge corpora to provide
robust information for idioms and expressions leading towards new lexicographical models of
idioms and fixed expressions as reflected in their coverage and treatment in current learners’
dictionaries designed for both decoding and encoding purposes (Moon, 1992b;1996, Gottlieb, 1992).
 67

Still, the scope of the study remains too broad for our purposes as it describes the formal and
pragmatic features of a wide set of fixed expressions under which the linguistic treatment of idioms
is subsumed due to the pronounced unavailability of large enough corpora whose search would
provide a sufficient description of idioms from a theoretical point of view.

Idioms have also constituted a subject of inquiry in the field of psycholinguistics, and for that
reason the following section aims to present some of the most significant models proposed in the
area of idiom acquisition and comprehension as precursors of idiom investigation in SLA.
 68

Chapter 5

The Psycholinguistic Approach

The classification of idioms as noncompositional, unitary word strings as outlined above, is also
evident in early psycholinguistic studies and models of idiom comprehension. During the past 20
years, a growing body of predominantly psychological and neuropsychological approaches has
begun to rigorously scrutinize the nature of idioms across languages giving rise to a series of
competing models with respect to idioms’ lexical representation and processing. Glucksberg
(1993:4) assigns them to two distinct classes, i.e. (a) Direct look-up models where idioms are
comprehended by direct memory retrieval of their arbitrary meanings, not by linguistic processing
and (b) Compositional models that will be discussed in separate section below. The Idiom List
Hypothesis (Bobrow, et al, 1973), the Lexical Representation Hypothesis (Swinney et al, 1979) and
the Direct Access Hypothesis (Gibbs, 1980, 1984; Schweigert, 1986) consist three of the most well-
known versions of the traditional non-compositional model of idiom analysis.

The Cognitivist Model will be additionally presented as an alternative approach that seeks to offer
another solution to idiom interpretation based on the principle of semantic decomposition.

5.1. The Idiom List Hypothesis (Bobrow and Bell, 1973)

The idiom list hypothesis, also known as the literal first hypothesis, was first introduced in 1973
by Bobrow and Bell in order to provide an explanation for the way idioms are lexically represented
in the mental lexicon and the way they are processed by native speakers. Based on experimental
evidence, the authors proposed a dual processing model founded on the assumption that idioms are
represented in a mental idiom list – hence the name for the hypothesis - that is, an idiom lexicon that
parallels the mental word lexicon but is distinct from it. These two distinct but separate modes of
idiom processing were assigned an active and inactive processing role respectively, i.e. a literal mode
– the normally active one – which processes the literal meaning of the idiom first and a figurative
mode – the normally inactive one – which becomes active and processes in a delayed fashion the
figurative meaning of the idiom when the literal meaning does not fit the context in which the
idiomatic expression is situated. This served as evidence that idiomatic meaning is always preceded
by the computation of the literal meaning which is substituted to a search of the idiom list when
linguistic analysis of idiomatic expressions fails to yield an interpretable result.
 69

The basis for the formulation of the hypothesis was an experiment that employed a perceptual set
paradigm in which subjects were first presented with either a set of four sentences which had only
literal interpretations or four sentences which contained idiomatic phrases followed by a
grammatically idiomatic (and therefore ambiguous) sentence. Subjects were asked to report which
meaning of the grammatical idiom string they first perceived in each of the conditions. Overall, when
these were compared to a baseline condition, the default mode of processing seemed to be the literal
interpretation of the idiom test sentence, whereas the idiom-processing mode seemed to be active
only when the participants were presented with the sentences containing idioms. Only after seeing
the set of sentences containing idioms did the participants interpret the test sentence figuratively. The
reasoning behind this conclusion was that if subjects were able to find or to avoid finding the
idiomatic meaning of a sentence by virtue of the presence of a processing set and – in the absence of
any specific semantic or syntactically necessitated choice – separate modes of processing must have
been called into play.

Despite the intuitive appeal of this first model, findings from later studies have led to the rejection
of the idiom list hypothesis as it was found that idioms are understood at least as quickly as
comparable literal expressions. In a series of experiments that timed the speed of recognition of
idiom meanings, Ortony and his associates (1978:475) provided evidence that idioms do not take
longer to comprehend than literal uses of those same expressions, while in some instances they also
seem to be processed more quickly than literal language. The data also suggested that the processes
required for the comprehension of figurative and literal uses of language are essentially similar in
terms of contextually generated appropriate schemata and their relatedness to the idiom meaning as
well as idiom familiarity itself. Criticisms against the validity of the literal first hypothesis became
strongest with the introduction of a new model proposed by the publication of the Swinney and
Cutler (1979) study.

5.2. The Lexical Representation Hypothesis (Swinney and Cutler, 1979)

The second model for idiom processing is most commonly known as the lexical representation
model, or else referred to as the simultaneous processing hypothesis. Introduced during the end of the
1970s, this model accounts quite nicely for the relative ease of understanding familiar idioms. Idioms
are represented simply as holistic long entries, together with all the ordinary words in the mental
lexicon. When a familiar idiomatic expression is encountered, linguistic processing proceeds
normally. Lexical access proceeds as part of linguistic processing, and lexicalized phrases such as by
 70

and large or kick the bucket are routinely found in the mental lexicon along with their constituent
words, by, and, large and others. Which of the two meanings - literal or idiomatic – is apprehended
first depends on the relative speed with which full linguistic processing and lexical/idiom access can
be completed. Normally, idiom access will be completed more quickly because it does not require
the lexical, syntactic, and semantic processing required for full linguistic analysis. Thus, familiar
idioms will be understood more quickly than comparable literal expressions.

In the timed experiments that offered empirical evidence for this hypothesis, participants viewed a
word string on the computer and had to decide whether or not the string formed a meaningful, natural
phrase in English. In addition to meaningful strings, the list contained in random order of appearance
idioms and literal phrases: for example, take him for a ride/take him for a beer or wrap it up/lift it up.
The rationale behind this Phrase Classification Task was that candidate strings be analyzed for
sensibleness as a unit. If idiomatic meanings were computed by reference to a special idiom list, via
some special mode of processing which is instigated following an attempt at literal computation, the
phrase classification decisions should take longer, or at least no less time, for grammatical idioms
than for non-idiomatic phrase controls. If, on the other hand, if the Lexical Representation
Hypothesis held, decisions made to idiomatic strings should be faster than those made to literal word
string controls.

Results showed that participants responded significantly faster to idioms than to matched control
phrases. Based on these findings, Swinney and Cutler (1979:525) were able to confirm their
hypothesis that ‘idioms are stored and retrieved from the lexicon in the same manner as any other
word’, rejecting thus the idiom list hypothesis by stating that ‘there is no special idiom list nor any
special processing mode…individual words are accessed from the lexicon and structural analysis is
undertaken on these words at the same time that the lexical access of the entire [idiom] string (which
is merely a long word) is taking place’. Subjects in an experiment on recall syntactically ill-formed
idioms in Portuguese tended to detransform such items and recall the base form instead as unitary
lexical items exhibiting more resistance to paraphrase than control, nonidiomatic, literal strings
(Botelho da Silva et al, 1993).

Thus, unlike the dual processing model proposed by Bobrow and Bell (1973) above, Swinney and
Cutler (1979) introduced the concept of a single phrase processor, which, while not rejecting entirely
Bobrow and Bell’s model, gave greater emphasis to a single mental lexicon. The assumption of a
single phrase processor that simultaneously computes both the literal and the figurative meanings of
 71

idioms when the first word of an idiom string is encountered provided an essential insight and gave
rise to the corresponding theory in the field of psycholinguistics that has become one the most
influential model of idiom processing and representation to date.
5.3. The Direct Access Hypothesis (Gibbs, 1980; 1984)
The 1980s saw the emergence of yet another model known as the direct access hypothesis that
was originally proposed by Gibbs in 1984. Though primarily aimed as an alternative to the previous
two models, the hypothesis is largely considered as an extension of the lexical representation
hypothesis as it posits that linguistic processing of an idiomatic expression may be entirely bypassed
if that expression is recognized immediately as an idiom. In essence, idiom meaning access may be
so rapid as to obviate any linguistic analysis at all. The literal meaning of idioms is not important for
their comprehension and it is always preceded by an analysis of the figurative meaning of the idiom.

In a series of experimental studies of psycholinguistic nature, Gibbs (1980;1984) provided

evidence in favour of the idea that people are automatically biased toward the nonliteral meanings of
some instances of figurative language such as indirect requests, sarcasm as well as idioms when used
in literal contexts. Thus, in Gibbs (1980), results showed, among other things, that the
conventionality of the idiom affects how easily it is understood and supported the researcher’s belief
that native speakers do not need to interpret the literal meaning of a common idiomatic expression
before deriving its figurative meaning. When presented with short vignettes that each set up the
context for either a figurative or a literal interpretation of a concluding idiom, subjects in the
experiment tended, most often than not, to choose the figurative meaning more quickly than they
chose the literal meaning leading Gibbs to observe that native speakers do not process idioms
literally by default. That processing of an idiom’s idiomatic meaning precedes processing of its
literal meaning is also supported by Schweigert (1986) who reported a significant idiom familiarity
by sentence type interaction with less familiar idioms requiring more reading time than familiar
idioms in both literal and idiomatic sentences (F (2,92) = 4.06, p< .02).

