1. Background
Few will take issue with the assertion that it is important for second language (L2) learners to build a sizeable vocabulary, and vocabulary is no longer by any means the neglected area in L2 research that it used to be (Meara Reference Meara1980). Vocabulary size has been found to be a major contributor to proficiency. Iwashita et al. (Reference Iwashita, Brown, McNamara and O'Hagan2008) found vocabulary use to be the strongest predictor of oral proficiency ratings, and correlation coefficients of .70 and higher have been reported between learners’ vocabulary size and their scores on reading and listening comprehension tests (Qian Reference Qian1999; Staehr Reference Staehr2009). It is estimated that good comprehension of a variety of (non-technical) texts requires knowledge of over 7,000 word families (Nation Reference Nation2006; Staehr Reference Staehr2009; Webb & Rodgers Reference Webb and Rodgers2009; Schmitt, Jiang & Grabe Reference Schmitt, Jiang and Grabe2011). That's a lot of word forms to be learned, because a word family consists of a lemma and its derivations. For example, argue, argues, argued, arguing, argument, arguments, arguable, argumentation and argumentative arguably make up one word family. It is sometimes assumed that if a learner knows one member of the family, its relatives will be understood as well. This, however, cannot be taken for granted (Schmitt & Zimmerman Reference Schmitt and Zimmerman2002).
The vocabulary learning challenge is also augmented by the fact that many words are polysemous. There's a strong argument for task-based learning and They haven't made up yet after their argument show different uses of the word argument. She embraced her husband and She embraced Dynamic Systems Theory illustrate different uses of the word embraced. Learning a ‘word’ typically involves learning more than one form–meaning mapping. Although learners may be familiar with the word mushroom in She doesn't like mushrooms in her omelette, there is no guarantee at all that this will enable them to accurately interpret its extended use in Second-hand shops have mushroomed across town (MacArthur & Littlemore Reference MacArthur, Littlemore, Boers and Lindstromberg2008). Polysemy is found everywhere, but high-frequency words such as prepositions and multipurpose verbs (e.g. have) make up a segment of vocabulary where it is especially rampant. The upshot of this is that the aforementioned vocabulary size goals – daunting as they are – disguise the true amount of work awaiting learners, not only because word families comprise several word forms, but also because a single word form is likely to have several meanings.
Given the vast number of word forms and meanings to be acquired – obviously far too many to be tackled in the classroom – we have to be hopeful that learners will pick up a lot of vocabulary incidentally, for example during extensive reading (e.g. Krashen Reference Krashen1989). Incidental learning has been found to be a very slow process, however (e.g. Laufer & Roitblat-Rozovski Reference Laufer and Waldman2011).
It gets worse. Vocabulary knowledge extends beyond single words. Since the advent of corpus linguistics, it has become increasingly evident that most words prefer the company of some other words over that of near synonyms. This Idiom Principle – as opposed to the Open-choice Principle (Sinclair Reference Sinclair1991) – shows up in a panoply of word partnerships and multiword units, such as collocations (make an effort, a warm welcome, utterly disgusting), compounds (peer pressure, love handles), multiword verbs (turn up, follow through with), social interaction routines (nice to meet you, how are you doing), clichés (live and learn, publish or perish), idioms (jump the gun, close ranks), and discourse organisers (on the other hand, having said that). The term ‘formulaic sequence’ is now often used as an umbrella term for such word strings that are believed to be acquired, stored and retrieved for usage by native speakers in a ‘holistic’ fashion (Wray Reference Wray2002). They contribute to fluent language use and their (appropriate) use is a distinguishing feature of native speaker discourse (see Schmitt Reference Schmitt2004, Meunier & Granger Reference Meunier and Granger2008, Barfield & Gyllstad Reference Barfield and Gyllstad2009 and Wood 2010, for collective volumes on formulaic sequences in L2 learning). Students’ (accurate) use of formulaic sequences correlates significantly with their proficiency, which suggests that learners stand to gain from adding such multiword items to their L2 repertoires (Boers et al. Reference Boers, Eyckmans, Kappel, Stengers and Demecheleer2006b). But this addition to the vocabulary learning challenge can hardly be taken lightly, given estimates that half of native speaker discourse reflects the workings of the Idiom Principle (Erman & Warren Reference Erman and Warren2001) and observations that the incidental acquisition of L2 multiword units is a very slow process as well, especially in non-immersion contexts (Stengers et al. Reference Stengers, Boers, Housen, Eyckmans, De Knop, Boers and De Rycker2010; Laufer & Waldman Reference Laufer and Waldman2011).
