Calloway's code

Calloway's code
O. Henryt
As a light interlude before the more theoreticalpart of this book, readersmay enjoy this short story. It is about a hundred years old and, printed here in its original form, the vocabularyand style seemvery dated.Surprisingly,however, it touchesdirectly on the content of this book. There is a short commentaryat the end of the chapter,but initially readers are invited to enjoy the linguistic skills of two journalists, Calloway (the encoder) and Vesey (the de-coder). Betweenthem, they put fixed collocations highly creatiyeuse. to The New York Enterprlse sent H. B. Calloway as special correspondent the to Russo-Japanese-Portsmouth war. For two months Calloway hung about Yokohama and Tokyo, shaking dice with other conespondentsfor drinks of rickshaws - oh no, that's something you ride in; anyway he wasn't earning the salary that his paper was paying him. But that was not Calloway's fault. The little brown men who held the strings of Fate between their fingers were not ready for the readers of the Enterprise to seasontheir breakfast bacon and eggs with the battles of the descendants the gods. of But soon the column of correspondentsthat were to go out with the First Army tightened their field-glass belts and went down to theYalu with Kuroki. Calloway was one of these. Now, this is no history of the battle of the Yalu River. That has been told in detail by the corespondents who gazed at the shrapnel smoke rings from a distanceof three miles. But, for justice's sake,let it be understoodthat the Japanese commanderprohibited a nearer view. Calloway's feat was accomplishedbefore the battle. What he did was to furnish the Enterprlse with the biggest beat of the war. That paper published exclusively and in detail the news of the attack on the lines of the Russian generalZassulitch on the sameday that it was made. No other paper printed a word about it for two days afterwards, except a London paper, whose accountwas absolutelyincorect and untrue. Calloway did this in face of the fact that General Kuroki was making his moves and laying his plans with the profoundest secrecy as far as the world outsidehis campswas concerned.The correspondents were forbidden to send out any news whatever of his plans; and every messagethat was allowed on the wires was censoredwith rigid severity. The correspondentfor the London paper handed in a cablegram describing Kuroki's plans; but as it was wrong from beginning to end the censorgrinned and let it go through.

Calloway's code ltg


ts rt


e g

e e

n a

o d





So there they were - Kuroki on one side of the yalu with forty-two thousand infantry, five thousandcavalry, and a hundred and twenty-four guns. on the other side, Zassulitch waited for him with only twenty-three thousand men, and with a long stretch of river to guard. And calloway had got hold of some important inside information that he knew would bring the Enterprise stafr around a cablegramas thick as flies around a park Row lemonadestand. If he could only get that message past the censor- the new censorwho had arrived and taken his post that day. calloway did the obviously proper thing. He lit his pipe and sat down on a gun cariage to think it over. And there we must leave him; for the rest of the story belongs to vesey, a sixteen-dollar-a-weekreporter on the Enterprise. calloway's cablegramwas handed to the managing editor at four o'clock in the aftemoon. He read it three times; and then drew a pocket mirror from a pigeon-hole in his desk, and looked at his reflection carefully. Then he went over to the desk of Boyd, his assistant was usually called Boyd when (he he wanted him), and laid the cablegrambefore him. 'It's from Calloway,'he said. 'See what you make of it.' The messagewas dated at Wi-ju, and thesewere the words of it: 'Foregone preconcertedrash witching goes muffled rumour mine dark silent unfortunate richmond existing great hotly brute select mooted parlous beggarsye angel incontrovertible.' Boyd read it twice. 'It's either a cipher or a sunstroke,'said he. 'Ever hear of anything like a code in the office - a secretcode?' askedthe ME, who had held his desk for only two years. Managing editors come and go. 'None except the vernacular that the lady specials write in,' said Boyd. 'Couldn't be an acrostic,could it?' 'I thought of that,' said the ME, 'but the beginning letters containedonlv four vowels.It must be a code of somesor1., 'Try 'em in groups,'suggested Boyd. 'Let's see- "Rash witching goes,,-not with me it doesn't. "Muffled rumour mine" - must have an underground wire. "Dark silent unfortunate richmond" - no reason why he should knock that town so hard. "Existing greathotly" - no, that doesn'tpan out. I'll call Scott.' The city editor came in a hurry, and tried his luck. A city editor must know something about everything; so Scott knew a little about cipher-writing. 'It may be what is called an invertedalphabetcipher,' said he. .I'll try that. "R" seemsto be the oftenest used initial letter, with the exception of ..m,,. Assuming "r" to mean "e", the most frequently used letter, we transposethe letters- so.' Scott worked rapidly with his pencil for two minutes; and then showed the first word accordingto his reading- the word 'scejtzez,. ,Great!'criedBoyd. 'It's a charade. My first is a Russiangeneral.Go on, Scott.'

I i I i


Calloway's code

that won't work,' said the city editor. 'It's undoubtedly a code. It's impossibleto readit without the key. Has the office ever useda cipher code?' 'Just what I was asking,' said the ME. 'Hustle everybodyup that ought to know. We must get at it some way. Calloway has evidently got hold of somethingbig, and the censorhas put the screwson, or he wouldn't have cabledin a lot of chop sueylike this.' Throughout the office of the Enterprise a drag-netwas sent, hauling in such membersof staff as would be likely to know of a code, past or present,by reason of their wisdom, information, natural intelligence, or length of servitude. They got together in a group in the city room, with the ME in the centre. No one had heard of a code. All began to explain to the head investigatorthat newspapers never use a code, anyhow - that is a cipher code. of coursethe AssociatedPressstuff is a sort of code - an abbreviation.rather -but.... The ME knew all that, and said so. He asked each man how long he had worked on the paper. Not one of them had drawn pay from an Enterprise envelopefor longer than six years. Calloway had beenon the papertwelve years. 'Try old Heffelbauer,'said ME. 'He was herewhen Park Row was a potato the patch.' Heffelbauer was an institution. He was half janitor, half handy-man about the office, and half watchman - thus becoming the peer of thirteen and one-half tailors. Sent for, he came, radiating his nationality. 'Heffelbauer,' said the ME, 'did you ever hear a code belonging to the office a long time ago - a privatecode?You know what a code is, don't you?' 'Yah,' said Heffelbauer, 'Sure I know vat a code is. Yah, apout dwelf or fifteen year ago der office had a code. Der rebortersin der city-room haf it here.' 'Ah!'said the ME,'We're getting on the trail now. Where was it kept, Heffelbauer? What do you know aboutit?' 'Somedimes,' said the retainer, 'dey keep it in der little room behind der library room.' 'Can you find it?'asked the ME eagerly.'Do you know where it is?' 'Mein Gott!' said Heffelbauer.'How long do you dink a code live? Der reborterscall him a maskeet.But von day he butt mit his head der editor, und - 'Oh, he's talking about a goat,' said Boyd. 'Get out, Heffelbauer.' Again discomfited, the concertedwit and resourcesof the Enterprise huddled aroundCalloway'spuzzle,consideringits mysteriouswords in vain. Then Veseycame in. Vesey was the youngest reporter. He had a thirty-two-inch chest and wore a number fourteen collar; but his bright Scotch plaid suit gave him a presence and conferred no obscurity upon his whereabouts.He wore his hat in such a position that people followed him about to seehim take it off, convinced that



cro rtu tol

am( thal

Ves 'co<

tou( Ves prot like 'It's 'The 'The

som Gee end

Vese cablr 'Let' 'I be Heu his c wisdr anott theor It to code 'I fe He's new Thu Fore Prec Ras Witr Goe Mul Rur

Calloway's code l2l

it must be hung upon a peg driven into the back of his head. He was never without an immense,knotted, hard-wood cane with a German-silvertip on its crooked handle.vesey was the bestphotographhustler in the office. Scott said it was becauseno living human being could resist the personaltriumph it was to hand his picture over to vesey. vesey always wrote his own news stories, exceptthe big ones,which were sent to the re-write men. Add to this fact that among all the inhabitants, temples, and groves of this earth nothins existed that could abashVesey,and his dim sketch is concluded. vesey butted into the circle of cipher readers very much as Heffelbauer,s 'code'would havedone,and askedwhat was up. Someone explained, with the touch of half-familiar condescensionthat they always used towards him. vesey reached out and took the cablegram from the ME's hand. under the protection of some special Providence,he was always doing appalting things like that, and coming off unscathed. 'It's . a code,'saidVesey.'Anybody got the key?' 'The office hasno code,'saidBoyd, reachingfor the message. veseyheld to it. 'Then old calloway expectsus to read it anyhow,' said he. ,He's up a tree, or somethingand he's made this up so as to get it by the censor.It,s up to us. Gee! I wish they had sent me, too. say - we can't afford to fall down on our end of it. "Foregone, preconcerted, rash,witching" _ h,m., vesey sat down on a table comer and beganto whistle softly, frowning at the cablegram. 'Let's haveit, please,'saidthe ME. ,We,vegot to get to work on it.' 'I believeI've got a line on it,' saidVesey.,Give me ten minutes., He walked up to his desk, threw his hat into a wastebasket, spreadout flat on his chest like a gorgeous lizard, and started his pencil going. The wit and wisdom of the Enterprise remained in a loose group, and smiled at one another,nodding their headstoward vesey. Then they beganto exchangetheir theories about the cipher. It took vesey exactly fifteen minutes. He brought to the ME a pad with the code-key written on it. 'I felt the swing of it as soonasI sawit,'saidvesey. .Hurrahfor old calloway! He's done the Japs and every paper in town that prints literature instead of news. Take a look atthat.' Thus had Vesey set forth the reading of the code: Foregone- conclusion Preconcerled- arrangement Rash - act Witching - hour of midnight Goes- without saying Muffled - report Rumour - hath it


Calloway's code

Mine - host Dark - horse Silent - majority Unfortunate - pedestrians Richmond - in the field Existing - conditions Great - White Way Hotly - contested Brute - force Select- few Mooted - question Parlous- times Beggars- description Ye - correspondents Angel - unawares Incontrovertible - fact 'It's simply newspaper English,'explainedVesey.'I've beenreporting on the Enterprise long enough to know it by heart. Old Calloway gives us the cue word, and we use the word that naturally follows it just as we use 'em in the paper. Read it over, and you'll seehow pat they drop into their places.Now, here'sthe message intendedus to get.' he Veseyhanded out another sheetof paper. Concludedaffangement act at hour of midnight without saying. to Report hath it that a large body of cavalry and an overwhelming force of infantry will be thrown into the field. Conditions white. Way contested by only a small force. Question the Times description. Its correspondentis unawareof the facts. 'Great stuff!'cried Boyd excitedly. 'Kuroki crossesthe Yalu tonight and attacks.Oh, we won't do a thing to the sheetsthat make up with Addison's essays, real estate transfers, and bowling scores!' FOOTNOTE Mr Veseyafterwardsexplainedthat the logical journalistic complementof the word 'unfortunate' was once the word 'victim'. But, since the automobile became so popular, the correct following word is now 'pedestrians'.Of course,in Calloway'scode it meant 'infantry'. 'Mr Vesey,'said the ME, with his jollying-which-you-should-regard-a-favour manner, 'you have cast a seriousreflection upon the literary standardsof the paper that employs you. You have also assistedmaterially in giving us the biggest "beat" of the year. I will let you know in a day or two whether you are to be discharged retainedat alarger salary.SomebodysendAmes to me.' or Ames was the king-pin, the snowy-petalledmarguerite,the star-bright looloo of the re-write men. He saw attempted murder in the pains of green-apple colic, cyclones in the summer zephyr,lost children in every top-spinning

urchir potatc his Br

Ames there fingerr the Ya brief n of the speech descrit which ffoops whatA founda he glee false an Army p

Only or ju. Call shouldl Ames ' descript by the effective guns thr the 'con the Enrc It was v censorbi dearlh o wonderfr with one part.

On the se repofier \ a coal-ho 'The old 'All right say - "S whole it c

Calloway's code 123

urchin, an uprising of the down-trodden massesin every hurling of a derelict potato at a passingautomobile.when not rewriting, Ames sat on the porch of his Brooklyn villa playing checkerswith his ten-year-old son. Ames and the 'war editor' shut themselvesin a room. There was a map in there stuck full of little pins that representedarmies and divisions. Their fingers had been itching for days to move thosepins along the crooked line of the Yalu. They did so now; and in words of fire Ames translatedcalloway,s brief message into a front-page masterpiecethat set the world talking. He told of the secret councils of the Japanese officers; gave Kuroki's flaming speechesin full; counted the cavalry and infantry to a man and a horse; described the quick and silent building of a bridge at Suikauchen, across which the Mikado's legions were hurled upon the surprisedZussulitch, whose troops were widely scatteredalong the river. And the battle! - well, you know what Ames can do with a battle if you give him just one smell of smoke for a foundation. And in the same story, with seemingly supematuralknowledge, he gleefully scoredthe most profound and ponderouspaper in England for the false and misleading accountof the intendedmovementsof the Japanese First Army printed in its issue of the samedate. only one effor was made; and that was the fault of the cable operator in wiju. calloway pointed it out after he came back. The word 'great' in his code shouldhavebeen 'gauge', and its complemental word 'battle'. But it went to 'conditions Ames white', and of course he took that to mean snow. His description of the Japanese army struggling through the snow-storm,blinded by the whirling flakes, was thrillingly vivid. The arlists tumed out some effective illustrations that made a hit as pictures of the artillery dragging their guns through the drifts. But, as the attack was made on the first day of May, the 'conditions white', excited some amusement. But it made no difference to the Enterpris e, anyway. I1 was wonderful. And calloway was wonderful in having made the-new censorbelieve that his jargon of words meant no more than a complaint at the dearth of news and a petition for more expense money. And vesey was wonderful. And most wonderful of all are words, and how they make friends with one another,being oft associated, until not even obituary notices them do pafi. on the secondday following, the city editor halted at vesey's desk where the reporter was writing the story of a man who had broken his leg by falling into a coal-hole - Ames having failed to find a murder motive in it. 'The old man says your salary is to be raised to twenty a week,' said Scott. 'Al1 right,' said vesey. 'Every little helps. Say - Mr Scott, which would you say - "we can state without fear of successful contradiction". or ,.on the whole it can be safely asserted"?'


Calloway's code

I am indebted to Jon Wright for drawing my attention to this splendid tale. Despite the laboured prose, some interesting commentson languageemerge: 1. Journalism is, and always was, riddled with its own particular lexis and collocations. Someof thoseusedin the story havestoodthe testof time, while otherssimply do not 'soundright'for us, a centurylater.So, maybethe fixed elementsof languageare not as fixed as we might like to think. 2.The managing editor is quite clear thatVesey has 'cast a seriousreflection on the literary standardsof the paper that employs you'. Like other educated people he had been taught that clich6 - a word drawn directly from the world of newspapers was bad 'literary style'. There is still an in-built prejudice against what is, in certain contexts, called clich6; but if non-native speakers write essays non-standardlanguage,they are told: that's not the way we say in lr. Apparently, while some of us need to learn to avoid clich6, others need to acquire alarger phrasal lexicon, in other words, more clich6s. 3. The brilliantly creative're-writer', Ames, needs'just one smell of smoke for a foundation'before he can write a graphic description of something he has not seen.What fires his account of the Japanese attack is a few central ideas - no, not ideas, but rather a few essential collocations provided by Calloway,once his message has been decoded. There is a message here for any teacherhelping studentsprepareto write essays;it is not enough to have some ideas, perhaps in your own language, what is essential is a few collocations central to the main themes.This reminds us of suggestions made by Deborah Petty and Graham Smith in the previous chapter.

l l

All chz hor the, da] enc rntr

Mat thar thec oft acql exte 10ti und< desc way

whil deve resea their of be

Part 2 - Background theory 125

Part 2
Background theory
All the contributorsto the first part of this book - both the authorsof complete chaptersand the many teachersquoted in Chapter 5 - repeatedly emphasise how their own understanding of collocation has developed step-by-stepas they have made small, and then increasingly radical changesin their day-today classroom practice. Again, a deeper understanding of collocation has encouraged them to extend and refine the modest changes they first introduced. Many other teacherswho have already begun to emphasisevocabulary rather than grammar would like to take their understanding further. Background theory is essentialfor such teachers. The first two chaptersofthe secondpart of this book provide a summary of recent research on language and acquisition - Chapter 7 on language, Chapter 8 on acquisition. Chapter 9 extendsthe comments on texts, corpora and concordances Part 1. Chapter in 10 takesup the questionof how examsmay changein the light of our changed understandingof the mental lexicon. Finally, in Chapter 11, the distinguished descriptive linguist Michael Hoey provides teacherswith a glimpse into the way research'may soontake us 'beyond collocation'. While PalrI 2 is more theoretical, its primary purpose is to help teachers develop their own understanding so they can initiate their own action research. provides frameworks within which they can evaluatethe results of It their observations,so that they can be sure the changesthey make really are of benefit to their learners.

i lll E

ilii s


Language in the lexical approttch

Language in the lexical approach
Michael Lewis
This chapter looks at the way descriptions of English have improved as a result of analysis of large amounts of natural spoken and written text on computers. It explains the breakdown of the old distinction between vocabulary and grammar, and emphasises how much of the language we use consists of multi-word phrases. It clarifies the terminology which is used in the other chapters of the book to discuss different kinds of rnulti-word phrasesl in particular, it explores the many different kinds of collocations. It will help readers new to the idea to develop their understanding of the different kinds of chunks of which lexis is composed. Finally, it rerninds us that the improved descriptions of English now available have radical implications for the language classroom.


ide Lo fig 30

lirr En

I 1 cor inc ten eml Do

7.1 Descriptions of English
In recent years, since the widespread availability of large computer-based corpora - collections of natural written and spoken text - which have been statisticallyanalysed,we have better descriptionsof English availableto us than ever before. Although intuition has an important role to play, many statementsabout how words are used, and the pattems which are typical of by The repofts sometypes of text can now be supported empirical evidence. of corpus linguists involved in this work sometimesconfirm our intuitions, but frequently provide overwhelming evidence contradicting some belief which is widespread.Their work is essentiallydescriptive,but it is selfevident that if their descriptions show that English is not used in the way traditional teaching has claimed, there are considerable implications for practice. classroom

wh ifv

say con mal, exa whe Iare

Coll exp( com like

than than

7.2 lnttition

and evidence

Every teacherhas said I've never heard that, which is usually interpretedto mean something very close to Nobody says that. But although every competent speaker Englishhasmet many millions of words of the language of while reading, watching TV, conversingand so on, the sampleof the language you personallyhave met is an absolutelyminute fraction of the ever-changing 'English'. Supposethere are 30 million native-speaker adults entity we call using English in Britain every day; and that they each speakfor a total of only 60 minutes at a normal speedof about 120 words a minute - that is 220billion words spoken every day in Britain by its adult population. Now pause to considerthe number of words of English producedworldwide every year.The

Alth help; by ar ofE diffe



3. Th fre thi

Language in the lexical approach 127

numbers are almost unimaginable, yet each of us believes we have a good idea of what is and is not possible in the language.The editors of the new Longman Grammar of spoken and written Engrish (LGSWE) give some figures which provide a perspective:a typical page of printed text is around 3001400 words, so a 300-page book is around100,000words; a million words is 10 books. In their recordings of conversation,speakers typically speakat a little under 120 words per minute, so a million words of informal spoken English correspondsto about 140 hours of conversationalinteraction. you could easily take parl in 15 million words of conversationa year. At my age, I have probably heard or spoken welr over half a billion words of conversationalEnglish. Despite thesehuge numbers,like everyoneelse, I am inclined to think that I rememberwhat I have and have not met before. we all tend to have confidencein our intuitions about language,but unfortunately the empirical evidencesometimesshows that our intuitions are seriously flawed. Do you recogniseToo many cooks spoil the broth as a'common'proverb? when do you think you last used it yourself? when did you last hear it? And if you did hear it, are you sure the person said an the words, or did they only sayTbomany cooks. . . 7 Do you recognise makea mistakeas a .common' to combination of words? If you answered res this time, you are yourself making a mistake. To make a mistakeis a dictionary-style generalisationfrom examplesyou may have heard such as I think we probably made a mistake when we. . . . The exactwords to makea mistakeare possible,but relatively rare.We needto ask what we mean by a 'common,phraseor word. collocation is about words which occur together more often than might be expected if words were produced randomly; collocations are ,common, combinations of words, so it is usefui to ask what we mean by expressions like 'they frequently occur together'. The number of words used every day is immense; the number of words and phrasesyou know is probably far greater than you would guess,so things you think are common are,in fact. much rarer than you think. Although this book is about collocation - the way words co-occur - it is helpful to think about how individual words occur.First, check your intuitions by answering these questionsabout a large corpus - say 10 million words of English including spoken and written language and representing many different geffes. 1. which is more common, a or the? (you can think this out logically.) 2. WhaI are the 10 most common words used in English? 3. Think of any one of the most common 250 words in English. How frequent do you think it is? If it occurs once in every x words, do you think x is closestto: a. 100 b. 1000 c. 5000 d. 50,000 e. 100,000


