What are collocations: Sandy beaches or false teeth?

This article was downloaded by: [University of Oklahoma Libraries] On: 17 March 2013, At: 08:39 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK English Studies Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/nest20 What are collocations: Sandy beaches or false teeth? Thomas Herbst a a Universität Erlangen, Version of record first published: 13 Aug 2008. To cite this article: Thomas Herbst (1996): What are collocations: Sandy beaches or false teeth?, English Studies, 77:4, 379-393 To link to this article: http://dx.doi.org/10.1080/00138389608599038 PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms- and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. WHAT ARE COLLOCATIONS: SANDY BEACHES OR FALSE TEETH? 1 Three views of collocation Looking at travel guides one could get the impression that England (or the Eng- lish language) is remarkably rich in sandy beaches. Butlins holiday camps are apparently always situated on beaches which are sandy (or excellent): (1) Ayr: It is an attractive place with excellent beaches and a fishing harbour. ... Butlins's ho- liday Centre has a glorious sandy beach with splendid views across the Firth of Clyde. (2) Bognor Regis: Bognor nestles between golden, sandy beaches and the green rolling downs of West Sussex. (3) Clacton-on-Sea: The miles of sandy beaches are dotted with picturesque little fishing vil- lages. An analysis of the sections on Cornwall, Devon, Kent and Sussex in the AA Guide to Britain's Coast produces similar results: Table 1: AA Guide to Britain's Coast: Cornwall, Devon, Kent, Sussex beach total: 195 sandy total: 75 without adjective: 103 â beach: 42 with adjective: 92 sandy: 42 splendid total: 7 street total: 18 view: 6 narrow: 14 This means that very often when we hear or read one word we expect another. Sometimes, this expectancy is exploited as a stylistic device - in P.G. Wode- house, for example, as was pointed out by Mackin (1978: 149), where in - (4) It being my constant policy to strew a little happiness as I go by, I hastened to point out the silver lining in the c.'s or (5) I could see at a g. that Jeeves had been right in describing her demeanour as despondent. it is perfectly clear that c and g stand for clouds and glance. What all this shows is something fairly obvious - namely that there is some- thing like a 'mutual expectancy of words', as J.RrT^irth (1957/1968: 181) has called it. The question is whether these are good examples of collocations - a English Studies, 1996, 4, pp. 379-393 3 7 9 0013-838X/96/04-0379/$12.00 © 1996, Swets & Zeitlinger En gl ish S tu di es 1 99 6. 77 :3 79 -3 93 . d ow nl oa de d fro m w w w .ta nd fo nl in e.c om term which is also closely connected with Firth.1 Discussing such cases as You silly ass Firth (1957: 195) says: 'one of the meanings of ass is its habitual collo- cation with an immediately preceding you silly, and with other phrases of ad- dress or of personal reference'. He also gives dark night as an example of a collocation and argues that the collocability of dark with night is part of the meaning oí dark and that oí night with dark part of the meaning oí night. How- ever, the fact that Firth regards collocation as part of the meaning of a word2 has to be seen against the background of a very wide concept of meaning in his approach - after all, Firth (1957: 192) also claims that it 'is part of the meaning of an American to sound like one'. This wide view of meaning is also revealed in a lecture in Berlin in 1956, where he describes the study of collocation as 'the study of key-words, pivotal words, leading words, by presenting them in the company they usually keep - that is to say an element of their meaning is indi- cated when their habitual word accompaniments are shown'.3 Thus Lyons (1977: 612) is certainly justified in saying that 'Exactly what Firth meant by col- locability is never made clear', but what seems clear are two things: - First of all, for Firth collocation was a co-occurrence relation between indi- vidual lexical items and not a relation between classes of items, as is appar- ent from the definition given above. - Secondly, the use of the term collocation is not restricted to combinations of two words. So Firth gives You silly ass or Don't be such an ass as examples. Despite the fact that Firth no doubt instigated the discussion of such co-oc- currence relations between individual lexical items, he, as Palmer (1968: 6) points out in his introduction to a collection of Firth's articles, never made an attempt to expand his ideas about collocation into a theory of semantic com- patibility. Also, it is not surprising that a certain vagueness in the use of the term by Firth has given rise to a number of different interpretations, which can prototypically be identified as three key positions, namely (i) a text oriented approach (ii) a statistically oriented approach and (iii) what might be called a significance oriented approach to collocation. The differences between these approaches can be illustrated by analyzing the following passage: (6) Oxford, the 'University' in the wider sense, embraces both the central body and the inde- pendent colleges within a loose federation of strong mutual interests. Members of colleges are members of the University. The University in the narrow sense is the original institution at the heart of this complex structure. This institution retains its traditional functions, of- fering lectures and seminars, maintaining libraries, SETTING EXAMINATIONS, and AWARDING DECREES. 