To make up for the generally slow pace of incidental vocabulary uptake, Laufer (Reference Laufer and Foster-Cohen2005) makes a plea for targeting more vocabulary explicitly in language-focused activities and exercises. Language-focused learning is one of four strands that Nation (Reference Nation2007) argues should be present in roughly equal amounts in a general proficiency programme. The other three strands are message-focused input, message-focused output and fluency development. (Nation actually uses meaning-focused input/output. I prefer message-focused, because Cognitive Linguistics (CL) treats all linguistic units, by definition, as meaningful.)
Message-focused input means ample exposure to comprehensible input (cf. Krashen's Reference Krashen1985 Input Hypothesis). It is input made up predominantly of familiar language, which aids the learner in making inferences about the meaning of as yet unknown items and it engenders (implicit) knowledge of the common usage patterns of items they are already acquainted with. The strand of message-focused output is where learners experience gaps in their knowledge which they may subsequently try to fill (cf. Swain's Reference Swain1993 Pushed Output Hypothesis). The fourth strand is fluency development. Fluency development activities foster smooth recognition, retrieval and usage of linguistic elements.
The language-focused learning strand is where learners’ attention turns from what is said to how it is said. This attention to the code rather than the content of a message may be at the learner's own instigation but it will often be instigated by a teacher (or another interlocutor) or by a materials writer. It is the strand that comprises procedures where a certain feature or segment of language is the object of a lesson as well as interventions where the student's attention is temporarily directed to the code as a brief intermezzo during message-focused activities. These have become known as forms-focused and form-focused procedures, respectively (Long Reference Long, de Bot, Ginsberg and Kramsch1991). As we shall see, the procedures used in the majority of the Cognitive-Linguistics informed intervention studies have so far been of the former type.
According to Nation, the language-focused learning strand should take up just about a quarter of a learner's opportunities for learning. When it is decided to invest time in vocabulary instruction, one should have good reason to believe that it is a kind of instruction that gives an optimal return on this investment: that it will accelerate the acquisition process in comparison with other types of instruction within the language-focused learning strand or in comparison with opportunities for incidental acquisition through the message-focused strands. It is in this context that we ask whether there is a niche in language-focused learning for ideas borrowed from the school of thought known as CL.
2. Why turn to Cognitive Linguistics?
One yardstick for estimating the likely effectiveness of a vocabulary-learning procedure is the degree of engagement with the lexical item(s) that it stimulates (Laufer & Hulstijn Reference Laufer and Hulstijn2001; Schmitt Reference Schmitt2008). It has been suggested that ideas from CL can provide such stimulation. A description of CL that does justice to this increasingly influential school is beyond the scope of this article. The interested reader will find Evans & Green (Reference Evans and Green2006) an accessible introduction to CL, and Geeraerts & Cuyckens (Reference Geeraerts and Cuyckens2007) is a recent collection that displays the wide array of projects under the CL umbrella. In the area of language learning, Robinson & Ellis’ (2008) edited volume brings together CL proponents and psycholinguists specialised in research on usage-based second language acquisition. An illustration that more alliances are in the making is Lantolf (Reference Lantolf and Hinkel2011), who proposes an integration of CL and Sociocultural Theory.
Due to constraints of space, what follows can only be a crude account of CL, focusing exclusively on CL ideas that have been tried in vocabulary teaching. The CL movement began mostly as a reaction to generative linguistics, which was felt to treat language too much as a special-purpose component, divorced from general cognition and the way language is actually used. CL treats language and its acquisition as usage-based and as reflecting the general cognitive abilities that operate in our interaction with the world (e.g. Langacker Reference Langacker1987). In this line of thought, linguistic phenomena are considered to be ‘motivated’. Some things are more likely to happen in language than others because they are more congruent with habitual human perceptual and cognitive experience. If language were not motivated that way, we might not find The table was being run around slightly odd. At least one of the reasons why we find this statement odd is its incongruence with the figure–ground organisation we habitually apply to our perception of scenes. We tend to select as figure of a scene the thing that draws our attention most, and that is typically something that moves or that is animate. If language were not motivated, we would not be momentarily puzzled by A glass of alcohol containing wine, please. That is because the specified information in this utterance is at odds with the way we organise categories around prototypical members, whose characteristics are clear-cut and therefore require no further clarification. It is the peripheral members of the category that invite modifiers, not the central members. If motivation did not play a part in language development, there would be no reason why Shall I give you a foot with that? has not become an institutionalised offer to help someone. Shall I give you a hand? has stood a better chance at becoming institutionalised because of our shared knowledge that we tend to use our hands rather than our feet to manipulate things. Such shared associations enable us to use many words as readily understood metonyms and to understand many underspecified messages effortlessly. If motivation did not play a part in language, we would probably not be puzzled by All this good news is getting me down. We have come to associate happiness with being up, not down. This is probably grounded in the recurring observation that fit and happy people tend to be up and about, and that smiling makes the corners of your mouth turn upwards. Lakoff & Johnson's seminal work Metaphors we live by (Reference Lakoff and Johnson1980) showed convincingly that a great proportion of everyday language is figurative, that is, language use is full of conventionalised metaphoric and metonymic expressions. The main tenet of what was subsequently to become known as Conceptual Metaphor Theory (e.g. Lakoff Reference Lakoff1987) is that it is metaphoric thought (i.e. the creation of analogies between distinct domains of experience) that enables us to reason and communicate about intangible domains of experience in the first place.