Language in the lexical approach

4. What percentage the words in the l0-million-word corpusdo you think of would occur only once in the corpus? a.50Vo b.25Vo c. l0% d" 57a e. less than 57o 5. We say that word A collocateswith word B if the two words co-occur 'frequently'. What do you think 'frequently' means here? On what percentage occasionsof occurrenceof word A, do you think word B of co-occurswith word A? Is it: a.90Vo b.50Va c.25Vo d..10Vo e.57o How confidentare you of your answers? Here are the answers: - the - occurs about twice as often as the indefinite 1. The definite article articles* a, an - combined. 2. The top ten are the, of, and, to, a, in, that, I, it, was according to Cobuild's published list. Interestingly, the proportions fall very fast. Here are the relative frequencies related to every 100 occurrencesof the most frequentword the: of a I 50 4 2 2 t i i n r and 50 3 2 2 1 to 44 that 22 was 18



use lim whr Chz real

nhrr r_"-

The 100th most common word, however,has a relative frequencyof less than 2.In a million-word corpusit would occur about 800 times. 3. Words which we think are common, such as set, given as an example by John Sinclair in Corpus, Concordance,Collocation, occur onlv about once every 4000 words. 4. About half the words in a corpus of 10 million words will only occur once.Words which we think of as rare or unusual,are very rarely used! 5. By now, you may have guessedthat the answer is less than 57o. Even very common words have relatively low frequencies; collocations involve two relatively rare things happening together, so they are even less frequent.Our intuitions are very unreliable.To quote Sinclair again: The languagelooks rather dffirent when you look at a lot of it at once. Words which we think of as common may only occur once in a million words; common collocations much rarer eventhan that, and commonexpressions are such as proverbs are rarer still. Chitra Fernando reports that when she consultedthe Birmingham Collection of English Texts in 1990,it contained 13 examplesof the phrasered herring in the then total of 20 million words in other words, this familiar expressionoccurred only about once in one-andhalf-rnillion words. LGSWE suggests that words which occur once in every thousandoccur about once every 8.5 minutesin speech, while a word which is only used once in every 100,000words will be heard once every fourteen hours(of non-stopconversation!). maturelanguage-user's A mentallexicon is much larger than we previousiy thought,and the non-nativeleamer'stask in mastering a sufficiently large lexicon correspondinglymore difficult.

largr proc Ling idiot this l one l these probl think Introt unhel well-r cours

The p of the rtems phrase into dl lot of catego goalke a Gree of thes descril a glver describ linguisr for dif express teachel underst

Language in the lexical approach



)cur that d B

limited duration. To do this, teachersneed to develop their understanding of which language really is useful. corpora and concordances [discussed in chapter 9f are a great help in testing your intuitions against the evidence of real languageuse.

nlte !to iast. the

7.3 Terminology
Although it has long beenrecognisedthat the mental lexicon had an important phrasalelement,it is now generally acceptedthat the phrasal elementis much larger than was previously recognised.perhapsinevitably, this realisation has produced a proliferation of terms, with the resulting potential for confusion. Linguistic terms with which the reader will certainly be familiar arephrase, idiom, fixed expression, phrasal verb, adverb, proverb, and, in the context of this book, collocation. unfortunately, all theseterms are used with more than one meaning by people writing about vocabulary and the mental lexicon. In






The problem of terminology arises as soon as we recognise that a large part of the immediatelyaccessible lexicon which we storeconsistsof multi_word

)ven taln: e. rrds; 10ns she ined dsandvery hich teen )n ls ;k in

of theseis 'really' me, or how I am 'best' classified.All of theseare coffect descriptionsof me, in different circumstances;which is more appropriateon a given occasion depends on your purpose in talking about me at all. we describe and categorisewith a particular purpose in mind; this is as true for linguistic description as it is for any other. The samepiece of languagemay, for different purposes, usefully be described as an utterance, a fixed expression,an idiom, a responseetc. From the classroompoint of view, teachers need to choose a limited range of terms, that learners "nrrr" understand them, and then use thoseterms consistently. But it is essential to


l-tnqunqe in the lexical approach

remember that leamers are learning a language, not information about language; describing the language is not itself language teaching. The sole purpose of description in the classroomis to ensure that leamers can notice featuresof the languagethey meet in ways which facilitate acquisition. They are languagelearners,not amateurapplied linguists. Some of the larger phrasal units are relatively uncontentious and familiar: quotations (A rose by any other name), proverbs (Too many cooks spoil the broth), and - providing we avoid the argument about prepositions and particles - phrasal verbs (The high cost put me off the idea). Some familiar terms do, however,need discussion,as different writers use them in different ways, so readers need to be constantly on guard against possible misunderstanding.

mosl f,rxed smal can s in its perhz pafi descr wnte



7.4 From idioms to idiomaticity
Most teachersprobably think they have a clear idea of what an idiom is (Uncle George hasfinally kicked the bucket), even if they are in some doubt about how the term 'collocation' is used. Many may be surprised thar, at a 'collocation'ate often seenas similar,even more theoretical level, 'idiom'and overlapping,terms. In order for us to get a clear idea of collocation, we need first to consider it as part, not of idioms, but of the wider concept of idiomaticity. Whether somethingis regardedas idiomatic or not is not decided by a single factor; most linguistsrecogniseat leasttwo factors- its position 'invariant' and 'variable' as its end points, and a on a spectrumwith 'fixed' or secondspectrumrangedbetweensemantic'opacity' and'transparency'. The compilers of The Oxford Dictionary of Current ldiomatic English use the following categoriesto describe a cline of idiomaticity from most to least fixed: Pure idioms such as blow the gaff are at the most fixed end; they are almost invariant and have lost any literal interpretation,so are semantically opaqueyou cannot guess the meaning of the whole from a knowledge of the meaningsof the individual words. Figurative idioms such as catch fire, a close shave are less opaque: in addition to their non-literal meaning, they are also still used in their literal sense;they are fairly, but not quite, fixed. Restricted collocations such as jog someone'smemot! have one element used in a non-literal senseand the other used in its normal meaning. Open collocations involve elements which are (more or less) freely combinable, with each element having its literal sense. These examplesmake clear that two factors are involved in what we loosely think of as idioms - a cerlain degreeof fixednessand a certain degreeof nonliteralness,leading to more or less difficulty in understandingthe meaning of the whole expressionfrom an understandingof its component words. What



...J gra gra the

Wirhix traditic

Tha Wel Ise Thkt plal sigrt goin ave take ahe heat

As exar show,ct observe Co wo

Chitra F academ lower en

Language in the lexical approach 131

tut Dle lce


the md liar :ent Lble

most teachersand studentsthink of as idioms are those which are both fairly fixed and non-literal. with a narrow definition like that, idioms are a fairly small part of the total lexicon, and from a languageteachingperspectivethey can safely be left to more advancedleamers.once we understandidiomaticity in its wider meaning - chunks which have some degree of fixedness, and perhaps some degree of non-literalness,it is clear that idioms are a central part of the lexicon and important for leamers at all levels. John sinclair, describing the early work by the team constructing the Cobuild dictionaries, wrrtes: The principle of idiom is that a languageuser has available to him or her a large number of semi-preconstructedphrases that constitute single choices, even though they might appear to be analysableinto segments....The overwhelming natureof [the corpus] evidenceleadsus to elevate the principle of idiom from being a rather minor feature, compared with grammar, to being at least as important as grammar in the explanation of how meaning arisesin text. ...Just as it is misleading and unrevealing to subject of course to grammatical analysis, it is unhelpful to attempt to analyse grammatically any portion of text which appearsto be constructedon the idiom principle. within this wider definition, all of the following, while not idioms in the traditional sense,exhibit some degreeof idiomaticity: That's neither here nor there. Well,I mustn't keepyou. I seewhat you mean. Thkeil or leave it. playing for time signed,sealedand delivered going bachuard and forwards a very cool reception take the earliest possible opportunity to . . . a heavy-handedapproach to the problem heavy rain As examples such as a cool reception, heavy rain, take the opportunity to show, collocation is part of the overall specftum of idiomaticity. As Sinclair observes: collocation illustrates the idiom principle. on some occasions words appear to be chosen in pairs or groups and these are not necessarilyadjacent. chitra Femando, whose book ldioms and ldiomaticity is perhaps the best academic survey of this area of language, states: "collocations are at the lower end of the idiomaticity scalebeing only weak realisationsof the idiom


ata jven Leed :of ded tlon rda , the east nost uethe ); m teral

reely rsely nonng of What


L . . ; , : t t i c s e i r t t l t e I e . r i c a la p p r o a c h

prrinciple." Very strongcollocations, where you can hardly imagine any other use of one of the parlner-words, are a kind of idiom: We had a blazing rov,/argumenf. Some collocations permit very limited choice: The whole (and not much else).But the vast storlevent was tinged with sadness/regret part of the spectrumof idiomaticity, are not majority of collocations, although so restricted.Partner-wordsoften combine freely with many other words, and the slots in a collocation can each be filled in many different ways. The same group of words may, therefore,be treated as both an idiom and a collocation but the focus of the two descriptionsis rather different. Idioms focus mainly on the meaning of the whole, while collocation is concerned with combinations of words which do or do not occur. 'wotds This distinction is used in this book, but our main focus is firmly on and the company they keep', although it is helpful to remember that this is part of the wider question of idiomaticity. We look now at how the term 'collocation' covers many different kinds of multi-word items.

Lear Thel teach to be synol but t intror to no indici

7.5 Collocation
Collocation is the way in which words co-occur in natural text in statistically significant ways. It sounds an innocent definition, but one very imporlant point needsto be made:collocationis aboutthe way wordsnaturallyco-occur 'used language'. Collocations are not in what David Brazll brilliantly called 'put together',they co-occurnaturally,and words which we, in some sense, the first task of the language teacher is to ensure that they are not unnecessarilytaken apafi in the classroom.If words occur togethet, leanters need to notice that co-occuffence and, if they are to be recorded in a vocabularybook, the words should be recordedtogether,a point already made by severalcontributors. In most classeslearnerswill already know many individual words, so in these 'putting them together' in circumstances, they may need to learn about standardcollocations,but this is part of the necessaryartificiality of language teaching.It would unquestionablybe better if learnershad acquiredthe words together as a single chunk - a single choice - in the first place. If you learn initial reaction (one item) it is easyto split the chunk apart, and acquire initial you must andreaction, two more items. If you learn the two words separately, also learn a third item, the correct collocation. Separatingcollocations into their component words is easy; it is considerablymore difficult to put words together to form natural collocations. Peter Howarth has pointed out that knowing which words do go with which, and which do not is a major problem for learners: It may be claimed that the problem facing the non-native writer or speakeris knowing which of a range of collocational options are restrictedand which are free. ...[the] significance[of the data] lies in the way in which specific collocations might be predicted by

When broke an ml becau The la of re-tr break even il terms Differi

If we d the del Cerlair recogn

7.ad 2. sttb 3. rad 4. exa 5. extr 6. reyi 7. the, 8.Top 9. a fet I0. tunt 17.awa 12.fire t 73.back 14.hooA 15. On tt 16.A sot

Language in the lexical approach 133

ner tn8 ole ,ast not md
Lme .10n

analogy, but are arbitrarily brocked by usage,and clearly they are the kind of phenomenonlikely to confound leamers. ...It is the gaps in collocability that are arbitrary.

LnIy vith rrds Lsis

introducingthe term 'blocked collocation'to leamers,and encouraging them to note such 'impossible'combinationsby asking them to record and then indicatetheir non-acceptability crossingthrough or ,cancelling,them. by

:a1ly rtant ccur : nOt and not

The larger the chunks are which learnersoriginally acquire,the easierthe task of re-producing natural languagelater. The message teachersis clear: don,t to break language down too far in the false hope of simplifying; your efforts, even if successfulin the short rerm, ate almost certainly counterproductivein lermsoI long-term acquisition. Different kinds of collocation If we define collocation as the way words occur together,it is easyto seethat the definition is very wide, and will cover many different kinds of item. certainly, all of the following are colrocations in the sensethat we readily recognisethat thesegroups of words are regularly found together: l. a dfficult decision(adjective+ noun) 2. submit a report (verb + noun) 3. radio station (noun + noun) 4. examinethoroughly (verb + adverb) 5. extremelyinconvenienr(adverb + adjective) 6. revise the original plan (verb + adjective + noun) 7. thefog closed in (noun + verb) 8. Toput it anotherlr;ay (discourse marker) 9. a few years ago (multi-word prepositional phrase) I0. turn in (phrasalverb) ll. aware of (adjeclive + preposition) 72.fire escape(compoundnoun) 73. bachnards and forwards (binomial) 14. hook, line and sinker (trinomial) 15. On the other hand (fixedphrase) 76.A sort of . . . (incomplete fixed phrase)

rna nade these :r' in luage vords learn 'nitial must s mto words .t that major r



Language in the lexical approach

77. Not half! (fixed expression) 18. Seeyou later/tomorrow/on Monday. (semi-fixed expression) 19. Too many cooks. . . (part of a proverb) 20. To be or not to be .. . (part of a quotation) From the languageteachingpoint of view, many of theseare familiar and have formed a regular part of classroomteachingmaterials.The contributorsto this book focus almost exclusively on those kinds of collocations which are relatively new in languageteaching and which are only now finding their way into materials. There is extensivediscussionof types I to 7 in the above list, with some referencesto types 8 and 9, but relatively little mention of the older, familiar types of multi-word item. George Woolard earlier suggested 'collocation' reasons for restricting the use of the term for leamers to the newer kinds. [Seep 29.] Lexical and grammatical collocations Some writers distinguish between lexical collocations such as suggeston alternative, an evasiveanswer, and grammatical collocations such as aware of, step into. In this terminology, lexical collocations combine two equal lexical components (open class words), while grammatical collocations combine a lexical word, typically a noun, vetb or adjective, with a grammatical word (one open class word and one closed class word). Within this framework, phrasal verbs are neither more nor less than grammatical collocations. The main focus in this book is on lexical collocations, though it is worth noting that learnerswould often be well advisedto record more than It simple two-word combinations. is better to record phrasessuch asput the meeting off until. . , so they include both lexical words and grammatical words which are often used together. Similarly, recording grammatical collocations such as aware of, interestedin is unsatisfactory as these combinations are never used without at least one more word, so it makesmore (collocational) senseto teachcombinationssuch as awore of the problems, interestedinfootball, choosing typical examplesof how the words are used in a slightly larger context. Throughout this book teachers are repeatedly urged to encourage students to record language in larger chunks, and to keep at least part of the context in which the word actually occurred as part of what is recorded. A comment of Svetlana TerMinasova's (Language,Linguistics and Lift) is typical: Foreign leamers must keep in mind that they should learn words not through translationsof their meanings(that is, referenceto bits of reality and concepts),but in their most natural, habitual contexts, typical of the target language. Collocations are often idiomatic Some collocations appear superficially 'logical' - open the window, play

tent thin mea earl mak collr metz othe

Did 1 is eas could strong

Becau when Howat have c combr: proble Sheha meanlr Swedis structu child; t closert baby is so that i learner.

Ter-Mir of Engll gate as the doa commer

Language in the lexical approach 135

lre ,aY rst, the ted the

tennis, breakyour leg -butmany, although very familiar and which we easily think of as 'obvious' or 'sounding right', are conventional. Notice how the meanings of the verbs in the following differ considerably from the three earlier examples:open a meeting (why not start), play some music (why not make), break the silence (why not interrupt or explode).In fact, very few collocations are truly self-evident or literal; there is a partially non-literal, metaphoricalor idiomatic element to most collocations.This meansthat, like other idioms, they are not fully predictable from their componentwords.

Think of three nouns which can follow the verb answer which intermediate leamers are unlikely to know and which they probably would not guess. Think of three nouns which can be used with the adjective strong but where the meaning of strong is quite different in each case. How many things can yor open where the opening is not like opening a door?

an are lual ons
1 A

thin ical

han the .ical din one ;uch :s of look
le ln

vord Ter-

Did you think of answeran enquiry,a letter,the door the chargethat . . .? rt is easy to seethat the translationsof theseexpressionsinto another language could very well involve a different verb in each expression.Similarly, with strong opinions, wind, coffee,cheese,or opening a bottle, a lette4 a meeting. Becausesomecollocations are so familiar, it is easyto think they are obvious when they are, in fact, highly idiomatic. In an article on phraseology,peter Howarlh refers in passing to 'completely transparent collocations such as have children' but although there may be little difficulty guessingwhat this combination of words means, it may present learners with considerable problems from a productive point of view. Note first the difference between she has a baby and she's having a baby; changing the grammar changesthe meaning of the verb. If you consider the expressionfrom the perspectiveof a Swedish leamer, for example, the first is Hon har ett barn mircoring the structure of the English exactly, although baru is closer to the English word child; bil the secondis Hon vcintabarn,literally she is waiting for (a) child, closer to English she's expectinga child, though even here she's expectinga baby is closer to the Swedish. Grammar and words are in complex interplay, so that apparentlytransparentcollocations hold many potential pitfalls for the leamer. Ter-Minasovagives anotherexamplefrom the perspectiveof Russianleamers of English: the introduction to the BBI Combinatory Dictionary sees open the gate as a free collocation, allowing substitutions such as lock the gate, open the door, etc which should cause few problems for leamers, but she comments:



Language in the lexical approaclt

Many word-combinations look deceptively free within their own language,and their non-freedom only becomesobvious when they have to be translatedinto anotherlanguage.The free, variable open the gate is, indeed free and variable within its own - English language. Howeverit looks much lessfree in the eyesof a Russian meaning....TheRussian the learnerwho tries to express equivalent equivalent of to open is presentedby quite a variety of verbs in Russian-English dictionariesopen,discover cleat bare, reveal . . . and the Russian word for gate has only the plural form. The point for teachers that even the simplestof collocationsmay contain is difficulties for learners, and some comment to make leamers aware of problems, including the 'blocked' collocations discussedearlier, may be necessary. may even be that unexpectedcombinationsof familiar words are It some of the most important and useful collocations from a pedagogicalpoint of view. George Woolard makes the point that it is helpful to ask leamers which words they are surprisedto find used together.Teachersneed to keep in mind that they may be surprisedat what surprisestheir leamers. We note in passing that this has one very important classroomimplication 'know' a particular collocation is quite definitely asking leamers if they testing,not teachingsince the idiomatic nature of many collocationsmeans they cannot be predicted with confidence from knowledge of the individual who forget this risk frustrating learnersby asking componentwords. Teachers questions which the learnercan only answerby guessing. One final potential source of confusion should be mentioned. In corpus linguisticsthe term 'collocation'tendsto be usedin a different way from the it way it is usedin this book. JeanHudsondescribes as follows: of In corpuslinguisticsit is more often usedin the abstractsense a in generaltendencyfor linguistic items to co-occur(not necessarily 'I didn't get that job, by the way. The immediate proximity): 'job' and 'application' application was in too late.' The words collocate quite strongly, whether or not they are adjacent. Much corpus work to date has in fact focused on reporting collocability and patterning, towards the ultimate goal of establishingthe most frequent collocatesof specific items, with information about the co-occurrenceprobabilities of words. This more abstractdefinition is used in this book onlv bv Michael Hoev. who is, of course,a co{puslinguist.

more term. coliig (gran parnc pronc respo

Some but fo (direc out, lt

SuchI it has that rl

man)' thanjL

Itisn StruCT gramn betwee degree are pc colloca colloca so gen( we cott Now th exceptl which explana

7.6 Colligation
Although many teachers are just beginning to incorporate the explicit is with the research also concerned teachingof collocationinto their teaching,

Instead languag fixedne rts own lexis, n principl are tr)/r generai they are pedagog English. explicit I

Language in the lexical approach 137

a1n of be are rtnt

more generalidea of colligation. Again, there are different ways of using the term. collocation is the way one word co-occurs with another word, colligation is the way one word regularly co-occurs with a particular (grammar) pattem, so, for example some verbs typically occur with a particular tense, or a noun might typically appear preceded by a personal pronoun, rather than an article (pass my/your driving test, It's my/your/our responsibility . . . , bil I'il take the responsibility . . .). to for Sornedescriptivelinguists use the term 'colligation'not for word + pattern, but for the more generalpattern + pattern. An example is (verb of motion) + (directional particle), which covers all combinations such as run away. rush out, hurry down etc. Such terminology and researchmay seema long way from the classroom,but it has a serious,classroom-orientated purpose.It is now generallyaccepted that the separationof vocabulary from grammar was an artificial one, and maintaining it causesa greatdeal of confusion and, more importantly, means many interesting and helpful featuresof how words are actually used- rather thanjust 'what they mean'- are overlooked. It is now generally acceptedthat languagedoes not consist of a few ,big' structures with slots which are filled with individual words. The whole grammar/vocabularydichotomy is invalid. All language lies on a specrrum between what is fixed and what is variable, and there are many different degreesof fixednessand, corespondingly, different degreesof generalisation are possible. colligation generalises beyond the level of individual collocations, so a bunch of grapes/bananas/flowers ate three separate collocations, but the last one can be generalisedto a bunch of (flowers), and, so generatea bunch of roses/daffodils/(anyother kind offlower). In general, we commit (crimes), and until quite recently suicide was a crime in Britain. Now the law has changed,so the collocation commit suicide has become an exceptionto the colligation commit (crime).It is just this kind of fossilisation which produces idioms which, over long periods, seem to defy ,logical' explanation. Instead of a few big structures and many words, we now recognise that language consists many smallerpatterns, of which exhibit varying degrees of fixedness generalisability, or eachbasedon a word; in a sense, eachword has its own grammar.It is this insight - that languageconsistsof grammaticalised lexis, not lexicalised grammar - which is the single most fundamental principle of the Lexical Approach. It is within this framework that researchers are trying to find accurate descriptions of English. obviously, the more generalisablethe patterns they find are, the more useful, at least in principle, they are. But teachersare right to be a little suspicious,for description is not pedagogy.It is not self-evidentthat the best, most accuratedescriptionof English, will be the most pedagogicallyusefur, any more than the fullyexplicit scientific descriptionsprovided by relativity and quantum mechanics

rep nely lns ual mg
lus the


icit the


Language in the lexical approach

would be appropriateto school scienceclasses. the other hand, science On teachersneed to be aware of these great theories, and of how and why the particular simplifications of them. content of their classesrepresents Language teachersneed to accept and fully intemalise the idea that dividing the language into a lot of individual words and a few big structuressuch as the presentperfect and the passiverepresentsa discrediteddescription of any language, and a dangerous distortion of the true nature of language. The emphasisof classroommaterials should move flrmly onto the middle ground of language,the grammar of words. That means taking colligation seriously as a real attempt to provide insights, initially at least for the teacher.It also meansmaking collocation a centralpart of languageteaching for all learners now. [For more on colligation,seep 233.]