1 On the history of the term collocation see Mitchell (1975: 134). 2 Cf. Lipka (1992: 166). 3 In that lecture Firth mentions die Insel Berlin as an instance of a collocation, an example, however, that recent events have rendered less permanent than one might have thought. 380 En gl ish S tu di es 1 99 6. 77 :3 79 -3 93 . d ow nl oa de d fro m w w w .ta nd fo nl in e.c om (i) A text oriented approach to collocation can be found in Halliday and Hasan's book Cohesion in English (1976). In a text such as (6), Halliday and Hasan would probably identify as collocations such words as Oxford, Universi- ty, colleges, lectures, seminars, libraries, examinations and degrees because they contribute to the cohesion of the text. Halliday and Hasan (1976: 287) speak of 'collocation' or 'collocational cohesion', which they describe as a cover term for the cohesion that results from the co-occurrence of lexical items that are in some way or other typically associated with one another, because they tend to occur in similar envi- ronments. However, the relations that in Halliday and Hasan's approach fall under cohe- sion are often explained in different terms in semantics, for example, through lexical fields or as relations such as synonymy, antonymy or hyponymy. For these and other reasons, Hasan (1984) herself rejects this use of the term collo- cation and modifies the term lexical chain. She now explains what was covered by collocation in the 1976 model by so-called identity and similarity chains, i.e. different types of lexical chain, but this is not relevant in this context. Two features mark the use of the term collocation by Halliday and Hasan (1976): - firstly, that collocations do not have to be immediately adjacent to each other, which is also in contrast to Firth's view and - secondly, that the elements of a collocation do not enter what dependency grammar would describe as an immediate dependency relation. Thus the way the term is used with reference to text linguistics is slightly awk- ward and perhaps not particularly helpful. What this view of collocation amounts to is not much more than saying that in a text about coastal walking there is a certain likelihood for words such as coast, sea, path, climb or steep to occur as well. This kind of likelihood of co-occurrence of lexical items, howev- er, seems to be determined to a greater degree by extralinguistic than by lin- guistic factors. (ii) This is an important difference to the second type of collocation, the sta- tistically oriented approach, which received an important impetus in the work of John Sinclair and the Cobuild project, now referred to as the Bank of Eng- lish.4 Whereas for Halliday and Hasan collocation is only relevant with respect to the creation of cohesion in texts, Sinclair (1966) has developed a theory of collocation as such. Jones and Sinclair (1974: 19) use a very wide definition of collocation and define it as 'the co-occurrence of two items in a text within a specified environment'. For Sinclair (1966: 415), the linear order of elements in a text is a relevant factor, as can be seen from his description of the structure of a collocation:5 4 Compare also the studies by Kjellmer based on the Brown Corpus. Kjellmer (1987: 133) defines collocation as 'a sequence of words that occurs more than once in identical form (in the Brown Corpus) and which is grammatically well-structured'. Cf. Kjellmer (1994: xiv). 5 Cf. also Sinclair (1991: 115). 381 En gl ish S tu di es 1 99 6. 77 :3 79 -3 93 . d ow nl oa de d fro m w w w .ta nd fo nl in e.c om We may use the term node to refer to an item whose collocations we are studying, and we may then define a span as the number of lexical items on each side of a node that we consider rele- vant to that node. Items set by the environment set by the span we will call collocates. Looking at text (6), this means that if you take the word sense as the node and a span of 5 words on each side of the node, then as collocates of sense you ar- rive at the words the, university, in, the and narrow as well as is, the, original, in- stitution and at. What is very important in Sinclair's theory is that he makes a distinction be- tween casual and significant collocation.6 Significant collocations are colloca- tions that occur more frequently than would be expected on the basis of the frequency of occurrence of the individual items. It is of course impossible to ar- rive at significant collocations in such short extracts as this, which becomes rather obvious when you consider that in this case sense and university would have to be seen as a significant collocation. However, a principal drawback of such corpus based statistical analyses lies in the design of the corpus and its representative nature. This already applies to syntactic analysis but the problems are much more difficult to overcome in lex- ical analyses. Halliday (1966: 159) says that for collocational analysis corpora of about 20 million words are needed; even the Brown corpus comprises only one million, however. By now more extensive corpora are