It cannot be overemphasised that ‘motivated’ in CL jargon is not synonymous with ‘predictable’. The fact that one can give plausible retrospective accounts of why a given way of packaging a certain message has become standardised in a language does not mean that one can with any confidence predict what will get standardised. If there were no room for arbitrariness, then all natural languages would look and sound identical. If the CL endeavour to describe linguistic phenomena as motivated holds a certain promise for language pedagogy, it is precisely because in many cases the motivation is not blatantly obvious to the learner.
Highly polysemous words, such as prepositions, are classic CL showcases for the workings of motivation, where it is demonstrated that seemingly distinct uses of the same word are related to a prototypical usage and together form a radial network of uses extending outwards from that central prototype (e.g. Brugman Reference Brugman1981). Some of the meaning extensions concern literal uses. For instance, if a preposition is often used to describe a situation of contiguity (the bottle on the table; the cloth over the table; the bulge under the blanket), then it is likely to develop a rotated sense (the picture on the wall; the veil over her face; the skin under his shirt) (Boers Reference Boers1996). Many other extended uses of prepositions are motivated by metaphor. For instance, the figurative use of beyond in Why she ever got married to him is beyond me and She'd changed beyond recognition probably derives from the notion of inaccessibility associated with the literal use of this preposition, as in The ball fell beyond the hedge (and so it was beyond our reach). Another type of multipurpose word that has attracted a fair amount of interest from cognitive linguists is modal verbs (Sweetser Reference Sweetser1990). The epistemic uses of such verbs (The light is on, so she must be home now) are believed to be extended from the deontic uses (You must come home before midnight) via inferences (if you oblige someone to do something, for instance, then it is highly likely that it will get done).
A particularly fruitful branch of CL has been Conceptual Metaphor Theory (CMT) (e.g. Kövecses Reference Kövecses2010), which demonstrates how we project our knowledge of concrete, familiar domains of life onto abstract domains in an attempt to come to grips with the latter. Along with the structure that the ‘source domain’ lends to the ‘target domain’ come words and phrases associated with the source domain, but which subsequently adopt an additional, figurative sense. For example, one of the source domains we sometimes resort to in our thinking and communicating about learning is that of eating/drinking (a version of a more generic learning-as-intake metaphor). This is reflected in utterances like Anna has a great thirst for knowledge; These students are expected to swallow a lot of nonsense; Good teachers try to spice up their classes; I cannot digest so much information and Don't just regurgitate what the teacher said.
Let's now turn to the ways in which the above kinds of portrayal of polysemy and figurative multiword items have been adapted by pedagogy-oriented CL advocates for purposes of vocabulary instruction.
3. Examples of CL-inspired proposals for vocabulary instruction
A considerable number of pleas have been made for introducing the aforementioned CL ideas (especially CMT) to L2 vocabulary teaching (e.g. Danesi Reference Danesi and Alatis1992; MacLennan Reference MacLennan1994; Scott Reference Scott, Barbara and Scott1994) and some authors have gone on to develop resources intended to be in accordance with CL tenets for language learners (Rudzka-Ostyn Reference Rudzka-Ostyn2003) and for teachers (Lindstromberg & Boers Reference Lindstromberg and Boers2008a).