Yet, this position faces the same problems as Swinney and Cutler’s (1979) lexicalization
hypothesis. Even if literal meanings could be entirely bypassed, the direct access hypothesis still
requires some mechanism for accessing idiomatic rather than literal meanings. What cues lead
listeners to seek idiomatic meanings in the first place? The direct access hypothesis also requires an
exact match between an input string and a stored idioms due to the possibility of idiom syntactic
variants that can yield the same idiomatic meanings (e.g. ‘Mary was just letting off some steam’,
 72

‘Some steam was let off by Mary’, and ‘Mary let some steam off’ can all be recognized as the same
idiom.

Based on his observations of lexical and syntactic phenomena in idiom use, Gibbs himself (1995)
soon adopted a more compositional view of idiom comprehension best expressed in his idiom
decomposition model, the most important aspects of which will be discussed below.

5.4. The Cognitivist or Decompositional Approach

Quite apart from the traditional non-compositional view of idiom analysis that treats idioms as
multi-word expressions with arbitrarily stipulated meanings which can simply be retrieved from a
special mental lexicon in order to be understood, the cognitivist approach seeks to offer another
solution to idiom interpretation based on the principle of semantic decomposition. A new wave of
studies on idioms have lately been undertaken that lend support to the idea that idioms do not form a
unique class of linguistic items, perhaps as ‘dead’ metaphors, but share many of the compositional
properties normally associated with more literal types of speech. Originally, the evidence came from
a series of psycholinguistic studies (e.g. Gibbs et al, 1989a, Nayak et al, 1990) indicating that
speakers have reliable intuitions about idiom analyzability or compositionality which can also
determine the syntactic behavior, ease of comprehension as well as the semantic productivity of
idioms. Geeraerts (1995:61) more eloquently redefines the notion of compositionality in terms of
isomorphism as follows:

…..a one-to-one correspondence between the formal structure of the expression and the
structure of its semantic interpretation, in the sense that there exists a systematic correlation
between the parts of the semantic value of the expression as whole and the constituent parts
of that expression’. (Geeraerts, 1995:61)

The idea that the syntactic behaviour of idioms can be determined by the internal semantics of
these figurative phrases found its best expression in the idiom decomposition hypothesis as proposed
by Gibbs, Nayak and Cutting in 1989. This proposal suggests that, on encountering an idiom, people
do not inhibit or shut down their normal language-processing mode. Instead, speakers make
assumptions about the degree of semantic analyzability or decomposability of idioms that directly
affects idioms’ syntactic versatility as well (e.g. Fellbaum, 1995). Their syntactic parser
automatically analyzes the grammatical structure of the words and phrases they hear or read; their
lexical processor accesses the lexical items in the mental lexicon and assigns a meaning to them; and
a semantic analysis is undertaken on the basis of the grammatical structure and the meaning of the
 73

lexical items of the phrase (d’Arcais, 1993:85). The process of idiom comprehension is complete
when ordinary linguistic processing combines with a pragmatic interpretation of the idiom in
discourse contexts (Gibbs, 1995; Tabossi et al, 1995).
Evidence from a series of reading-time experiments (Gibbs et al, 1989) showed that where
participants had to decide whether a given word string formed a meaningful English expression,
subjects needed significantly less time to process decomposable idioms, i.e. idioms in which the
figurative and literal meanings are close (e.g. hit the jackpot), than to process nondecomposable
idioms, i.e. idioms in which the literal meaning offers no clue for the construction of its figurative
meaning (e.g. kick the bucket). These data suggest that people try to analyze an idiomatic expression
compositionally during processing, much like they analyze a literal expression. They try to assign
independent idiomatic meanings to the individual parts of the idiom, which then can be combined to
form its figurative interpretation as a whole.

However, this assumption implies that access of the meaning of an idiom is dependent on the
extent to which an idiom can be compositionally analyzed, i.e. meaning access is dependent on the
compositionality of an idiom. Thus, according to the idiom decomposition hypothesis, there are three
classes of idioms based on their degree of semantic compositionality. (a) Normally decomposable
idioms, for instance pop the question. Idioms of this class are normally analyzable because their parts
– in this instance pop and the question – directly refer, respectively to ‘suddenly ask’ and ‘marriage
proposal’. (b) Abnormally decomposable idioms, such as carry a torch for someone. This idiom is
characterized as abnormally analyzable as the word torch in this idiom does not refer directly but,
rather, metaphorically to ‘warm feelings’. (c) Nondecomposable idioms such as for instance kick the
bucket cannot be analyzed in such a way that the individual parts refer directly or indirectly to the
individual parts of the act of dying.

Normally and abnormally decomposable idioms are processed faster than literal expressions due
to the familiarity of the idiomatic word combinations as compared to the compositional analysis of
literal phrase, which are not familiar as well as than noncompositional idioms whose meaning can be
recovered only by a ‘double-take’ action, which leads to more processing time. Thus, when
processed normally and abnormally decomposable idioms yield some meaning-either figurative or
literal-for the individual parts can be modified and moved like the parts of any literal expression. For
instance, internal modification does not impair idiomatic processing of a decomposable idiom (e.g.
He left no legal stone unturned will be processed idiomatically. In this case of noncompositional
idioms, though, the initial compositional analysis fails because their individual parts are semantically
 74

empty and their stipulated meaning must be recovered from the mental lexicon. Parts of a
noncompositional idiom can be neither modified nor moved. For instance, the noncompositional
idiom kick the bucket cannot be internally modified by an adjective without losing its idiomatic
meaning - *He kicked the empty bucket – hence, being syntactically frozen. However, when
modification is semantically external - usually, if not always in the form of viewpoint modifiers such
as from a political point of view, visually, army-wise, etc - the idiomatic processing of
noncompositional idioms (e.g. Greenpeace is keeping ecological tabs on the French government’s
tests in the Pacific) is not impaired (Nicolas, 1995; Van de Voort et al, 1995).

In later work, Gibbs (1993; 1995) also suggested that this idea of analyzability in idiom
comprehension might open up the possibility that the internal semantics of idioms be correlated in
systematic ways with the concepts to which they refer, claiming that the figurative meanings of
idioms might be partially motivated by people’s conceptual knowledge. By studying the relationship
between metaphors and idioms, Gibbs and his associates (1990) advocated that people must have
strong conventional images for many idioms and that the regularity in people’s knowledge of their
images for idioms is due to the conceptual metaphor motivating the figurative meanings of idioms.
Uncovering speakers’ tacit knowledge of the metaphorical basis of idioms through an examination of
speakers’ mental images for idioms, the researchers found a high degree of consistency in
participants’ descriptions of their mental images for idioms with similar meanings because of the
constraints conceptual metaphors (e.g. the MIND IS A CONTAINER, IDEAS ARE PHYSICAL
ENTITIES and ANGER IS HEAT) impose on the links between idiomatic phrases and their
nonliteral meanings. Conceptual metaphors bring into correspondence two domains of knowledge:
one is the familiar, physical or source domain and the other is the less familiar, abstract or target
domain. Knowledge of the metaphorical links between different source and target domains provides
the basis for the appropriate idiom interpretation and use in different discourse contexts.

Thus, by using the idiom spitting venom, the domain of venom is used to understand the domain
of hatred/speaking ill of someone, i.e. hatred is comprehended via the concept of venom (see Moon,
1994:170 for more examples of conceptual metaphors). In this sense, idioms appear to be closer to
metaphor as their meanings are complex semantic configurations motivated by a small set of
conceptual metaphors. For instance, the figurative meaning of an idiom such as spill the beans
indicates something about the cause of the secret revealed, the intentionality of the person doing the
revealing and the manner in which the revelation occurs based on two different conceptual
metaphors, i.e. the conceptual of the mind as a container and that of ideas as physical entities. The
 75

availability of these conceptual metaphors in semantic memory renders the expression spill the beans
easy to understand and also semantically analyzable, i.e. people can apprehend the relation between
the literal meaning of the idiom’s constituents and its figurative meaning. Though the contribution of
mental imagery, as outlined above, appears crucial in idiom processing, its role still remains fairly
controversial with respect to the kinds of images – the concrete literal or the abstract figurative – and
to the extent each of these types are reflected in the mental images people produce in idiom
comprehension. Recent research findings on the issue (Cacciari et al, 1995) seem to indicate that the
images tend to reflect the concrete actions and events denoted by an idiom’s constituent words and
phrases, not idiomatic meanings or the conceptual metaphors that presumably underlie those
meanings.

