The Treatment of Collocations and

The Treatment of Collocations and Idioms in Learners' Dictionaries A. P. COWIE University of Leeds INTRODUCTION The collocation wage freeze is recorded as a main entry (variant wages freeze) in the Longman Dictionary of Contemporary English (1978), and at wage 1 (with cross-reference to a fuller treatment at freeze) in the Oxford Advanced Learner's Dictionary of Current English (1974). In these different ways the foreign student has his attention drawn to a widely attested combination in current use, and to an established figurative sense of one of its elements. Of course, this is incomplete information. In the sense of 'control' or 'restriction' freeze collocates more widely: the lexicographers might also have recorded price(s) freeze, incomes freeze and salary freeze—whether at one point of entry or several in each case. The fact that none of these attested collocations appears in LDOCE or is given special prominence in ALD reflects the dilemma of makers of general pedagogical dictionaries when brought up against the known collocabilities of given headwords. How completely and how explicitly should they be treated? There is no simple answer to this question. Explicitness of treatment (i.e. listing the items themselves) depends in part on whether the potentiality which items have to co-occur with a given headword in certain syntactic relationships can be defined in general terms, i.e. in terms of semantic features which can be assigned to the lexical items (Matthews 1965). It seems that the attested range of items functioning as modifiers of freeze are of this kind, since (with the exception of prices) they all denote earned or unearned income. As a result, and leaving other factors out of account, the entry for freeze can include a reference to the feature 'income'. One advantage of indicating collo- cability in such a general way is that it may enable a dictionary user to predict the possible occurrence of dividends freeze or pensions freeze. Other factors, however, bear on the question. It is clear, for example, that some collocational ranges whose members have shared features of meaning are more extensive than others (Cowie 1978, Aisenstadt 1979). While, for example, the set of noun modifiers which can collocate with freeze probably number no more than five or six, the list of semantically related nouns which can co-occur as direct objects with a verb such as run (in the sense of 'direct' or 'manage') is virtually open-ended. In such cases the pedagogical lexicogra- pher can provide an informal feature specification—thus, run . . . direct, manage (institution, organization)—or a few representative examples—run a business/a theatre/a bus company—secure in the knowledge that many colloca- tions acceptable to native speakers can be extrapolated from such minimal Applied Linguistics. Vol. II, No. 3 224 COLLOCATIONS AND IDIOMS guidance by the foreign learner. The compiler of general EFL dictionaries is in any event driven towards some such procedure by the need to account for openness of collocation within a limited space. In the case of the more limited collocability of freeze, on the other hand, he can entertain pedagogical argu- ments for specifying all, or several, of its attested modifiers. Whether in fact he does so may be influenced by yet another factor, which cuts across the open-restricted continuum. This is the tendency of a given dictionary headword to collocate more frequently with some items than with others, possibly as measured by informant responses in controlled experi- ments (Greenbaum 1970, 1974). Outside the limited range of textual studies reported by Greenbaum (1970) and the results of his own work with infor- mants, there is little empirical evidence of collocational frequencies in English available to pedagogical dictionary-makers (though cf. Backlund 1976, Mackin 1978). Where, as in the case of freeze, say, a judgement has to be made as to the relative frequency of collocation of wages, prices, incomes, etc., lexicographers in practice balance the evidence of a corpus against their own intuitions and those of colleagues. The admitted inadequacy of such methods partly explains the discrepancies which are often found between treatments of the same collocation in different dictionaries. At this stage, it might be agreed that the decision to record wage(s) freeze as an invariable collocation reflects judgements of the relative frequency of co- occurrence of wage(s)—as compared, say, with price(s) or incomes—rather than the wish to illustrate a representative member of a semantically related series. Conversely, the decision to indicate at freeze certain of its possible modifiers, as in ALD and the Collins English Dictionary (1979), lays emphasis on its (limited) collocational range. A further point has to do with the relationship of collocations to idioms, and with how the assignment of individual examples to one or other of these categories may