One common thread in CL-style instruction is the attempt to make non-basic uses of polysemes more memorable by (re-)establishing associations with the basic uses that they are derived from. This may simply involve informing or reminding learners of the literal counterpart of a conventionalised figurative use of a word, when that word is met in a text: the literal sense of rake when students encounter The banks were raking in a lot of profits, the literal sense of dodge in The PM dodged all the questions about the war, and so on. The literal counterparts can be clarified verbally, but their meaning can often be made more memorable through enactment (Lindstromberg & Boers Reference Lindstromberg and Boers2005) or the use of pictures and drawings (Boers et al. Reference Boers, Lindstromberg, Littlemore, Stengers, Eyckmans and Lindstromberg2008). Cognitive engagement can further be stimulated by asking learners themselves to make an educated guess at the meaning of a metaphorically used word, such as fledgling in A fledgling democracy, aided by information about its literal counterpart (fledgling in the sense of a young bird). These are very brief interventions – and by no means unique to CL-informed teaching. CL-style treatment of highly polysemous words, however, has typically involved more extensive teaching, which starts with the identification of a prototypical meaning, followed by a step-by-step explication of meaning extensions from that prototype until a semantic network comprising most of the uses of the word is established (e.g. Tyler & Evans Reference Tyler, Evans, Achard and Niemeier2004, for an account of the preposition over, and Lindstromberg Reference Lindstromberg2010, for a pedagogical treatment of English prepositions at large).
A second common thread in CL-style pedagogy is the attempt to make idioms and phrasal verbs easier to learn by showing how these instantiate common conceptual metaphors. Expressions such as Simmer down, He blew up at me, He's hot under the collar, She was fuming, He's blowing off steam and Don't add fuel to the fire can all be categorised as instantiating the metaphors the body is a container for the emotions and anger is heat. (It is customary in Conceptual Metaphor Theory to use small capitals to refer to conceptual metaphors.) Idioms can also be categorised according to their source domain, that is, the experiential domain in which they are/were used in their literal sense. For example, Clear the decks, On an even keel, Give someone a wide berth, Walk the plank, Take on board, Out of your depth and A leading light can all be traced back to seafaring. Additional stimuli for cognitive engagement can be given. Learners can be asked to propose reasons why a certain conceptual metaphor ‘makes sense’. For instance, it seems natural to talk about emotions such as anger in terms of heat, because of the physiological changes that coincide with passionate emotions. Learners can also be asked to make comparisons with their L1 to see if the metaphors are shared (Deignan, Gabrys & Solska Reference Deignan, Gabrys and Solska1997). They can be asked to propose reasons why a certain source domain is drawn upon strikingly often in the target language. For example, the abundance of seafaring expressions in English can easily be motivated with reference to Britain's history as a seafaring nation. Learners can be asked to group figurative expressions under the headings of conceptual metaphors or to identify their common source domains themselves. For instance, Set the stage for something, Be waiting in the wings, Take centre stage, In the limelight, Play to the gallery, Behind the scenes and The curtain comes down should suffice for learners to recognise the theatre as a source domain. Again, the meaning of the idiom can be made more memorable via pictorial elucidation of its literal reading (Boers et al. Reference Boers, Piquer, Stengers and Eyckmans2009; Szczepaniak & Lew Reference Szczepaniak and Lew2011) – provided it is congruent with the figurative meaning, of course. Cognitive engagement can further be stimulated by asking learners themselves to make an educated guess at the meaning of an idiomatic expression, such as Being on the ropes, aided by a hint about its source domain – boxing.
CL proponents’ expectation that the above kinds of teaching initiatives will benefit learners’ retention of the targeted words and phrases rests on three premises. Firstly, the attempt to stimulate cognitive engagement with target vocabulary ties in with the Levels of Processing model of memory (Cermak & Craik Reference Cermak and Craik1979), which holds that ‘deep’ processing of information (including vocabulary) is conducive to retention. Deep processing, in this model, is processing that involves semantic elaboration (Craik & Tulving Reference Craik and Tulving1975). Secondly, the type of semantic elaboration that is prompted by CL-style teaching typically involves mental imagery. The expectation that this will aid retention ties in with Dual Coding models of memory (Paivio Reference Paivio1986; Sadoski Reference Sadoski2005), according to which associating an abstract lexical item with a mental image renders the item more concrete and hence more memorable. Thirdly, if CL-style presentations of lexis succeed in convincing learners that the L2 lexicon is more systematic than they had hitherto realised, then this may instil confidence in them. This is because the task of learning a large number of items is likely to appear more manageable if the items are organised somehow, for example when different phrasal verbs are shown to reflect the same conceptual metaphor and when diverse idiomatic expressions are shown to share the same source domain.
4. Does it work?
Table 1 is an inventory of published quasi-experimental intervention studies in which the retention of words or phrases by groups of learners who received CL-style instruction was compared to that of groups who received a different type of instruction about the same items. The rightmost column of the table indicates whether the post-treatment test scores were significantly better for the group(s) to whom the lexical items were presented along CL lines: ‘yes’ means statistically significant at least at p 0.05, unless indicated otherwise. This was the case in the vast majority of the studies, and in none of the studies (apart from one of four trials in Condon's Reference Condon, Boers and Lindstromberg2008 study) did comparison groups outperform the CL groups.