Th To Th

We spe


rter lim

7.7 Other multi-word expressions
There are many kinds of fixed and semi-fixed expressionsin addition to collocations. Here, we look at some which may be new to readers but we begin with a simple extensionto a familiar term. 1. From adverbs to adverbials In traditional grammar 'adverb' was (with noun, pronoun, verb, adjective, preposition,conjunction)one of the word classes. Even then it was the most problematical class as it was used as the dustbin for all the words which did not seem to fit into the other, neater classes.As Stig Johansson puts it: 'Adverbs are no doubt the most heterogeneousof the traditional word just, neve4 classes."It is difficult to seewhy all of frankly, almost, extremely, carefully all belong in the sameclass,especially as they have clearly different functions in examplessuch as: Frankly, my dear, I don't give a damn. He almost failed, so he was extremely pleased that, in the end, he just managedto scrape through. I never did anything more carefally in my lfe. Adverbs were recognised as having different functions such as qualifying a verb (do it carefwlly), intensifying an adjective (extremely pleased), or qualifying a whole sentence(Frankly, . . .). When we think of multi-word phrases- adverbialsrather than adverbs- we need to have different uses in mind in a similar way: On the other hand, there is a significant saving with the older model. I'll get it in thepost as soon &spossible. He expectedto pass easily, and he did pass but only by the skin of his teeth. Observationof real languageshows our mental lexicons are particularly rich in multi-word sentence adverbialsthat areusedfor structuring what we say or write, what David Brazll neatly called 'talk about talk', as in theseexamples:

The mol Tea teac alm and

simi prep such have to dr 2. Nr

Anne spea spea expre largej 'state

chara the u: This r simila speak

The cl . func . lexic


. seml These sort of

Her da separa

Language in the lexical approach


tnce 'the ding has any The rund usly also ners

The above examplesall seemto suggest. . . Tb go back to the point I made earlie4 . . . That may be so, but the point I particularly want to emphasise/stress/remind yow oJ . we immediately notice that the first seemsto be usable in both (academic) speech and writing, while the second two are typical of speech and not appropriate for writing. secondly, becausewe are dealing with multi_word items, thereis the possibility of one or more ,slots,whichlay ue filled by a limited number of alternativewords or phrases. The multi-word lexicon is more complex, with more possibilities, and often more restrictions, than traditional vocabulary teaching has recognised. Teachers need to be aware that unless they consciouJy avoid the trap, teaching learners 'new words' too often meansignoring the adverbial lexicon almost completely.If leamers are to have the ability to structure what they say and write, the importance of developing their multi-word adverbial lexicon cannot be over-emphasised. passing we may note that (In leamers need a similar extension to their prepositional rexicon; there are very few prepositionsof place and time, but an enormousnumber of multi_word items such as immediately opposite the . . . , earry in the New year, someof which have a strongtendencyto collocate with particular verbs.Again, teachersneed to draw thesemulti-word items to reaniers'attention, context.t in 2. Negotiating Ianguage Anne williams (Arena,Issue 19) has analyseda substantial corpusof native speakers doing simulations of negotiations.unsurprisingly, shehas noted that speakers regularly rely on a large repertoire of fixed and semi-fixed expressions.These have two main advantages:firstly their meanings are largely conventionalisedso they ensureall parties to the negotiation know the 'state of play' at any moment, and secondly,as they are largely pragmatic in character producedasprefabricated and wholes,they free processing-space in the user's brain so (s)he can concentrateon the content of the negotiation. This may itself have a high percentageof standardcollocations which servea similar purpose. Such prefabricated language is frequently used by native speakerswho are in relatively complex situations. The chunks can be divided into three catesories: functional stems such as If we were to . . . , would be prepared to . . . lexical chunks linked to the specific subject matter [which she describes as 'of little interest'.Seecommentsbelow.Ed.l ' semi-lexicalchunkssuch as what sort of, somethingthat we. These last, she points out, are frequently used with interesting notns: what sort of p ercentage/time care/benefit/dis ownt/mov s c p ement/re io; . Her data, therefore, reinforce the view expressedthroughout this book that separatmggrammar from vocabulary violates the nature of language, and

nto rwe

t1ve, nost idid s rt: vord



nga ,or vord )s ln

eth. rich ry or rles:


Language in the lexical approach

makes the conversion of input into intake more difficult. She goes on to discussthe implicationsfor teachers: "Chunks may be prefabricated, but the chunks in my data are not grammaticallysimple: Whatsort of volumeare you lookingfor? If we were to give you that commitment, what sort of discountwould yowbe talking abowt? Even when the language is complex, however, it appears to require little processing- the chunks are simply used additively. The interest for teachersis that if we can teach studentsa restricted range of extremely flexible chunks, we can provide them with a tool to aid fluency when they are focussing on their negotiating objectives. Such chunks are of interestbecause: . They are short and manageable. . They are high frequency becausethey are often invariable forms (infinitives, modals, were to, eIc). . They can standaloneor be embedded, Thereis onepoint that I'd like eg to clarifi. . They combineflexibly: If you were to . . . , we'd be preparedto . . . . They allow avoidanceof grammatical decisionswhile concentratingon a meaning,yet do not suggest simplified grammar,eg We'd be lookingfor somesignif cant movement. . They encourage focus on the richnessof de-lexicalised eg language, It's a a side issue/ real concent." dismissalof lexical The only point which requiresa commentis her apparent chunks. Her article is specifically a discussionof the essentiallypragmatic chunks which facilitate fluency in negotiations. The lexical chunks are, of the course,primarily usedto express all-importantcontentof what is said, so lexicon of this kind too. Her point is simply that all learnersneed an adequate on too often languageteaching concentrates content languageand ignores, or the at least under-emphasises, lexical chunks which all language users, including native speakers,rely on to provide fluency. Using these frees processmgcapaclty, so you can think about what you are doing, rather than how you can say it. [SeeDeborahPetty's commentsp 96.] 3. Utterance launchers Anne Williams' examples are taken from an apparently specialisedarea of spoken English, but the Longman G:rammarof Spokenand Written English shows that a very similar phenomenonexists in ordinary conversation.The grammar lists about 100 five-word clusters - seebelow - which are common English.Most of thesebegin with 1, and most in their colpusof conversational of the expressionsare what they term utterance launchers - expressions gives usedto introducethe contentof what you want to say.A small selection an immediate feel for this kind of language:

Idc Idc

I rh

I v't yoLt

Self will effer then expr lmm are l unna lmpc will awar in thr


Coml studii rtem. groul punct cluste kinds I dort what' in the in the

The ri conve provid text t)' this t1' for anc chapte if lean level.

Some identifi Ironica even b

Language in the lexical approach


I don't know how you I don't think you can I'm going to have to I was going to say you seewhat I mean self-evidently, anyone racking an adequaterepertoire of these expressions will have trouble taking part in natural conversation. Note, too, that the effective use of these expressionsis intimately bound up with pronouncing them as units [as Jimmie Hill suggested, p 55]. Although see presenting such expressions is no guarantee that learners will acquiie and use them immediately - on the contrary,teachersare well awareof the factthatleamers are more likely to rely on tried-and-tested expressions even if these are unnatural - there is a strong case for presenting them, explaining their importance, and engaging in a little controlred pru.ii"" - not so that leamers will immediately add them to their productive language, but as part of the awareness raising which does seemto contribute to tuming input into intake in the medium term. 4. Clusters computers are exfemely good at doing mindless sorting. Recent corpus studies have therefore provided us almost by accident *ltt u new kind of item, the cluster (called lexicar bundles in LGSWE). clusters are small groups of words which appear consecutively in text without regard fbr punctuation marks, or even changesof speaker.Needless to say, many such clusters are of little interest for the classroom,but it is interesting to note the kinds of phrasethat do occur. Here are a few reported in LGSWE: I don't know what I thought that was I think I might what's the matter with how do you know going to be a in the case of the it should be noted that on the basis of in the present study the way in which the extent to which

I don't know what to I don't know whetheryou I rhink it might be I won't be able to you won't be able to


Language in the lexical approach

may not be immediately relevant to the classroom, they demonstrate for teachershow easy it is to be so focussed on the content of a text, that the phraseologywhich holds it all together- what we may call the text*grammar or discourse-grammar- is all but invisible to both teachers and learners. Increasedawareness the languageof which the text is composedwill help of teachersto guide learnersmore effectively towards the languagethey need to notice, and which will almost certainly be unnoticed without teacher mterventron.



inc are


7.8 Words
The single most important insight provided by the new corpus-based descriptions of English is that the whole vocabulary/grammar dichotomy needsto be replacedby a spectrumof pattems which exhibit different degrees of restriction and generalisability.Words are used in patterns which learners need to notice; structures are subject to constraints which were frequently ignored in traditional EFL teaching; the new descriptions mean we need to revise our views of both vocabulary and grammar - words and structures. 'new words' in the same way is a It is increasinglyclear that treating all wholly inefficient way to expand leamers' mental lexicons; indeed, even 'new words' is unhelpful. While learners do need to thinking in terms of acquire new words, particularly through extensive (pleasure) reading, they also need to expandthe phrasal element of the lexicon, acquiring hundredsof useful combinations of familiar words. We have already seen that an adequatelexicon, in addition to individual words, involves large numbers of collocationsand idiomatic expressions, adverbialand prepositionalphrases, colligationalpatterns. Although the emphasis in this book is on the co-occurrence of words in collocations,it is also becoming increasinglyclear that regardingall single words as frindamentally similar is a grossly misleading over-simplification. questionwords etc Words from the closedclasses pronouns,prepositions, 'grammar', rather than have traditionally been regarded as belonging to the corpora the 'vocabulary' of the languagebut analysis of computer-based quite the way that has reveals that these words are not different in kind in usually been assumed. Once again, the traditional vocabularylgrammar dichotomy breaks down; there is a spectrum upon which all words can be placed; how the words behave is a matter of degree, rather than different words belonging to clearly different categories. At one end are rare words - mostly nouns - which caffy a lot of meaning, and which have small collocational fields; at the other end are the most frequent words of the language which caffy very little meaning in themselves,but which are elementsin many different pattems- which is, of course,why they are among the most frequent words.

Ag rev( true coli wor mu( fielc and Do thes of it wor Trar higt

Cor 1.7

the mea is vr

Language in the lexical approach


) for t the rmar ters. help :d to cher

The Cobuild data clearly reveal that most of the most frequent 100 words in English are what were traditionally thought of as 'grammar' words; they include as, which, these,our; most.Yery few of the 100 most frequent words are those traditionally thought of as 'vocabulary': time, people, man, little, good.

We can explore this idea from a slightly different perspective. Write down 10 adjectives that you think are likely to be used fairly frequently with each of these nouns: wit, plea, guest, position, idea. For which word is the task easiest, and for which almost impossible?

ased omy rees



15a SVen

dto they is of tan rs of and ls in ngle t1on. than pora . has

A glance at a dictionary such as The LTP Dictionary of Selectedcollocations revealsthat some words have many more collocatesthan others. It is broadly true that the less frequent a word's overall frequency, the smaller its collocational field, so it is easier to think of collocates of more common words. Many nouns which are frequent and which do not themselveshave much meaning - situation, idea, position, way - have huge collocational fields. It makeslittle sense with suchwords to ask exactly what they ,mean', and even less to ask learnersif they 'know' the word; it is similar to asking Do you know the word 'to'? Although traditionally thought of as vocabulary, these words are more part of the grammar of the languagethan they are part of its lexicon. In each case there is a great deal to be leamed about how the word is used - the collocations and expressions of which it forms part. Traditional vocabulary teaching has largely overlooked the central role these highly frequent words play in the language.

Different words exhibit different kinds of pattem, which implies different kinds of treatment in the language classroom. write two or three sentences for each of these words - try to think of examples which are 'typical'uses of the words: telescope,cat; speak, beautiful, strange, sometimes. which of the words do you think are easiest for learners to acquire and which are most difficult? Which do you think they are most and least likely to ask vou about in class?

nbe lrent and uent . but they

Commentary 1. Telescope.Perhapssurprisingly, the rarest is the easiestfor the learner and the teacher;once the leamer has understoodthe word - and in this casethat meansknowing the correspondingword in the leamer's own language- there is very little more to be said abouttelescope.


Language in the lexical approach

2. Car. But a relatively common noun like car is a different matter; it would be useful to introduce it togetherwith some of its common adjective and verb collocates suchaspowet'ul, second-hand, family, hire, start, (not*begin), the car broke down. Nouns which are more common have larger collocational fields, so some collocatesshould be introduced from the earliest possible point in courses. 3. Speak. This verb posesdifferent problems;it is one of a group of verbs with similar meanings - say, speak, tell. The difference between these does not need to be explained, so much as explored, because the difference between them lies not so much in their meaning, more in the way they are used - in other words their different collocational fields. They should be introduced with a small family of real examples which show some typical collocations and their families compared and contrasted.[Compare the lists on pp 34 and 61.1It is important for the teacherto draw attention to patterns: say (actwalwords) : H ellolThank yoilS orry : t ell John/me/ omeone/theclass to s (do something).The teachershould also provide the 'negativeevidence'of what is not actually possible: *say me/John/someone . . . . Say is not to followed by a name, person or personal pronoun in this structure (although compareHe said John wowldhelp.) 4. Beautiful. This adjective is only the opposite of ugly in a small range of examples. The oppositedepends the following noun. on 5. Strange. Here is a word with at least two very different meanings. Like most adjectives,as soon as it is taken out of context, much of the information about how it is used is lost. Such adjectivesshould be introducedas part of naturally occurring collocations, and possible alternative collocations should be explored with learners immediately and attention drawn to important blockedcollocations. soonas simple word-for-wordtranslations in the As are learner'smind, acquisition of the patternsin which they are actually used will be impeded. 6. Sometimes. This apparently simple word presents a different problem again.One currently popularcoursebook teaches on a scale: it DVo............. never rarely not often 50Va........... sometimes oJten usually l00%o always

It sc with sotT teac avol gran violz frust

Thin slmp and c notrc

So, r class in co after learni

One r dealt phras


Looki one fe the'u

but is this helpful? Write down two or three natural sentences which contain one of neve\ sometimes or always. Now try substituting one of the other words.Try theseexamples: I sometimes wish I lived in France. xI never/alwayswish I lived in France. I've always wantedto live in France. (?)I've sometimeswantedto in France. live Sometimes think I shouldmove. I *Never/AlwatsI think I should move.

Tradrt that o, many relatec at bes posse roof, t) produc the Kir

Ci Ci


Language in the lexical approach


)uld /erb , the rnal ible :rbs loes are lbe ,ical lists rns:
is to

It soon becomesclear that the words are rarely substitutablefor each other without the sentence seeming odd, absurd or plain wrong. And does sometimes ever suggest50Ea,and if it does, 50vo of what? As long as the teaching of grammar and vocabulary are separated,this kind of problem is avoided until the learner actually tries to use the language.vocabulary and grammar interact at every level, in language of all kinds; separating them violates the nature of language; it also helps introduce confuslon, eror and frustration into the whole processof learning a secondlanguage. Thinking about thesefew words clearly shows that telescope canbetaught by simple translation, but in all the other cases,some exploration of collocation and context is essentialif important featuresof how the word is used are to be noticed by leamers. so, different kinds of words require different kinds of treatment in the classroom,but rnost need to be met and acquired together with other words in collocations, expressions other chunks.words taught on their own will, or after all, have to be put with other words if they are ever to form parl of the leamer's active vocabulary. one word above all others demonstrates both that different words need to be dealt with differently and the importance of acquiring words in complete phrases.


not ugh :of .ike
l10n tof ruld lant the ivill lem

7.9 The central role of o/
Looking at languagethough a narrow grammatical perspectivehas obscured one feature of English of staggeringimportance - the central role played by the 'word'of Traditional grammar has very few word-classes,so it was perhapsinevitable that of was classified as a preposition. Sinclair points out, however, that in many exampTes aware of the problem, much of the time - of is closely related to the word which precedes it rather than the word that follows it, so at bestthe term 'preposition'is highly inappropriate. Nor is it typically about possession, althoughin a few casesthere is a deceptivesimilarity: the car's roof, the roof of the car. rn most cases,however,this kind of ,transfbrmation' produces bizane results;ffy it, for examplewith these:a breachof the peace. the King of Sweden,the price of a ticket.



Make a list of a dozen expressions containing the word of. Can you find different pattems of use? Can you find one particularly important pattern? Can you see why of might be one of the most useful and most frequent words in English?