available; Sinclair's own Cobuild corpus is a very remarkable case in point.7 Nevertheless it could be argued that frequency of co-occurrence alone is not a very satisfactory crite- rion for the significance of a combination. If, for instance, Berry-Rogghe (1972), on the basis of the analysis of a corpus of 72,000 words, arrives at the conclusion that the most frequent collocates of a word such as house include the determiners the and this and the verb sell, this is neither particularly surprising nor particularly interesting. This example illustrates very clearly the major problem of the statistically oriented approach - namely that even the so-called significant collocations need not be terribly significant, at least not, if you com- pare it with a view of significance that is not based purely on statistics. (iii) The third position has been stated in a very pronounced way by Haus- mann (1984: 398) in his reference to collocations as 'Halbfertigprodukte der Sprache' - semi-finished products of language - , or (1985: 178) as 'typische, spezifische und charakteristische Zweierbeziehungen von Wörtern' - typical, specific and characteristic relationships between two words. According to this view, one would call set examinations and award degrees collocations in text (6). Such a narrow view of collocation does not seem to be very common in British linguistics. However, collocation is used in this sense in the Comprehensive Grammar of the English Language, if only in a footnote, and also by Cowie (1994)8 and Robins (1971: 63), who defines collocation as the 6 See also Altenberg/Eeg-Olofsson (1990). 7 Cf. Sinclair (1993) and Svartvik (1993). 8 Cowie (1994: 3169) mentions 'arbitrary limitation of choice at one or more points' as a char- acteristic of collocations. 382 En gl ish S tu di es 1 99 6. 77 :3 79 -3 93 . d ow nl oa de d fro m w w w .ta nd fo nl in e.c om 'habitual association of a word in a language with other particular words in sen- tences' and gives such examples as white coffee, green with jealousy or maiden speech. Robins distinguishes between collocations such as bright day and dark night on the one hand and word groups such as bright night and dark day on the other, which shows that he does not use collocation for any kind of combi- nation of lexical items. It is not clear, however, whether the frequency of a com- bination alone is taken as a defining criterion for collocations - i.e. what Sinclair would call significant collocation. Thus, as with so many other terms in linguistics, one could refer to the way the term collocation is used as a kind of Humpty Dumpty situation. In the fa- mous passage in Through the Looking Glass Humpty Dumpty says 'When I use a word, it means just what I choose it to mean, neither more nor less' and the puzzled Alice replies: 'The question is whether you can make words mean so many different things' to which Humpty Dumpty answers: 'When I make a word do a lot of work like that, I always pay it extra'. Collocation must be one of the top earners in linguistics; the question is whether one shouldn't cut down its income a little. ⢠The use of collocation employed by Halliday and Hasan can probably be ne- glected; Hasan herself has shown that the usefulness of such an approach is limited. ⢠It must also be doubted whether there is much point in using collocation for any kind of co-occurrence of two lexical items. On the one hand, the term collocation is given an unnecessarily general interpretation; on the other hand such terms as 'co-occurrence' or 'word group' seem perfectly sufficient to de- scribe the phenomenon. This leaves us with two rivalling approaches - one where the significance of the combination is seen purely on the basis of frequency counts and one where significance is seen in such terms as unpredictability of the combination on se- mantic grounds. A purely statistical view of collocation - as advocated by Sinclair - seems prob- lematical for a number of reasons: ⢠Firstly, because of the general problems involved in any kind of corpus analysis, especially regarding the representative nature of the material analysed. Howev- er, computer assisted analyses may help overcome this problem. ⢠Secondly, positional statements as produced by Sinclair are of limited value if one disregards the context. Greenbaum (1974/1988: 115) has illustrated, for example, that the occurrence of particular adverbs is determined by a num- ber of factors. Thus, much, for instance, occurs in negative sentences such as (7a) I don't like him much; in declarative sentences, however, it cannot occur without a premodifier, thus not (7b) *I like him much 383 En gl ish S tu di es 1 99 6. 