An inventory of quasi-experimental intervention studies on the effect of CL-informed treatments on vocabulary retention.
Notes:
a. Li's (Reference Li2009) publication of the author's Ph.D. dissertation (2002) reports five quasi-experimental studies. Studies 2 and 3 are not included here because they compare variants of CL-informed instruction rather than comparing CL instruction to a non-CL treatment.
b. In Yasuda (Reference Yasuda2010) one of the aims was to assess the learners’ ability to autonomously apply the CL insights they had acquired during instruction about one series of phrasal verbs to work out the correct particle of phrasal verbs in another series. While the CL groups did not outperform the comparison groups on the ‘taught’ items, they did outperform them on the items assumed to be new to the students. Kövecses & Szabó (Reference Kövecses and Szabó1996) reported a similar finding. An evaluation of the possibility of learners using CL techniques independently, for example outside the language classroom, is made in Boers (Reference Boers2011).
There is a fair amount of diversity among and within the studies, of course. The number of participants ranged from just 24 (Boers, Demecheleer & Eyckmans Reference Boers, Demecheleer, Eyckmans, Bogaards and Laufer2004) to 127 (Li Reference Li2009, study 4). Some studies measured whether participants could remember the meaning of the items (e.g. Verspoor & Lowie Reference Verspoor and Lowie2003), others measured whether they could reproduce them (e.g. Boers Reference Boers2000a, b), and still others measured both (e.g. Boers Reference Boers2001).
I have included in the inventory only studies concerned with vocabulary. For CL approaches to other kinds of targets, I refer the reader to chapters in De Knop & De Rycker (Reference De Knop and De Rycker2008), Holme (Reference Holme2009), Littlemore (Reference Littlemore2009), De Knop, Boers & De Rycker (Reference De Knop, Boers and De Rycker2010) and Littlemore & Juchem-Grundmann (Reference Littlemore and Juchem-Grundmann2010). It is in the area of vocabulary that most of the empirical work on the effects of CL-style instruction has been done so far, though, and the inventory indicates that by and large that work has been favourable towards CL.
It is worth mentioning that this inventory represents but a segment of the existing research literature on CL-style vocabulary instruction. Excluded are quasi-experimental studies that do not investigate effects on retention, but rather focus on the benefits of CL-style pedagogy for vocabulary comprehension (e.g. Lindstromberg & Boers Reference Lindstromberg and Boers2005; Guo Reference Guo2007). Also excluded are studies in which variants of CL-inspired instruction are weighed against each other (e.g. Skoufaki Reference Skoufaki, Boers and Lindstromberg2008) and investigations of the influence of learner characteristics (such as cognitive style) on the effectiveness of CL instruction (Boers, Eyckmans & Stengers Reference Boers, Eyckmans, Stengers and Foster-Cohen2006). Nor are studies included that did not involve a comparison group (Tyler Reference Tyler, Robinson and Ellis2008), that is, a group that engaged with the target items for a similar amount of time as the CL-instructed group but was guided by a different kind of instruction.
I cannot guarantee that my inventory is complete, as some published studies may have escaped my attention. Besides, there are certainly several unpublished studies (such as Ph.D. and M.A. dissertations), which are not represented here.
Also, some caution is in order when interpreting the results of individual studies. Firstly, it could be argued that some of the studies (e.g. Boers Reference Boers2000a) lack sufficiently precise pre-test measures. Secondly, some of the studies (e.g. Boers 2000b, study 1) involved only an immediate post-treatment test: no delayed post-test to ascertain whether the benefits for retention were durable. Thirdly, in some of the studies the superior performance of the CL groups may be ascribed to more than the CL ingredient in their input. For example, the comparison group would in some studies (e.g. Boers 2000b, study 2) receive target vocabulary in the format of one longish list, whereas the same input would be segmented in the CL treatment to reflect metaphor categories. Better retention in the latter condition is thus not necessarily the product of metaphor awareness per se, but perhaps (also) a side effect of having cut up the information into more easily digestible bits (something I call ‘goulash versus nouvelle cuisine’ design). Another example is when the recall test followed a treatment where the CL group acquired a better understanding of the lexical items, which is likely to have given them an edge over the comparison group (e.g. Beréndi et al. Reference Beréndi, Csábi, Kövecses, Boers and Lindstromberg2008). Potentially confounding variables are also found in studies where the CL treatment engaged the participants not only differently but also in more ways than the comparison treatment (e.g. Li Reference Li2009, study 5) or where the CL group devoted more time to the lexical items that would be the object of the post-test (e.g. Tyler, Mueller & Ho Reference Tyler, Mueller and Ho2010).