Language in the lexical approach

In fact, o/is the secondmost common word in English, second only to the. This immediately suggestsit either has many different roles in English, or it studiesshow that it Sinclair's corpus-based has a use which is all-pervasive. does have different uses,but that its frequency is largely a result of a single use, unemphasised in large academic glammars, and almost completely ignored in pedagogic grammars and teaching materials. It is the single most important way of building a pafiicular kind of multi-word noun phrase, and therefore central to any considerationof collocation' Most traditional grammar lessonsinvolve patternsof the verb phrase,loosely 'the tenses'.Traditionally, little or no attention has been paid to the grammar (nonof the noun phrase. However, examination of naturally occurring narrative) texts shows that one of the defining features of such texts is the : of preponderance complex noun-phrases of in developments the management financially sensitive Recenttechnologichl the importanceof finding ways of controlling information have demonstrated the meansof accessto such information. Knowledge of data managementis essentialfor graduates of any discipline who hope to work in those areas of the economy which currently have the greatestchanceof growth during thefirst half of the next decade. Look back at those examples; does one word jump off the page? The (6). examplescontain 65 words, the most frequent of which areof (9) and the There it is, staring us in the face, the most common word in the examples the second most common in the whole language, hardly mentioned in traditional ELI grammar teaching; o/ is the key to the construction of noun phrases English. in Sinclair gave a clear explanation of the function and importance of of in Collocation" Corpus,Concordance, The simple structureof nominal groups is basedon a headword which is a noun. Determiners, numerals, adjectivesetc. come in front of the noun and modify its meaning in various ways. Prepositional phrases and relative clauses come after the noun and add further strands of meaning. The function of of is to introduce a second noun as a potential headword: this kind of Problem the axis of rotation the botrle of Port Each of the two nouns can support pre-modifiers' 'zones' As I write this, there is a heated debateabout the name of one of the of London's Millennium Dome; religious gloups did not like the name spirit that zone,so it was changedto thefaith zone.This promptedsomecomplaints

it sor on th will l probl beca pre-n

Althc ways that i categ they I

spe qua con nou

They conta glves
AS (t

in t) tnil in rJ at tl

Itisu to ear more l they n probal


Sincla may st


the fa langua lexis althou of this

Althor approl linguis catego

Language in the lexical approach 147




)st nd

lar )nhe

it soundedas if one faith had priority - the faith, so the organisers have settled on the ungrammaticalnamefaith zone,although when anyoneasksfor it, they will presumably say Where is the faith zone? The ambiguity, and hence the problem, could have been solved by calling it the zone of faith, precisely becauseo/ separates the two nouns so that each is separately available for pre-modification without ambiguity. Although LGSWE rightly points out that noun phrases are made in many ways, and that such noun phrasescan be very long, it also endorsesthe view that different kinds of phrases containing of are one of the largest subcategoriesof noun phrase.Here are a few of the dozen or so types of phrase they list: thesekinds of questions a set of books the brutal murder of a child a mouthful of food They also list well over a hundred short phrases- lexical bundles - which contain of, and which are typical of academic writing. This small selection gives a flavour of how central such phrasesare to this kind of writrns: as a result of as afunction of from thepoint ofview of in the case of in tenns of in theformation of in the direction of in the case of a in a number of ways in the context of the similar to that of with the exceptionof at the time of the at the level of at the time of writing It is worth noting that this languageis precisely the kind of languagereferred to earlier which is likely to be invisible to learners,whose attention is much more likely to be focussedon difficult content words. If they are to write well, they need to add both kinds of lexical item to their mental lexicons. This will probably not happenwithout proactive intervention by the teacher. species nouns: quantifying collectives: comparable genitives: to nouns with -/z/:

rhe he
6) S -

1n run 1n

7.10 Grammar
Sinclair has arguedthat once we have sufficient corpus-based evidencewe

of this book. es'
LrLt hat

Although there is considerabledisagreementabout what categoriesare most appropriate for different purposes,the consensusof opinion among applied linguists is that the separation of grammar ,and vocabulary as distinct categories simply wrong. As long ago as 1990,Sinclair claimed: is


Language in the lexical approach

becomingavailablecastsgravedoubtson the wisdom of The evidence postulating separatedomains of lexis and syntax. In similar vein, the authors of the monumental LGSWE observe in their introduction: Syntax and lexicon are often treated as independent components of English. Analysis of real texts shows, however, that most syntactic structurestend to have an associatedset of words or phrasesthat are frequently used with them. There is generalagreementthat there is a spectrumbetweenwhat is particular and what is general; single words, sffong collocations,certain idioms and fixed expressions are invarianl, or at least almost invariant, while open collocations, colligations and traditional grammar structures represent varying degrees of generalisability. In other words, we do not simply rememberevery bit of languagewe have ever met and list it; we also sort it in aboutit. Michael Hoey has commented: someway and make generalisations Grammar is the product of the colligationsyou have noted in the language and Sinclair claims: Grammatical generalizationsdo not rest on a rigid foundation, but are the accumulation of the pattenxsof hundredsof individual words and phrases.Implicit in thesecommentsis the importanceof learners meeting large amountsof input which they can use as the basis for their own generalisations. Equally, of course, this view denies that when learners 'rules'the learnerhas producecorrect sentences theseare basedon abstract been taught;the rules are neither more nor less than various provisionaland partial generalisations,based on understanding and breaking down in different ways and to different degrees,input which is essentiallylexical. Grammar often over-generalises to We have seen Different 'levels' of languagegeneralise different degrees. other words at random, somewords regularly cothat words do not occur with prove occur with other words - lexical collocationssuch as stronglysuggest, grammatical featuressuchas articlesor prepositions conclwsively. Sometimes may be part of a pattem - grammatical collocations such as take the opportunity ro. Colligations are even more general - they relate word to pattern,or patternto pattern:I'll seeyou (time expression); There'sno need b (asUwarn/tell/remind etc) (John/your mother/me/etc). Grammar sffuctures generalisefurther, but here a word of warning is necessary. have already We met the idea of blocked collocations; although such collocations are grammatically well-formed and could be sanctioned by the native-speaker community, they are not. A sourceof frustration to learnersand teachersalike, they are arbitrarilydeemed'wrong': Wedon't say that.At leastthis is familiar to teachers, there has been a tendencyto believethat the generalisations but of grammar really are true generalisations;Chomsky claimed that a grammar should produce 'all and only' the coffect sentences a language. of We now

realis there

Althc 'pote least

Manr, gener that n - the set ol prona in sec actua gener mefia peopl begin: bring The c Peter Iti
t1fi InC


Dou-e of dif man)'


Withir empha at the i langua fixed genera

A mor

We mr realise EFL r colloc

Language in the lexical approach 149

f Ltreir )f c 'e

realise, however, that grammar frequently over-generalisesin the sensethat there are many possible grammatically well-formed sentencesto which the native speaker'snatural responseis You could say that, bwt you wowldn't. Although these sentencesmay be 'correct', and are what we may call 'potentialEnglish', they do not seemto be part of the languageas it is, or at leasthas been,actuallyused. Many recent studiesof used languagehave shown that some of the long-held generalisations provided by grammar are in practice over-generalisations, and that many supposedlygeneralpattems are subjectto restrictions of somekind - the pattem is only used in a particular genre,or is typical of only a restricted set of verbs, or is almost invariably used with, for example, a personal pronoun, occurs naturally in first person sentences speech,but hardly at all in in secondor third person, or some other similar restriction. Looked at as an actual utterance rather than a sentence,it is difficult to make grammatical generalisations from I'll seeyou tomorrow -varrations such as You/He'll see me/you/him tomorrow, although 'coffect' are not entirely convincing as what people do, rather than what they might, say. This is true of many utterances beginning I'll, where the 'equivalent'with loz seemslikely only if tagged:I'll bring it on Monday.Yow'llbring it on Monday,won't you? The current theoretical position is represented thesetypical commentsby by Peter Skehan(A Cognitive Approach to Language Learning): It is likely that the growing existence of large corpora together with ...increasingly sophisticated text analytic software ...will for the first time reveal ways in which there is a lack of regularity and rule much more of the time than previously thought. DouglasBiber and his colleagues have analysedmany different sub-corpora of different registers and draw the general conclusion that: [These] show many generalisations have limited domainsof applicability.

:ular and open )sent mply titin nted: 'uage rigid idual ,mers own

:r has L1 and
/Il llt


) seen ly coprove ;itions :e the rrd to t need ctures lready

7.ll Lexis
Within the traditional grammar/vocabularydichotomy, teaching collocation emphasises vocabularyrather than grammar,but this is the wrong way to look at the issue.The dichotomy is invalid; languageis fundamentally lexical, and languagepattems are arrangedon a spectrumfrom thosewhich are absolutely fixed and non-generative, to those which provide a high degree of generalisation, though usually with somerestrictions. A more'grammatical' approach We may summarisesimply by saying that words are more generativethan was realised earlier, and structural pattems are more restricted than traditional EFL rules acknowledged. Every word has its own grammar; teaching collocation means giving attention to a much wider range of patterns which

peaker ;alike, rmiliar iatlons lmmar re now


Language in the lexical approach

of any suffound individual words; this meansmany mole patternsthan those 'gfammatical' respect, it is a more traditional structural syllabus. In this it avoids approach than the traditional structural syllabus precisely because the problems which arise from separatinggrammar from vocabulary"

But il leame taken Howa
Ot co

7.L2 Collocation and testing
than we have It is now clear that any learner's mental lexicon is larger items rather previously recognisedand that much of it consistsof multi-word level, the more this is true' As than individual words. The higher the student's know, and a Michael Hoey remarks in his chapter,there are a lot of words to is more lot to know about eachword. If the size of the leatnet's mental lexicon the learner knows than we to do with the number of phrasesand collocations the ways in have previously recognised,it is clear that we need to reconsider is which vocabularY tested. is being As Peter Hargreaves' paper clearly demonstrates, collocation examining boards as an element in recognised increasingly explicitly by the Cambridge a assessing leamer's overall proficiency. Examinations such as (cAE) and First certificate (FCE), Certificate of Advanced English Proficiency(CPE),nowencoulagethestudyofcollocationaSpartofthe exampfeparation'Here,forexample'a.tresomecommentsfromtheFCEand CAE examiners: Learningcollocationsisausefulprepalationfor[Paper3]sincethese arefrequentlytested....Itisalsovaluabletostudythedifferences trip and between words of a similar meaning such as voyage,iowrney, excursrcn. just an It is in fPaper 3] that knowing a whole phrase rather than to know individual word is particularly important. candidates need or gerund for example) and what structurefollows a verb (an infinitive whatprepositionitgoeswith.Collocationsarealsoanimportant goeswith to feature of Part 3. It is useful for candidates know thatreach or agreement tell wrth lies, for example' Recommendationfor candidatepreparatlon This is particularly relevant to Question 1 [in CAE] in which an knowledgeof collocationis a greatasset' extensive recognisethe Bearing in mind the enormousvocabulary load which we now leaming individual collocations in leamers need, it is clear that rather than whole idea' preparationfor the exam, learnersneed a clear introduction to the -rt they also tt"y are to gain maximum benefit from any languagethey meet, and record collocations. Testing need to acquire the ability to notice suchas: collocationis superficiallyeasy,using questions is The government trying to closethe """' betweenrich and poor' C' distance D' door A. sPace B.gup


m1 of kn

In Ch testc(

Teach simpl comb learne testltr colloc both t 1. Wt


h C. A

3. Hc w(



Even betwe gener know


This < in the

Language in the lexical approach 151

of any Latical' avoids

But if such testing is to avoid arbitrariness, simply demonstrating that the leamer either does or doesnot know a particular collocation, stepshave to be taken to ensure that the test items are devised in principled ways. As Peter Howarlh has observed: One drawback (of finding out how many subjects know a given collocation) is the difficulty of establishing the validity of any predehned list of target collocations, since, as many EFL teachers might agree,this componentof a learner's linguistic competenceis one of the least predictable, making it hard Io generalizefrom subjects' knowledge or ignoranceof a small number of individual items. In chapter 10, Peter Hargreavestakes up the question of how difficult it is to test collocation in principled and sustainableways. Teacherswho devise their own tests need to be conscious of the danger of simply making the lexicon seem like a confused mass of arbitrary word combinations. Two possible ways forward are to use test items which ask leamers to recogniseboth acceptableand unacceptable collocations, or to use test items which require learnersto write either a specifiednumber or asmany collocates as they can for a particular word. Here are example test items of both types: 1. Which of the following are acceptable: My a) car b) health c) income d) marriage e) holiday was damaged. 2. Which of the following contain correct usesof the verb break? a. I have broken my watch. b. It broke his confidence. c. A suddencry broke the silence. d. You said you'd help.You can't break your promise! e. Shall we break the meeting now? 3. How many nouns do you know which can go immediately after these words: a. market (research,price, penetration,etc) (rise, stability,promise,etc) b. price c. football (match, stadium, player, etc) (timetable, ticket, crash, station, times etc) d. train Even with items of these types, however, care must be taken to distinguish between testing what has been formally taught in the course, and more general testing. Care must also be taken to ensurethat the words chosen are known to be appropriateto the learners'level.

e have ; rather rue.As t. and a 1S more han we n,aysin ; being nent m ntrridge E) and t of the rCE and hese

r and st an (now ) and )fiant with

gnisethe ations in Lole idea. they also Testing

7.13 Necessityfor change
This chapterhas been about language,and in particular about recent changes in the way we describelanguage.corpus linguistics and computer corpora are


Language in the lexical approach

powerful tools, and regularly produce new, and unquestionably better, descriptions of English than we have ever had before. Competent teachers need to be up to date with thesedescriptionsin the sameway that a competent doctor needs to know about new drugs and treatments. But it is not selfevident that improvements in description necessarily imply changes in methodology. In the earlier years of this descriptive revolution, Henry Widdowson wrote: Swchanalysis provides wswith facts, hitherto Ltnknown,or ignored, but they cartj any guarantee of pedagogic relevance.Guy Aston do not of themselves has pointed out that basic factors such as availability, teachability and classroom needs are at least as important in designing language courses as strictly linguistic criteria. Ron Carter, deeply involved in new descriptionsof has similarly observedThereis a tensionbetweentruth the spokenlanguage, pedagogicjudgement. He goes on to argue,for example, to the langwageand that there may be a case for specially designedlanguage teaching materials 'real' language;such materials may which simulate some of the features of be more effective than real material, or materials which replicate exactly the linguistic featuresrevealedby accuratedescription. In an article assessing importance of corpora and the descriptive insights the they offer for all languageteachers,Ivor Timmis writes: Where corpora seem to be telling us something about the nature of languageor the nature of languageprocessing,I feel it incumbent upon us to take this on board. If the evidence is accumulating that it is not tenableto make a sharpdistinction betweenlexis and grammar and that speech involves, to a large degree, the assembly of prefabricated chunks, this must affect the way we approachour teaching. Maggie Baigent, writing of her experiencein trying to implement a lexical various activities Vol. 8 No. 2) suggests approach(Modern English Teacher, of and with a very balancedassessment both the opportunities and concludes difficulties presentedby our presentunderstandingof the nature of language: Michael Lewis says in Implementing the Lexical Approach that what may not changevery much as a result we actuallydo in the classroom of our own change in thinking about the centrality of lexis in the languageand the languagelearning process.Unfortunately, to an extent by and he is right, aslong aswe are constrained coursebooks syllabuses based primarily on a list of discrete grammar points with vocabulary sectionswhich are little more than lists of single words. However,good teachersalways adapttheir coursematerials and I would like to believe in the 'cumulative effect' of small but unremarkable changes in classroom practice which Michael Lewis also hopes for in Implementingthe Lexical Approach.I would like to think that we can bring about a gradual changein learners'perceptionof the language

at, clr m

Teacl radict chan5 better

Magg vocab learne single


Wen descri man), thin-ss precls

compl theonchapte in con is fund princig compa on dis

their tr

It is nc or cou action the ho providi before. change

This bc We mt teacha we har. shall s method

Language in the lexical approach 153 )ttel,

hers )tent selfsln -ote: they ston and

and lead them towards greater autonomy in identifying multi-word chunks in language they are exposedto, and making these more and more part of their communicativerepertoire. Teaching colrocation does not mean a radical upheaval, but it does mean a radical change of mindset for the teacher, *irirt proar"", many small changesin the activities they focus on in the classroo-. L"urrr"rs do become better at noticing' storing and using lexical chunks. We must hope that in due coursesyllabusesand textbookswill change and the constraintsmentionedby Maggie Baigent will become less. Lexis the merging of gru--u. unc vocabulary, or better still, a refusal to separate them _ has more to ofl,er learners than any sylrabus based on a limited list of structures and lists of single words.

sof rtth

7.14 Summary
We may summarise the position: given that we now have much better descriptionsof English than we have everhad before, and,thatthisrevealsthat many of the 'rures' previously taught are either wholiy pur,iuirv inaccurate, things in the classroommust change- no changer, ", un option. This is noi precisely the position proposed sorneyears ago in The Lexicar Approach, but in a recent article Scott Thombury (Moderu English Teacher;yor. 7,No. 4) complains that The Lexicar Approach does not have a coherent leaming theory. In many ways, this is true and the implications are taken up in the next chapter, a lexical.viewof language but doespoint to a numberof deficiencies in conventionalsyllabuses. Karr Fopper As hu, ,o conclusivelyshown,there is fundamental asymmetry betweenproof and disproof; we can never,even in principle, prove general statement, to be true, but we can, often with comparativeease,demonstratetheir untruth. change una p.ogr;rs are based on disproving the currently accepted view, prompting new theories and experiments,which will provide further evidence urri n"i" theories,which in their turn will be disproved. It is not by any meansclear how best to mcorporatelexical views into books or courses;at the sarnetime, teachersneed to be wilring to engagein mini_ actlon researchprogrammes,which the the holy grail of a ,comprehensive l, providing leamers with a more effecti, before. The improvement may be subr changeis not an option. This Lrookis specificallyabout teaching, rather than describing,collocation. we must, then, ask what rnay o. -uy not make material more or less teachable,and more importantly still, what aids or impedes leaming. So far we havelooked at the useof new descriptions of English but, rike Timmis, we shall suggest that there are considerable imprications for classroom methodology.That is the subject of the next chaoter.

rals nay the ghts I

i t I

ical tres and
I5 " . ' ' Oe

t t t

I t t


Language in the lexical approach

Discussion Questions
Do you usually teach new words alone, in collocations or in complete contexts?Why do you follow the procedureyou do? Do you think a different proceduremight be more effective or more efficient? Which of the following do you regularly draw to leamers' attention: new words, traditional (opaque) idioms, fixed expressions.grammar structures, collocations, clusters? 'Description is not pedagogy.'What implications, if any, do you think the phrasal nature of languagehas for your classroom?

Micl This

what meetl learn recog abilit cruci impot surpr Finall chanl chanX


Over us bet we n( impin thoug agarn the a 'teach know 'learn

and ar separz sortir and ct consic feedbr underr who v

Teachi experi challer know why sr

Learning in the lexical approach 155

rplete' :erent new lures, k the

Learning in the lexical approach
Michael Lewis
This chapter considers what we know what circumstances learners are most meet, both in class and outside. It intr learning. It emphasises the importanc recognised on many-.teacher trainirrg .oo.."r,' ;"Jil;r;;*'ri"'r"u"t ability to select and direct rearners'atiention to particurar r.i.ra. or exampres is ".,, ideas of syllabus and level. It emphasises the I mor: than once, and, perhaps to the r,e unimportance of controlled practice. understanding of learning suggests real and how collocation is central to these

8.1 Introduction
over the last ten yearsor so, the analysis of computer-based corporahas given us better descriptionsofEnglish than have ever been availabrebefbre but, as we noted in the previous chapter,it is not self-evident that thesedescriptions impinge directry on the languagecrassroom, and if they do, it requires careful thought to determine-howbest to modify cunent crassroomprocedures.once again,it is herpfur to begin by thinking ctearty about terminology, particularly the area covered by the broad terms 'knowredg"' unJ-:i"arning, and r two kinds of knowledge _ declarative

input from intake; identifythreJ

and complexity - which contribute tL the overall idea of ,level,. we shall also consider the non-linear nature of acquisition, and the rmptications of feedback for the acquisition process. Fa, f.om unnecessarytheory, a clear understandingof these ideas provides an essential framework for teachers who wish to develop their own understanding. Teachers sometimes dismiss theory on the grounds that they know from experiencethat something works. Henry widdowson trus irsr"o rwo potent challenges ro this posirion - firstry, even if-something *";il,;"* do you know something else wourd not wtrk better? Secondli ,f y; io not know why somethingworks, you may be unable to replicate the success, shareit or

sortin and cribing di, ;h ;hr'J;?:fi:r':x?11""; g, des we "., rinsri 1"",1?

separate processes involved*t



Leaming in the lexical approach

why someclassroom with others.So thereis a strongcasefor understanding activities seem to be more effective than others. Everything that happensin class should be consistent with what we know about the nature of both languageand learning; equally importantly,nothing which happensin class should violate the nature of either. The previous chapter looked at what we now know about language; this chapter looks at what we now know about learning i.n general,and languageleaming in particular. do It is comparativelyeasyto study what teachers in classand specify what particular activities are intended to achieve.It is, however, difficult to know what use learners, individually or collectively, make of the language they meet in class and almost impossible to evaluate the effect any particular activity has on learners' long-term language acquisition. However much teachers dislike the idea, the relationship between teaching and learning remainsmysterious.Some,like Krashen,even questionthe value of explicit leaming. Less controversially, it is clear that no teaching can guarantee it, has so eloquently expressed we acquisition.As Diane Larsen-Freeman that teachingdoesnot causeleaming. constantlyneedto remind ourselves We also need to lemind ourselvesthat teaching is never an end in itself; its sole purposeis to facilitate acquisition. Is traditional languageteaching likely to achieve this end? If not, what teaching strategiesare likely to be rnore successful? Teachertraining coulses often examine what the teacherdoes,but if we want to understand what is most effective in the language classroom, it is with learning,not teaching,that our analysisshouldbegin.

wror rna

Proc addr knol discr proc( we ci to lea decla such whicl in the slowl balan canno

8.2 Two kinds of knowledge
We look first at knowledge in its widest sense. Two different kinds of knowledge have long been recognised - declarative and procedural knowledge. The first is knowledge that, and involves stating facts or rules the the date of the Declaration of Independence, exchangerate of the pound go. The second knowledge how to, is againstthe dollar, the pastparticipleof - selve at tennis, drive a car, give a short, the ability to actually do something witty speechof welcometo a group of visitors. The two kinds of knowledge are different in important ways. With declarative knowledge,you either know it or you don't; you can rememberit (correctly or incorrectly) or forget it; you need to look it up or be told it, directly or indirectly, by someone else; there is nothing to understand, it is simple from eachother item. Importantly, the lack of information, eachitem separate a single piece of such knowledge may be frustrating, or make you look slightly silly, but it will not render you unable to do what you want - if you say Ever since the Declaration of Independence,whenever that was, America has . . . ; or I think the companyhas goed on the seventeen whatever,

The tr we ac Indep at thal is dec knowl excha the cu but un a grafit this la: or you


H kr

Declarz With a vocabu system modifie positior comple gramma



Learning in the lexical,approach


sroom ens in I both t class rat we about '' what know e they ticular much ammg xplicit rantee
It, we lng. "^lf; its likely I mole o want s with

wrong line this year . .. , your message may be badly expressed, you still, but in a more global sense, achieveyour purpose. Procedural knowledge is about global ability; each bit of learning is not added to what you already 'knew', but is integrated into your earrier knowledge, modifying it in some way. procedurai knowleoge rs not simple discreteitems, but, as its name suggests, sets of comprexpiocedures.New proceduralknowledge,once properly acquired,is not;forgttten' in the way we can forget a date or a new word. Lack of procedurarknowredge is likely to leave you unable to do something;if you cannot ride a bike, no amount of declarativeknowledge about how to ride a bike will help. you cannot look up such knowledge,and no one else can tell you or explain it to you in a way which ensures you wil|know'it. watching someone elseride may help; but in the end, you have no altemative but to get on, try, fall off, try again and, slowly, you will acquire an integrated set of skills which are to do with balance,speedand so on. once you have acquiredthe ability to ride, you cannotforget it - it is yours,part of you. The two kinds of knowredgeare not totally separated, the but ways in which we acquire them are. In the examples earlier, the date of the Declaration of Independence declarativeknowledge, why America decrared is independence at that time is complex procedural knowledge; learning why in High School rs declarative;understanding why and relating that undersianding to a wide knowledge of American history and politics is procedural. similarry, the exchangerate on any day is a matter of fact; understanding the movementsof the currency market is procedurar;the past participre of go is a simple fact, but understandinghow it is usedfluently and accuratelyis procedural. Stating a grammar rule is declarativeknowledge; the ability to useit is procedural. As this last examplemakesclear,you can 'know,the rule but be unableto use it or you may have masteredthe point without being able to state the rule.

rds of :dural rrles pound ow to, : short, arative ffectly ;tly or simple lack of r look if you I was, on the

Can you give three examples of procedural knowledge you have and explain how you acquired the knowledge? How does this knowledge differ from the ways you reamed decrarative knowledge?