77 :3 79 -3 93 . d ow nl oa de d fro m w w w .ta nd fo nl in e.c om but only (7c) I like him very/too/so much. It must be doubted whether a purely statistical kind of analysis is able to grasp the complexity of such factors. ⢠Finally, there is the problem of the limited power of statistical statements. As already indicated, one must ask what sort of statement is really being made. Is - to quote one of Firth's famous examples - dark night a significant collo- cation because nights tend to be dark and not bright? In other words: wouldn't it be true to say that the fact that certain words tend to co-occur must be attributed to certain facts of the world - together with the way this world is conceptualised in language? Is sit on a chair more interesting lin- guistically than chop up a chair because one tends to sit on chairs rather than chop them up? Not really - but this is exactly the sort of result statistically oriented research on collocation is likely to produce. Again, we are faced with the problem of the relationship of language and the world. Is it surprising that in the travel guide analysed significantly more occurrences of pretty village could be found in the counties Cornwall, Devon and Kent than in Tyne and Wear? Does the frequent occurrence of sandy beaches in advertis- ing material allow any conclusions other than saying that beaches can be called sandy in English, and that this is done quite often - simply because it makes the beach more attractive for the potential customer? Thus statistical significance alone is hardly a suitable criterion for determin- ing the significance of linguistic combinations. It seems pretty obvious that there are different degrees of affinity between the words in such combinations as (8) sandy beaches or (9) sell a house on the one hand and such combinations as (10) set an examination (11) weak tea (12) flock of sheep (13) false teeth . and (14) I-entirety agree on the other. The problem is that it is much easier to make such a statement than to find precise criteria for delimiting collocations in this sense. 384 En gl ish S tu di es 1 99 6. 77 :3 79 -3 93 . d ow nl oa de d fro m w w w .ta nd fo nl in e.c om 2 The significance-oriented approach It is perfectly obvious that this area of linguistic description is a classic example of gradience. Any attempt to define collocation in this narrow sense can thus only be aiming at defining a kind of prototype of collocation, recognizing the gradience character of the distinction between collocation and free combina- tions. It is also perfectly obvious that there are considerable differences between in- dividual words with respect to what Mclntosh (1961) has called range. A few examples from a novel by David Lodge9 may serve to illustrate this: (15) The modern conference resembles the pilgrimage of medieval Christendom in that it al- lows the participants to indulge themselves in all the pleasures and diversions of travel while appearing to be austerely bent on self-improvement. To be sure, there are some pen- itential exercises to be performed - the presentation of a paper, perhaps ... Resemble in the first line has practically no selectional restrictions whatsoever: a conference can resemble a pilgrimage in the same way as Humpty Dumpty can resemble an egg. Indulge in line 2 is more restricted in that it requires a sub- ject with the feature '+ human'. Finally, presentation in the last line of the pas- sage is subject to even more serious restrictions since (16a) *the presentation of a lecture/a talk is not acceptable, whereas the verb give in (16b) give a paper/lecture/talk could be used equally with all three nouns. Apparently there are words which have a very wide range, others, where the selectional restrictions can be described through general semantic features, and words whose range is restricted to certain other words. Distinguishing between the last two groups causes a lot of descriptive problems, also to applied linguists - the only dictionary, for instance, that provides really reliable information of this kind is the admirable Oxford Dictionary of Current Idiomatic English by Cowie, Mackin and McCaig. This area is also one of the crucial features of a current valency dictionary project (cf. Herbst 1987). The gradience character in the nature of collocation is also recognised by all scholars who distinguish between different types of collocation: Sinclair (1966) distinguishes between casual and significant collocation, Cowie (1978) between open and restricted,10 Aisenstaedt (1981) between free and restricted, and Carter (1987) draws a four level distinction, namely between unrestricted, semi- restricted, familiar and restricted collocation. However, one looks in vain for precise or even coinciding criteria for restricted collocations. 9 David Lodge, Small World, Harmondsworth, 1984: vii. 10 Cf. also Gläser (1990: 39). 385 En gl ish S tu di es 1 99 6. 77 :3 79 -3 93 . d ow nl oa de d fro m w w w .ta nd fo nl in e.c om Maybe in this case it might make sense to put the cart before the horse and to ask which combinations are not acceptable and for what reasons. In his book on Semantics, Palmer (1981) uses the term collocational restrictions in this con- text and identifies different types of such restrictions. There are two points, however, where one might object to Palmer's analysis: In the first case collocational restrictions can be attributed to the meanings of the words in question. His example is green cow. In my view, this is confusing meaning and properties of the referent or of the real world. If green cow is a somewhat unusual combination, then that is because most native speakers of English have very little experience of green cows. In German, this is quite dif- ferent, of course, if you think of the famous lila Kuh introduced by Suchard in its television commercials. This shows that restrictions of the kind found in green cow are not really linguistic in character. Palmer also