Pilgrims HomeContentsEditorialMarjor ArticleJokesShort ArticleIdeas from the CorporaLesson OutlinesStudent VoicesPublicationsAn Old ExercisePilgrims Course OutlineReaders LettersPrevious EditionsLindstromberg ColumnTeacher Resource Books Preview

Copyright Information

Would you like to receive publication updates from HLT? You can by joining the free mailing list today.


Humanising Language Teaching
Year 5; Issue 2; March 03

Major Article

My good-bye to the Lexical Approach

Seth Lindstromberg

If this article interest you, Pilgrims offers courses in this area. Click here for more information.

  1. The Lexical Phrase Hypothesis
  2. Some characteristics of lexical phrases
  3. Collocations
  4. The essence of the Lexical Approach
  5. Applying the Lexical Approach
  6. So is there a way to learn all this vocabulary, collocations and stuff?
  7. So maybe the LPH has been widely over-stated
  8. Good-bye!

1 The Lexical Phrase Hypothesis

A major element in the theoretical basis of the Lexical Approach to teaching/learning an additional language is the Lexical Phrase Hypothesis1 (LPH), that is—

Many clumps of words—e.g., as well = 'too' and by and large = 'in general'—are not only used but also stored in memory and processed as if they were single words.

This proposition derives from theoretical linguistics. In applied linguistics, and in the lore of TESOL, it is often framed so as to include not just fixed multi-word expressions such as the ones just mentioned but also ones that are only relatively fixed (see Section 3, 'Collocation').

The LPH has strong and weak forms—

Strong—A very large proportion of language-in-use consists of lexical phrases.
Weak—The proportion is significant but not overwhelming.

Lewis (1993) advocates an extremely strong form which I summarize as follows—

Language-in-use is not primarily individual words combined/parsed according to rules of grammar); rather it is primarily recurrent word combinations, or lexical phrases that are stored, accessed and processed as if they were single words).

Part of the rationale for the LPH is that speakers of a language gain substantial benefits from treating strings of words as if they were units in themselves—

  • increased fluency (when speaking and writing)
  • faster comprehension of when reading and listening generally
  • better comprehension, especially when listening to fast speech.

In other words, by storing multi-word expressions as chunks, we can recall and use them without having to mentally construct them from individual words each time we want to use them or mentally parse (grammatically analyze) them each time we read or hear them. Similarly, when we hear or read a lexical phrase, we interpret it more swiftly if we take it 'as a chunk'. An alternate term for lexical phrase, which suggests these functional benefits, is pre-fabricated language.

One bit of evidence for the LPH is that, in the course of acquiring their mother tongue, children pass through a stage where they use phrases as if they were single words.

Another is that—if we examine what people say and write—we do not find that they are combining words (into phrases) in all the meaningful ways allowed by supposed rules of grammar. Instead, much the same phrases tend to occur over and over again—e.g., take umbrage at but not feel umbrage on account of..

So far as I know there is as yet no certain way (e.g., looking at images of brain activity) of saying for sure that a given phrase really and truly is stored and processed as a single lexical item by any particular person, let alone people in general.2 Therefore, when people (me, for instance) say that a particular group of words is a lexical phrase, what they mean (or ought to mean) is that these words might be a lexical phrase; they might, that is, be a phrase which the mind in some significant respect treats as if it were a single word. And that is what I will mean from now on when I speak of lexical phrases even though, for the sake of readability, I will omit the might.

Some characteristics of lexical phrases

A lexical phrase is a group of words which forms a grammatical unit of some kind and which exhibits a degree of 'inflexibility'. As to the last feature, some lexical phrases are totally 'frozen' (unchangeable) while others are rather variable:

  • Invariable phrases: by and large, as well, let alone, so be it
  • Somewhat variable phrases: Don't rock the boat, She's rocking the boat…[rock the boat = 'disturb the institutional status quo']

It is often claimed that certain long clichés are remembered like single items of vocabulary, e.g., There's no time like the present; Never a dull moment; It never rains but it pours, God only knows. If so, there can be lexical sentences. (Multi-word lexical unit would therefore be a better term than lexical phrase.)3

Some lexical phrases are highly idiomatic (i.e., unguessable from component words)—e.g., by and large (= 'generally'). Others are not—e.g., pick up a bad habit (whose meaning can easily be guessed by a learner who knows a common meaning of each word in the phrase).