Learning in the lexical approach

intrinsically procedural. Any discrete bit of language which is leamed purely additively cannot contribute to, indeed is not part of, the leamer's 'known', it is not availablefor use. mental lexicon; althoughin some sense This immediately brings to mind Stephen Krashen's distinction between leaming and acquisition. Learning and acquisition ln The Natural Approach Krashen introduced the distinction between language learning, which is conscious, and language acquisition, which is unconscious. He has controversially claimed that only language which is unconsciously acquired is later available for spontaneoususe. He claims and leaming hasno value,as what is leameddoesnot acquisitionis essential, contribute to what is acquired.We shall examine this view in somedetail, but 'learning' is used only in the senseof what for the rest of this chapterthe term 'acquired' is confined to language is consciouslylearned;similarly the term to which the learner has immediate accessfor purposesof comprehensionor productiveuse. If Krashen is right, formal teaching, which is explicitly directed at conscious learning, is effort wasted. Even if he is wrong, and formal presentationand practice of specific items doesaid acquisition, our new awareness the sheer of problems. Any suggestion that size of the mental lexicon raises immense 'teach' a lexicon which runs to many tens of teachers could formally many hundredsof thousands)of thousands(or, for competentnative speakers, items is clearly unrealistic. If each of 20,000 items took 2 minutes to teach, that is already over 600 classroomhours, more than the total duration of many learners' entire formal languageinstruction. This new understandingof the size of the learners'lexical task implies radical changesto the teacher'srole. Either teachersmust selectand teacha restricted lexicon - but on what criteria, for studentsof generalEnglish? - or they must adapt classroomactivities so that, rather than teaching individual items, they provide learners with strategies which ensurethe leamersget the maximum benefit from all the languagethey meet in and, mote importantly, outside the formal teaching situation.

unde exam - lan menti meetr 'knov

so tha

Not r' a less that 1, you \\ and u'

Comn fundat languz empha the orr right i except produc all tim mlsun( produc

Acquis knowle make ,

routini: invoive lexicon From i

8.3 Acquisition and noticing
The basic position of all the contributors to this book partly agrees with Krashen's position, and partly modifies it (in a way he would not accept). Acquisition is acceptedas of central importance, but it is suggestedthat the conscious noticing of features of the language that learners meet does facilitate acquisition. These ideas need to be explored in more detail. Acquisition and input Krashen's claim that we acquire language in one and only one way, by

In addi further The cer only on there m languag to for s1 linguisti 'noticinl

is perhq

Learning in the lexical approach 159


: use. ween

ween ch is ch is laims )snot 1,but what ruage on or clous r and sheer L that rs of ls) of each, nany Ldical rcted must they mum le the

understanding messages,provides a clear starting point from which to examine our presuppositionsabout how leamers do learn - in the loose sense - language.His position is that a learner's interlanguage(the learner's total mental representationof the target language at any moment) is modified by meeting new language which lies on the edge of what the learner already 'knows' in such a way that it is incorporated into the learner's interlanguage so that it is availablefor spontaneous use. Not very long ago languageteaching emphasisedgrammar structuresand to a lesserextent vocabulary('new words'). The fundamentalassumption was that you first neededstructuresand, having masteredsome central structures, you would move from accuralebut halting production, to more fluent speech and writing. communicative approaches rightly tumed this system on its head. The fundamental emphasis of communicative approacheswas and remains that language is about the expression and communication of meaning. This emphasison 'communicating'inevitably values fluency above accuracy,so the order of priorities is reversed.while this is unquestionably a step in the right direction, it has one unintended side-effect, unless the teacher is exceptionally careful - it places great emphasison the languagethat learners produce, so it has a tendencyto encourageproduction, particularly speech,at all times, even in the earliest stagesof a course. At the risk of being badly misunderstood, I must point out that you cannot acquire a language by producingit. Acquisition involves taking in new material and incorporating it into the knowledge or skills you already have. Producing language - speech- may make you more confident or may make your speech more automatic or routinised, but that is not the sameas expandingyour languageresources; that involves integrating new language into your intergrammar and mental lexicon. From input to intake In addition to introducing the leaminglacquisition distinction, Krashen has further claimed that there is only one way in which learnersacquirelanguage: The central hypothesis of the theory is that language acquisition occurs in only one way: by understanding messoges. (The Naturar Approach). while there may be much truth in this, it is also true that if you wish to turn the languagelearnersmeet - input - into languagethey acquire and have access to for spontaneoususe - intake - they almost certainly need to notice the linguistic wrapping in which the messageis delivered. Exactly what this 'noticing'might involve, and what may help or hinder input becoming intake, is perhapsthe most important of all methodological questions.

with :ept). lt the does

y, by


Learning in the lexical approach

Every teacher knows that some of what you teach seemsto be acquired very easily by leamers but some things that you teach again and again still cause problems for learners. What factors do you think influence whether input becomes intake? Do you think it depends mostly on the input language or mostly on the leamer's current knowledge? How important do you think factors such as motivation, tiredness, age or the temperature of the room play?

mp lan

Iri h"lt is tl pro inta


In this chapterwe are mostly concernedwith the kind of languageinput which is needed. It is important to realise that the input which is used in the comprehensionof the messagemay differ from the input which is the raw material for the acquisition of language. Many applied linguists and most teachersbelieve that, at least to some extent, focussing leamers' attention explicitly on some aspect of the linguistic form of the input is helpful in acceleratingthe acquisition process.We need to examine this belief in detail. You probably make a daily journey from your home to your place of work; the route is completely familiar, and you could give someoneelse directions for the journey. But do you know the names of all the streetsyou drive or walk down? In all probability you know the route, but you have simply not noticed the names of some of the sffeets- they are irrelevant when you can achieve your global purpose without attending to such details. The global purpose of languageis the communicating of messages; the medium for but - words and phrases - which may need to be doing it is language items noticed if they are to be acquired. In normal language use, we are usually so predisposed to focus on the message,that the languagein which it is delivered is frequently ignored, or, if presentedin writing, transparentto the point of being invisible.

Sec prec broa therr the voca For r the r, are c stude such mem

Notic the te also t some occas Awarr wary l in the expen

What sorl of language which would be useful from an acquisitional point of view do you think your students might fail to notice unless you provided guidance?

Experiments have shown that even quite advanced and motivated learners often do not seethe difference between their own effective but inaccurateor unnatural language and a similar correct, natural version which expresses exactly the samecontent. If they do not notice - seeor hear - the differences between the languagethey used to expresssomething and the correct natural version expressing the same content, then that input cannot contribute to intake. Activities which encourase learners to notice certain features of the

Given helpfu (witho unders kinds < This is langua

Discus Freeme

Isev will ltem valu

Learning in the lexical approach


mput probably contribute to the value of the input specifically from the language acquisition point of view. It is essentialto remember,however, that the belief that deliberate noticing helpsis by no meansan established certainty;the current-uin.t."u- position is that it probabry has at least a facilitative, helpfur effect. Explicit noticing is a necessary, nor sufficienr condirion bur ro ensurethat input becomes ,or?J.::tt

8.4 Noticing
rhich r the raw most rtron ul in etail.

irons /e or i not r can lobal l for obe L the i, or,

Second Language Acquisition researchers are somewhat divided over preciserywhat factors influence what part of input becomesintake. There is a broadconsensus ranguage that that is not noticeddoesnot b""o-" intake,but there is no agreementon the precise meaning of the word ,rroir."o,. Even in the most traditional grammar-orientated crassrooms, leurn".s acquire vocabularywhich must result from accidentar, or at reastincidental,noticing. For example, while ostensibly studying a structure,rearnersacquire some of the vocaburaryused to exemplify and p-ractise the grammar.Equally, teachers are only too aware that formally teaching a number of *ordr, or requiring studentsto 'learn these words for homework' is not sufficient to ensure that such items w'l be committed even to short-term, much ress long_term, memory. Noticing is not quite the straightforward matter it might seemon first meeting the tem' In everydayuse,the word can refer to both accidentalawareness and also to the results of.deliberate focussing of attention. It is also the casethat sometlmeswe are able to recall what we accidentally noticed, ,'t il" on oth", occasronswe cannot recall something to which we paid a"tiu.rut" attention. Awarenessof the potentially wide mJaning of the word, shoutc mate us very wary about attaching too much importanc! to any particular kind of noticing in the languageclass.As always, caution and an open-mindedwillingness to experiment, and revise our vlews on effective methodology,is essential. Given the presentstageof our knowledge of acquisition,it is likely to be helpful to make learners explicitly u*urJ of the Lxical nature of language (without using that terminology). This means helping leamers deverop an understandingof the kinds of chunks found in ttre teits ,rr"y -""r, and the kinds of prefabricatedgroups of words which are the prerequisite of fluency. This is one part of the teacher'stask in encouragingrearnersnot to break the languagethey meet down too far. Discussing value of instruction,of the which noticing is a part, Diane Larsen_ Freemancomments: [severalresearchers] have pointed out that explicit grammar instruction will not likery result in immediate mastery of specific grammatical items, but suggestnevertheless that explicit instruction does have a value, namely, facilitating input.


le or

nces lural eto : the


Learning in the lexical approach

'grammar' instruction, they Although her comments relate specifically to Sulelyapplyequallytoinstructionwhichensufesleamersnoticeanykindof patteming in the inPut theY meet. Sorting and describing the A word of caution, however. There is a world of difference between you notice?Being teacheraskingDid you notice . . . ?, andasking What did different from able to describe - verbalise - what you noticed is completely


{ 1 I

Aft< mln I
I ll



giving the (supposed)rules. It would be a tragedy if further time was wasted u"rtatitlttg complex descriptionsof lexical patteflrs,especially when there is no evidence that such descriptions would help acquisition. It could well be that concentrating on such descriptions is an activity which may appeal to teachers, but which is of no benefit to, and indeed may intimidate and confuse, learners.Noticing language helps; sorting it into categoriesor pattems may help (see below); describing the categories almost certainly doesnot. Directing learners' attention Despite any doubts about precisely how noticing helps, it is safe to say that learners frequently do not notice the precise way an idea is expressed' sometimeseven if their attention is drawn to it' Some training in the sorts of chunks which make up the texts they read or hear increasesthe chance of them noticing useful language, rather than many other features which are irrelevant from an acquisition point of view. In A Cognitive Approach to LanguageLearning, PeterSkehanobserves: Input contains many alternative features for processing, and the learner's task is to extract relevantfeatureswhich can then be focussed on fruitfully. ...Instructioncan work ...by making salientless obvious aspectsof the input, so that it is the leamer that doesthe extracting and focussing, but as a function of how he or she has been prepared' Reporting a major study of noticing by Richard Schmidt, Skehancontinues: of The consequence Schmidt receiving instruction was that what had been unstructured, undifferentiated input (but whose nonunderstanding had not impeded understanding very much) became noticeable and analysable,leading to future progress' in In his article (The Role of Consciousness SecondLanguage Learning, Applied Linguistics Vol. 11 No.2) Schmidt points out the crucial difference

l( 1e r(

fc tn is


bt bi in

In su teach whic learn input


Noti langr

Learning in the lexical approach


they rdof.

r the ieing from iome new you I you Jless at ls, lsted :re is li be al to and )s or ainly

information that is noticed: When reading, for example, we are normally aware of (notice) the content of what we are reading, rather than the syntacticpecutiaritresor the writer's style,.ihe style of type in which the text'is set, music playing on the radio in the next ioom. ...However, we still perceive thesecompeting stimuri and may pay attention to them if we choose. After a long discussion the role of conscrousness of he concludes (the bold is mine): I have claimed that subliminal language learning is impossible and that intake is what rearners notice. This'requirement of "ottr"io.,Jy noricing is meant to equalry to all aspect, or rungffi (lexicon, lqlv phonology, grammatical form, piagmatics)-, and can u-" ii"orpo.ur"o into many different theories of second language acquisition. ...What learners notice is constrained by a number of factors, but incidental leaming is certainly possible when task demands focus attention on relevant featuresof the input. ...Incidental leaming in another sense,picking up target language forms from input when they do not carry information crucial to the task, appears unlikely^foradults. paying attention to Ianguage form is hypothesised to be facilitative in aillases, and -t;-";""essary for adult acquisition of redundant grammatical features. ...Recent psychologicaltheory suggesrs that implicit learning t best characterized as the gradual accumulation forriOr", Urt i, of associations between frequentry co'occurring features, rather than unconscious induction of an abstractrule system. rut which they do not understand,but rg leamers toward the input language onal point of view; the more aware text is made, the more likely that the (e.

between information that is perceived and

,that ssed, rts of :e of I are :h to te :d

8.5 The importance of examples
Noticing examples of language in context is central to the acquisition of language, which raises the difficult question of what we mean by good examples. Tasx You arepresumably fairly confidentthat you can identify a chair whenyou seeonebut canyou define,uihuir,Z Do you think your definition is precise, so that it includesat chairs and excludesbenches, stoolsand other things you might sit on?

rd rues: rd n-

ning, rence


Learning in the lexical approach

'chair'than othersand Somechairs seemto be better examplesof the category between you almost certainly found while doing the task that the boundaries closely at what are ,ilnlt* categories alwaysfttzzy.It is worth looking more 'good examPle'. we meanby a categorisationwas until the mid-2oth century, it was generally assumedthat aproblem-freeprocedure'butitisnowrecognisedasaprincipalsourceof the later work confusion and error in many disciplines. Philosophers,notably Previously it was of Wittgenstein, amply demonstrated why this is so' a group of characteristics; assumedthat all the members of a category shared or class was thus a these could be listed, and membership of the category could be made' matter upon which definite and indisputable decisions considering the concept of that this was not so by wittgenstein demonstrated 'a game'. What is 'a game'? patience,solitaire' chess'poker, golf, Most of us happily agreethat SoCCer' games. It is difficult, baseball and Tomb Raider (a computer game) are all satisfy all the probably impossible, to make a list of criteria so all the games 'games'' to call all the activities criteria; despite this difficulty, we are happy the members of a The difficulty arises from our initial assumption,that all In fact, a class is classmust shareall the defining characteristicsof the class. morelikeafamily;allthememberssharesome,perhapsmost,ofalistof d e i i n i n g c h a r a c t e r i s t i c s : i t m a y e v e n b e t h e c a s e t h a t t h e'game" t s h a r e o n e o r ymus for with to more or the characteristics count as a member of the class. are rules, it is played between example, criteriainclude: you can p|ay it, there score' a two individuals or teams,it is played in a specialplace' there is a one result, and a winner. We immediately see that some games' which no more of thesecriteria. The implication doubts are games,do not fulfil one oI is that not all members of a category or class are equal; some are better 'game" perhapssurprisingly, exemplarsof the classthan others.In the caseof the only criteria all games seem to fulfil is purely linguistic, you play all of themaccording to rules; the only necessarycharacteristicsare collocational! In general,a classis a collection of items all of which shareenough of the list of defining characteristicsto count as members of the class; but some are better examples than others, some of the characteristicsmay be essential, some important but not essentialand some of relatively marginal importance' The most important fact we need to note is that membershipof a class is not the simple matter earlier analysessupposedit to be' In the early days of the Cobuiid project Sinclair wrote: do grammatical generalizations not lest on a rigid foundation, but are of the patterns of hundreds of individuai words and the accumulation phrases....The main simplification that is introduced by conventional grammar is merely the decoupling of lexis and syntax'

Pete corI bef toa gen( lmp( be tr wide wec exan coml exan cateE

The i justil of bc

Lang Sincli pailer


and ri differr

Unfor suppo more needa often :

1 2 3 4. 5. \\

Learning in the lexical approach 165

s and vr/een, what t was ce of work was stics;
lUS a

nade. :pt of

Peter skehan endorsesthis view when he suggests that the more we examine corpora, the more we realise that the generalisations which we thought could be formulated as 'rules' tend to be more parlial or more restricted - perhaps to a particular genre - than has usually been believed. colligations are parlial generalisations; collocations exhibit pattems,but the blockeJcollocationis an important reminder that even quite modest linguistic generalisations need to be treated with caution. Description of a category - a rule rarely apply as widely as you first think. Learners (and teachers)need to acceptthat the best we can hope for are helpful but provisional descriptions. Similarry,,typical, examplesneed to be given with both care and caution. The languageis more complex, with more sub-palterns than we like to think, so truly typical examples are elusive, unless they are typical of only a very small subcategory. The difficulty of choosingtypical examples, shouldnot, however,be usedto justify inventing examples- as Sinclair has remarked, one does not study all of botany by making artificial flowers.

golf, icult, ll the mes'. ; o fa

far aspossible,teachers shouldtry to avoid inventingexamples in class." you agree? Do What justification can you offer for inventingexamples? Language examples for learners Sinclair has pointed out that, parado patterns are, in fact, difficult to ident frequent.This is because words ar the and what is a good example from on different types of texts. Unfortunately, languageteachinghas a history of inventing examplesto fit the supposed rules.The more we know abouthow ranguag"i, a"tuaity used,the more unsatisfactory this processseems.It is increasinglyclear that teachers need a sensitivity to examples,so they can direct l"arrreisi attention to natural, often spoken,exampleswhich are most rikery to promote acquisition. 'As

ist of
Ine oI

:" for .ween
lle, a

) one

better ingly, all of ional! helist Ie are :ntial, lance. ls not of the

Write down each of the following: i. a good example of the present continuous. 2. a good example of the present perfect. 3. a common expressioncontaining I'm going to. 4. some verbs that go with the noln exam. 5. some examples to show the difference between speak, talk, say. What makes you think that your examples are good examples?

nd aal

166 Learning in the lexical approach

In doing the task, did you go and find real examplesor did you invent them? After Sinclair's warning, surely you didn't invent them? If you did, what makes you so sure that they are good examples?It is almost certain if you that they are very poor invented examplesconsisting of a one-clausesentence examples,particularly those which exemplify grammar structures.It is also very likely that for the structure examples you chose untypical vocabulary, and for the vocabulary examples you chose untypical grammar. If this judgement seems harsh, use a computer concordance and compare your examples with their authentic examples. Even with authentic examples, howevet, choosing exampleswhich are the most appropriatefor a particular group of leamers is not easy.You will become better at it with experience particularly if you pay attention to the co-text which surroundsthe language point you are interestedin. The importance of real examples also has implications for what language and how this shouldbe done. shouldbe recordedin learners'notebooks Recording examples From a classroom point of view, we have to remind ourselves that collocations are not words which are put together, they are words which 'build' collocations in the language naturally occur together; when we classroom the process is artificial, the reverse of how language is used in This meansthe teacherneedsto be alert to the fact that normal circumstances. the larger the unit she can identify, and which leamers can be encouragedto notice and fecord, the more likely it is that this languagewill becomepart of the leamers'intake complete with certain grammatical features, accessible for future use. If the intake is accurately noticed and stored, certain grammatical eIroIS afe likely to be eliminated, or at least their number reduced,in the learners'own output. When language is used naturally, grammar and words interact in complex ways; both teachersand learnershave a sffong tendencyto take new words out of context - the emphasis is on understandingthe meaning. But taking words out of their natural context leavesbehind important information, not concerned with what the word means, but with how the word is used. Every word has its own grammar, and, therefore, separatinggrammar from vocabulary is unwise. Sinclair puts the position clearly and strongly: Many of those people who are professionally engaged in handling languagehave known in their bonesthat the division into grammar and vocabulary obscuresa very central area of meaningful organisation. The decoupling of lexis and syntax leads to the creation of a rubbish 'idiom', 'phraseology','collocation'and the like. ... dump that is called The evidence now becoming available casts grave doubts on the wisdom of postulating separatedomains of lexis and syntax'

Peo mu( co-c regu


mad easil ASto ofm Exal

If tei autor route
+ noi teachi

gramr then r much Itisd leame somet kind 1 would There langua that u,i acquirr immed rush ht Stop fc have an idea is product somebo which I Assumi languag (abstrac generali has me observa