identifies a type of collocation which cannot be explained in terms of meaning. A similar argument has been put forward by Mackin (1978: 150), when he says: Unfortunately for the foreign learner of English, there is no way in which he can be led to con- struct the collocation 'weak tea' rather than 'feeble tea'. What Mackin is driving at is that if one considers weak and feeble to be syn- onymous, then the fact that weak tea is acceptable and feeble tea is not cannot be explained semantically, weak tea is thus not predictable and thus must be seen as a typical phenomenon of usage, or, in the terms of Coseriu (1979), of the linguistic norm. However, it is almost impossible to escape the danger of circular arguments here since the transparency of a combination does indeed depend on the mean- ing attributed to its constituents. Thus whether or not a combination is clas- sified as a collocation depends on the underlying analysis, which is also pointed out by Palmer (1981: 77), when he discusses such examples as white coffee and white wine, where the use of white appears arbitrary in that you would not de- scribe the colour of these drinks as white: It would ... be a mistake to draw a clear distinguishing line between those collocations that are predictable from the meanings of the words that co-occur and those that are not. ... For one can, with varying degrees of plausibility, provide a semantic explanation for even the more restricted collocations, by assigning very particular meanings to the individual words. ... We can thus ex- plain white coffee, white wine ... by suggesting that white means something like 'with the lightest of normal colours associated with the entity'. This is not quite satisfactory, however, since a combination such as white tea seems odd in English. The argument becomes even more obscure in the case of collective terms such as (17a) flock of sheep/goats/birds (17b) herd of cows/cattle/elephants (17c) school of fish/whales/dolphins etc 386 En gl ish S tu di es 1 99 6. 77 :3 79 -3 93 . d ow nl oa de d fro m w w w .ta nd fo nl in e.c om where Palmer (1981: 78) excludes a semantic explanation for these collocations when he says It is difficult to see any semantic explanation for the use of the collective terms. The only differ- ence between herd and flock is that one is used with cows and the other with sheep. This analysis is perfectly correct - with the exception of the italics for cows and sheep. The point is that herd and flock are not used because of the occurrence of the WORDS cows and sheep, but that they are used when referring to the ex- tralinguistic objects cows and sheep. One can easily imagine a sentence such as (17d) That is a nice flock over there but only of course when the speaker is referring to sheep and not to cows. If one doesn't want to perform the trick of analysing (17d) as a collocation whose second element has been deleted in ellipsis, one will have to see the fact that flock is used with respect to sheep as part of the meaning oí flock. In the same way it must be seen as part of the extensional meaning of adjectives such as white, black or weak to which drinks they can refer since sentences such as (18) Do you prefer white or red? (19) Do you drink it black? (20) It is rather weak. are perfectly feasible. Of course, it must be clear from the context whether it is wine or coffee, but it is definitely not the WORDS wine or coffee that lead to the occurrence of white, black or weak. For this reason I would reject the ap- proach taken by Palmer and Mackin to analyse such cases as flock of sheep or white wine as combinations which are not transparent semantically. These are instances of the arbitrary conceptualization of the world through language and not of an arbitrary combination of linguistic signs. I would only agree with Mackin insofar as there is no way for the foreign learner to predict that the equivalent of German schwacher Tee is weak tea and not feeble tea just as there is no way to predict that the equivalent of Bachstelze is wagtail and not budgeri- gar. This is not to say that there are no collocations of the kind described by Palmer. Thus it should be difficult to find semantic criteria to explain why you can (21a) have lunch/take lunch (21b) have tea/take tea but (21c) have dinner but not 387 En gl ish S tu di es 1 99 6. 77 :3 79 -3 93 . d ow nl oa de d fro m w w w .ta nd fo nl in e.c om (21d) »take dinner." This also applies to an example taken from Leech (1981: 17), who says 'one trembles with fear, but quivers with excitement', although in OALD one of the example sentences for tremble is tremble with excitement and in LDOCE one finds quiver with fear. Whichever way the restriction may work, tremble and quiver would present good examples of non-transparent collocations. However, should transparency be made the defining criterion for colloca- tions? I think not - mostly for two reasons: ⢠First of all, Palmer is - irrespective of any criticism of individual examples - perfectly correct when he points out that some cases can either be analysed as being opaque or be seen as transparent by more or less constructing mean- ings for the elements of a collocation. In this area one may well reach the lim- its of semantic analysis. Thus a verb such as commit takes objects such as crime, sin, suicide, but