In terms of form, lexical phrases are categorized in various ways. Here is one—

  • polyword - at any rate, by and large, as well [= 'also']
  • frame or slot - the [adj.]-er the [adj]–er, as [adj]….as,
  • so [adj]…that… ,
  • Little did…realize that…
  • sentence head - Could you....., God only knows wh-…
  • sentence tail - …, if you would., …and so on.
  • cliché There's more than one way to skin a cat.

Some lexical phrases have 'speech-functional meaning'. For example, Could I…? has the function in speech of introducing a request. Others have lexical meaning, which is to say that they are directly definable. Phrasal (or multi-word) verbs are like this. For instance, bump into is definable as 'meet by chance'.

As with vocabulary generally, lexical phrases may be formal (Would you mind if…?) or informal (Put up or shut up, Gimme… = 'Give me…', D'ya wanna…? = 'Do you want to…').

In short, believers in a strong version of the LPH consider 'vocabulary' to include not just thousands of individual words but lexical phrases in their hundreds or, more probably, thousands. (Has anyone tried to count them?) But there is more. There are collocations too, thousands and thousands of them. The LA says that vocabulary in this greatly expanded sense should be the prime language focus of our teaching.


It seems to me that people in TESOL, in their use of the term collocation, sometimes give it wider scope than lexical phrase. That is, the term collocation refers to both fixed lexical phrases such as kith and kin (= 'relatives') and relatively loose associations of words such as heavy rain. (This linkage is loose in the sense that there are more or less synonymous alternatives to heavy rain. So while strong rain is unnatural, hard rain is possible; rain can also be pounding, drenching, driving, torrential and so on.) But sometimes collocation is used to refer just to more or less loose associations. Just to be clear, I will adopt the latter practice.

It is sometimes said that a word may collocate left or right. For instance, the noun rain can be followed by a great many words but there are fewer that come to its left. It is therefore combinations with these words that are more frequent and, we may suppose, more habit governed. All this is meant by saying, rain 'collocates left' (like this: heavy <= rain).4 By the same reasoning, so does umbrage , take<=umbrage. Depend, on the other hand, can be preceded by many words but it is generally followed by on. So depend collocates right, like this: depend => on. And so does similar: similar=>to.

Sometimes collocation seems to be more or less evenly bi-directional (e.g., rock <=>the boat).

It is often pointed out that computer-run analyses of multi-million word 'corpora' (a 'corpus' is a large collection of texts) have made it relatively easy to search out large numbers of collocations. Before the advent of the computer this was an exceedingly time-consuming enterprise. This is why the LA is recent.

The essence of the Lexical Approach

A common view of language competence is, partially and roughly, as follows:

Speakers and writers intend meanings which they express through the appropriate choice and combination of words. These words may, additionally, be nuanced by means of, for instance, grammatical inflection.

To put this in an even smaller nutshell:

  • intended meanings => words + grammar/morphology=> meaningful output

    The converse (for listeners and readers) is:

  • input=> grammar + words=> understood meanings

    Before the LA came on the scene, it was very commonly held that a learner with a moderately substantial knowledge of grammar (including morphology) and only a few thousand words could be a strikingly effective communicator. Consequently, there was near unanimity among methodologists, materials writers and teachers that learners, right up to upper-intermediate level, should spend a lot of time on grammar.

    There is something theoretically elegant about this view of things; and it had such promise of efficiency! Indeed, it did not always fail as a method of instruction.

    The premise of LA, in contrast, is that language learning is achieved largely by the brute ability of the human mind to learn, store and process individual lexical items (=words and recurrent word combinations).

    To elaborate—

    A common non-LA view is that our ability to learn vocabulary is not particularly outstanding. What is outstanding is our ability, in the instant before speaking, to speedily combine single words into meaningful strings and, when listening, to parse and understand at lightning speed. In this view, our memories are relatively small but our on-line processors are very powerful.

    The LA view is that our ability to learn vocabulary is massive but our ability to do on-line grammar isn't. We are just not that quick at combining single words into meaningful strings and we tend, as much as we can, to avoid using grammar to build utterances word by word. As listeners and readers, the speed with which we understand goes up when we are confronted with text that is full of lexical phrases and familiar collocations. In short, our memories are vast but our on-line processors are rather weak.