Learning in the lexical approach


nem? what, I you poor r also rlary, ' this your rples, icular lence Iuage luage

People are most comfortable in the company of friends and acquqintances, much less so surroundedby strangers;words are very similar. words do not co-occur in random ways, but in pattems which exhibit varying degrees of regularity and generalisation.Around each word there is, in sinclair,s phrase, 'meaningful organisation'. This meanslanguage doesnot needto be _ indeed should not be - 'cleaned up' before being recorded by learners and, if for some reason the teacher manufactures an example, every effort should be made to ensureit is a natural example which simulatesreal use,rather than an easily-madebut highly artificial EFL example with which we are all familiar. Astonishingly, the best known students'grammar appearsto consist entirely of manufacturedexamples. Examples given by the teacher If teachers find it necessary to create an example in class, it should be automatic for them to create a context for the use, probably following the route from word (usually a noun) to collocation - perhapsverb + (adjective) + noun - to a complete phrase which evokes a situation, and then for the teacher to draw attention to the most typical group of words, including grammatical collocations such as prepositions and articles, which leamers then record. Storing phrasesin this way builds the learners'mentallexicons much more systematically than simply learning ,new words'.

that vhich luage ed in rt that ;ed to art of ssible rs are 'own act m e new ;. But Lation, used. 'from Ing Lnd ish

rush had died down.

somebody to do something ne generalisations from real-world examples which become abstractions which most of us have never actually heard. Assuming that such generalisationscan be taught as the basis for innovative


Learning in the lexical approach

8.6 Acquisition is non-linear
If understandinginput and noticing elementsof the input sufficiently often are the two defining characteristicsof the way we improve our linguistic ability, there are radical implications both for syllabus specification and classroom methodology.We consider first how the nature of acquisition challengesthe conventionalidea of a sYllabus. Non-linear phenomena was, tool mathematics, andparticularlyof its essential The history of science, This was until20 yeals ago, exclusivelythe story of solving linear equations. inevitable, as linear equations could be solved by analytical methods, while non-linear equations could not. The advent of computers meant non-linear equationscould be solved and a whole range of new questionscould therefore be tackled. Many phenomena such as population changes' the spread of diseasesand predicting the weather ale now recognised as non-linear; nonlinearity seemsto be nature's notm. Before the computer revolution, mathematiciansproduced idealised versions of a problem so that the equationswefe linear, and could therefore, at least in principle, be solved. Often the answels were good enough to be of enofmous practical use, but, and this is the crucial point, although they were powerful and useful approximations, they wele not accurate descriptions of nature' Mathematics professor Ian Stewart describes the situation in mathematics until very recently as follows: In classical times, lacking techniquesto face up to non-linearities, the process of linearization was carried to such extremes that it often were being set up. Few askedtbemselves occurredwhile the equations what the long-term future might be for a method which - to be brutal 'Give me an answer!'is the demand.So solyesthe wrong equations. linear theory obliges, hoping that nobody will notice when it's the nonscienceshowsthat natureis relentlessly wrong answer....Today's linear. ...So ingrained becamethe linear habit that by the 1940s and knew little else' and engineers 1950smany scientists Traditional teaching and syllabuseswere linear. Later the syllabusesbecame 'cyclicaf in the sensethat some gfamma.rwas explicitly visited more than once,but this is only a different form of linearity - a predeterminedseqluence, followed by the teacheron the assumptionthat the leamers are following the same sequence.Like the scientists of the 40s and 50s, language teachers wanted a syllabus, and could hardly imagine anything other than some form of linear syllabus; after all, teaching is linear, going from lesson to lesson, week to week, term to term. But what if acquisition, which after all is another natural process,is fundamentally,irredeemablynon-linear?


The whi bec by tl syst syst proc

Cha and surp pred

with wee sma

It nc wha what whic Ever forgr whal linea

Som such I don bir f 'undr


Amr coult unde unde alonl Simu word past l these existi and t< leamr mem(

Learning in the lexical approach


Feedback means acquisition is non-linear
ln ale

bility, room :s the

, was,

while iinear refore ad of nonrsions )astm mous l'erful ature. natics :he ien

lSo rhe )nmd lcame e than lence, ng the achers : form esson, nother

The simplest way to get an idea of non-linearity is to think of systems in which feedback is an intrinsic part of the phenomenon- objects slow down becauseoffriction; the speedis affectedby friction, but the friction is affected by the speed,which is affectedby the friction and so on. The behaviour of any systemwhich has in-built feedbackis difficult to analyseand predict, but such systemsare all around us, and languageacquisition is self-evidently a natural processin which feedbackplays a centralrole. chaos theory is the modern disciprinewhich studiesnon-linearphenomena, and it provides some surprisescomparedwith classicalscience. Among its surprises is one of particular interest to us - although you may be able to predict the macro-behaviour a system,you may not be able to predict of the micro-behaviour of the samesystem.we can, for example,predict the climate with considerable accuracy, but we cannot predict the weather fbr next weekendwith anything like the samecertainty; the big picture is clear, but the smaller,more local picture is much less subjectto accurate prediction. It now appearsincontrovertible that acquisition is a non-linear phenomenon_ what you acquireis a function of the intergrarnmaryou have alreadyacquired, what you notice in the languageyou meet, which modifies your intergrammar, which affectswhat you notice and so on; feedbackis intrinsic to the process. Even this description is oversimprified, for we must add factors such as forgetting, misunderstanding, successfulguesseswhich were not based on what the learnerhad fully acquired,and so on. If acquisitionis non-linear,no linear syllabuscan be adequate. Sometimesteachingis conductedon the (covert) assumptionthat examples such as: It could have been a lot worse, if it had happenecl during the night. I don't understand how it could have takenthreeweeksare assembled bitrby_ bit from a knowledge of bits of grammar such as ,uses of could, and 'understanding the present perfect' but this assumptionis, at best, very questionable A more plausible explanationis that the individual learner meetsa number of could have + past participle examples and understands,or partially understands examples of that colligation used in context. Slowly, an understanding of the nature of colligation itself begins to develop, and alongside this, increasing awarenessthat it can be broken down. simultaneously, the leamer has a developing understandingof other multiword chunks which involve could, and other chunks which contain have + past participle.The leamer begins subconsciously analysesome or all to of these;over a period, with both increased understanding and backslidingcoexisting, the leamer acquires the ability to analyse several different chunks, and to syntactisise, that, in due course,there is a permanentchangeto the so learner's interlanguage.Gradually, languagewhich was part of the formulaic memory-based element of the leamer's knowledge is transferred to the


Learning in the lexical approach

analytic, rule-basedpafi, thus becoming available to genelatenew language basedon syntactisisation.At this stagereal acquisition has taken place. This non-linear model, which is almost certainly still an over-simpliflcation, is much more likely to representacqursition than any linear, oI even cyclical 'everything affects everything else', progression.It could be summarisedas gradual ro ii is as far aspossiblefrom any linear model. This kind of model of in many but permanentchangein complex phenomenais curently discussed fields. In the study of language itself, it is now a commonplacethat when a language is used, it changes; the phenomenon can be summarised in the phraseWhenyou play the 7ame,you changethe rules' As Ian Stewart reminded us, fol centuries mathematics solved the wrong in equationsbecauseit clung on to linear models of phenomenawhich were' fact, non-linear. It now seemsincontrovertible that acquisition is a non-linear phenomenon, so only a non-linear model of acquisition has any chance of The implication is that linear teaching can neverbe representingit adequately. with non-linear acquisition, explaining Larsen-Freeman'sdictum "ong-"rrt quoied earlier that teaching does not cause learning. She discusses the relationship of non-linear systems and language acquisition in an extended paper (Ciaos, Complexity,Science and Second Language Acquisition in the Applied Linguistics,vol. 18,No.2). Her commentslargely endorse above: The purpose of this article is to call attention to the similarities u-ong complex non-linear systems occurring in nature and languageand languageacquisition. while the value of the analogy 'you don't seesomethinguntil may be only metaphoric, sometimes you have the right metaphor to perceive it'' It is my hope that learning about the dynamics of complex non-linear systems will discourage reductionist explanations in matters of concem to secondlanguageacquisition researchers' Further, learning linguistic items is not a linear process learners do not master one item and then move on to another.In fact, the leaming curve for a single item is not linear either. The curve is filled with peaks and valleys, progressand backsliding' 'rules'often represent As we saw in the previous chapter,traditional grammar noted by gestaltpsychology' similar to the phenomena over-seneralisations,

toti lgn atte oftt rev Litt



The po\r des


SylJ rntr( undr the r

Our com lean


onr Wirl help

Itm prac wou appr strea appr cont(

'see' the above In order to process information, you will be tempted to diagrams as a circle and a square,but careful observationshowsthey are both incomplete. In order to process information we have a natural tendency to

Com form gene No., says

Learning in the lexical approach l7l

uage tron, :lical :Ise' , Ldual nany ten a I the 'fong re, 1n inear ;e of er be ctum s the rnded )n tn

totalise, and thereby simplify our perceptions.This is highly efficient, but it ignores details and implicitly assumes that the details are unimportant. In an attempt to explain everything with a few big generalisations,grammar rules often encourageus to ignore variations which corpus linguistics increasingly revealsare important featuresof how the languageis actually used.In Corpus Linguistics, Doug Biber, reporting the results of a massive corpus-based programme,observes research : ... a finding that is common in corpus-based research:that overall generalizations of a language are often misleading, because they average out the important differences among registers. As a result, overall generalizationsare often not accuratefor any variety, instead describinga kind of language that doesn't actuallyexist at all. The kind of over*generalisation familiar in grammar is very reminiscentof the powerful, but ultimately unsatisfactory mathematical simplifications describedabove by Ian Stewart.

8.7 Which is fundamental- lexis or structure?
Syllabuses were traditionally structural, and later the multi-syllabus was introduced, usually basedon grammar,vocabulary and skills, but our current understandingof languageand learning suggestswe may need to re-evaluate the role of grammar in the syllabus. Our present understanding of the sheer size of the mental lexicon of a competentuser of English is deeply dispiriting from the point of view of the leamer (or teacher) of English as a second language. It seems the leamer needs not several thousand words, but at least tens of thousands of combinations of words and mini-patterns.The task seemsoverwhelming, the more so when we considerhow much language productionseems be based to on memory, rather than the ability to generate from a few general rules. Within this framework it is essential to re-evaluate what is both oossible. helpful and efficient in the classroom. It must be immediately apparent that any attempt to formally teach and practise the lexicon item by item is impossible, and any attempt to do this would completely overwhelm learners.Fortunately,this is not what a lexical approachsuggests. Traditional grammar teaching, with a strong behaviourist streak, emphasisedrepeated practice as a way of fixing pattems; a lexical approach suggeststhat it is repeated meetings with an item, noticing it in context, which convertsthat item into intake. Communicative competence is dependent on two parallel systems, a formulaic exemplar-basedone and an analytic one, based on generative generalisations 'rules'. Scott Thornbury,(Modern EnglishTeache4Vol. 7, or 4), joining the debate on how best to implement the Lexical Approach, No. says:

resent iogy.

above e both 1cy to


Leaming in the lexical approach

A lexical approach provides a justification for the formulaic, unanalysedffeatment of a lot more language than has been the case since the advent of the high-analysis era. ...Clearly, the Lexical needsto be undeftaken, more research Approachis work in progress... to the part memory plays in second-language particularly with regard learning, and whether (and under what conditions) memorised language. becomesanalysed language He is correct in asserting both that little is yet known about what turns unanalysed language into analysed language, and that this is an important question.What does seemclear is that any analysisperformed by the learner is basedon inductive generalisationon the basis of languagewhich is already part of the learner's unanalysed intake, rather than formal descriptions or rules. Earlier in the samearticle, he criticises the Lexical Approach: Phrasebook-type learning without the acquisition of syntax is is ultimately impoverished....Fossilization likely to occur when the on learnerbecomesdependent lexicalisedlanguageat the expenseof engaging the syntactisization processes. ...In short, the Lexical Approach lacks a coherenttheory of learning and its theory of language is not fully enough elaborated to allow for ready implementation in terms of syllabusspecification.


ml qu



of sta cot tha en{ the cor


Not acq


Imagine an intermediate learner talking to a native speaker who has no experience of teaching her native language' Now imagine yourself talking to the same learner. How would the language you use be different from that used by the inexperienced native speaker? Which would be of most benefit to the learner? Why? If you have been teaching for some time, do you think you have got better at changing your language to make it more useful to learners? If so, how have you changed your language? Do you think the language you use to learners is very precisely targeted or do you use a rather broad range of structures and vocabulary? How important is paraphrasing and recycling new words and collocations in helping turn the input you provide into intake?

The natr the Acc

It is an act of faith to assumethat it must be possible to specify a syllabus in linguistic terms; indeed, because of the non-linear nature of acquisition discussedabove,I do not believe it is possible to do so in any other than very broad terms - the input needsto be (largely) comprehensibleand the learners need to be engagedrather than intimidated by the languagethey meet. It is, I fear, difficult to be more specific than that, bearing in mind that learnersin the

Perl chal rulei the fluer prefz adeq a prc work a sul fluen long achie their comtr the gr
-mal accel

Learning in the lexical approach 173

l I

sameclass are all at different levels and that they make differential use of any input. This suggestsKrashen is broadly right in suggesting that it is the quantity of roughly-tuned input which is the key to acquisition, and that this is in itself the best we can do in specifyinga linguistic syllabus. It is also worth reminding ourselves that most learnersof a secondlanguage never progressbeyond some sort of intermediate level. In other words some of their languageis, and always will be, fossilised- a repertoireof useful standard prefabricated items together with other language which may be communicatively effective even if it contains 'grammar mistakes'. The fact that their syntactisisation processeshave not been fully and successfully engaged,rather than invalidating the Lexical Approach, merely acknowledges the inevitable. combinationof islandsof reliability'which canbe usedwith A confidence [see p 1751, and language which is communicatively effective even if defective, is surely better than anything any structure-based approach has been able to offer. Not only does the lexical nature of language and the non-linear nature of acquisitionchallengereceivedviews of syllabus,however,it also challenges many widespreadideas about methodology.It is to thesethat we now turn.




:ady sor

8.8 The lexical challengeto methodology
The challenge to conventional syllabusesis based mainly on the non-linear nature of acquisition, while the challengeto methodology is basedmainly on the lexical nature of language. Accuracy and fluency Perhapssurprisingly, the lexical nature of languagerepresentsa considerable challenge to conventional language teaching. Traditionalists value grammar rules and accuracy,believing more or less explicitly that fluency results from the ability to construct first accurately, then accurately and increasingly fluently. An acceptance and understanding of the enormous number of prefabricated chunks of different kinds, implies fluency is based on an adequatelylarge lexicon, and that grammar 'rules' are acquiredby learnersby a processof observing similarities and differencesin the way different chunks work. This reversesour traditional understandingcompletely - first you need a sufficiently large number of words and larger chunks; this allows some ('rules'). This situationcontinues a very fluency and somegeneralisations for long period, and thoserelatively few second-language learnerswho do finally achievea very high standard- that is, they achieve 'accuracy'- do this late in their leaming careers as a result of being able to break down chunks into components, reassemble in novel ways.This involvesboth respecting and this by the generalisations represented the rules and avoiding over-generalisations - many sentences which are grammatically well-formed are not sanctionedas acceptableby native speakers,so accuracy involves knowledge of what is

ls ln t10n very ners is, I r the




Learning in the lexical approach

'ought' to be but, in fact, is not. The inevitable sanctioned and what conclusionis that accuracy is based on fluency, not, as was believedfor so long, the other way round.



8.9 What do we mean by 'level'?
Fluency and accuracy have traditionally been seen as the two components '1eve1'.Communicative approaches,which are the within the idea of unquestioned standard in many places, rightly recognise that accuracy is inevitably late-acquired, so learners are encouraged to communicate effectively, albeit defectively,and then to set about improving their production so that it becomesmore accurate. 'proficiency' suggests, however, Recent,more preciseanalysisof the idea of proficient leatner uses language which exhibits three rather than the more two distinct characteristics:accuracy,fluency and complexity. The recognition of complexity gives teachersa framework within which to ask How can I best help my learners to improve their language? It is immediately clear that different emphasesmay be appropriate for different leamers. of Firstly, awareness the phrasalnature of the mental lexicon also modifies the idea of a leatner's level. An increasedlexicon involves different elements: . Adding new words. . Expanding knowledge of the collocational field of words already known (including awareness blocked collocations)' of 'a . Increased that word'may havemore than one meaning,or be awareness usable as, for example,both a noun and a verb. This correspondsto that a word may have severaldifferent, overlapping or even awareness independentcollocational fields. . Knowledge of more colligational pattems - that is, greaterknowledge of the grammar of the word, and a correspondingly greaterability to use it fluently, accuratelyand in more complex pattems. Secondly, the ideas of prefabricated language in speechand complex noun phrases in writing are parlicularly helpful in improving the complexity in quite different ways at different levels. Improving complexity in speech Modern computer-basedstudies of spoken and written language confirm, consists what we havelong suspected that fluent speech indeedemphasise, largely of rapidly produced short phrases, rather than formally correct 'sentences'. This is as true of relatively formal, educatedspeechas of any other variety; it is characteristic of all (unscripted) speech,and in no sense substandard.Many of the phrases are relatively fixed, prefabricated lexical items. Access to a comprehensivemental lexicon of such prefabricatedunits

sp u4 of chr

Fo coi
cei eve


the inti lear

and wn botl rES meI clas

Knc the com fluer Imp

Atr tertii also cons saw numl most bya pass

N w th be int


Learning in the lexical approach 175

rble rso

)nts the
/ 1 3

)ate tl0n

han rto


.the n rbe

is the basis of fluency in speech.This means learnersmust be exposedto an adequateamount of natural spoken language.Fluency needs to be based on spoken input, and it is the quantity and quality of that input, not language which learnersthemselves produce- which is the basisof an adequate lexicon of essentialphrases,providing, as we have already seen, that the language chunking is noticed. Formulaic chunks have been called 'islands of reliability' by several commentators.Chunks which leamers are sure are accurateand convey the central meaning of what they wish to say are immensely reassuring, especially when contrasted with the intimidating prospect of constructing everything you want to say word-by-word, on every occasion.Initially, then, the prospect of the lexicon being much larger than we previously thought, is intimidating for learnersand teachersalike. However, if teacherscan reassure learners,and encouragethem to seethe value of larger chunks - (semi-)fixed expressions,sentenceheads or frames for the pragmatic element of speech, and collocations for the central information content of both speech and writing - theseislands of reliability provide important psychological support both in helping leamers express themselves within their present linguistic resources, and, equally importantly, as starting points in expanding their mental lexicons. The activity described on page 91 provides a detailed classroom procedurefor building on theseislands. Knowledge of fixed items also means additional brainspaceis available, so the leamers are more able to processother language,which enablesthem to communicate more complex messages, simple messageswith greater or fluency or accuracy. Improving complexity in writing

t oun /in

trTn, ;1StS rect any )nse ical nlts

At more advanced levels good writing, in particular the kind required of tertiary level students,is characterised not only by accuracy and fluency, but also by complexity. This is largely dependent on the writer's ability to construct noun phraseswhich are high in informational content. This, as we saw in the previous chapter, implies the text will contain a relatively high number of nouns, which in turn implies frequent use of the word of, the single most powerful tool in constructingnoun phrases.This is clearly demonstrated by a few sentencesfrom Brian Magee's confessions of a philosopher; rhe passage contains172 words,including no fewer than 13 usesof the word of Near the heart of the mystery of the world must be something to do with the nature of time. Neither the time of common-sense realism nor the time of Newtonian physics is given to us in experience,nor could it be, since it stretchesforward and backward to infinity, and nothing infinite can ever be encompassed observation or experience,Any in such time has to be an idea, something thought but never observedor experienced,a construction of our minds, whether it be a mathematical

l7 6

Learning in the lexical approach

calculation or an imaginative assumption presupposed by the The sameconsiderationsapply to space:the deliveranceof our senses. space of common sense, stretching as it does to infinity in all or directions,is not a possibleobject of observation experience it too is a projection, a construct of some kind, as must also be the threedimensionalEuclidean spaceof Newtonian physics...The time and spaceof our experiencedreality are forms of our sensibility, and it is in that capacity that they appearas dimensionsof experience. Notice that most of the afphrases would be entirely absent from or overlooked in regular EFL classes,including those for English for Academic on Purposes. Traditional grammarhas concentrated the verb-phraseto such an extent that the construction of complex noun phrases has been largely ignored. It must be self-evident that for many more advancedlearners, the study of noun phrasesand expressionswith of of the kind discussedin the previous chapterare at least one of the keys to learnerswriting both more fluently and at a greaterlevel of complexity. Competence 'performance'in 'competence' and discussing Chomskyintroducedthe terms how the human mind masters and produces language. Performance was language actually produced by people, and thus subject to empirical investigation. Competence was a rather mysterious ability involving knowledge of the rules of (the grammar of the sentencesof) a language. Chomsky claimed all performance was basedon competence,an abstraction reminiscent of the 'pure forms' of Greek philosophy, and by definition not directly accessibleand not subjectto empirical investigation.It is astonishing that he got away with inventing a supposedlyscientific distinction, one half of which was, by definition, not a scientific concept.(To be scientific,an idea must be testable,and therefore, at least in principle, falsifiable.) At the beginning of 'the communicative age', Hyams pointed out that language was not an abstract system to be studied by grammarians, but a purposes. real, non-linguistic, symbolicsystemused by real peopleto achieve - it is a tool Language was a communicative tool. Both words are important in the sensethat its purposelies outsideitself, it is a means,not an end; the end is communication.If you can communicateanything you wish to on every occasion and do not in the process also communicatethings you do not This is a intend, you may be said to possesscommunicativecompetence. feature not of the language,but of people or, in classroom terms, learners. Communicative competencecan be analysed,specified and form a basis for pedagogicaldecisions.It replaces Chomsky's rarefied abstractionwith a concept which is entirely concreteand practical. We can now go one step further, and ask What is the basis of communicative


bY (' wish prod invol lexic calle

Asu one f lnput (syntl langu mput than ti alread idea t 'reass

descril quote(

sugges by evet to hum theorie wrong. Iangua specul thrn-es mean\ compr agams 'Acqut

new \r breaki into u' not ls i and thi


The P much I Obsen this bo

Learning in the lexical approach 177

) I

competence This is an alternativeversion of the question what do we mean ? by (very) advanced learners - people who can communicate anything they wish, without communicating things they do not intend. Such learners produce languagewhich is fluent, accurateand stylistically appropriate.This involves the learner having a sufficiently large and sufficient phrasal mental lexicon, where many single choices are multi-word items. Jimmie Hill has called this ability collocational competence. As we have already noted, proficiency in a language invorves two systems, one formulaic and the other syntactic, and unless the learner can break down input language (analyse) and re-assemble it in novel combinations (synthesise,or syntactisise,using a knowledge of syntax), the rearner's languagewill remain 'impoverished'. The argumentin favour of collocational input is that it is easier to break down groups and learn to reassemblethem, than to start from isolated words which then have to be combined.As we have already seen,however,languageacquisitionis a non-linearprocess,so any idea that we 'start' with groups, and then 'break them down' and then 'reassemble' them, is very unlikely to be anything other than a partial description of part of acquisition. In the article on The Lexical Approach quoted earlier, Thornbury complains that Lewis does not have a comprehensivelearning theoty, and to some extent the criticism is fair. I suggest, however,that few peopledo. StephenKrashenhas,but it is rejected by everyoneto at least somedegree.It is also the casethat the greatestdamage to human understandinghasnot been causedby thosewithout comprehensive theories, but by those with comprehensivetheories which turned out to be wrong. A certain humility is required - we simply do not know precisely how language is acquired, and no so-called comprehensivetheory is more than speculation.we do, however, have partial theories and evidencethat certain things do not work. As Thombury suggests,we need more research; meanwhile we must avoid turning the limited knowledge we do have into a comprehensive theory prematurely. we must guard even more carefully againstturning theory into dogma. 'Acquisition' almost certainly involves a non-linear combination of acquiring new words, acquiring new multi-word items, becoming more proficient at breaking large wholes into significant pafis (words into morphemes,phrases into words etc), combined with developing awareness whether explicit or not is a subjectof heateddebate- of the 'rules' of permittedre-combinations and the restrictions which exclude certain re-combinations.