In a series of related studies, Cacciari and her associates (Cacciari, et al, 1991; Cacciari, 1993)
attempted to provide further evidence for the availability and use of the constituent word meaning in
discourse. Based on her findings, not only do the meanings of the individual words in idioms
contribute directly in their immediate comprehension but they also play a significant role in
determining their lexico-grammatical flexibility in terms of semantic productivity, as when people
use the semantics of an idiom’s elements to create variants of familiar idioms (e.g. no matter how
terribly they tortured him, he didn’t spill a single bean!) and of discourse productivity, when the
semantics of idioms’ elements used in a conversation may constrain its further use and provide the
bases for elaborations and responses. In a conversation between Tom and Joe:

Tom: Did the old man kick the bucket last night?
Joe: Nah, he barely nudged it.

The verb phrase ‘kick the bucket’ permits responses and elaborations that play on the semantics of
the verb ‘to kick’. Accordingly, the researchers expect that different types of idiom might function
differently in conversation depending on the nature of elements comprising the idiom and how those
elements contribute to the meaning. In this sense, idioms are assigned to four different categories, i.e.
(a) totally opaque idioms (e.g. kick the bucket), (b) retrospectively transparent idioms, i.e. transparent
once you either know the meaning or are reminded of the episode or setting that originated the idiom
(e.g. break the ice), (c) directly transparent idioms, the sense of the words leading to the idiomatic
meaning of the string, perhaps by means of the recreation of an analogical or metaphorical mapping
(e.g. spill the beans) and (d) figuratively transparent or quasi-metaphorical mapping, i.e. idioms
composed of other idioms, or parts that appear in other idioms or as metaphorical vehicles (e.g. carry
coals to Newcastle). According to Cacciari and Tabossi (1988), idiom meanings are associated with
 76

certain word configurations – hence, the Configuration Hypothesis - and become available when
sufficient input renders these configurations recognizable.

The individual words that participate in the configuration are the same lexical items that are
accessed ordinarily during comprehension (e.g. take in a sentence like ‘The boy took the book and in
an idiom like ‘take the bull by the horns) but some words are more important than others for
detecting the configuration. Such a word is called a key and, according to the hypothesis, it is
suggested that an idiom can be recognized only when the (final) key has been accessed – until that
moment the word string is processed literally. However, questions related to the identification of an
idiom’s key as well as its subsequent activation in the process of idiom comprehension are largely
unaccounted for in the model, though recent research findings from spoken language processing
(Tabossi et al, 1992; Tabossi et al, 1993) show that when the key is the first content word in a string,
idiomatic meaning becomes available soon after it.

Idiom compositionality in terms of conceptual processes as established by experimental work in

psycholinguistics has also been the premise of research in the area of cognitive semantics. Kovesces
and Szabo (1996:330) characteristically claim that the meaning of idioms is motivated and not
arbitrary, by pointing out that:

An idiom is not just an expression that has meaning that is somehow special in relation
to the meanings of its constituent parts, but it arises from our more general knowledge
of the word (embodied in our conceptual system).
(Kovesces & Szabo, 1996:330)

According to the researchers, conceptual metaphors are largely seen as conceptually motivating the
use of words such as spark off, burn the candles, fan the flames etc. in the idioms in which they
occur. Given these conceptual metaphors, it becomes clear why the idioms have the general meaning
that they do; that is, why they have to do with anger, love, imagination, etc. respectively. The reason
seems to be that these conceptual metaphors exist and they serve as links between two otherwise
independently existing conceptual systems. Because of the connections they make in out conceptual
system, the conceptual metaphors allow us to use terms from one domain (e.g. fire) to talk about
another (e.g. anger and love). The idioms that employ these terms (such as those of fire) will be
about certain target domains (such as anger) as a result of the existence of conceptual metaphors
(such as ANGER IS FIRE).
 77

However, conceptual motivation for idioms does not arise only from the existence of conceptual
metaphors. Conceptual metonymy and conventional knowledge are considered as two further
cognitive mechanisms that are also involved in the process. Use of conventional knowledge refers to
the shared information that people in a given culture have concerning a conceptual domain, like the
human hand. This shared everyday knowledge includes standard information about the parts, shape,
size, use, and function of the human hand, as well as the larger hierarchy of which it forms a part
(hand as a part of the arm, etc). In this sense, the meaning of an expression such as with an open
hand comes from the image of a person physically giving objects to another with an open hand
implying the knowledge that nothing is held back and everything can be taken. This stands in marked
contrast to the knowledge about the image of a person who gives with his fist held tight.

On the other hand, metonymy is distinguished from metaphor in such as way that metonymy is
characterized as typically involving one conceptual domain, rather than two distinct ones and
involving a ‘stand for’ relationship between two entities (within a single domain), while metaphor
involves an ‘is’ or ‘is understood as’ relationship between two conceptual domains, such as anger
and fire. However, the researchers suggest that in most cases, motivation of idiom meaning comes
from a combination of both conventional knowledge and metonymy with the latter being the
dominant force and cognitive source. In the case of hold one’s hand the meaning arises in large
measure as a result of the metonymy THE HAND STANDS FOR THE ACTIVITY but it is also associated
with the further knowledge of holding one’s hand: when we hold our hands (i.e. when we arrest the
movement of the hand), we have temporarily stopped an activity waiting to see whether to continue
or how to continue the activity we have been engaged in. Thus, the metonymy THE HAND STANDS FOR
THE ACTIVITY and some further conventional knowledge jointly produce a large part of the
motivation for the idiomatic meaning of the expression hold one’s hands. This way, idiom semantic
transparency is more explicitly specified as a pedagogical term that plays a decisive role in both
teaching and learning process of idioms in L2. In line with Irujo (1993:217), the researchers suggest
that the notion of transparency or motivation of idioms arising from the cognitive mechanisms of
metaphor, metonymy and conventional knowledge provides students strategies for dealing with
figurative language by figuring out the meaning of an idiom by themselves based on a link from the
idiomatic meaning to its literal words.

Along similar lines, Baranov et al (1996) also asserts that the product of semantic derivation of
actual meaning in idioms derives from specific cognitive structures and cognitive operations in the
form of frames (or ‘conceptual structures’) designed to represent a stereotypical situation or scripts,
 78

i.e. particular types of frames whose specificity comes from the fact that it is oriented towards the
sphere of action described in an algorithmic form, or towards such types of knowledge which are
procedural in nature (i.e. composed of separate procedures or small actions). According to the
authors (Baranov, et al, 1996: 417-428), in the process of semantic derivation complex
transformations arise on the basis of elementary transformations and lead to a modification of frames
and to a generation of actual meaning of idioms. Such complex transformations include, for example,
the combination of several frames into a single frame, the integration of one or more frames into the
same conceptual area, the transposition of structural characteristics from one frame to another. This
typology of slots transformation covers the bulk of the procedures of conceptual summarization
which result in ‘the packing or compressing of knowledge in linguistic structures, i.e. in the
ontologization of knowledge in language systems’. Idioms differ from all other lexical phenomena
only to the extent that their inner form allows one to reconstruct the cognitive procedures involved in
paradigmatic knowledge ontologization, which in itself is a reflection of their mutability (Baranov et
al,1996:428).

Pursuing the claim that idioms are non-arbitrary, transparent expressions, Nunberg et al
(1994a:497) propose to distinguish idiomatically combining expressions (e.g. to take advantage, to
pull strings), the meanings of which, while conventional, are distributed among their parts, from
idiomatic phrases (e.g. to kick the bucket, to saw logs), which do not distribute the meaning of their
components. This proposal was meant to prove that the majority of phrasal idioms are in fact
semantically compositional and that the very phenomenon of idiomaticity is fundamentally semantic
in nature. In contrast to previous taxonomies offered by Gibbs et al (1989) and Cacciari et al (1991)
that describe differences in one semantic dimension of idiomatic expressions, i.e. compositionality,
Nunberg et al (1994a) argue that idioms differ along three orthogonal semantic dimensions, namely,
compositionality, conventionality and transparency. In an effort to eliminate the confusion among
these three dimensions in studying idioms, the researchers claim that idiomatic combinations differ
from idiomatic phrases in terms of degree of compositionality present in the case of idiomatically
combining expressions and absent in the case of idiomatic phrases. Idiomatic combinations and
idiomatic phrase may nevertheless be more or less conventional and transparent. According to their
analysis, compositionality refers to the ease with which literal word meanings of idiomatic
expressions can be mapped onto components of idiomatic meaning once the idiomatic meaning has
been apprehended and is clearly known. In idioms such as grease the wheels, this mapping may be
readily apparent on first encounter but for an idiom such as get your goat, the mapping may only
become apparent with accumulated experience with the idiom.
 79

Idiom conventionality, on the other hand, reflects the degree to which a speaker would be able to
recover an idiomatic meaning in the absence of all contextual information, i.e. understand an idiom
for the first time, in a particular communicative environment, irrespective of how idiom parts map
onto idiom meanings. To illustrate the difference between compositionality and conventionality,
Nunberg et al (1994a) discuss the difference between an idiomatic combination such as to pull
strings and an idiomatic phrase like kick the bucket. Although the idiomatic meaning to die would
not be predictable given the word sequence kick the bucket for a naïve language user, individuals
who are members of a particular linguistic community would have no trouble understanding the
sequence given its conventionality within that linguistic community. In contrast, when we hear a
sentence like: John was able to pull strings to get the job, there is an element of conventionality
involved in the interpretation of this idiomatic phrase. Besides context that enables us to conclude
correctly that pull strings means something like ‘exploit personal connections’ even though we might
not have been able to predict that the phrase had this meaning if we had heard it in isolation,
conventionality also helps through a sense of collocation between the constituents of the expression:
strings can be used metaphorically to refer to personal connections when it is the object of pull, and
pull can be used metaphorically to refer to exploitation or exertion when its object is strings.
However, conventionality does not directly map onto idiom frequency. Although idioms have highly
conventionalized relationships with their figurative meanings, they may also differ in frequency of
occurrence in a language.

