Review of Alison Wray (2002), Formulaic language and the lexicon. Cambridge UK: Cambridge University Press. 332 pp + xi. US$65 Hardback.


By Tom Cobb

Dépt de linguistique et de didactique des langues

Université du Québec à Montréal

For Canadian Journal of Applied Linguistics.

It is almost 20 years since the "lexical phrase" burst onto the applied linguistics stage, with a chapter by Pawley and Syder (1983) in an otherwise forgotten volume for language teachers. The concept has implications which, if taken seriously, could revolutionize our views of language use, acquisition, and pedagogy, and possibly even human cognitive architecture. Phrase theory stands on its head the notion that normal language use involves mainly the assembly of primitive linguistic units through the application of grammatical rules, re-describing it as the production and reception of precast lexical strings of various lengths with only occasional recourse to grammar operations. The proof of the phrasal pudding, so to speak, is twofold. First, it is doubtful whether people have the memory resources needed for online language processing on a grammatical or analytic basis alone, i.e. without recourse to many largish chunks of language that are accessed whole, like words. Second, while a grammar may make indefinitely many word combinations possible, only a fraction of these will ever see the light of day. Would you like to become my spouse? and Will you marry me? are equally acceptable, grammatically, but one of them is almost always used, the other almost never. There has long been interest in the role of lexicalized phrases in language use, of course, but until recently no means of proving that it was more than a marginal phenomenon. It was only with the computer analysis of large corpora, for example by applied linguists working on the COBUILD and related projects in the late 1980s, that the extent of our reliance on precast, formulaic language became clear. With the phenomenon thus noted (by Pawley & Syder) and the extent of it validated (by the corpus studies), the next task was presumably to work out its implications, establish methods of investigating it, and propose hypotheses about what it means, and this was the task Alison Wray set herself in her book-length treatment.


As Wray argues in a compendious review of the lexical phrase research, phrase theory has implications for language use at all levels. As mature native speakers of a language, we apparently produce and interpret “ready made surface structures” (p. 13) for nearly all of our communicative functions (burst onto the stage, otherwise forgotten, stands on its head, proof of the pudding, see the light of day), retrieving sometimes quite lengthy strings from memory as single lexical units, while using our “live grammar and lexicon” (p. 33) sparingly, mainly for stitching the precasts together. We thereby reserve our main energies for idea generation and interpretation, and of course for an occasional novel construction should the need arise (novel constructions commonly being where the meat of an utterance lies and requiring some effort to interpret, particularly if bearing a mixed metaphor or other sign of on-the-fly assembly). As language learners, we apparently learn our first languages largely through hearing, storing, and reproducing recurring extended whole sequences corresponding to recurring extended whole contexts and situations, presumably on an associative rather than instinctual basis, committing these to analysis only on an as-needs basis. As cognitive systems, we are apparently more reliant on massive and possibly redundant information storage than we are on streamlined computation from primitive units, as we used to think when Chomskyans ruled the roost.


Thus Pawley and Syder's chapter on phrases and second language pedagogy had implications far beyond its brief, surely a case of the tail wagging the dog. Despite this, the implications of phrase theory still remain to be worked out for language pedagogy itself. Outstanding questions include these:


  1. Does phrase learning function for second language learners as it does for first?
  2. To what extent should lexical phrases be included in a language learning syllabus?
  3. Does whole-phrase learning eventually lead to grammatical analysis of phrases and re-use of constituents?

It is not only in second language studies that the (re)discovery of the lexical phrase has introduced a new set of difficult issues. Also affected and disrupted to varying degrees are linguistics proper, the modeling of normal and abnormal language functioning, cognitive theory, and possibly others. It is predictable, then, with the lexical phrase being approached from several perspectives that terminology and methodology might both stand in need of a tidy-up, and this is where Wray's ambitious task begins. Her goal is nothing less than to organize and synthesize recent work on the lexical phrase, and following that to offer an explanatory model that puts it all together and secures the way forward for future researchers.