determine their dictionary treatment. A collocation is by definition a composite unit which permits the substitutability of items for at least one of its constituent elements (the sense of the other element, or ele- ments, remaining constant). According to these criteria, run a business and wages freeze are both collocations, given the assumption of substitutability in both cases. By contrast, an idiom is 'immutable in the sense that its parts are unproductive in relation to the whole in terms of the normal operational processes of substitution, transposition, expansion, etc' (Mitchell 1971:57). It is important for the lexicographer to note that according to this view of idiomaticity, wage freeze is an idiom (and thus calls for prominent treatment in a general dictionary, possibly as a main entry) if it is regarded as lexically invariable. The foregoing discussion of descriptive variables which are relevant to the treatment of collocation in general dictionaries has also shown that lexicogra- phers must be constantly aware of more practical constraints. There are first of all considerations of space. Compilers of general pedagogical dictionaries must account for the linguistic structure of items at various descriptive levels, a constant preoccupation being to keep the various parts in proper balance. A more extensive treatment of collocabilities would need to be matched by corresponding reductions elsewhere (Cowie 1978, 1980). There is also the question of the known preferences of learners for using dictionaries for some A. P. COWIE 225 study purposes rather than others. The results of the investigation carried out by Bejoint, which appear elsewhere in this issue, suggest that certain categories of advanced student use EFL dictionaries primarily for decoding activities (principally reading), and continue to neglect the encoding informa- tion (for example on syntax) which is already plentifully supplied. Now a specification of the collocational range of dictionary headwords, whether in detail or in more abstract terms, is aimed precisely at fostering the active use of language, and specifically at helping the foreign learner to construct sen- tences which are 'lexical' (i.e. lexically acceptable) as well as 'grammatical' (Halliday 1966). Yet it is doubtful whether, in the face of continuing user conservatism, lexicographers will undertake an ambitious treatment of collo- cations in general pedagogical dictionaries. It seems that, for the present, efforts must be concentrated on extending and refining the kinds of guidance already being provided in specialized dic- tionaries of various kinds, where a reduction in the overall word-stock makes possible fuller and more explicit statements of collocational range (see, for example, Cowie and Mackin 1975). In so-called dictionaries of idioms, especially complex problems of analysis can arise, not least because an initial distinction must be drawn between collocations and idioms, as categories, for purposes of deciding which complex items to include. It is certainly also true that detailed analysis of these categories in dictionaries of limited coverage is the best preparation for their eventual treatment in general dictionaries. For these various reasons I now turn to consider the description of collocations and idioms in specialized dictionaries, with special reference to work on the two volumes of the Oxford Dictionary of Current Idiomatic English.1 1. COLLOCATIONS AND IDIOMS IN A DICTIONARY OF 'IDIOMATIC' ENGLISH The lack of a standardized terminology in this area continues to bedevil the work of lexicologists, though Greenbaum (1970) and Mitchell (1971) have helped to spread an awareness of terms already familiar to many linguists in Britain. There is, for instance, no generally accepted term under which both collocation and idiom can be subsumed, though Mitchell usefully introduces 'composite element' as a label embracing idioms, collocations and compounds (1971:57), and I have adopted it here. Then again, and in contrast to the situation in Eastern Europe (where 'phraseology' is a recognized sub-branch of lexicology), there is no name in general currency for the study of composite lexical units (Glaser 1980). Unfortunately, 'phraseology'—like 'fixed phrase'— has the disadvantage of blurring a distinction which it is important to preserve in collocational studies, between lexical units of various kinds on the one hand and the more abstract clause and phrase structures in which they function on the other. Confusion, or uncertainty, extends to the use of 'idiom' and 'collocation' also. The former is still used by linguists to refer to compo-- site units of differing degrees of variability, while the latter is not yet widely used outside a broadly Firthian tradition of linguistic analysis. Because of the terminological uncertainty, I shall approach the categories directly, attempting to define the various types of composite unit which need to be recognized by the lexicographer, and tying in the terminology proposed 226 COLLOCATIONS AND IDIOMS by various scholars as I proceed. Throughout, I shall try to show how the various distinctions which I draw are reflected in dictionary design. Idioms and collocations are found in a wide range of syntactic constructions; to avoid unnecessary complications examples will be chosen from one structural type (verb + direct object). 