In addition to conventionality and compositionality, the transparency of idiomatic expressions

refers to the degree to which the original motivation of these phrases is immediately accessible. To
illustrate transparency, consider the idiom jump the gun. According to the researchers, this idiom
would be classified as transparent insofar as its idiomatic meaning, start ahead of time, is directly
related to a literal interpretation of the phrase in which jumping the gun at the start of a race refers to
beginning a race before one is supposed to. In this way, jump the gun is both transparent and
compositional because both the motivation for the expression and identification of how the
individual parts map onto the expression are easily discernible. Nunberg et al (1994a:497) suggest
that transparency is orthogonal to compositionality such that idioms like saw logs are transparent (i.e.
it is clear how this expression motivates the idiomatic meaning to sleep) but not compositional (i.e. it
is not clear how the individual components of the phrase map on the idiomatic meaning to sleep).

The proposal that parts of the idiom should be assigned interpretations, contributing to the
interpretation of the whole idiom, was supported by the claim that these parts can be modified (e.g.
 80

leave no legal stone unturned), quantified (e.g. we could pull yet more strings), emphasized through
topicalization (e.g. those strings, he/she could not pull for you), omitted in elliptical constructions
(e.g. my goose is cooked, but yours isn’t) and can be antecedents for anaphora (e.g. although the
F.B.I. kept tabs on John, the C.I.A kept them on Lily) (Nunberg et al, 1994a:500-502). The
suggestion here seems to be that passivisation, the possibility of internal modification and
quantification, and the possibility of pronominal reference to or ellipsis of a part of a phrasal idiom
are loosely correlated properties clustering around the notion of lexically analysed, idiomatic
combinations (cf. Abeille, 1995 for French idioms). As for the syntactic versatility of idioms,
Nunberg et al (1994b:109) claim that it can also be accounted for in semantic terms as ‘a function of
how the meanings of their parts are related to one another and to their literal meanings’. In this sense,
the approach predicts a strong correlation between semantic analyzability and ‘transformational
productivity’ by positing a bifurcation between idiomatic phrasal constructions and idiomatically
combining expressions, with only the latter type permitting transformational processes. Thus, to the
extent that compositional semantic analysis of an idiomatic expression is possible, a lexical analysis,
i.e. an analysis that posits interpretationally interdependent words combining by general syntactic
principles, is to be preferred, from the point of view of the language learner. English topicalization,
for instance, the meaning which involves some sort of highlighting of the interpretation of one
constituent, cannot be used with an idiom unless the parts of the idiom carry identifiable parts of the
idiomatic meaning. Therefore, the claim here is that semantic analyzability is correlated with
passivization, internal modifiability, quantifiability, and possibility of anaphoric reference.

Based on the premise that the syntactic properties of idioms are predictable from their semantic
properties, Nunberg (1978) categorizes into two major types, i.e. non-analyzable and analyzable
idioms. Non-analyzable or ‘semantically non-decomposable’ idioms are those whose individual parts
do not contribute individually to their figurative meanings such as: kick the bucket, beat around the
bush, trip the light fantastic, give someone the cold shoulder, etc. In the processing of such idioms,
semantic and syntactic analyses are non-functional. Accordingly, spic and span, by and large should
be regarded as neither syntactically nor lexically flexible; therefore, such idioms can be treated as
special lexical entries, no different than ordinary long words where the individual morphemes do not
comprise the meaning of the word itself. The word meaning in these idioms do not map onto the
stipulated meaning of the idiom but they constrain both interpretation and use. Thus, the semantics of
the verb ‘to kick’ constrains both interpretation and discourse productivity. Our understanding of
what it means to die guides and constrains how the idiom kick the bucket may be used. People cannot
die sharply, and even though one can kick sharply, one cannot say: ‘He/she sharply kicked the
 81

bucket.’ Again, kicking is a discrete act, and so, even though one can say, ‘He/she lay dying all
week’, one cannot say, ‘He/she lay kicking the bucket all week’ (Wasow et al, 1983). This is because
one can kick the bucket all week if he/she kicks it over but one cannot die over and over again. Such
idioms were called compositional-opaque by Cacciari and Glucksberg (1991).

On the other hand, the processing of decomposable idioms requires that the semantic and
syntactic analyses be functional. In this type of idiom, some relationships between the idiom’s
elements and its meaning can be discerned. When the mapping from word meanings onto the
stipulated meaning of the idiom is direct, the idiom is classified as compositional and transparent and
belongs to the sub-group of normally decomposable idioms (see also Gibbs et al, 1989). In this case
there are one-to-one relations between the idiom’s words and components of the idiom’s meaning,
usually because of metaphorical correspondences between an idiom’s words and components of the
idiom’s meaning. For example, in the idiom pop the question, the noun phrase the question is
normally used to refer to a marriage proposal in contexts in which no other question is salient and the
verb pop is normally used transitively to mean ‘cause to emerge suddenly’ only with an explicit
adverbial, like out or up. Another example is the idiom break the ice, where the word break
corresponds to the idiomatic sense of changing a mood or feeling, and the word ice corresponds to
the idiomatic sense of social tension. On the other hand, is such mapping is indirect, the idiom is
classified as opaque but transparent belonging to the sub-group of abnormally decomposable idioms
(see also Gibbs et al, 1989). In this case, the relation between an idiomatic constituent and its referent
is not semantic, but rather is in some sense indirect. For instance, in the idiom spill the beans, there is
no direct semantic relation between beans and secrets. Nunberg explains that some idioms are
abnormally decomposable as each part does not by itself refer to some component of the idiomatic
referent but only to some metaphorical relation between the individual part and the referent. For
example, people understand the hitting of certain buttons in hit the panic button as a conventional
metaphor of how we act in extreme circumstances. On the other hand, some idioms of this type can
be classified as quasi-metaphorical when the literal referent of the idiom is itself an instance of the
idiomatic meaning; for example, giving up the ship is simultaneously an example of the act of
surrendering and a phrase that can refer to any instance of complete surrender. Included in this class
of idioms are such metonymic phrases as bury the hatchet, where the act of burying the hatchet was
once an actual part of the ritual of making peace, but is now used to refer to any instance of peace
making in its entirety.
Nevertheless, the compositional approach has not gone unchallenged as it leaves many aspects of
the phenomenon of idiomaticity unaccounted for. One important issue refers to the availability of the
 82

conceptual structures hypothesized: a conceptual structure may be available in semantic theory yet
may not be accessed for either production or comprehension. In this sense, although many idioms are
transparent and can readily be seen as corresponding to certain conceptual information (e.g. blow
your stack), other idioms have opaque metaphorical roots (e.g. trip the light fantastic) and people are
unlikely to recognize why these idioms mean what they do. The transparency of an idiom seems to
be very much a matter of degree with more transparent phrases being most easily tied to people’s
conceptual knowledge. The class of what Fillmore et al (1988:505) categorizes as extragrammatical
idioms with missing verbs and nouns (e.g. first off, all of a sudden, by and large, so far so good)
along with opaque idioms are also not particularly amenable to a conceptual analysis as their
structures as their structures, though grammatical are not intelligible by knowledge of the familiar
rules of the grammar and how those rules are most generally applied. Keysar and Bly (1999:1576)
also claim that idioms cannot be used to argue for the existence of conceptual structures suggesting
that the reasons for the perceived transparency or lack of transparency of idioms may have less to do
with conceptual motivation than with the nature of interpretive strategies used in our effort to explore
what possible meanings are transparent for specific idioms and what strategies are invoked to make
the different links.