ror jmic han gely yof 10us and

;sing was rical ving lage. ctlon 1 not ;hing rlf of idea that

r tool 1; the every c not
i l s a

rners. is for rith a :ative

8.10 Teaching paradigms
The Present-Practise-Produce (P-P-P) paradigm remains a central part of much teacher training. rn The Lexical Approach I suggestedthe alternative observe-Hypothesise-Experiment paradigm,and many of the suggestions in this book endorseand expand this paradigm.


Learning in the lexical approach

is Present-Practise-Produce intrinsically incoherent as a learning theory. Presentation is done by the teacher; Practice by leamers, moving from controlled to free practice under the direction and time-consffaints imposed by the teacher; Production is wholly within the learners' domain. The implicit assumption is that teaching does cause leaming, a view which we have already seenis mistaken. Any paradigm should, as a minimum requirement,statewhat the learners are doing at different phasesof the process.The non-linear nature of acquisition means that different parts of the plocess may be occurring before' after, or simultaneously with other parIs, with different parts of the process being applied by different learners to different parts of the input at the same time. Within this framework, any adequateparadigm is a simplifled idealisation of how any individual leamer may be dealing with somepart of the input in any given phaseof the lesson.However unnerving that is for teachers,or teacher trainers who urge teachersto plan lessons- hence the attraction of the P-P-P paradigm - that diversity is what is happening at any moment in a language class; any theorising to the contrary simply ignores the nature of either languageor acquisition. With those explicit caveats,it is clear that learners do experiencea sequence which may be summarisedas meet-muddle-master. This is essentially the Observe: new languagemust be same as Observe-Hypothesise-Experiment. this point that we differ from Krashen's met and noticed. It is precisely on Natural Approach. Hypothesise: means sorting the input on the basis of appafently significant similarities and differences- as we have seen,this can be done without necessarilybeing able to describe the categoriesor sorting processexplicitly. Experiment: involves using the languageon the basis of the learners' current intergrammar (that is, his or her current best hypothesis),thereby stimulating new input at the appropriatelevel to provide examples which confirm or contradict some part of the learners' current hypothesis. Mastery happens - if ever - when new input serves only to confirm the leamers' intergrammar' Within this paradigm, the importanceof noticing the difference betweenwhat 'correct'is clear'Leamers effective,and what is formally is communicatively who believe their output is completely successfulwill seeno leason to modify any of their current intergrammar. Unnoticed deviancy may confirm rather than modify leanters' culrent intergrammar. In addition' teachers have a valuable role to play in predicting problems and providing the negative evidence necessaryfor effective hypothesis formation, as we see when we consider how the teacherintervenesto direct leamers in ways which they are unlikely to use without helP. Teacher intervention The fact that text - spokenor written - consistslargely of multi-word chunks

of dil unde indir

1.It word is the

2.Itt alrea<


3.Itr rnten.

Consi violatr saw j

under to app

The le the fr one lar unless are liki than th If lean individ havebr

This is (Lisrett The


bear paIt(

We are phrasal comme in whic pattems are thus

This cor store thz this boo

Learning in the lexical approach 179

lory. from osed The nwe 's are

)r, or lelng tlme. on of n any acher ]_P-P luage either

of different kinds, doesnot stop teachersaskingAre there any words you don,t understand?This question- at least implicitly directing leamers' attention to individual words - is misguidedfor threereasons: 1. It encourageslearners to believe that language consists of structuresand words, and that single words have unique meanings (and implicit within that is the idea that word-for-wordtranslationis possible). 2. rt treatsinput as if comprehensionis the whole story, although as we have already noted, input for acquisitional purposes may differ from simple message-catrying input. 3. It means learners frequently do not invite teacher intervention when that intervention would be immensely useful. considering the first point, nothing done in the languageclassroomshould violate the nature of either languageor learning,and this questionis, as we saw in the previous chapler, based on an out-dated and misguided understandingof the nature of language.It also implicitly encourages learners to approach both text and 'leaming'in an unhelpful way. The learners' intuitive belief that single words are the units of meaning and the frequently mistaken belief that if there is a single word for something in one language,then there must be a single word for it in another,means that unless noticing chunks is explicitly taught, learners left to their own devices are likely to break the text into single words or into chunks which are smaller than the optimal units neededto convert input into maximally useful intake. If learnersbreak the text down into individual words, which they then store as individual words, they make re-encoding much more difficult than it would have been if they had storedthe languagein larger chunks from the start. This is also true in the areaof pronunciation; twenty years ago Gillian Brown (Listening to SpokenEnglish, p 49) suggested: There is a certain amount of evidence that native speakersrely very strongly on the stresspattern of a word in order to identify it. It is suggestedthat we store words under stresspatteflls, so if a word of a given stress pattem is pronounced, processing in this word, we bring to bear our knowledge of that part of the vocabulary which bears this pattern. we are now more likely to speak of phrasesthan individual words, and of a 'vocabulary', phrasalmentallexicon,ratherthan someone's but otherwiseher commentsremain valid and relevant.It is very likely that one important way in which we store and accesslexical items is by the ,shape'of their stress patterns. Short phraseswhich have patterns and can be stored with a ,tune,, are thus likely to be more memorable than patternlessmonosyllables. This counter-intuitive insight, that larger chunks are more useful and easierto store than small chunks, lies at the heart of every idea or activity discussedin this book.

ty the ;st be ;hen's sis of
15Can ofimg Lsisof

best rovide
uffent n1y to n what )amers nodify rather
lave a

:gative Ien we ley are



Leaming in the lexical approach

Moving on to the second point, as every classloom teacher knows, it is difficult to move the lesson forward if learners do not undefstand, so understandingis impoftant. For languageto contribute to acquisition it must be (at least pafily) understood.But the purpose of input is for it to become intake, and that in turn, must be available for productive use. The ultimate pulpose of input is learner output. From this pelspective, it is clear that understanding, though necessary, is not sufficient; in addition to understanding the input the leamer must notice the chunks which carry the meaning.Each chunk is a single choice of meaning; if chunks are not noticed as chunks, they cannotbe storedin the way which facilitates their availability as output. If input is to become optimal intake, understanding and noticing the chunks are both necessary (though perhaps still not sufficient) conditions. Which brings us to the third point - teacherinterrrention.Understandingthe lexical nature of language makes it increasingly clear that some aspectsof language learning ale countel-intuitive: phrasesare easier to remember than make single words; breakingthings into smallerpiecesdoesnot necessarily just seen, understanding is not enough to them simpler, and, as we have ensurethat input becomesintake. This meansteachersneed to be proactive in helping learnersdevelop an increasing understandingof the lexical nature of the language they meet and be more directive over which language is particularly worth special attention. This point was exemplified earlier by MorganLewis[p l8l. When a patient visits the doctor, the task of diagnosis is exclusively in the hands of the professional. The patient's role is to describe his or her symptoms honestly and clearly, and to take responsibility for following the doctor's advice by, for example, taking the prescribed medicine at the appropriatetimes, perhapseven choosing between alternativetreatments,but the choice of medicine, the choice of which alternatives are offered to the patient is exclusively the responsibility of the professional. So it should be with the languageclassroom- by all meansadopt a leamer-centredapproach, encouraging learners to take responsibility for their own learning, and allowing choices, but teachers cannot, or at least should not, abdicate responsibility for the syllabus, for deciding which languageis worthy of the learners' attention in a particular piece of input, which should in turn have languageactivity, been chosenas suitablefor that class.After teacher-directed else you would like to ask it may be appropriate to ask Is there anything should about?,but the questionAre there any words you don't understand? from the classroomsof competentteachers. be banishedfor ever Negative evidence Because generalisations may be subject to restrictions, language which 'wrong', may be non-standard learnersmay think is possible,if not actually

or ma accep some under Swan of ther but otl skiing,

Simila possib is the which
meetill teache this ne those t


Both c Where consci< help. \\ the con as chun prefabr prefabr the lear miss mr that the the Nat that for (except Curren

If there case tha usually controll stateme some te others st after the a numbe

Learning in the lexical approach lg1

lst ne Ite rat to he ed rty ng 1r) he of an ke to

or markedly 'foreign'. It follows that leamersneed to know not only what is acceptable, but also what is not. Knowing the restrictionswhich apply to some pattem or generative rule is an important part of exploring and understandingthat rule. At the beginning of many paragraphs of Michael Swan'sPractical English (Jsage givesexamples typical mistakes. he of Some of theseare traditional grammar mistakessuch as *one if my friencl is a pilot, but othersare more like collocational errors: +I oftenfellyeste"rday when I was skiing. *Let's have one drink and then I'Ir bring yow back home. Similar problems apply to collocation; learners need to know both what is possibleand what is not. As we saw in the previouschapter, a major problem is the blocked collocation,the one learnersthink ,ought'to be correct,but which is not in fact acceptedby the native speakercommunity. No amount of meetingEnglish naturallywill provide evidence what is not sanctioned, of so teachers need to predict problems and once again be proactivein providing this negative evidence,warning learnersof blocked coilocations, particularly those they anticipate by (false) analogy with their own language.

8.11 The Lexical Approach and the Natural Approach
Both of these approachesvalue large quantities of comprehensible input. where they differ is that in the Natural Approach Krashen claims that conscious activity doesnot aid acquisitionand therefore that noticing doesnot help. we believe that noticing featuresof the input, in particular the nature of the component chunks of the text, has a facilitative vaiue. Noticing language aschunks,aids storage chunks.It thereforeaids acquisition, some as as ofthis prefabricated language is then available to the learner both for use as prefabricateditems and as raw material for syntactic analysis,raw material for the learners' intergrammar.without guidance from a teacher, learners may miss much that is of value from this acquisitional point of view. Ensuring that they do notice certainlanguage may (but will not necessarily) help. Both the Natural Approach and the Lexical Approach are in complete asreement that formal practice what is noticeddoesnot contribute acluisition of to (exceptthat it may, of course,incidentallyresult in noticing). Current practice If there is such a thing as standardpractice at the moment, it is probably the case that most teachers believe in presenting a particulu, gru---u, pornt, usually a feature of sentencegrammar; then encouraging some sort of both controlled and free practice. opinion is divided on whether some fbrmal statement a 'rule' is helpful, and among those who do believe of rules help, some teachersgive a rule and examples as part of the presentation, while others summarise, ask leamersthemselves summarise, or to refospectively, after they have practisedthe point in question.Theseprocedures are basedon a number of more-or-less explicit assumptions, including at leastthese:


he ler he .he

)ut .he be rh, nd ate "he

tsk rld

ich rd


Learning in the lexical approach

1. Important featuresof what learnersneed to know are often exemplified within a single sentence. 2. It is possible to selectwhat (all of a group of) leamers are going to learn next. 3. Practising a particular bit of languagehelps learnersretain it. 4. A formal description of a languagepattern helps learnersto seeit. And for those who believe in rules: 5. A formal description of a pattem helps learnersretain the pattem in such a way that it becomesavailable for their own use. It is now clear that all of these assumptionsare either untrue, or at best very limited half-truths. If classroom procedures are to be changed, we need to understand why these assumptionsare wrong, and establish a clear set of principles, the basis for a leaming theory.

of tl teac muc ackn 3 , A nece diffu

8.12 Towards a learning theory
1. Grammar and vocabulary are not separablein the way assumedin most language teaching materials. Observation of real language data reveals that languageconsistsof many more pattems than was previously believed. Many boundaries of thesedependon the genre of the text and extend over sentence - they are discourse grammar, not sentence grammar. Many pattefirs are intimately bound up with specific vocabulary - they are word grammar, not sentencegrammar. One of the implications of the new descriptions we have at our disposal is that simply lumping together half a dozen examplesof, for example, the present perfect, because they are examples of the present perfect, may violate the nature of language,bringing togetheritems which are At not similar in the way the teacher assumes. the same time, this emphasis on structure may discard information which is an essentialpart of the real pattern. Taking language out of context in order to teach it, may, paradoxically, make it less available for acquisition. In general,learnersare more likely to acquirenew languagein such a way that it is available for spontaneoususe if it is incorporated into their mental lexicons as an element of some comparatively large frame, situation or schema. Presenting larger units in class,such as a completetext, an episode of a soap opera or a self-containedpart of a dialogue increasesthe possibility of learners transferring items to their mental lexicon within these global organising schemata.In contrast, it is increasingly clear that the tradition of presentinglexis as individual words, or practising individual decontextualised sentencesis at best inefficient, and at worst actively unhelpful. Grouping those sentences accordingto somearbitrary linguistic feature seemsalso to be counterproductive, as it in no way mirrors what we know about the organisationof the mental lexicon. 2. Acquisition is not a linear process; at any moment learners in the same group have different intergrammars;different learnerswill make different use

makt term focur empl langr langr exam and , transl leamr listen Succ not ol produ wond was perfor nottce mole The P frame the tra book c produ< they dr In this langua activiti mottva leamer lnstruc view n accural player procedt conditir

Learning in the lexical approach 183

iuch very rd to ei of

3. Although using language may help the learner retain it, this is not necessarilyso. Using language is stressful - in speech, it involves the difficulty of articulation, working under time constraints, the possibility of error and many other factors which take up brainspacein ways which may make the brain less able to processlanguage,so that it is moved from shortterrn to long-term memory. communicative approacheswere intended to focus on meaning, but have often been interpreted in ways which have emphasised production, particularly speaking, from the earliest stages of language learning. This runs directly counter to what we know about first

most i that t4any Iaries s are r, not have f, for 'esent :h are rhasis : real may, y that rental ln oI ,isode ,bility ;lobal on of alised upmg rtobe Lt the same nt use

learner's confidence; they do not acquire new languageby speaking,but by listening and reading, subject to making good use of the input they meet. successful production may, indeed probably does, help retention, but input not output is the key to long-termimprovementin learners'ability.Successful production does not even guarantee retention, as anyone who has hit a wonderful tee-shotwhile learning to play golf knows. However good the shot was - perfect production - it does not ensure that you can reproduce the performance ever again. classroom activities which ensure that leamers notice input in ways which convert it, or help to convert it, into intake, are more likely to be valuable in the long-term. The Present-Practise-Produce (P-p-p) paradigm is unrealistic unlessthe time frame is weeks or months rather than a single lesson or day. Teachersused to the traditional P-P-P paradigm may feel uneasywith the concentrationin this book on awareness raising and noticing activities, and the comparativelack of productivepractices,perhapsasking - so when they've noticeclit, what do they do with it then? In this book all the contributors accept that helping learners to notice useful language accurately, helping them avoid wasting their time on unhelpful activities, guiding their choice of materials and activities, and maintaining motivation may be the principal contributions the teacher can make to learners'acquisition.This is what is meant by seeingthe teachernot as an lnsftuctor, but as a leaming manager.Teacherswho feel unhappy with this view may like to consider the way a sports coach operates- it is precisely by accurate observation of a player's performance and the ability to make the player more aware of his or her performance. Sports, like language,involve proceduralknowledgeand the ability to 'put it all together'underreal-world conditions and time-constraints, so that there is often too little mental


Learning in the lexical approach

processingspaceleft to observeyour own perfomance - hence the need for coaches.Knowing what is important, to that playerflearner,and ensuring that the playerflealner notices what is most likely to benefit him or her at that particular time is a real and valuable role. 4. As we have already noted, understandingthe messageis not sufficient to ensure that input becomes intake; learners can fail to notice the chunks of which a text is made. Ensuring that they are familiar with the idea of chunks, and developing their ability to identify the chunks they need to expand their lexicons at that particular point in their leaming, will help turn input into intake. is no1. It however. necessary give a formal description anypattem to of - it is sufficient that learnersnotice the words, in the correct chunks. 5. Explicit description of the pattems is not necessaryand indeed a great deal of time could be wasted labelling patterns. We can all recognise a huge number of colours and sort them into families without necessarilybeing able to name the families in any precise way. Sorting into finzy-edged categories is both all that is needed,and probably all that is possible.The mental lexicon stores items in patterns; more than the few traditional EFL structures,fewer than a vast number of separatelexical items we know - a large number of comparatively restricted pattems. This is done by noticing similarities - we earlier quoted Michael Hoey's observationthat grammar is the product of the colligations you have observed.Sorting, consciously or unconsciously,is essential;the ability to describethe sorting categoriesis not.

The disc facil thei conS the I wel mor( sylla


Is th stron

Is tht


What pomt

Thinl b) les

8.13 Summary
In this chapterwe have noted severalimporlant featuresof acquisition: . Meeting and (at least partially) understandingthe samenew languageon severaloccasionsis a necessarybut not sufficient condition for acquiring the new language. Noticing the languagechunks which make up the text is again a necessary but not sufficient condition for tuming input into intake. Noticing similarities, differences,restrictions and examplesarbitrarily blocked by usageall contribute to tuming input into intake, but formal description of the categoriesinto which input languagemay be sorteddescriptive'rules'- probablydoesnot help the processof acquisition,and may hinder it by intimidating some, perhapsmany, learners. Acquisition is not basedon the application of formal rules which generate correct examples,but on an accumulation of examplesabout which everchanging provisional generalisationsmay be made by the individual learner.These generalisationsmay be the basis for the production of languagewhich is novel for that learner,but all such production is ultimately the product of previously-met examples,not formal rules. No linear syllabus can adequatelyreflect the non-linear nature of acquisition.

. .



Learning in the lexical approach 185

for that that rt tO sof nks, heir rnto tem deal ruge able rries tcon )weI rof 'we the /, rs

These factors, together with the lexical description of language already discussed,give a clear indication of the teacher's role - it is to constantly facilitate the accurateobservationby learnersof appropriateparts of the input they meet. Put simply, teaching should encourage learners to search constantly for many different small patterns,rather than repeatedlypractising the few large patternsof traditional grammar.This servesto remind us that, as we saw in the previous chapter, the Lexical Approach is in important ways more grammatical - that is, pattern-centred - than traditional structural syllabuses.