1.1 Many lexicologists agree to recognize a maximally variable type of com- posite unit, of which run a business, cited earlier in this paper, is an example. Variously referred to as 'free constructions' (Weinreich 1969), 'free phrases' (Arnold 1973) and 'free word-combinations' (Aisenstadt 1979), such units are characterized by the openness of collocability of each element in relation to the other or others. Not only, as has already been suggested in regard to the above example, are nouns freely collocable with run (in the sense of 'manage'), but verbs have a similarly wide privilege of occurrence with business. This view of the collocability conditions which collocations must satisfy in order to be 'open' or 'free' prompts two observations. The first is that it does not provide the analyst with a procedure for identifying those collocations which, while relatively open, nonetheless constitute problems for the foreign learner. These begin to emerge if one asks how the collocational potential of run, say, compares with the closely related manage and direct (cf. Greenbaum 1970:9). Whereas all three verbs have a shared collocational range in business, company, institution, etc., run has a wider collocational spread than manage, and manage than direct, among nouns of more specific reference. Consider all three in relation to bank, sauna, airport, polytechnic. The second point has to do with the treatment of fine collocational differences such as these in specialized dictionaries. I have already referred to the difficulty of treating them on any but a modest scale in general dictionaries. At the same time, they cannot be handled in 'idiom' dictionaries (but cf. 3, below). Leaving aside theoretical objections, the sheer bulk of such information would be overwhelm- ing. There is clearly a place for separate dictionaries of 'open' collocations such as those discussed elsewhere in this issue by Tomaszczyk. 1.2 In regard to sheer range of collocability, we may compare explode a bomb, say, with explode a myth. Whereas once again, and subject to the qualifications which have been made, explode and bomb exhibit openness of collocability in relation to each other, explode in its figurative sense has a very limited collocational range: myth, belief, idea, notion, theory. Thus, while ex- plode a bomb goes the way of run a business, as an open collocation, explode a myth clearly represents a different category, which merits inclusion in a dic- tionary of the kind being discussed. Like command respect, escape someone's attention and canvass someone's opinion, explode a myth belongs to a numerous class of composite units char- acterized by extreme restriction on collocability (at one point, as in these examples, or at several points). I have discussed elsewhere the essential arbi- trariness of such limitation, and the problems it poses for the foreign learner (Cowie 1979). Here I should like to take up two further points. The first is that the figurative meaning of one element (the verb in the examples just above) is an important determinant of limited collocability at the other (Ais- enstadt 1979:73). The second point is that in some studies determination is A. P. COW1E . 227 viewed the other way about. Since explode in the sense 'show to be false, or no longer true' occurs in no lexical context other than that already shown (myth, belief, etc.), one can say that the choice of the specialized meaning of the verb is contextually determined (Weinreich 1969:42). This approach to the definition of an important category of collocations characterizes the work of several Russian lexicologists (helpfully mediated to non-Russian-speaking students by Weinreich and Arnold). Speaking of substi- tutability in these 'semi-fixed combinations', Arnold comments 'we are not only able to say that such substitutes exist, but fix their boundaries by stating the semantic properties of words that can be used for substitution, or even listing them' (1973:143). Here there is a reference not only to limited choice but to the possible semantic relatedness of the collocates (a possibility which is demonstrated by explode a myth). It is important to note that in all semi-fixed combinations the unit which emerges from analysis and is recorded in the dictionary has a paradigmatic as well as a syntagmatic dimension. A complex such as (not) entertain (the) idea notion suggestion proposal doubt suspicion consists simultaneously of the collocational range (at idea) of entertain and of the various collocations (entertain the proposal, entertain