A second point of criticism of the conceptual metaphor approach to idioms revolves around the
universality of conceptual mappings across different language and cultures. If the theory works, then
it would be possible, even for non-native speakers to make appropriate connections on the
conceptual level to interpret some of these expressions, even if encountered for the first time, and
even if they do not have the full lexical and structural equivalents in their own language, or have no
equivalent in their language at all. To this end, there is still not adequate evidence in the form of
relevant cultural inputs that help shape and compare the conceptual systems on a cross-linguistic and
cross-cultural basis. Despite its weaknesses, more recent research efforts strive towards an integrated
model of idiom representation and processing combining compositional and non-compositional
approaches on the belief that such as model
‘…..incorporates the word-like aspects as well as the metaphorical and compositional
derived aspects of idioms and is likely to be more generalizable across many classes of
idiomatic and other nonliteral sequences.’ (Titone et al, 1999:1672)

Though the advantage of such a model seems appealing in terms of an account of similarities and
differences in representation and processing of many classes of idiomatic sequences as well as other
non-literal sequences, we will not follow it in our analysis of idiom interpretation in this study. The
 83

working definition of idiom that we propose is wholly on traditional, non-compositional accounts

that most clearly delineate and offer distinct linguistic dimensions for the delimitation of the idiom
classes distinguished from the wealth and breadth of phraseological data in the English language. In
addition to the fact that the notion of conceptual metaphor and imagery as means in the process of
idiom interpretation appears to be rather fuzzy and subjective terms to be used as determinants of the
semantic transparency of an idiom we also believe that it has nothing to add to a definition of idiom
that is strictly based on mainly linguistic criteria.

5.5. A Working Definition of Idiom in English

This review of theoretical discussions of the nature of idioms shows that attempts to provide
categorical, single-criterion definitions of idioms are always to some degree misleading and ad hoc
as they fail to capture the phenomenon of idiomaticity as usually understood in its totality. In actual
linguistic discourse and lexicographical practice, ‘idiom’ is applied to a fuzzy category defined on
the one hand by such prototypical examples as kick the bucket, take care of NP, keep tabs on NP and
on the other by implicit opposition to related categories like formulae, fixed phrases, collocations,
clichés, sayings, proverbs and allusions. The problem is exacerbated by the fact that terms like
‘idiom’ itself inhabit the ungoverned area between lay metalanguage and the theoretical terminology
of linguistics. Idiomaticity seems to be a phenomenon too pervasive to be correlated with a specific
form of morpho-syntactic structure or the presence or absence of syntactic constraints. In a personal
communication with Fernando & Flavell, Randolph Quirk states:

‘The problem of idiomaticity is rather that most phenomena in language respond very well to
treatment by the procedures that have evolved for handling ‘syntax’ on the one hand and
‘lexicon’ on the other. But languages include entities that are hermaphroditic in sharing the
properties of both and responding satisfactorily to the descriptive procedures of neither’.
(Fernando & Flavell 1981:22)

To define idiomaticity we need multiple criteria, and each criterion must represent a single property

which could in turn take the form of a scale rather than a simple yes/no distinction. This would result in

varying degrees of idiomaticity correlating with different types or categories of idioms ranging along a

scale. Undoubtedly, the most satisfying and almost universally recognized criterion for idiomaticity is the

semantic one as is obvious in many idiom definitions presented in our summary in table 5.1. below. The
 84

ordinary non-specialist native speaker’s first thoughts about the definition of an ‘idiom’ are that it does

not mean what the individual words in it would lead you to expect. However, a definition based only on

semantic transparency as one dimension does not uniquely identify idioms with respect to compound

words for instance (Davies 1982:70). There are still other expressions that can be defined as

‘syntactically complex expressions whose meanings are not determined by the semantic properties of

the constituents and modes of combination’, for example, combinations such as carpet sweeper and

vacuum cleaner. In both cases, the meaning is not the sum of the meanings of the constituents and

their modes of combination alone: a carpet sweeper sweeps carpets, but a vacuum cleaner does not

clean vacuums; instead, it cleans by means of a vacuum. Yet, most experts would be reluctant to call

these phrases idioms. For this reason, a more complex definition of idiom is needed.

The five properties of idiom that have been most regularly invoked and summarized briefly here

by Flavell (1981:47) are the following:

1. The meaning of an idiom is not the result of the compositional function of its constituents.
2. An idiom is a unit that either has a homonymous literal counterpart or at least individual constituents
that are literal, though the expression as a whole would not be interpreted literally.
3. Idioms are transformationally deficient in one way or another.
4. Idioms constitute set expressions in a given language.
5. Idioms are institutionalized.

Acknowledging the multifarious character of idiomaticity in language, we believe that it would be

best to opt for a multi-dimensional definition with each property clearly delineated with the intention

to reach a narrow as well as a strictly linguistic definition of ‘idiom’ as a distinct lexical unit. Given

below is a summary of the defining features adopted for the characterization of a pure verbal phrasal

idiom as it is currently used for the purposes of the present study:

 85

A pure verbal phrasal idiom is:

1. lexically complex – i.e. it should consist of at least more than one lexical items that can freely

occur as single words outside the idiom.

2. grammatically well-formed, and for our study we have chosen to restrict ourselves to the

syntactic V + NP type (e.g. kick the bucket).

3. semantically not the result of the compositional function of its constituent parts, although not

necessarily totally opaque as some idioms may contain a metaphorical element that contributes to

their partially transparent meaning.

4. syntactically stable allowing the minimum of transformational potential in its structure.

5. a ‘double exposure’ due to the existence of a literal homonymous counterpart which

complements the fact that the syntax of an idiom is non-correlative, i.e. a pure idiom has both

a literal and a non-literal ‘face’.

My working definition of a verbal phrasal idiom is therefore:

A pure idiom is a non-literal, grammatically well-formed set expression that consists of more than
one familiar lexical items and allows the minimum of syntactic transformational potential in its
structure and whose meaning is not a compositional function of its syntactic constituents, but
which always has a homonymous literal counterpart.

In essence, our definition of idiom stresses two key features in particular: a non-correlative

syntax, i.e. the grammatical structure of an idiom does not contribute to the meaning of idiom

through a one-to-one literal translation of its word components, thus resulting in non-literalness and

homonymity. While Weinreich (1969) and Makkai (1972) choose the property of ‘disinformation’ or

‘potential ambiguity’ underlying our criterion (5) as the key feature of an idiom, we regard (3) and

(5) as complementary and of equal importance in our definition of pure idiom in view of the

delimitation of the term per se as well as of its use in the present study. On the one hand, the
 86

consequence of the two features of semantic non-compositionality and literal homonymity in a pure

idiom is that the mismatch between the apparent (literal) and the intended signification (often

metaphorical) is much greater than is the case with other figures of speech in general.

The fact that figures of speech such as metaphors, similes and metonymies and fixed phrases

manifesting peculiarities of encoding lack homonymous literal counterparts renders such expressions

structurally different from what we are calling here pure idioms. In fact, Glasser (1998; 1988)

considers idioms to be the prototypical class that lies in the centre of the phraseological system of

modern English due to some unique feature of form in the sense of fixed wording that is unaltered

and deficient in possible lexical substitution that might affect their idiomatic interpretation or

meaning in the sense of semantic non-compositionality. In this sense, idioms are differentiated from

other fixed yet non-idiomatic phrases in that the latter are characterized only by having fixed content

and order determined by custom which dictates that they be expressed by a particular selection and

order of words they have transparent meaning and include terminological word groups (e.g.

unconditional surrender), onymic entities, i.e. phrases which are proper names (e.g. the Black Sea,

the Golden Twenties), clichés, (e.g. an eloquent silence), (para) phrasal verbs (e.g. to take a walk),

and other set expressions (Gates 1973:70). Other non-idiomatic sentence-like phraseological units

that lack in the feature of non-compositionality of meaning include phrases such as: (i) fragments or

reductions of proverbs (e.g. a new broom sweeps clean), (ii) Proverbial sayings (e.g. to throw out the

baby with the bath-water), (iii) Irreversible binomials (e.g. bread and butter, alive and kicking), (iv)

stereotyped comparisons or similes (e.g. as proud as a peacock), (v) literary allusions and quotations

(e.g. a Jekyll and Hyde), (vi) commonplaces or truisms (e.g. boys will be boys), (vii) routine formulas

(e.g. mind the step), (viii) slogans (e.g. value for money) and (ix) commandments and maxims (e.g.

thou shalt not kill).