    An implication of the latter view, given acceptance of the LPH, is that a lot of time must be devoted to learning vocabulary (including thousands of recurrent word combinations) so that learners will be able to—

    • to compose and construe utterances quickly, as they must do if they are to be competent in face to face communication
    • to speak and write with natural phrasing, e.g., to avoid collocations such as strong rain which are grammatically correct but not natural.

    And, the stronger your version of the LPH, the more hundreds or thousands of vocabulary items you think your students have to learn.

    Applying the LA

    For the teacher, the biggest question relating to the LA is, “How do you translate it into practice?” Personally, I don't think that advocates of the LPH have answered this question in a way that adds much to current mainstream practice as reflected in, say, recent offerings of UK publishers. (But read Lewis 1997 and see what you think.) Further, I reckon there is room to doubt whether it is indeed possible to translate into practice any version of the LA which is based on a strong form of the LPH.

    For one thing, there does not yet seem to be any means satisfactory to education bureaucrats or to the merely curious of deciding what vocabulary to teach at what level, in particular, what vocabulary to teach beyond elementary level.

    1. No comprehensive vocabulary frequency list made within the past 50 years says which meaning of a word is most frequent…and therefore (perhaps) most deserving of priority in the classroom. Think of such polysemic words as on, right and sound and you will see what a problem this might pose, since I assume you will not want to make a habit of teaching all meanings at the same time.
    2. Lexical phrases are sparsely included in such lists, if at all.
    3. So far as I know there exists no frequency list which includes pedagogically useful information about the commonest thematic contexts of this or that unit of vocabulary.
    4. In any case, publishers (who have been particularly active in the funding of list compilation) try to keep their frequency lists secret.

    In short, any immediate, detailed application of the LA in syllabus creation would be premature.

    Things are only a bit better with respect to implementation in the classroom. To start with, any ideal Lexical Method which rests, ultimately, on a strong version of the LPH would enable students (who in general cannot spend more than a few hours a week in class) to learn vocabulary items (including recurrent word combinations) at the rate of several score per hour. But, the claims of Suggestopedia notwithstanding, no such method is available.

    So doubt forms. Is a truly effective method for mega-vocabulary learning a realistic hope? Let's look at some of the suggestions that have been made.

    Again and again over recent years I have come across articles in which authors recommend using concordances in the classroom.5 But having students do this can, at best, lead to a minutely tiny increment in their vocabularies because—

    • Concordance analysis is a highly bookwormy type of activity and few people are willing bookworms. It's hard enough to get students, particularly kids, to read stuff that is intrinsically interesting. Getting them to pore over collocations so that they may discover how this or that bit of vocabulary patterns is minimally realistic. It is, like the Soviet collectivization of agriculture, a great idea if you ignore people's likely reactions to it.
    • It is far from invariably easy to see from a very limited co-text (which is precisely what a concordance is) what this or that unit of vocabulary might actually have been intended to mean in its situation of use. True, one might (this is at least theoretically possible) go back to the corpus to see the larger context: but who except an out and out bookworm would do this?
    • Concordance study strongly tends to lead to in-depth learning about the near co-text of particular vocabulary items rather than the learning of items in significant numbers per se. But the gist of a Lexical Method is that quantity is key.

    All in all, collocation study is likely to be a grotesquely inefficient use of class time if one does more than ask learners (unless they plan to become linguists) to investigate the collocation of a few highly frequent words. OK, if we are talking about raising language awareness, then this activity could be useful to some…you never know. But if we are talking about any significant boost to vocabulary learning, then I think advocacy of concordance study in the classroom is deep in the realm of baseless hope…but still not as deep in as the hope that a significant number of students would do this kind of thing on their own time, learner training or no learner training.

    Michael Lewis, who has advocated the LA with particular zeal (e.g., Lewis, 1993), is a particularly interesting case. Although he talks revolution, he concedes (p. 193) “that no one methodology represents the way forward”.

    A key conclusion he draws from a strong version of the LPH is that learners need to devote a lot of time to study-learning lexical units. The instructional/learning pattern he recommends is that students should first observe data, that they should then hypothesize about it, and then experiment.

    It is not that I object to any of this at root. But what I do not think this plan of action promises is the requisite increase in the rate of vocabulary learning for if the assertion that language is mostly lexical chunks means anything, it means that students who wish to be comprehensively communicatively competent have to learn tons and tons more vocabulary than anyone used imagine even in their dreams.