the doubt, etc.) which can be recovered from it. This complex structure must be reflected in the design of dictionary entries, and editorial policy in such cases is to use the boldface 'headphrase' to indicate one of the possible collocations and the place of substitution, while the limited set of substitutes is listed in the body of the entry. (For reasons to be discussed later, the syntactic func- tion of collocates is given in all cases.) (not) entertain the idea, etc [V + O pass]... O: idea, notion; suggestion, propo- sal; doubt, suspicion. Within the general category of'restricted collocations' (to use an alternative term proposed by Aisenstadt 1979), there are interesting departures from the pattern in which a single item in a figurative sense is determined by (i.e. has no other privilege of occurrence than) a limited set of items used in a 'literal' sense. We may find, for example, that choice operates at the place of the figurative item, as in catch capture seize grip seize < s b s i m a g i n a t i o n Here the whole set of figurative synonyms are uniquely determined by the item imagination, as in that sense, and in that relationship to each other, they can have no other collocate as direct object. The form of entry is as follows, 228 COLLOCATIONS AND IDIOMS where the 'danger sign' is a warning against extension of the range by the foreign learner: catch etc sb's imagination [V + O pass]... V: catch, A capture, seize, grip ... That entry invites comparison with one for catch etc sb's fancy, where catch could also be glossed as 'rouse' or 'stimulate'. Despite the semantic similarity, however, catch in the second collocation is a member of its own uniquely determined set, thus: catch take tickle sb's fancy (cf. ?capture sb's fancy, ?tickle sb's imagination). A further variation in the general pattern is represented by examples in which the determining context is once again a set of items in general use, but where the direction of determination is from the verb to its object, thus: cause create make a stir One final variation is the limiting case of contextual determination, where one lexeme in a given specialized meaning can co-occur with only one other lexeme (Weinreich 1969). Examples include foot the bill, catch one's breath and curry favour, where in each case it is the verb which has the figurative sense. At this point, it will be noted, collocability is at an end: we are no longer dealing with collocations in the strict sense. 1.3 Composite units such as foot the bill and curry favour form a bridge category between collocations and idioms in the strict sense. On the one hand, they contain an element in a specialized meaning (foot, curry) and display contextual determination (though by a single item in each case—bill, favour). On the other hand, they are quite invariable, "foot the account and *'curry support being equally unacceptable; for that reason they fall under one possible definition of idiomaticity (Healey 1968, Mitchell 1971, Cowie and Mackin 1975). Although we lack evidence of the problems posed by particular types of composite unit, it seems likely that because of the unfamiliar sense of foot and the stability of the whole, foot the bill may be as difficult to interpret for many learners as fill the bill (which is altogether opaque). Certainly, com- posites of this type must be included in a dictionary which deals with idioms in the narrower sense. It is examples such as kick the bucket, spill the beans and blow the gaff- expressions which are as immutable as they are semantically opaque—which most commonly feature in any non-technical discussion of idioms. They are also the kinds of examples most usually cited in formal treatments of idioma- ticity within a transformational framework, where interest has chiefly focused on the theoretical problems raised when attempting to account for the inter- pretation of idioms and their syntactic properties in terms of a particular model of grammar (Katz and Postal 1963, Fraser 1970, Newmayer 1972). At A. P. COWIE 229 both these levels of discussion, then, there is a tendency to limit consideration of a spectrum of related categories to only one—the class of 'pure' idioms. As I have tried to argue, there are compelling reasons for treating parts of this wider spectrum of categories in an 'idiomatic' dictionary for foreign learners. Definitions of the idiom by linguists working within various traditions curiously echo that part of the standard dictionary definition of the term which fastens on the impossibility of interpreting the whole in terms of the known meanings of the parts. Compare: idiom ... 1 ... a group of words whose meaning cannot be predicted from the meanings of the constituent words ... (Collins English Dictionary 1979). ... any group of words whose meaning cannot be deduced from the meanings of the individual words (Healey 1968: 71). ... series of constituents for which the semantic interpretation is not a composi- tio