 87

Our definition has several consequences for what we may find with respect to (i) identification

and (ii) comprehension of unknown target idioms. Being more semantically covert than most single

words and other fixed expressions (e.g. clichés, collocations, catchphrases, etc), a pure idiom can

also be more difficult to identify by a non-native speaker of English. If we accept that the essence of

idiomaticity rests on an asymmetry between syntax and sense, then our argument is that the presence

of a homonymous literal counterpart complements such asymmetry both structurally and

contextually and assigns to idioms the status of being readily ‘mistakenly identified’ or deceptively

transparent lexical items that constitute a major stumbling block in the course of reading

comprehension in L2 contexts. Furthermore, where unknown idioms are finally identified, our

definition has potential consequences for their subsequent codebreaking and interpretation in text

context. In this respect, we claim that the existence of a literal counterpart, the regularity of the

syntax and the multiword nature of the idioms provide far more internal cues than are available in the

case of single unknown words for L2 learners to use in their guessing attempts along with the cues

supplied by the surrounding text context.

According to The New Webster’s Encyclopedic Dictionary of the English Language, the

equivalent term of ‘idiom’ in Greek, i.e. ιδιωματισμός (‘idiomatismos’), derives from the lexeme

idios which means ‘proper or peculiar language to one’s self’. In this sense, idiomatismos is most

commonly used in Greek textbooks in a similar way to our definition of idiom adopted and utilized

in our materials within the framework of the present study as referring to:

‘A mode of expression peculiar to a language or to a person; a phrase or

expression having a special meaning from usage, or a special grammatical
character; the genius or peculiar cast of a language; a peculiar form or variety of
language; a dialect.
(The New Webster’s Encyclopedic Dictionary of the English Language, 1980:420, as cited in Liontas,
2003b:2)
 88

Based on this definition, idiomatismos is to be differentiated from the term idioma which basically

refers to either regional or dialectal changes in the phonology, morphology or vocabulary of a

language or to stylistic expressional means used by writers as a way of conveying their message to

their audience.
 89

Table 5.1. Overview of Definitions on Idiom in the Literature

Definitions of Idiom
Source Definition
Roberts (1944:300) ‘Idiom is the idiosyncrasy of permutation which a given language exhibits in
contradistinction to all other languages or a given period exhibits in contrast to all
previous periods.’
Katz & Postal (1963:275) ‘The essential feature of an idiom is that its full meaning…is not a compositional
function of the meanings of the idiom’s elementary grammatical parts.’
Healey (1968:73) ‘[An idiom is]…..a group of two or more morphemes and an equal of greater number
of tagmemes whose meaning as a whole is not deducible from the meanings of its
component morphemes and tagmemes or any subgrouping thereof.’
Weinreich (1969:89) ‘…….idiomaticity – a phenomenon which may be described as the use of segmentally
complex expressions whose semantic structure is not deducible jointly from their
syntactic structure and the semantic structure of their components.’
Fraser (1970:22) ‘An idiom is a constituent or series of constituents for which the semantic
interpretation is not a compositional function of the formatives of which it is
composed.’
Makkai (1972:29) ‘[An idiom is]…..a peculiarity of phraseology approved by the usage of the language
and often having a signification other than its grammatical or logical one. Idioms are of
two types: (i) idioms of encoding that can be used and understood literally and (ii)
idioms of decoding. Idioms of decoding are those used non-literally (e.g. red herring,
take the bull by the horns) and are those that allow for the possibility that the hearer
will decode the idiom in a logical yet sememically erroneous way.’
Newmeyer (1972:294) ‘[An idiom is]….a collocation of two or more lexical items whose meaning cannot be
predicted from the meaning and structure of those elements contained in it.
Fernando and Flavell ‘A pure idiom is a non-literal set expression whose meaning is not a compositional
function of its syntactic constituents but which always has a homonymous literal
(1981:48)
counterpart.’
Strassler (1982:79) ‘An idiom is a concatenation of more than one lexeme whose meaning is not derived
from the meanings of its constituents and which does not consist of a verb plus an
adverbial particle or preposition. A concatenation as such then constitutes a lexeme in
its own right and should be entered as such in the lexicon.’
Wood (1986:95) ‘An idiom is a complex expression which is wholly non-compositional in meaning and
wholly non-productive in form.’
Fernando (1996:31) ‘[Idioms are]…….those expressions which become conventionally fixed in a specific
order and lexical form, or have only a restricted set of variants…’
Moon (1998a:4) ‘[Idioms are]……semi-transparent and opaque metaphorical expressions….’
 90

Bibliography

Abeille, A. (1995) The flexibility of French idioms: A representation with lexicalized tree adjoining
grammar. In M. Everaert, E. van der Linden, A. Schenk and R. Schreuder (Eds.) Idioms:
Structural and Psychological Perspectives. (55-42). Hillsdale, NJ: Lawrence Erlbaum.
Aisenstadt, E. (1979) Collocability restrictions in dictionaries. In R.R.K. Hartmann (Ed.)
Dictionaries and their Users. (71-74). Exeter Linguistic Studies, Exeter:University of Exeter.
Aisenstadt, E. (1981) Restricted collocations in English lexicology and lexicography. ITL Review of
Applied Linguistics 53 (53-61).
Alexander, R. J. (1984) Fixed expressions in English: reference books and the teacher. English
Language Teaching Journal 38 (2) (127-134).
Alexander, R. J. (1987) Problems in understanding and teaching idiomaticity in English. Anglistik
und Englishunterricht 32 (105-122).
Alford, D. (1971) Kicking the habit. Linguistic Inquiry 2 (4) 573.
Awwad, M. (1990) Equivalence and translatability of English and Arabic idioms. Papers and
Studies in Contrastive Linguistics 26 (57-67).
Austin, P. (1988) Constraints on coordinated idioms. Te Reo: Journal of the Linguistic Society of
New Zealand 31 (3-17).
Baranov, A. N. and D.O. Dobrovolskij (1996) Cognitive modelling of actual meaning in the field
of phraseology. Journal of Pragmatics 25 (3) (409-429).
Barkema, H. (1992) Morphosyntactic flexibility: the other side of the coin. In M. Everaert and E.
van der Linden (Eds.) Proceedings of the First Tilburg Workshop on Idioms. (23-39).
Tilburg:ITK.
Barkema, H. (1993) Idiomaticity in English NPs. In Aarts, J., P. De Haan and N. Oostijk (Eds.)
English Language Corpora: Design, Analysis and Exploitation Papers from the Thirteenth
(13th) International Conference of English Language Computerized Corpora. (257-
278).Amsterdam: Rodopi.
Barkema, H. (1994) Determining the syntactic flexibility of idioms. In U. Fries, G. Tottie and P.
Schneider (Eds.) Creating and Using English Corpora: Papers from the Fourteenth (14th)
International Conference on English Language Research on Computerized Corpora,
Zurich, 1993. (Language and Computers, 13) (39-52). Amsterdam:Rodopi.
Barkema, H. (1996a) Idiomaticity and terminology: A multi-dimensional descriptive model. Studia
Linguistica 50 (2) (125-160).
Barkema, H. (1996b) The effect of inherent and contextual factors on the grammatical flexibility of
idioms. In Synchronic Corpus Linguistics. Papers from the Sixteenth (16th) International
Conference on English Language Research on Computerized Corpora, Toronto 1995. (69-
83). Amsterdam: Rodopi.
Bobrow, S.A. and S.M. Bell (1973) On catching on to idiomatic expressions. Memory and
Cognition, 1 (3) (343-346).
Bolinger, D. (1976) Review of A. Makkai. Language 52 (1) (238-241).
Bolinger, D. (1977) Idioms have relations. Forum Linguisticum 2 (2) (157-169).
Botelho da Silva, T. and A. Cutler (1993) Ill-formedness and Transformability in Portuguese
idioms. In C. Cacciari and P. Tabossi (Eds.) Idioms: Processing, Structure and
Interpretation. Hillsdale, NJ: Lawrence Erlbaum Associates (129-143).
Bressan, D. (1979) Idioms and Second Language Learning. Teanga 1 (31-40).
Carter, C. (1987) Vocabulary. Applied Linguistics Perspective. Routledge.
Cacciari, C. and P. Tabossi (1988) The Comprehension of Idioms. Journal of Memory and
Language, 27 (668-683).
 91