    The stress Lewis lays on learner training is more to the point. With all this vocabulary to learn, people are going to have to learn most of it on their own time. But the idea of learner training has been around for some time and as worthwhile as it may be—provided it is seldom overt (being explicitly trained to learn is not something that fascinates students for long—there is no revolution here, not as it has been described. All in all, an examination of the Lewis's specific recommendations leads me to conclude that, for all his fiery talk of fundamental change, what he is advocating in the way of method is an assortment of activities and task types familiar not only from Communicative (but not particularly Humanistic) methodology but also, in too large a part, from the methodology of dry-as-dust workbooks.

    But this is still not the weightiest objection to a strong version of the LA (i.e., a version based on a strong form of the LPH).

    So is there a way to learn all this vocabulary, collocations and stuff?

    In order for a strong version of the LA to be more than just plausible and thought-provoking, it must be implementable in a way that reflects these propositions—

    A normal foreign/second language learner is endowed with the ability to—

    1. retain in memory and quickly access and process a staggeringly vast number of vocabulary items of all sizes
    2. learn vocabulary, including lexical phrases and other recurring word patterns, at a very great rate.

    In other words, what the LA seems to cry out for is a purpose-built Lexical Method, or two, or three, or more. But so far, no such method is anywhere in sight.

    This is not to deny that, now and again, some very good relevant material comes onto the market. Examples that spring to mind are—

    • the learners' dictionaries put out by (especially) UK publishers (all rich in information about collocation…though all still disappointing with respect to prototypical lexical phrases).
    • English Vocabulary in Use (McCarthy, Michael and Felicity O'Dell. 1994. Cambridge)

    Such bright spots not withstanding, not only has there been no lexical revolution in classroom (as opposed, e.g., to lexicographic) methodology; it is not even clear what one would look like. And no revolution either in learner training, I suggest.

    It merely, as before, seems likely that we can help our students learn vocabulary by providing—

    • interesting comprehensible input that is rich in recurrent word combinations.
    • interesting activities that help students notice and recall chunks of these texts.
    • opportunities for students to speculate, to ask and to hear about why it is that some words do or do not recurringly combine with certain other ones. (I will return to this point in the Section 7.)

    We could, therefore, do worse than …

    • use song lyrics a lot
    • often use other kinds of short authentic texts
    • (See especially,
      • Davis, Paul and Mario Rinvolucri. 1987. Dictation. Cambridge U. Press.
      • Maley, Alan. 1994. Short and Sweet. Penguin.
      • Maley, Alan. 1995. Short and Sweet II. Penguin.)
    • re-double our efforts to entice students to read for pleasure (e.g., by combining the reading of a novel with showings of a filmed version.)
    • apply Suggestopedia not to the lame dialogues normal to the method but instead to material of quality, e.g., sections of the filmscript of My Fair Lady, the lyrics of 'Mr Tambourine Man'…whatever seems right for your group.
    • involve students in conventional amateur dramatics.

    So maybe the LPH has been over-stated

    Yes, it has been. Proponents of a strong version of the LPH under-play the importance of semantics at the word level. For instance, they maintain that many high frequency words, such as spatial prepositions, are 'delexicalized', which is to say that they are often used with no meaning at all. The implication of supposed delexicalization is that students shouldn't try to understand what prepositions mean; instead, they should somehow commit to memory many hundreds, or thousands, of lexical phrases which contain prepositions. I have two things to say about this.

    1. That would require a lot of phrase learning, much more learning by far than would be involved in adequately learning the meanings of 70 or so spatial prepositions.
    2. This view of prepositional meaning is false. Rather, prepositions are rarely delexicalized as I will now briefly argue. (Fuller arguments have been made in Lindstromberg 1996, 1997 and 2001.)

    Let's take for example something I wrote above in Section 5: students would do this kind of thing on their own time. I am quite sure that many LA enthusiasts would identify on [their] own time as a lexical phrase, more particularly, as a 'frame' which students should commit to memory as is. But this view is unable to account for these facts—

    • One could also say in [their] own time.
    • In [their] own time has a different meaning from on [their] own time.
    • The difference in the meaning of the two phrases derives from the difference in meaning of the two prepositions (which therefore cannot be considered as being delexicalized).

    The explanation is as follows:

    It is common to speak and think of unwelcome events, states and concerns as burdens that one can be 'under' (just like you can be under a heavy burden) or which can be 'on' whoever is experiencing them (just like a heavy burden can be on your back) —e.g.,

    take on a [burden of] responsibility, the straw that broke the camel's back, be under strain, The problem weighed on his mind, Her car died on her.