not *commit burglary or *commit theft. Is this a case of idiosyncratic collocates or can the restrictions be described by a feature such as '+ serious crime'? If so, at what point does a crime become serious? Obviously a train robbery isn't - neither would you get *commit the train robbery nor *commit insult. ⢠Secondly, at least intuitively, there seems to be a closer semantic affinity be- tween flock and sheep or weak and tea than between, say, sell and house. This is difficult to prove, however. One could of course argue that there is a clos- er link between set and examination than between sell and house since a ques- tion such as (22) Did you set anything this year? in the appropriate context can only refer to examinations, whereas (23) Did you sell anything? allows a wide range of answers. In this meaning, the selection restrictions of set can indeed be stated quite specifically since the range of set comprises all hyponyms of examination, whereas for sell no selection restrictions can be specified at all since you can or could sell practically anything, from your house and your children to your soul. Nevertheless, one will probably not be able to say much more than that the collocational character of a combination decreases the more general the selectional restrictions are or the wider the ex- tension is. Thus it seems that when it comes to calling something a collocation there is a lot to be said in favour of a false teeth rather than a sandy beach approach. In such a narrow view, collocation comprises two types, which, although dif- ferent, may not always be easily distinguished, namely 11 One possible criterion that was in fact suggested was that take implies 'small meals', but this should then also apply to breakfast. 388 En gl ish S tu di es 1 99 6. 77 :3 79 -3 93 . d ow nl oa de d fro m w w w .ta nd fo nl in e.c om ⢠firstly such combinations, in which one element - on the basis of its meaning - is only applicable to a very restricted number of extralinguistic objects and thus admits only a restricted number of collocates, such as a flock of sheep, a flock of lambs or also a flock of birds, ⢠secondly such combinations in which one element only enters a limited num- ber of combinations and where this does not seem in any way motivated by meaning. Examples of this kind are commit a crime, false teeth or possibly quiver with excitement. One of the most prominent features of a collocation (one which also applies to idioms)12 is that they allow a certain degree of predictability, as Greenbaum (1974/1988) has shown in his work on verb-intensifier collocations in the Survey of English Usage. It is remarkable that in completion tests, in which the infor- mants were presented with the first part of a sentence, there often was a clear tendency in a certain direction. For instance, in the case of / entirely 82% of Greenbaum's informants produced a sentence containing the verb agree, in the case of a stimulus / completely 50% continued with the verb forget. Thus describing the character of collocation merely in terms of acceptability would be as inappropriate as relying exclusively on statistical findings. Apart from the acceptability of a collocation its frequency will also have to be re- garded as an important characteristic. 3 Collocation and foreign language competence The case for such a narrow view of collocation is mainly that it reveals a cen- tral characteristic of language. Since, as Cowie (1994: 3168) puts it, 'native-like proficiency of a language depends crucially on knowledge of a stock of prefab- ricated units', there can be no doubt that competence in a language involves knowledge about collocation.13 This becomes apparent when one compares the reactions of native speakers and foreign learners of English in this respect. This was done in a series of tests consisting of 100 test items which was carried out with 100 students of English at two German universities, Augsburg and Erlan- gen, and 58 English students from the universities of Birmingham, London, Newcastle and Reading. The tests comprised translation, a kind of cloze-test where the first letter was given, and a completion test. The results show quite clearly that word associations of the collocational type are much more uniform with native speakers than with the German students. In the left-hand column of table 2, the number of different answers obtained from the English students, in the right-hand column those of the Germans is given: 12 Palmer (1981) and Robins (1971) see idioms as a special type of collocation, whereas Mitchell (1975) and Cowie (1981) make a distinction between collocation and idiom. The distinguish- ing criterion is that an idiom constitutes a single unit, both structurally and semantically, whereas a collocation, in Cowie's (1981: 224) words 'is a composite unit which permits the substitutability of items for at least one of its constituent elements (the sense of the other element, or elements, remaining constant)'. 13 Cf. Svartvik (1993: 26) or Alexander (1992: 242) as well as Sinclair's (1991: 110) idiom prin- ciple. 389 En gl ish S tu di es 1 99 6. 