It must also be conceded that in two of the studies the gains between pre-test and post-test obtained through the CL-style instruction were arguably too small to warrant pedagogical recommendations. While the CL groups outperformed the comparison groups to a statistically significant degree, this was probably due to the poor performance of the latter. For example, in Tyler et al.'s (Reference Tyler, Mueller and Ho2010) study (on modal verbs), the students who had received the CL-inspired instruction gained on average just 2.6 out of 32 test items between pre- and post-test. The comparison group made no progress. In Cho's (Reference Cho, Knop, Boers and De Rycker2010) study (on prepositions) the CL group gained on average only 1.5 on a 26-item test. The comparison group actually regressed. Cho (Reference Cho, Knop, Boers and De Rycker2010) is an exception among the studies reviewed here in that a statistical measure of effect size is reported (partial eta squared 0.17 in the delayed post-test). I have not been able to work out an average effect size across the collection – as has become customary in meta-analyses – because standard deviation data is absent from too many of the (older) publications.
In short, there are undeniably a number of problems with the individual studies listed in the above inventory. At the same time, it is also undeniable that, collectively, the studies begin to constitute a body of evidence that is hard to dismiss out of hand. As always, more research would definitely be welcome to confirm and fine-tune the findings. In addition, it would be worth finding out if the effectiveness of CL-informed interventions could be enhanced by aligning them better with insights from ‘mainstream’ approaches to L2 vocabulary. That is the question we will turn to next.
5. Talking to the ‘mainstream’
5.1 The desirability of distributed learning
The CL-style presentation of figuratively used words and phrases used in the intervention studies often involves groupings under the conceptual metaphors or source domains that they are derived from. Whether presenting vocabulary items in sets facilitates learning is a matter of some debate, however. Several studies have revealed that learning new words that are semantically related (e.g. words for professions, words for clothes, words for personality traits) takes longer than learning semantically unrelated words (Tinkham Reference Tinkham1997; Waring Reference Waring1997; Finkbeiner & Nicol Reference Finkbeiner and Nicol2003; Erten & Tekin Reference Erten and Tekin2008). Apparently, the semantic commonality of the co-presented words increases the risk of confusion. This risk is especially high when the words show formal resemblances as well. Teaching the words lemon and melon together in a set of words for fruit is not advisable. The risk of confusion is reduced if one of two confusable words has had the time to settle down in long-term memory first: if lemon is already well established in its allotted semantic space, then melon, as a new kid on the block, is less likely to compete for that space. In short, the obvious way of reducing the risk of erroneous cross-associations between related words is to learn one word well first and only learn the other word later.
The CMT-type groupings of figurative expressions seem different from the semantic sets that have been found unadvisable, however. Let's take the following set of idioms from the source domain of boxing: flex your muscles, lower your guard, take it on the chin, be on the ropes, and throw in the towel. Each occupies a slot in a scenario or ‘frame’. This is comparable to a set of words such as castle, dark, haunted and scream, which also conjure up a frame (a ghost story). This is the kind of organisation of vocabulary which is believed to be helpful for learning (Nation Reference Nation2000). Still, even in this kind of grouped presentation, it would probably be unwise to present a lot of (unfamiliar) items in one go. Items (e.g. stick your neck out, be in a tight corner and be down for the count) can be added when the time is ripe.
A radical adherence to distributed learning is not easy to marry with the exigencies of vocabulary building that we lamented in the introduction to this article, so we may have to settle for a compromise. But what are we to make of the way highly polysemous words, such as prepositions and multipurpose verbs, have been presented in the CL intervention studies? This typically involved lessons on the diverse uses of a couple of words and how these uses form a semantic network around a core sense. This may seem memory-friendly at first sight, but digesting information about a dozen different uses of a word (e.g. on the bus; on Sunday; on the phone) and their distinction from neighbouring words, each also with a dozen different uses (e.g. in bed; in the morning; in trouble; at work; at night; mad at him) is a case of massed learning, just the same. It is not surprising that it is precisely the studies which presented participants with this kind of learning challenge that produced the smallest effects.
A systematic approach to polysemy need not in pedagogical practice coincide with an intensive ‘let's now learn everything there is to know about this word’ procedure. As with idioms, teachers can draw students’ attention to a novel usage of a ‘known’ polyseme as it happens to come up in context and so trace the connection to its basic sense as the opportunity presents itself. It seems to me that this ‘teach as the need arises’ approach could make CL-style instruction more palatable.