Wray begins at the beginning, looking first at the problem of determination. How do we know when a word string is a lexical phrase, accessed whole rather than grammatically generated? A good deal of hard thinking has gone into this question in recent years, and interesting approaches have been explored, including the examination of speech rate (lexical phrases run fast and slur their consonants), pausing (lexical phrases have fewer pauses), and corpus frequency studies (lexical phrases can be counted by software that extracts all strings of length>x and frequency >y). As already mentioned, phrase frequency was an early proof in the phrase argument (Will you marry me? outnumbering alternate formulations in any large corpus). Wray, however, is not merely summarizing the phrase research, but also moving it forward, and the frequency issue presents a good example of this.


Phrase recurrence, I was interested to learn, does not in itself indicate that a phrase is being processed as a single lexical item. To know if this is the case, you need to know the pragmatic intent behind a particular utterance, and the art of corpus tagging has not yet advanced to this point. For an example (mine, not hers), Shut your mouth is probably a lexical unit if the intent is to make someone stop talking, but a generated sentence if the dentist is signalling a time-out from oral surgery. In other words, the same string may function as a unit in some contexts but not in others, so that what counts as a lexical phrase can only be characterized dynamically, and the lexicon must be considered multi-representational.


Wray examines the phrase issue in a number of research contexts--first language acquisition, adult first language functioning, second language acquisition, and impaired language functioning (having already published widely in all these areas). Her text, while complex, is readable, mainly because of the lively examples supporting the main points. To sample one or two, the multi-representationality just mentioned comes to life in examples from aphasic patients, such as one who frequently resorted to the phrase son of a bitch while unable to identify his own son in a photograph; or from normal language users, who typically cannot tell you what Rice Krispies are made of, since their representation of the lexicalized unit does not necessarily make contact with rice and crisp stored elsewhere in the lexicon. The text is also studded with syntheses of research findings at an appropriate level of detail and with an always obvious relevance. The reader moves easily between examples, findings, and big picture topics, in much the same way the author proposes language users move between on-line computation from primitive units and wholesale dealing in larger chunks, according to the need.


And what is the need for formulaic language in human communication? Wray considers several possibilities, such as the easier online language processing already mentioned, and finds none of them adequate to account for the extent of the phenomenon. The most novel part of her treatment is to propose a unifying explanation for the prominence of formulaic language, which, unexpectedly, is unrelated to language processing per se. Lexical phrases, she argues, are used mainly for  signalling group membership and specifically for "the promotion of self." When we want to get our needs met, issue orders, or manipulate others, we do not trust to novel constructions, which may go awry 'twixt speaker and hearer, but instead to precast whole constructions known in advance to both parties.


While Wray argues the self-promotion explanation long and well, in the end I found myself unconvinced. For one thing, any explanation positing a single drive as the basic motivator of human behaviour (like Freud's "sex drive" or Marx's "mode of production") is vulnerable to Popper's charge of unfalsifiablilty. As already mentioned, there seem to be basic problems with empirical testing of several of the most interesting ideas about phrases, and this is especially true where the goal is unification and model building. Still, whatever the eventual fate of this particular explanation, Wray's attempt to gather the pieces together and make sense of them is bound to be the point of departure for the next major expedition into phrase territory.


As a second language specialist, I noticed that when Wray deals with second language research it is not particularly with the goal of producing a set of pedagogical implications for language teaching. Her goal is mainly to provide psycholinguistic explanation, and second language learning is just one of her several data sources. Nevertheless, most of the questions about phrases and language teaching that I set out above receive some sort of answer along the way. Unfortunately, none of the answers serve to make second language learning or teaching seem any easier.