77 :3 79 -3 93 . d ow nl oa de d fro m w w w .ta nd fo nl in e.c om Table 2:" Test series Al A2 A3 A4 Test series Bl B2 informants: A (Please fill in the gaps) By the Act of Union, which came into in 1536, ... The Act a provision that ... no one could an official position in Wales ... Parliament an act in 1844 ... B: Complete the following sentences they all greatly My friend entirely English (58) 11 15 14 6 11 19 German (88) 23 29 24 28 41 51 These results indicate that the tendency to use a particular collocation is much stronger with the English native speakers than with German students of Eng- lish. This also becomes very clear when you look at the completion tests. In order to have a statistically sound basis of comparison, some of my tests were identical with the ones used by Greenbaum. As you can see from table 3, for the English native speakers my own results are rather similar to Green- baum's (presented in the first column), whereas the reactions of the German students (given in the third column) differ remarkably. Table 3 Test Cl C2 C3 C4 series C (Complete the I entirely I badly My friend entirely Some people much following sentences) AGREE DISAGREE FORGET NEED WANT AGREE PREFER English Greenbaum 82% 9% 1% 65% 28% 56% 56% 78% 3% 3% 69% 12% 62% 31% German 23% 7% 8% 11% 6% 11% 1% The results given here represent only a small sample of the tests carried out. Not all of the results are as extreme as that, but still these tests show quite clearly that the German students' knowledge concerning both the acceptability and the 14 Since the number of German and English informants is not identical it is difficult to have a statistically satisfactory basis of comparison - this is why table 2 contains absolute figures rather than percentages. 390 En gl ish S tu di es 1 99 6. 77 :3 79 -3 93 . d ow nl oa de d fro m w w w .ta nd fo nl in e.c om frequency of collocations shows obvious deficiencies compared with those of native speakers. Other types of test show similar discrepancies: only 27% of the German students translated schwacher Tee as weak tea, 33% as light tea. In the case of a completion exercise such as A number of objections were but ..., only 30% of the German students were able to use the collocation raise objections. And finally, test series G 'Which words occur to you when you read the following adjectives' produced differences such as 90% versus 7% for tor- rential rain and 57% and 2% for scattered showers, which may partly be ex- plained by the fact that German students did not know the meanings of torrential and scattered, of course. 4 Conclusion These tests can be seen as a further argument for restricting the use of the term collocation to a narrow interpretation, where significance in the sense of a cer- tain lack of semantic predictability or transparency is taken as the defining cri- terion. Such a view of collocation would seem appropriate in three different areas: ⢠firstly, as is obvious from the tests reported on, in foreign language teaching15 and ⢠secondly, closely connected to that, in lexicography. Admittedly, a purely sta- tistical approach to collocation provides insights that can be usefully applied to lexicography, with respect to example sentences, for instance. For foreign users of a dictionary it is however of vital importance that collocations in the narrow sense should not only be contained in the dictionary, but be recog- nizable as such and be presented in such a way that the user can actually find them. It is true that the development especially of English foreign learners' dictionaries has made enormous progress in recent years, if one thinks of such dictionaries as the fifth edition of the OALD, of the third edition of LDOCE or the second edition of Cobuild, where corpus access and improved layout techniques16 have given more prominence to phrases and collocations. Nevertheless, there is still room for improvement: a dictionary that gives the collocation weak tea under weak is of relatively limited usefulness to the for- eign learner; indeed none of the dictionaries mentioned list weak tea under tea, or severe gales under gale; under injuries LDOCE3 gives minor (but not serious) injuries, OALD5 and Cobuild2 give serious (but not minor) injury/in- juries. ⢠thirdly, and perhaps most importantly for theoretical linguistics, such a nar- row view of collocation seems appropriate because it illustrates a central characteristic of language. Collocation in this sense is something idiosyncrat- ic, it concerns knowledge about the individual word that cannot easily be de- scribed in general rules. The extent of this knowledge - which is part of Sinclair's (1991) idiom principle - should not be underestimated considering 15 For the importance of collocation in foreign language teaching see Bahns (1993ab). 16 Cf. Herbst (forthcoming 1996). 391 En gl ish S tu di es 1 99 6. 