    Do something on one's own time is in this group. That is, when I say that students would do it on their own time, the nuance that on supplies is that students' own time (their free time) suffers through being used up by teachers' demands on it. The expression in their own time has no such meaning because the preposition in has no conceptual relation to the notion of burden. Rather (because in is about containment), in their own time means merely 'during their own time'.

    There is a similar difference in meaning between have something on your mind and have something in [your] mind. The former is a concern or problem, i.e., a metaphorical burden. The latter is not.

    There is a virtual infinity of well-formed expressions like on your mind and in [your] mind. Learners can either try to memorize them one by one with comparatively little regard to the meanings of the individual words that make them up (this would be the advice of an LA fan), or, more efficiently…

    • learn some of them one by one (e.g., the straw that broke the camel's back)
    • but, for the rest,
    • become aware of operative metaphors
    • learn the meanings of individual words such as under, on and in.

    In short, over-reliance on the LA would lead away from the discovery of real and usefully generative word meanings and into inefficiency.


    There is a great deal of evidence for a weak form of the LPH—a version, that is, which does not radically under-state the importance of grammar and of word-level meaning. But accepting this qualification raises the question—Is the Lexical Approach a useful construct? Given that it suggests no new marching orders, my view is that it is not. Discard it, I say, but at the same time incorporate a weak form of the LPH into the set of beliefs underlying whatever pedagogical approach you have already been operating with. It is perhaps too bad though if you have been a zealous LA practitioner. Certainly, pedagogy based on a strong version of the LPH should be avoided since that would mean having a view of language which (1) suggests that language learners have an astronomical amount of vocabulary to learn but (2) offers no promising tips on how they can speed up enough to do it. Fortunately then, a strong version of the LPH is (3) inaccurate.

    I am very grateful Rick Cooper for his comments on a late draft of this. And thanks too to Michael Lewis for pointing out that terms in a key passage on page one were in the wrong order.


    1. Also underlying the LA is the belief that fluency is a more important global learning aim than accuracy, grammatical accuracy in particular.
    2. Pinker (1999: 295-99) discusses the potential of four techniques of 'functional neuroimaging' (positron emission tomography, functional magnetic resonance imaging, event related potential and magnetoencephalography) to reveal how much of language comes from rules and how much is listemes. He says there is hope that good evidence may be in the offing.
    3. Pinker (1999, p. 26) suggests the name listeme as a blanket term for all the items (words and multi-word units) that we can find listed in our mental dictionary, pre-assembled.
      Justifying statements about direction of collocation is not necessarily a straightforward matter. For instance, my colleague Rick Cooper observed, on reading this passage, that rain collocates to the right with dogs (i.e., rain dogs is in a Tom Waits song), Man (as in the movie Rain Man), with check and forest, for instance. I might counter his counter-examples by saying that maybe rain serves the function of adjective in them, not that of noun; but I would not be terribly confident that my reply is to the point.
    4. Re the practice of concordancing, see, e.g., Stevens, Vance (1995) at http://www.ruf.rice.edu/~barlow/stevens.html Cobb, Tom (1997) at http://www.er.uqam.ca/nobel/r21270/cv/Hands_on.html
      Re concordances pure and simple (e.g., if you just want to know what they are), check out: http://www.nsknet.or.jp/~peterr-s/concordancing/ http://ec.hku.hk/macomp/malangd.html http://www.rjcw.freeserve.co.uk/

    For a rosier view of LA see
    Olga Moudraia at http://www.cal.org/ericcll/digest/0102lexical.html

    Lewis, Michael. 1993. The Lexical Approach: The State of ELT and a Way Forward. LTP.
    Lewis, Michael. 1997. Implementing the Lexical Approach: Putting Theory into Practice. LTP.
    Lindstromberg, S. 1996. 'Prepositions: meaning and method'. ELT Journal, 50/3: 225-36.
    Lindstromberg, S. 1997. English Prepositions Explained. John Benjamins.
    Lindstromberg, S. 2001. 'Preposition entries in UK monolingual dictionaries'. Applied Linguistics 22/1: 77-103 Pinker, Steven. 1999. Words and Rules: The Ingredients of Language. Phoenix.

    Back to the top