  1. Does phrase learning function for second language learners as it does for first?

    A recurring theme in the analysis is that lexical phrases are mainly learned prior to the acquisition of literacy, because written language makes the separation of individual words very clear. Since most people attempting to learn a second language have already acquired literacy in their first, and are likely to use literacy as a primary aid in acquiring their second, it is unlikely that phrases will play the same roles in second language acquisition and functioning as they did in first. Second language learners are much more likely to be forced to resort to a classic Chomskyan grammar-and-lexicon mode of language use as their principal option, with all the psycho- and socio-linguistic difficulties this entails.


  1. Should lexical phrases be included on the language learning syllabus?

    Many pedagogical applied linguists assume that a syllabus of selected phrases should be taught to L2 learners as a sort of vocabulary, (e.g., Nattinger & deCarrico, 1992, p. 32, and of course Syder & Pawley, 1983). However, Wray seems sceptical about this. For one thing, as noted in the discussion of literacy just above, post-childhood learners tend to be at a stage where analytic approaches to learning are somewhat inevitable.

    A practical problem with teaching phrases is the sheer number of them: the number of identifiable phrases that can be formed from, say, 3000 high-frequency words is truly stupendous (ten phrases per word, according to Mel'čuk, 1998, cited p. 283). Given that not all learners manage to learn 3000 basic words qua words, the idea of building a second phrasicon through deliberate instruction seems problematic.

    Still more problematic is the possibility that, if we accept the self-interest explanation Wray offers, then we really would have no way of knowing what an appropriate target for a second language phrasicon might be. An extremely fluid and dynamic picture of phrasicon-lexicon interactions is built up over the pages of this book, with “formulaic sequences as just one of many solutions which arise for an individual on a particular occasion in the course of protecting his or her interests” (p. 211). This implies that we really have no native-speaker standards for any performance components below the level of overall success in using the language to promote one’s interests, and therefore we have no way of grading or sequencing a phrase-based syllabus for second language learners.


  1. Does whole-phrase learning eventually lead to grammatical analysis of said phrases and re-use of constituents?

    Few would deny that beginning language learners might usefully be provided with some amount of formulaic language, for initial communicative purposes, that goes beyond what their interlanguage grammars could independently generate at that stage (Could you tell me the time? etc.) Less obvious is whether it can be assumed that such constructions will later be analysed (so that could you recombines as could he, you could, and so on). There are some problems in principle standing in the way of this happening. One stems from the fact that the grammatical analysis of phrases is not always possible. For example, implications need to be worked out can be fully regrammaticalized, e.g., rendered in active voice; rule the roost can receive only minor modification, such as person, tense, and extra modifiers; 'twixt X and Y while relying on a remote literary allusion will accept any two nouns, but no modifiers; the tail wagging the dog can hardly be modified at all, except in humour; and by and large is unmodifiable, if not ungrammatical. In other words, the second language learner's task is not only to commit some holistic phrases to analysis, but before that to decide when this is and is not possible—with, presumably, insufficient means for doing so.


As noted above, pedagogical applied linguists rediscovered the lexical phrase without apparently doing much with it, perhaps for the good reason that there is not much that can be done with it. This is despite the fact that non-idiomaticity is normally the final issue for advanced learners (Granger, 1998), or maybe it only is for their teachers.


 This review has turned out longer than I expected, but I have hardly sampled from the book's revelations and revolutions, and then only from my own point of view. Readers involved in any aspect of language as communication should read this book, which is bound to become a classic of our field that will be cited for years to come. It may even be re-issued, at which time its publishers might consider completing their work on the names index, where at present one can attach page numbers to only two names, Baudelaire and Field Marshall Montgomery, but to none of the host of language specialists extensively cited--including Chomsky and the author herself.




Granger, S. Ed. (1998) Learner English on computer. London: Longman.


Nattinger, J., & DeCarrico, J. (1992). Lexical phrases and language teaching. New York: Oxford University Press.


Pawley, A., & Syder, F.H. (1983). Two puzzles for linguistic theory: Nativelike selection and nativelike fluency. In J.C. Richards and R.W. Schmidt (Eds.), Language and communication (pp. 191-226). New York: Longman.