77 :3 79 -3 93 . d ow nl oa de d fro m w w w .ta nd fo nl in e.c om that competent native speakers clearly seem to associate a very small number of lexical items with a word such as entirely, and this, as Greenbaum has shown, in relation to particular contexts. Thus the analysis of collocation in this narrow sense also throws light on such theories within linguistics that state as their primary aim the discovery of very general, possibly universal rules of language. Frank Palmer once ended a lecture at Oxford saying something like: 'We'll just have to accept the fact that language is much more of a muddle than we'd like it to be'. Even if one agrees entirely with such a view of language, a significance oriented view of collocation may provide a way out of at least part of the muddle. Although one would generally prefer sandy beaches to false teeth, seeing collocations as the false teeth of language means identifying them as something at the heart of the matter. Universität Erlangen THOMAS HERBST REFERENCES AA Illustrated Guide to Britain's Coast, Basingstoke (1984). AISENSTAEDT, ESTER: 'Restricted collocations in English lexicology and lexicography', ITL 53 (1981: 53-61). ALEXANDER, RICHARD J.: 'Fixed expressions, phraseology and language teaching', Zeitschrift für Anglistik und Amerikanistik (1992: 238-49). BERRY-ROGGHE, G.L.M.: 'The computation of collocations and their relevance in lexical stud- ies', in A.J. Aitken/R.W. Bailey/A. Hamilton-Smith (eds): The computer and literary studies, Edinburgh (1973: 103-12). ALTENBERG, BENGT/MATS EEG-OLOFSSON: 'Phraseology in Spoken English: Presentation of a Project', in: J. Aarts/W. Meijs (eds.): Theory and Practice in Corpus Linguistics, Amsterdam (1990: 1-26). BAHNS, JENS: 'Lexical collocations: A contrastive view', ELT Journal 47/1 (1993a: 56-63). BAHNS, JENS: 'Kollokation kontra Kontext. Wider ein zu weites Verständnis des Kollokations- begriffs', Praxis 40 (1993b: 30-7). CARTER, R.: Vocabulary, London (1987). Collins Cobuild English Dictionary (Second edition ed. by JOHN SINCLAIR), London (1995) [Cobuild2]. COSERIU, E.: 'System, Norm und Rede', in: E. Coseriu: Sprache, Strukturen und Funktionen, Tübingen: (31979). COWIE, A.P.: "The place of illustrative material and collocations in the design of a learner's dic- tionary', in: P. Strevens (ed.): In honour of A.S. Hornby, Oxford (1978: 127-39). COWIE, A.P.: 'The treatment of Collocations and Idioms in Learners' Dictionaries', Applied Lin- guistics 2.3 (1981: 223-35). COWIE, A.P.: 'Phraseology', in: R.E. Asher (ed.) The Encyclopaedia of Language and Linguistics, Vol. 6, Oxford/New York (1994: 3168-71). FIRTH, J.R. (1956/1968): 'Descriptive Linguistics and the Study of English', in F.R. Palmer (ed.) (1968:96-113). FIRTH, J.R. (1957/1968): 'A synopsis of linguistic theory, 1930-55', in F.R. Palmer (ed.) (1968: 168-205). FIRTH, J.R.: 'Modes of Meaning', Papers in Linguistics 1934-51, London (1957: 190-215). GLÃSER, ROSEMARIE: Phraseologie der englischen Sprache, Leipzig (21990). GREENBAUM, SIDNEY (1974/1988): 'Some verb-intensifier collocations in American and British English' in: S. Greenbaum: Good English and the Grammarian, London (1988: 113-24). 392 En gl ish S tu di es 1 99 6. 77 :3 79 -3 93 . d ow nl oa de d fro m w w w .ta nd fo nl in e.c om HALLIDAY, M.A.K.: 'Lexis at a Linguistic Level': in C.E. Bazell/J.C. Catford/M.A.K. Halli- day/R.H. Robins: In Memory of J.R, Firth, London (1966: 148-62). HALLIDAY, M.A.K./R. HASAN: Cohesion in English, London (1976). HASAN, RUQAIYA: 'Coherence and cohesive harmony', in J. Flood (ed.): Understanding Reading Comprehension (1984: 181-219). HAUSMANN, F.J.: 'Wortschatzlernen ist Kollokationslernen', Praxis des neusprachlichen Unter- richts 31 (1984: 395-406). HAUSMANN, FRANZ JOSEF: 'Kollokationen im deutschen Wörterbuch', in: H. Bergenholtz/J. Mugdan (eds.): Grammatik im Wörterbuch, Tübingen (1985: 118-29). HERBST, THOMAS: 'A Proposal for a Valency Dictionary of English', in: R.F. Ilson (ed.): A Spec- trum of Lexicography, Amsterdam (1987: 29-48). HERBST, THOMAS: 'On the way to the perfect learners' dictionary: A first comparison of OALD5, LDOCE3, Cobuild2 and CIDE', International Journal of Lexicography (forthcoming). JONES, S./J. SINCLAIR: 'English lexical collocations. A study in computational linguistics', Cahiers de lexicologie 24/25 (1974: 15-61). KJELLMER, GÃRAN: 'Aspects of English collocations', in: W. Meijs (ed.): Corpus Linguistics and Beyond, Amsterdam (1987: 133-40). KJELLMER, GÃRAN: A Dictionary of English Collocations, Oxford (1994). LEECH, GEOFFREY: Semantics, Harmondsworth (21981). LIPKA, LEONHARD: An Outline of English Lexicology, Tübingen (1992). Longman Dictionary of Contemporary English (Third edition; ed. by DELLA SUMMERS), Harlow (1995) [LDOCE3]. LYONS, JOHN: Semantics 2, Cambridge (1977). MACKIN, R.: 'On Collocations: "Words shall be known by the company they keep" ', in: P. Strevens (ed.): In honour of A.S. Hornby, Oxford (1978: 149-65). MACINTOSH, ANGUS: 'Patterns and Ranges', in: Language 37 (1961). MITCHELL, T.F.: Principles of Firthian Linguistics, London (1975). Oxford Advanced Learner's Dictionary of Current English (Fifth edition; ed. by JONATHAN CROWTHER), Oxford (1995) [OALD5]. PALMER, FRANK R.: Semantics, Cambridge (21981). PALMER, FRANK R. (ed.): Selected Papers by J R Firth 1952-59, London/Harlow (1968). ROBINS, R.H.: Genera! linguistics. An introductory survey, London (21971). SINCLAIR, JOHN: 'Beginning the Study of Lexis', in: C.E. Bazell/J.C. Catford/M.A.K. Halli- day/R.H. Robins: In Memory ofJ.R. FIRTH, London (1966: 410-30). SINCLAIR, JOHN: Corpus Concordance Collocation, Oxford (1991). SINCLAIR, JOHN: 'Text Corpora: Lexicographers' Needs', Zeitschrift für Anglistik und Amerikanistik (1993: 5-14). SVARTVIK, JAN: 'Lexis in English language corpora', Zeitschrift für Anglistik und Amerikanistik (1993: 15-30). 393 En gl ish S tu di es 1 99 6. 77 :3 79 -3 93 . d ow nl oa de d fro m w w w .ta nd fo nl in e.c om