It would also make the approach more faithful to the usage-based nature of language. It is remarkable that in the majority of the studies reviewed here, the target vocabulary was poorly contextualised, so the input was lacking in cues regarding common usage patterns: just the kind of cues that fuel L1 acquisition.
Unfortunately, there is still a dearth of research into the effects of distributed (‘as the opportunity presents itself’) CL-style interventions.
5.2 Issues of selection
Of course, researchers sometimes choose low-frequency words as target items in learning experiments in order to control for prior knowledge, and this also holds true for some of the studies reviewed here. Still, one may wonder about the utility of the type of lexical items that are targeted in some of them. A question I have often been asked is why we would want to spend precious class time on idioms. This scepticism is certainly warranted, as there are so many more frequent words and formulaic sequences that deserve prioritisation in a vocabulary programme. Still, there are some reasons why a blanket dismissal of idioms as targets for learning would not be in the best interest of at least some groups of learners. Firstly, idioms pose comprehension problems, even in contexts that are rich in clues (e.g. Boers, Eyckmans & Stengers Reference Boers, Eyckmans and Stengers2007). Littlemore et al. (Reference Littlemore, Chen, Koester and Barnden2011), for example, report that university lecturers’ use of figurative phrases is a major source of misunderstandings for international students. Cross-cultural differences throw up extra obstacles for L2 idiom comprehension (Hu & Fong Reference Hu, Fong, De Knop, Boers and De Rycker2010). Misunderstanding idioms can put a strain on communication especially because idioms fulfil crucial socio-pragmatic functions, such as conveying an evaluation and steering the course of conversation (O'Keeffe, McCarthy & Carter Reference O'Keeffe, McCarthy and Carter2007: 80−99).
A second reason why idioms cannot be discarded out of hand as items worthy of learning at least for some groups of learners is that, as a class, they are not as rare as has often been assumed. Boers & Lindstromberg (Reference Boers and Lindstromberg2009: 67), for example, found a higher frequency of figurative idioms (e.g. on the same wavelength; off the hook) in a popular crime novel than they found strong verb-noun collocations (e.g. make a decision; tell the truth) (ibid: 42). The irony is that I have seldom heard anyone question the usefulness of learning collocations, even though they are generally considered less likely to cause comprehension problems. Still, it is self-evident that efforts need to be made to align targets for learning with the needs of the learners, and CL pedagogy as described so far does appear most at home when this alignment is made at a relatively high level of proficiency, when learners are ready to turn their attention to the less frequent uses of words.
The quest for non-arbitrary phenomena in vocabulary goes on, however. One novelty is the exploration of motivation at the level of phonology rather than semantics. For example, there is robust evidence that phonological repetition (alliteration, rhyme, etc.) plays a part in the formation of word partnerships (Boers & Lindstromberg Reference Boers and Lindstromberg2009, Ch. 6). In a bank of 5,667 multiword units sampled from the Macmillan English Dictionary, no fewer than 737 (13%) display alliteration (e.g. time will tell; peer pressure; cut corners). If rhyme and assonance are added to the mix (e.g. brain drain; small talk), the proportion of phonologically repetitive multiword lexis reaches almost 20%. Some types of formulaic sequences, such as binomial phrases (e.g. part and parcel) and similes (e.g. busy as a bee) are particularly prone to these phonological repetitions (32% and 54%, respectively). It seems safe to say that in the development of word partnerships, sameness attracts. Importantly for the purpose of this article, experimental evidence suggests that alerting language learners to alliteration, rhyme or assonance in the word strings they meet positively influences their retention (Lindstromberg & Boers Reference Lindstromberg and Boers2008b, Reference Lindstromberg and Boersc). Another addition to the bag of tricks for engaging learners with vocabulary exploits the occasionally felt ‘fit’ between the form and the meaning of a word. For example, when asked whether the rare word harageous means kind or brutal, many a respondent will (correctly) guess it is the latter. The form or sound of words is not a reliable clue to guess the meaning of a word, of course, but asking students to judge whether the meaning of a newly learned word fits its form/sound does appear to help retention (Deconinck, Boers & Eyckmans Reference Deconinck, Boers and Eyckmans2010).