Cacciari, C. and S. Glucksberg (1991) Understanding idiomatic expressions: The contribution of

word meanings. In G. B. Simpson (Ed.) Understanding word and sentence. (Advances in
Psychology, 77) (217-240). Amsterdam, New York, Oxford and Tokyo: North Holland.
Cacciari, C. (1993) The place of idioms in a literal and metaphorical world. In C. Cacciari and P.
Tabossi (Eds.) Idioms: Processing, Structure and Interpretation. (27-55). Hillsdale,
NJ:Lawrence Erlbaum.
Cacciari, C. and S. Glucksberg (1995) Imaging idiomatic expressions: literal or figurative
meanings? In M. Everaert, E. van der Linden, A. Schenk and R. Schreuder (Eds.) Idioms:
Structural and Psychological Perspectives. (43-56). Hillsdale, NJ: Lawrence Erlbaum
Associates.
Cermak, F. (1988) On the substance of idioms. Folia Linguistica 22 (3-4) (413-438).
Cermak, F. (1991) Universals, Typology and Idioms. In B.Palek and P. Janota (Eds.) Proceedings
of LP ’90 Conference on Linguistics and Phonetics: Prospects and Applications. (261-269).
Cermak, F. (1994) Idiomatics. In P.A. Luelsdorff (Ed.) Prague School of Structural and
Functional Linguistics. (185-195). Amsterdam and Philadelphia: John Benjamins.
Cermak, F. (2001) Substance of idioms: perennial problems, lack of data or theory? International
Journal of Lexicography 14 (1) (1-20).
Chafe, W. (1968) Idiomaticity as an anomaly in the Chomskyan paradigm. Foundations of
Language 4 (109-127).
Coulmas, F. (1979a) Idiomaticity as a problem of pragmatics. In H. Parret, M. Sbisa and J.
Verschueren (Eds.) Possibilities and Limitations of Pragmatics. (139-151). (Proceedings of
the Conference on Pragmatics, Urbino 1979). Amsterdam: John Benjamins.
Coulmas, F. (1979b) On the sociolinguistic relevance of routine formulae. Journal of Pragmatics 3
(239-266).
Cowie, A. P. (1981) The treatment of collocations and idioms in learners’ dictionaries. Applied
Linguistics 2 (3) (223-235).
Cutler, A. (1982) Idioms: the colder the older. Linguistic Inquiry 13 (317-320).
Drew, P. and E. Holt (1995) Idiomatic expressions and their role in the organization of topic
transition in conversation. In M.Everaert, E. van der Linden, A. Schenk and R. Schreuder
(Eds.) Idioms: Structural and Psychological Perspectives. (117-132) Hillsdale, NJ:
Lawrence Erlbaum Associates.
Ernst, T. (1981) Grist for the linguistic mill: idioms and ‘extra’ adjectives. Journal of Linguistic
Research 2 (3) (51-68).
Fernando, C. (1978) Towards a definition of idiom: Its nature and function. Studies in Language 2
(3) (313-343).
Fernando, C. and R. Flavell (1981) On Idiom: Critical Reviews and Perspectives. (Exeter
Linguistic Studies, 5). Exeter: University of Exeter
Fernando, C. (1996a) Australian Idiom. In J. Clark (Ed.) The Cultivated Australian: A Festschrift
in honour of Arthur Delbridge. (349-359). Hamburg:Helmut Buske.
Fernando, C. (1996b) Idioms and Idiomaticity. Oxford: Oxford University Press.
Fellbaum, C. (1993) The determiner in English idioms. In C. Cacciari and P. Tabossi (Eds.)
Idioms: Processing, Structure and Interpretation. Hillsdale, NJ: Lawrence Erlbaum
Associates. (271-295).
Fillmore, C. J., P. Kay and M. C. O’ Connor (1988) Regularity and Idiomaticity in grammatical
constructions: the case of ‘let alone’. Language 64 (501-538).
Flores d’ Arcais, G. B. (1993) The comprehension and semantic interpretation of idioms. In C
Cacciari and P. Tabossi (Eds.) Idioms: Processing, Structure and Interpretation. (79-99).
Hillsdale, NJ: Lawrence Erlbaum Associates.
Fraser, B. (1970) Idioms within a transformational grammar. Foundations of Language 6 (1)
(22-42).
 92

Fraser, B. and J. R. Ross (1970) Idioms and Unspecified NP deletion. Linguistic Inquiry 1 (1)
(264-265).
Gates, E. (1973) Idioms: the wild creatures of our talk: genus and species. Indiana English
Journal (70-76).
Geeraerts, D. (1992) Specialization and reinterpretation in idioms. In M. Everaert, E. van der
Linden, A. Schenk and R. Schreuder (Eds.) Proceedings of IDIOMS. Tilburg:ITK. (57-73).
Gibbs, R.W. (1980) Spilling the beans on understanding and memory for idioms in conversation.
Memory and Cognition, 8 (449-456).
Gibbs, R.W. (1984) Literal meaning and Psychological theory. Cognitive Science, 8 (275-304).
Gibbs, R.W, N. P Nayak, J.L. Bolton and M.E. Keppel (1989a) Speakers’ assumptions about the
lexical flexibility of idioms. Memory and Cognition,17 (1) (58-68).
Gibbs, R.W., N.P. Nayak, C. Cutting (1989b) How to kick the bucket and not decompose:
Analyzability and Idiom Processing. Journal of Memory and Language, 28 (576-593).
Gibbs, R. W. (1990) Psycholinguistic studies on the conceptual basis of idiomaticity. Cognitive
Linguistics 1-4 (417-451).
Gibbs, R. W. (1993) Why idioms are not dead metaphors. In C. Cacciari and P. Tabossi (Eds.)
Idioms: Processing, Structure and Interpretation. (57-77). Hillsdale, NJ: Lawrence Erlbaum
Associates.
Gibbs, R. W. (1995) Idiomaticity and Human Cognition. In M. Everaert, E. van der Linden, A.
Schenk and R. Schreuder (Eds.) Idioms: Structural and Psychological Perspectives. (97-
116). Hillsdale, NJ: Lawrence Erlbaum Associates.
Glaser, R. (1988) The grading of idiomaticity as a presupposition for a taxonomy of idioms. In W.
Hullen and R. Schulze (Eds.). Understanding the Lexicon. (264-279). Tubingen:Max
Niemeyer.
Glaser, R. (1998) The stylistic potential of phraseological units in the light of genre analysis. In
A.P.Cowie (Ed.) Phraseology: Theory, Analysis and Applications. (Oxford Studies in
Lexicography and Lexicology). Oxford: Oxford University Press. (125-144).
Glucksberg, S. (1993) Idiom meanings and allusional content. In C. Cacciari and P. Tabossi (Eds.)
Idioms: Processing, Structure and Interpretation. (3-26) Hillsdale, NJ: Lawrence Erlbaum
Associates.
Gottlieb, H. (1992) Idioms into Danish: Principles and Design of a bilingual dictionary of current
idiomatic usage: the English-Danish Idiomatic Dictionary. In J.E. Nielsen (Ed.) Words that
Teem with Meaning. (56-81).Copenhagen Views on Lexicography Copenhagen:Museum
Tusculanum Press.
Gottlieb, H. (1994) Idioms in corpora:types, tokens, frequencies and lexicographical implications.
In K. Hyldgaard-Jensen and V. Hjonager-Pedersen (Eds.) Symposium on Lexicography VI.
(85-91). (Proceedings of the Sixth (6th) International Symposium on Lexicography, 1994).
Tubingen:Max Niemeyer.
Gotz, D. (1976) Idioms and real texts. Arbeiten aus Anglistik und Amerikanistik 11 (1) (83-93).
Greim, B. D. (1983) The elliptical idioms: Change from icon and index to symbol. In I. Rauch and
G. F. Carr (Eds.) Language Change. (1-9). Bloomington:Indiana University Press.
He, Z. (1989) Creative Idiomaticity. In K. Hall, M. Meacham, R. Shapiro (Eds.) Proceedings of the
Fifteenth (15th) Annual Meeting of the Berkeley Linguistics Society, February 18-20 1989:
General Session and Parasession on Theoretical Issues in Language Reconstruction. (150-
160). Berkeley.
Healey, A. (1968) English Idioms. KIVUNG 1 (2) (71-108).
Katz, J.J. (1973) Compositionality, Idiomaticity and Lexical Substitution. In S. Anderson and P.
Kiparsky (Eds.) A Festschrift for Morris Halle. New York: Holt, Rinehart and Winston.
(357-376).
 93