5.3 Different kinds of learning, different kinds of knowledge
In the introduction to this article, we mentioned the processing advantage that formulaic sequences afford native speakers: they facilitate fluency in processing and in speech. Native speakers enjoy this advantage thanks to familiarity with the conventionalised word strings they have been exposed to time and again. According to Wray (Reference Wray2002), though, this is not the only explanation for the fact that natives process formulaic sequences faster than non-natives. Put simply, in her model the native speaker's resource of formulaic language consists largely of sequences that were acquired (in childhood) holistically, so they are ready to be retrieved from the lexicon as prefabricated units. Adult L2 learners, she believes, are much more inclined to adopt an analytic, word-by-word mode of processing. It has been suggested (e.g. Lewis Reference Lewis1997) that L2 learners should try to bypass the word-by-word learning mode and learn formulaic sequences as integral chunks. It should be clear from the examples of CL interventions given throughout this article that, instead of steering learners towards a more holistic mode of processing multiword lexis, CL pedagogy actually intends to harness learners’ analytic abilities (e.g. Liu Reference Liu2010). CL encourages learners to wonder why certain words collocate, how the meaning of an idiom is connected to its original use, how different uses of a word are interrelated, and so on. Let's face it: these are mental pleasures that native speakers seldom indulge in.
There is no absolute reason, of course, why L2 learning should mimic L1 acquisition. Given the different matrices of L1 acquisition and adult L2 learning, it is to be expected that even highly advanced L2 learners’ performance will be fuelled by knowledge of a different nature from that of the native speaker. If the cognitive engagement brought about by CL-style instruction helps to compensate for the far fewer opportunities for incidental learning that the adult L2 learner enjoys in comparison with L1 acquisition, then, surely, this is to be cherished. As long as one appreciates that the (explicit) knowledge it generates is not the ideal fluency facilitator. (See Barcroft, Sommers & Sunderman, Reference Barcroft, Sommers, Sunderman, Trofimovich and McDonough2011, concerning another mnemonic technique – one outside the CL camp – the Keyword Method.) Let's put it this way: if successful, a CL approach may serve as a high-speed train to Word City, but its terminus is in a suburb. That's okay, as long as one realises it's still a bit of a stroll downtown, where the real action is.
That is why the language-focused learning strand needs to be complemented by other opportunities for learning, provided by message-focused input, message-focused output and fluency development activities (Nation Reference Nation2007). One of the great dilemmas for language instructors is, time and again, to decide whether to ‘teach’ a linguistic item/feature or to ‘leave it alone’. It is decision-making informed by the chances of incidental uptake of the item, and those chances will depend on a complex interplay between the profile of the learners (including likely interference from their L1), the samples of target language they are exposed to and in what mode (reading or listening), and the frequency therein of the linguistic item/feature concerned (e.g. Alejo, Piquer & Reveriego Reference Alejo, Piquer, Reveriego, De Knop, Boers and De Rycker2010, on phrasal verbs in the Spanish EFL programme). How much leeway is given to the language-focused learning strand inside the classroom will obviously depend on the extent to which the other strands are present in the students’ lives outside the classroom, but the general need to strike a balance inevitably calls for a judicious use of instructional techniques, including CL-inspired ones.
6. By way of conclusion
Is there a niche for CL in L2 vocabulary instruction? I believe there is. Non-arbitrary features of language can serve as stimuli for learners’ cognitive engagement with at least some L2 words and phrases, and this engagement aids vocabulary retention. In the face of the daunting challenge of vocabulary learning this is a welcome resource, even if the knowledge it generates is qualitatively different from that acquired by native speakers.
More empirical studies (especially longitudinal ones) would be welcome to fine-tune the evidence for the pedagogical merits of CL applications and to explore ways of enhancing their effectiveness. With regard to the latter, the incorporation of insights about distributed, in-context learning seems a way forward. Last but not least, the recent interest in motivation at the level of linguistic form (such as the role of alliteration as a match-maker) looks promising as a means of extending the scope of pedagogy-oriented CL.
Acknowledgements
I would like to thank the five anonymous reviewers for their useful feedback on an early draft of this article. I would also like to express my gratitude to the organisers of the 2011 Conference of the American Association for Applied Linguistics for inviting me to give the plenary address on which this article is based.
Frank Boers is an associate professor at the School of Linguistics and Applied Language Studies of Victoria University of Wellington. His initial interests were lexicology and metaphor, but most of his recent research concerns second language acquisition, especially the teaching of L2 vocabulary and phraseology. His work has appeared in journals such as Applied Linguistics, Language Teaching Research and System. He is co-author (with Seth Lindstromberg) of Optimizing a lexical approach to instructed second language acquisition (Palgrave Macmillan, 2009) and co-editor of two collective volumes in Mouton de Gruyter's Applications of Cognitive Linguistics Series.