Katz, J.J. and P.M. Postal (1963) Semantic Interpretation of Idioms and Sentences containing
them. M.I.T. Research Laboratory of Electronics Quarterly Progress Report 70 (275-282).
Keysar, B. and B. Bly-Bridget (1999) Swimming against the current:do idioms reflect conceptual
structure? Journal of Pragmatics 31 (2) (1559-1578).
Kovesces, Z. and P. Szabo (1996) Idioms: A view from cognitive semantics. Applied Linguistics
17 (3) (326-355).
Lattey, E. (1986) Pragmatic classification of idioms as an aid for the language learner. IRAL 25 (3)
(217-233).
Makkai, A. (1969) The two idiomaticity areas in English and their relationship:A stratificational
view. Linguistics 50 (44-58).
Makkai, A. (1972) Idiom Structure in English. The Hague:Mouton.
Makkai, A. (1977) Idioms, psychology and the lexemic principle. In R.J. DiPietro and E.J. Blansitt
Jr. (Eds.) The Third (3rd) LACUS Forum. (467-478). Columbia, SC:Hornbeam Press.
Makkai, A. (1978) Idiomaticity as a language universal. In J.H. Greenberg (Ed.) Universals in
Human Language Vol.3: Word Structure. (401-448). Stanford, California:Stanford
University Press.
Makkai, A. (1987) Idiomaticity and phraseology on post-Chomskyan linguistics:The coming-of-age
of semantics beyond the sentence. Semiotica 64 (1-2) (171-187).
Makkai, A. (1993) Idiomaticity as a reaction to ‘l’Arbitraire du Signe’ in the universal process of
semeio-genesis. In C. Cacciari and P. Tabossi (Eds.) Idioms: Processing, Structure and
Interpretation. (297-324). Hillsdale, NJ: Lawrence Erlbaum Associates.
Malkiel, Y. (1959) Studies in Irreversible Binomials. Lingua 8 (113-160).
Meier, H.H. (1975) The state of idiomatics. Dutch Quarterly Review of Anglo-American Letters 1
(3) (163-179).
Mittwoch, A. (1971) Idioms and Unspecified NP deletion. Linguistic Inquiry 2 (2) (255-258).
Machonis, P.A. (1985) Transformations of verb phrase idioms:Passivization, particle movement,
dative shift. American Speech 60 (4) (291-308).
McCarthy, M.J. (1992) English Idioms in Use. Revista Canaria de Estudios Ingleses 25 (55-65).
Michiels, A. (1977) Idiomaticity in English. Revue des Langues Vivantes XLIII (2) (184-199).
Moon, R. (1988) Time and Idioms. In M. Snell-Hornby (Ed.) Zurilex ’86 Proceedings. (107-115).
Tubingen:A. Francke.
Moon, R. (1992a) Textual Aspects of Fixed Expressions in Learners’Dictionaries. In P.J.L. Arnaud
and H. Bejoint (Eds.) Vocabulary and Applied Linguistics. (13-27). London:Macmillan.
Moon, R. (1992b) There is a reason in the roasting of eggs: a consideration of fixed expression in
native-spaker dictionaries. EURALEX '92 Proceedings II. (492-502). Tampere: Studia
Translatologica ser, A vol. 2.
Moon, R. (1994) Textual Aspects of Fixed Expressions in Text. In R.M. Coulthard (Ed.) Advances
in Written Discourse Analysis. (117-135). London:Routledge.
Moon, R. (1996) Data, description, and idioms in corpus lexicography. In M. Gellerstam, J.
Jarborg, S. G. Malmgren, K. Noren, L. Rogstrom and C. Roder (Eds.). Euralex ’96
Proceedings. (245-256). Goteborg:Goteborg University.
Moon, R. (1997) Vocabulary Connections: Multi-word items in corpus lexicography. In N. Schmitt
and M.J. McCarthy (Eds.) Vocabulary:: Description, Acquisition and Pedagogy. (40-63)
Cambridge:Cambridge University Press.
Moon, R. (1998a) Fixed Expressions and Idioms in English: A Corpus-based Approach. (Oxford
Studies in Lexicography and Lexicology). Oxford:Oxford University Press.
Moon, R. (1998b) Frequencies and forms of phrasal lexemes in English. In A.P. Cowie (Ed.)
Phraseology:Theory, Analysis and Applications (Oxford Studies in Lexicography and
Lexicology). (79-100) Oxford: Oxford University Press.
 94

Moon, R. (1999) Needles and haystacks, idioms and corpora:gaining insights into idioms using
corpus analysis In T. Herbst and K. Popp (Eds.) The Perfect Learners’ Dictionary. (265-
281). (Lexicographica Series Maior 95). Tubingen:Max Niemeyer.
Nayak, N and R. W. Gibbs (1990) Conceptual knowledge in the Interpretation of Idioms. Journal
of Experimental Psychology: General, 119 (3) (315-330).
Newmeyer, F. (1972) The insertion of idioms. In P.M. Peranteau, J.N. Levi and G.C. Phares (Eds.)
Papers from the Eighth (8th) Regional Meeting of the Chicago Linguistic Society. (294-
302).
Newmeyer, F. (1974) The Regularity of Idiom Behaviour. Lingua 34 (327-347).
Nicolas, T. (1995) Semantics of idiom modification. In M. Everaert, E. van der Linden, A. Schenk
and R.Schreuder (Eds.) Idioms: Structural and Psychological Perspectives. (233-252).
Hillsdale, NJ: Lawrence Erlbaum Associates.
Nuccorini, S. (1990) From transparency to opaqueness: The case of fixed expressions. In C. De
Stasio, M. Gotti and R. Bonadei (Eds.) La Rappresentazione Verbale e Iconica:Valori
Estetici e Funzionali. (417-428). (Proceedings of the Eleventh (11th) AIA Congress)
Bergamo e Milano Guerini Studio.
Nunberg, G. (1978) The syntax of idioms. In The Pragmatics of Reference. (117-135).
Bloomington:Indiana University Club.
Nunberg, G., I.A. Sag and T. Wasow (1994a) Idioms. Language 70 (3) (491-538).
Nunberg, G., I.A. Sag and T. Wasow (1994b) Idioms: An Interim Report. In S. Hattori, I. Kazuko,
(eds.) Proceedings of the XIIIth international congress of Linguists, Tokio (102-115).
Ortony, A., D.L. Schallert, R.E. Reynolds and S.J. Antos (1978) Interpreting Metaphors and
Idioms: Some Effects of Context on Comprehension. Journal of Verbal Learning and
Verbal Behaviour, 17 (465-477).
Pawley, A. (1986) Lexicalization. In D. Tannen and J. E. Alatis (Eds.) Language and Linguistics:
the Interdependence of Theory, Data and Application. (98-120). (Georgetown University
Round Table on Languages and Linguistics 1985). Washington, DC: Georgetown
University Press.
Roberts, H. M. (1944) The Science of Idioms: A Method of Inquiry into the Cognitive Design of
Language. Modern Language Association of America 68 (291-306).
Rose, J.H. (1978) Types of idioms. Linguistics 20 (3) (55-62).
Ross, J.R. (1970) Two types of idioms. Linguistic Inquiry 1 (1) 144.
Ruhl, E. (1977) Idioms and data. In R.J. DiPietro and E.L. Blansitt Jr. (Eds.) The Third (3rd)
LACUS Forum. (456-466). Columbia, SC:Hornbeam Press.
Schenk, A. (1995) The syntactic behaviour of idioms. In M. Everaert, E. van der Linden, A. Schenk
and R. Schreuder (Eds.) Idioms: Structural and Psychological Perspectives. (253-271).
Hillsdale, NJ:Lawrence Erlbaum Associates.
Schweigert, W.A. (1986) The Comprehension of Familiar and Less Familiar Idioms. Journal of
Psycholinguistic Research, 15 (1) (33-45).
Strassler, J. (1982) Idioms in English: A Pragmatic Analysis. Tubingen:Gunter Narr.
Swinney, D.A. and A. Cutler (1979) The Access and Processing of Idiomatic Expressions. Journal
of Verbal Learning and Verbal Behaviour, 18 (523-534).
Tabossi, P. and F. Zardon (1992) The activation of idiomatic meaning. In M. Everaert, E. van der
Linden, A. Schenk and R. Schreuder (Eds.) Idioms:Structural and Psychological
Pespectives. (273-282). Hillsdale, NJ:Lawrence Erlbaum Associates.
Tabossi, P. and F. Zardon (1993) The activation of idiomatic meaning in spoken language
conversation. In C. Cacciari and P. Tabossi (Eds.) Idioms:Processing, Structure and
Interpretation. (145-162). Hilldale, NJ:Lawrence Erlbaum Associates.
Titone, D. A. and C. M. Connine (1999) On the compositional and noncompositional nature of
idiomatic expressions. Journal of Pragmatics 31 (2) (1655-1674).
 95

Van deVoort, E.C., C. Marties and W. Vonk (1995) You don’t die immediately when you kick an
empty bucket: a processing view on semantic and syntactic characteristics of idioms. In M.
Everaert, E. van der Linden, A. Schenk and R. Schreuder (Eds.) Idioms: Structural and
Psychological Perspectives. (283-299). Hillsdale, NJ: Lawrence Erlbaum Associates.
Wallace, M. (1979) “What is an idiom? An applied linguistic approach”. In R.R.K. Hartmann (Ed.)
Dictionaries and their Users. (63-70) (Exeter Linguistic Studies, 4). Exeter:University of
Exeter.
Wasow, T., I. Sag and G. Nunberg (1983) Idioms:An interim report. In S.Hattori and K.Inoue
(Eds.) Proceedings of the Thirteenth (13th) International Congress of Linguistics. (102-
115). Tokyo:Comite International Permanent Linguistes.
Weinreich, U. (1969) Problems in the analysis of idioms. In J. Puhvel (Ed.) Substance and
Structure of Language. (23-81). Berkeley, California:University of California Press.
Wood, M.M. (1986) A Definition of Idiom. Bloomington:Indiana. Indiana University Linguistics
Club.