Discussion Questions
Is there one idea in this chapter which is new for you and with which you strongly agree? Is there one point aboutwhich you stronglydisagree? How often, in your experience,do leamershave to meet a new bit of language before you can be fairly sure they will have acquired it? What role, if any, do you think controlled practices,particularly of grammar points, play in aiding acquisition? Thinking of your own teaching, to what do you think you will give a) more b) less emphasisas a result of reading this chapter?



and rate
r l -


Materials and resources

Materials and resourcesfor teaching collocation
Michael Lewis
This chapter explores the importance of choosing texts with the right type of collocational input for particular groups of learners. It also provides a simple introduction to language corpora and concordancing for teachers new to these tools. Although recommending the use of real data, it suggestscaution is needed, particularly if learners are to be exposed to raw data. Finally' the chapter comments briefly on dictionaries, particularly collocation dictionaries, as a resource for learners.

builr rem( rnpu for I

9.1, Choosing texts
Collocation is to be found in texts of all types, but different kinds of text have radically different collocational profiles, so two of the teacher's most important skills in the teaching of collocation are choosing the right kinds of text, and then guiding the leamers' attention so that they notice those items likely to be of most benefit in expandingthose particular leamers' lexicons. In general, fewer words of written English are needed to expressthe same content than are needed in the spoken mode. This is partly becausewritten English containsmany more complex noun phrasesand phrasesusing o/such as The choiceof textsof dffirent typesis conditioned. . . . Collocationsof a small number of key nouns tend to re-occur throughout discursiveprose text such as academic writing or a magazine or newspaper article. Newspaper reports (as opposed to articles) contain large numbers of often quite large collocational groups, but many tend to be largely confined to journalism. to Although there are motivational advantages using stories, narratives such novels or readers are much less collocationally dense, so the use of as narrative texts is often an inefficient way of expanding learners' mental lexicons. Speech,naturally richer in semi-fixed expressionsand multi-word adverbials, contains comparatively few of the verb + (adiective) + noun combinations which learnersneed if they are to write essaysor reports, and indeed recent research (see particularly the Longman Grammar of Spoken and Written English - referred to below as LGSWE) suggestsmajor differencesnot only between speech and writing, but between different genres of speech, or different geffes of writing. If learnershave immediate specific needs,far from needing a 'balanced' diet of different types of text, the texts to which they are exposed should be skewed in the direction of their needs. For learners of general English, a balance of different text-types is of major importance in


Hot'( toba< ls gr( we \\ contr

If du acce Oster the cr Calai Text

It rnta and t anrmi knou' huma Text

The I its fu distin 'imag

other infon not g( but tl partic

Materials and resources lg7

building their mental rexiconsin a balancedway; no one .type, of English is remotely adequateto representthe whole. when choosing texts for leamer_ input, it is important to choosenot only from an interestpoi.rt of ui"*, but also for linguistic, and specificallycollocational,reasons.

re of nple hese ded, pter asa

Readeachof the following texts and: . Ask what kind of text it is. . Underlineany items which you think are probably storedand produced multi-worditems. as . Now, refine that selectionto include only thoseitems suitablefor drawingto the attentionof a particularclassthat you know well. Text 1


nost ls of ems ns. ame tten ;uch ofa text lper rrge tsm. luch :of ntal ials, tons lent tten rnly

Hoverspeedretail director David King said: .,with duty rates on alcohol and tobacco continuing to rise in the uK, the market for cross-channel shopping is growing all the time. our first store in Scandinaviaalso demonstrates that we will be looking to expand in other markets where differential tax regimes continue to provide a major incentive for travel retail opportunities.', If duty-free sales are abolished within the EU next year, the firm plans to accelerateits expansion into shops located near ports. plans for outlets in ostend' Belgium and Fredrikshavn,Denmark are already well-advanced, and the company expectsto expand its retail operation, with additional outlets in Calaisand Dieppe. Text2 It was a bitter winter. The stormy weather was followed by sleet and snow, and then by a hard frost which did not break till well inio February. The animals carried on as best they could with the rebuilding of the windmill, well knowing that the outside world was watching them and that the envious human beings would rejoice and triumph if the mill were not finished on time. Text 3 The key to understandingthe modem analysis of advertising is to understand its functions as a purveyor of messagesand information. it is important to distinguish between adverts which provide specific information and those 'image adverts' which present what may be termed non-informational messages. Specific information may relate to price, physical characteristics or other aspects of goods or services mentioned. Such adverts are obviousry information-providing. some adverts, however, such as Marlboro Man, are not generally perceivedas providing information except in the broadestsense, but they may still have an important role to play in the marketplace, particularly in relation to the competitive process.

rom are sof


Materials and resources

Text 4 After an unsuccessfulattempt to win the vice-presidentialnomination on the ticket of Adlai Stevensonin 1956, Kennedy beganto plan for the presidential election of 1960. He assumedthe leadershipof the Democratic party's liberal wing and gathered around him a group of talented young political aides, including his brother and campaign manager,Robert F. Kennedy.He won the nomination on the first ballot and campaigned with Senator Lyndon B. Johnson of Texas as his running mate, against Vice-President Richard E. Nixon, the Republican nominee. The issues of defense and economic stagnationwere raised in four televised debatesin which Kennedy's poised and vigorous performancelent credenceto his call for new leadership.

' gri . lot d^

All o such obser

lr€ pa po

9.2 Genre
If you instantly identified theseextracts,it can only be becausedifferent types of text - genres - have markedly different linguistic profiles, even if it is difficult to say exactly what makes the profiles different. If learners are to acquire effective and balanced mental lexicons, the range of types of input text to which they are exposedis clearly of great importance. Text 1, with its mixture of reporting and quotation, luse job + name- retail of - and explicit detail, is typical of newspaperreporting. director David King Text 2 is a shofl paragraph from Animal Farm. I am indebted to Chitra Femando for drawing attention in ldioms and ldiomaticiQ to something George Orwell wrote in 1946 in Politics and the English Language: This invasion of one's mind by ready-made phrases (lay the foundations, acquire a radical transfonnation) canonly be preventedif one is constantly on guard against them, and every such phrase anaesthetisesportion of one's brain. a Considering Orwell's strictures about the use of clich6, it is amusing to note that in as unusual a work as Animal Farm he cannot avoid collocations and fixed expressions:bitter winter, stormy weathet a hard frost, carried on as best they could, the outside world, finished on time. The point is simply that theseitems, despite apparently consisting of severalwords, are in fact single choices in any mature native speaker'smental lexicon - even Orwell's. In short, there is no other convenientway of expressingtheseconcepts,however 'creative'you may wish (or claim) to be. Even the famouslycreativeopening line of l9B4: It was a bright, cold day in April and the clocks were striking thirteen, while undoubtedly creative, is based on the collocation the clock struck. Collocation is a feature of all kinds of text, both spoken and written, though different kinds of text contain different kinds of collocation. Text 3 is obviously from an information-bearing, text. It includes: . lexical collocations: the modern analysis of advertising, the competitive process

Text i It con write i

lent on can anl The sa are larl highesl

LGSW that a inadeqr of a pie reportir

Effectir more n particul very p corresp creative of word or at le writing, informat standard prerequi

9.3 Sr

Everyon realises 1

Materials and resources lg9

r the ntial leral Ldes, r the rB. dE. rmic rised

. grammatical collocations: what may be termed, such adverts . longer expressions:The key to understandinp. . . , It is important to distinguish between. . . All of these linguistic features are typical of the genre, as are collocations such as modfu the theory. In an article on collocation peter Howarth observes: lexicar collocations of the type transitive verb + object areJ a smalr part of the whole framework, but of considerabre imptrtance from the point of view of the propositional content of an academicafgument. Text 4 is from a computerencyclopaedia entry of only just over 100 words. It contains at least the folowing useful colrocations foi arryon" who needsto write aboutpolitics: assumethe leadership of . . . the liberal wing lend credenceto vi gorows/poi sed p erfo rmance on thefirst ballot callfor change campaign with . . . as (his) running mate an unsuccessJul attempt to win the nomination The samplesclearry show that the language of fiction and newspaperreports are largely unsuited as input for leamers, except, perhaps those at the very highest levels. Factual, information-bearing texts a'e much more suitable. LGSWE, in which fiction is one of the four geffes, crearly endorsesthe view that a simple distinction between written and spoken English is wholly inadequateas a basisfor choosingmaterials; at leasi u, i-po.i*t is the genre of a piece of writing. Factuarwriting, including n"*rpup", u.ticles rather than reporling,is most suitable.Fiction is the least suitable_ it is easyto seewhy. Effective language is a combination of familiar, prefabricated chunks and more novel, innovative combinations created by the language_user for a particular occasion.The advantages prefabricated of chunks"are"that they are very precise, require little processing by the listener/reader, and comespondinglycarry rittre danger of being misunderstood.unsurprisingly, creativewriting such as novels may contain many non-standardcombinations of words, as the author literally creates character and mood by creating new, or at least unfamiliar, combinations of words. In contrast, in academic writing, where the focus is almost exclusively on accuratecommunication of information' among colleagueswith a shared backgroundin a particular topic, standard words, phrases, colocations and other chunks are an essentiar prerequisite effective for communical ion.

lt ls :e to

nD0 .' '

hitra hing e if e note and n0s that ngle s. In

nmg kirug 'lock

9.3 Subject-specificIanguage
Everyone who changes jobs or has to read in an unfamiliar disciprine quickly realisesthat every discipline has subject-specific language.Sometimessuch


Materials and resources

'jargon'but in fact it is a preciseand language dismissivelycharacterised is as necessarytool for anyone who uses English for a specific purpose; such language should be criticised only if used to non-specialists.The following specification of businessgoals and strategiesfrom a company document is in no sense jargon, but note that the whole content of both the goals and strategies expressed 2- or 3-word collocationswhich are both concise is in and precise. Goals Increaseprofit margins Improve cashflow
Reduce overheads

conc fictic
ncrf l

Strategies products Introduce more added-value Tighten credit terms; reduceaveragecollection period Review stock levels; outsourceseruices

The lonq _-^_ cont( subje conc( almo


Put simply, it is clear that no amount of English literature will help you write 'academic a good paperin economics, and,lessobviously,a coursein general English' will be of very limited use to studentswho have to read and write in a particular academicdiscipline. In a fascinating paper at TESOL Intemational1999, Patricia Watts described the difficulties of teaching studentsfrom different cultural backgrounds the that the useof sourcematerial'in your field in the US'. Firstly, sherecognises conventionsdiffer from subjectto subject; secondly,that the conventionsmay be US-specific. Two other obserrrationsare, howevet, of particular interest. Shepoints out that in her experiencenon-US studentshave a sffong tendency to over-quotation, so their assignmentsare dominated by source materials rather than the student's own ideas. Fufihermore, that what is plagiarism in one culture may be a compliment in another- as one of her studentsobserved: exactwords. In my country it is a complimentto the author to usesomeone's Most interestingly from our point of view, however,is the observationthat all learnershave trouble introducing quotation or citation. This is becausethey have a wholly inadequate repertoire of the kind of fixed and semi-fixed expressions which are used in these - often subject-specific - academic genres;it is a highly specific lexical, collocational deficiency. (Language, Linguisticsand Lfe) observes: Svetlana Ter-Minasova Teaching communication for special purposes must be based on the previous linguistic analysis of special texts resulting in recommendations for teaching those grammar forms and structures which are most characteristicof thesetexts. The store of units of this kind (prefabricated blocks) form a certain stable system of linguistic means which are constantly in use in the processof scientific communication, which form a sharedcode of the participants the communication. of Despite the difficulties faced by applied linguists in Russia, particularly during the Soviet period, they recognisedlong before EFL in the West that the

Althc much plone the bi three of tei langu poren Corpr

Cobui corpui data-d fully z which additic langua provrd

Other 1 a basis and exr million that its has lea part ot impresr While I teachin

The pr descript referenc which r

Materials and resources l9l

ise and :; sugh ilowing :nt is in rls and ;oncise

conceptof 'academicEnglish'might be no more than a language-teachers' fiction, and that something more specific to the kind of texts studentsof a particular discipline met regularly, may be required. The time when teacherswere restricted by the texts in their text books has long gone. It is now easy and cheap to find material which will presentfully contextualisedexamplesof collocations relevant to writing about a particular subject.As Michael Hoey points out in detail, any factual text is a disguised concordanceof someof its key words; such texts are now readily availableon almost any subject,at the click of a mouse [Seebelow].


9.4 Language corpora
u wrlte ademic vrite in icribed rds the hat the 15may lterest. rdency Lterials rsm in ;erved: r,vords. :hat all ;e they i-fixed Ldemic Although the idea of collocation in linguistics goes back at least 60 years, much of the current interest for languageteachersbegan about 1990 with the pioneering work of John Sinclair and his team at Cobuild. Sinclair describes the background to this work in Corpus, Concordance, Collocation and the three ideasare closely linked. All three are well worth exploring in the context of teaching collocation; they provide potentially powerful tools for the language teacher, but unless used with proper discrimination, there are potential pitfalls which may confuse rather than help leamers. Corpus size and balance Cobuild's purpose in developing the first really substantial computer-based corpus of real texts, both spoken and written, was lexicographic - to make a data-driven dictionary, where every example was taken from the corpus as a fully attestedexample of 'real English'. It was essentialto build a cotpus which was sufficiently large to give useful information about rarer words. In addition, many rarer words are almost exclusively confined to the written language,so large amounts of written data were neededif the corpus was to provide adequateevidenceof the use of such words. Other publisherslater established their own corpus-based projects,usually as a basis for their own dictionaries. Any corpus which is to be used to define and exemplify words in a dictionary, needsto be very large - at least several million words - if it is to provide sufficient examplesof rarer words to ensure that its information and examples are both comprehensiveand typical. This 'based has lead to publishersproclaiming on a corpusof x million words' as part of their promotional material. This has left many people with the impression that the larger the cotpus, the better the information it provides. While this may be true if your purpose is lexicographic, from a language teaching perspective,this is, at best, a rather misleading half-truth. The primary purpose of such massive corpora was to be the basis of descriptionsof English which in turn would form the basis of comprehensive referenceworks. It is the purposefor which they were devised,and this alone, which meant the corpora neededto be so large.

the 1n

a1n the the

:ularly rat the


Materials and resources

In a similar way, the lexicographic corpora are often describedas 'balanced', but a moment's thought reveals how careful the reader needs to be when coming across the term. It is again essential to ask for what purpose the corpus was designed; any idea of 'balance' is intimately bound up with the purpose.Supposethat 90% of all English produced is spoken and l)Vo is written; how should a balancedcorpusbe designed?Perhaps90% spokenand 70Vowritlen to mimic the relative frequency of the two modes of language use? Or would 50Vo-50Vo a better balance so each mode is equally be represented? you want a balanced cor?us of spoken English, how would If you decide on the relative proportions of, say, informal conversation,formal lectures and business meetings; and what about the age groups to be represented?Corpora need to be constructedfor specific purposes,and for languageteaching purposesa huge balancedcorpus is, surprisingly perhaps, often not the most appropriate,as Biber points out below. In 1990, in Corpus, Concordance,Collocation Sinclair statedunequivocally: We are at a very primitive stage of understandingthe character of corpora. Since then, considerable progress has been made and small scale corpora investigating particular genres,or comparativecorpora such as those used as the basisfor LGSWE have been developedand analysed.What is essentialfor language teachers,however, is an understandingthat only when the corpus, the way it is analysedby the software and the purpose for which it is being used are 'in harmony', willit be the powerful tool it can and shouldbe. Corpora for learners Learners are not amateur applied linguists and raw unedited corpus data is likely to overwhelm many ordinary learners. If teachers are going to use cor?us data with their leamers, they may need to edit by making a suitable selection of examples. At the same time, they must not edit the examples they do include. The whole point of using naturally occuring examplesis to ensurethatthe examplesreally are examplesof how the word is used; editing or 'cleaning up' destroys the value of using corpus-based evidence.The difficulties of choosingexampleswas discussed earlier.[seep 163] One warning is necessary whenever you are reading about corpus-based evidence.In the talk at the 1998 IAIEFL conferenceon which Chapter 11 is based, Michael Hoey consistently said: My corpws shows On one occasion he explicitly reminded the audience and in my corpus - it is important to emphasise that it is only the evidence of my cotpus Unfortunately few lecturers or authors are as careful as this. A typical (deliberatelyanonymous)example is: . . . will show a simple numerical listing of the most commoncollocations of the word you are interestedln. This quote is taken from a serious commentary on corpus-based materials. I have explored the materials in question, and can assurereadersthat they did not provide 'the most common collocations' of the words I was interestedin. A

con prol bes off parl conl obse Gen

Bib€ 1998 the u corp{ parn( T]

mr an




It is cl to sele likell,t

Teache - alat corpus so the l that a : and col once aE corpusl Corpor

So, doe purpose students appropn each grc patterns called 'a

Materials and resources 193

rnced" ) when )se the ith the l)Vo is :enand lguage :qually would formal to be md for rrhaps, ocally: )rpora. )orpora rsed as Ltialfor ro{pus, ; being

Genre-specific corpora

data is [o use uitable amples es rs to editing e. The ;-based : r 1 1i s )n one -itis typical ' listing s quote I have lid not din.A

Corpora for specificpurposes


Materials and resources

under the Think of the subjectmattel - chemistsregularly describeprocesses conffol of the experimenter,while studentsof fine alt tend to describe longterm historical plocessesoutside the control of any individual; a supelficial functional similarity - describing plocesses- masks a significant difference of lexical and glammatical content. Such differences are reflected in the languagewhich is typical of the particular discipline, so the differencesfrom subject to subject are much more important than previously recognised. This question can now be investigated empirically. All that is needed is a corpus of the kind of texts the chemistry comparatively small computer-based studentswill need to understandor produce during their course, and a similar corpus for the fine arts students. Both corpora can be balanced to take account of what the students will need to do - read academictexts, attend lecturesin their special subject,write an extendeddissertationor whatever.A concordancingploglam will then reveal in a matter of minutes, the words the learnersmost need and the patterns in which those words typically occur in texts related to the particular subject. It will also reveal whether the needs 'academic of the two groups of studentscan be best met using general English' materials or whether the groups would benefit considerably more from materials particular to their own field of study' Biber and his colleaguesleport just such a study, comparing researcharticles in ecology and history. They introduce the results by commenting: To this point, the analyses in this book have made use of existing corpora to analyzegeneralregister categories,such as conversationand academicprose. For studies in ESP, however, a broad sample of texts from academicprose is too general. Their analysis shows that both academic article-types differ considerably from general fiction, and both share certain features compared with fiction, but it also revealsmany important linguistic differences,for example: history articles are more narrative than ecology articles. This meanshistory texts use more past tenses and ecology articles mole plesent simples (for stating generalisations).Ecology articles use mole impersonal style, but certain parts of history articles also use impersonal style. Their analysis reveals many more details, but in summary it shows that good ESP coursesdo need 'academic English' is indeed a language and that to be subject-specific, fiction. teachers' It would be professionally incompetentto offer studentsstudying English for specific or academic purposes a diet of general English. It is equally incompetent (though, sadly,it remains not uncommon) to offer them material 'popular' science, archaeology or whatever. A from magazines devoted to At or decade more ago John Sinclair statedunequivocally: present,selections ESP textsl are made on an intuitive basis, and there is no guarantee that tof of afragment of a text is representative the book or paper it camefrom. Quite

often text. beca that. more acad prepa

Profe contn comp youn( T}

ta no Tet scr net

An er that th about use au the on closel3 themse Focus

Mark I specia their sl This n spondy easetht as $tre|

and tht importa to corrx suppliet Buildin

The so1 compari disciplir

Materials and resources 195

)n q"'D

icial )nce the

l S a

rstry rilar lake tend :r. A ; the rr in eeds
)mlc nole Lcles

often, what appears to be introductory motter is offered as typical technical text-He was criticising the useof introductoryratherthan bodj text, precisery becausethe two are linguisticalry so different. It is seriously worrying to find that, a decadelater, popular joumalism on academicsubjeciswhich is even more different from real academic writing than the introductory part of an academic article - is being offered to students of a disciprine as suitabre preparationfor their studiesof authentic discipline-specific Lxts. Professor Steve Jones who is both a highly esteemed scientist and a regular contributor of scientific articles to the Daily Telegraph newspaper,advises competitors (Daily Teregraph,Dec g 1999) in the paper's competition for young writerson science: The rules of sciencewriting differ utterly from those of writing about science. To a scientist, alr that matters is to convince an audience trained to pick holes in his argument.Every sentence must be weighed not for style, but for accuracy, every "if' matched with a ..but,,. Terseness all and elegancemuch frowned on. is ...Writing about sciencedemandsskills that most of the subject'sactuar practitioners neverbolherto acquire. An expert at both sciencewriting and writing about science,he makes clear that the two genresare radically different. Any teachers temptedto use writing about scienceas input material for sciencestudents need to ensurethat they use authentic sciencewriting instead.It should now be abundantly clear that the only suitable material for such studentsis material which resembres as closely as possibre the kind they will have to understand or produce themselves. Focus on osub-technical'collocationfor ESp learners Mark Powell has pointed out, from extensiveexperience with learners with specialistbackgrounds,thal they frequently know the technical vocabulary of their subject in English, but may welr not know the sub-technical vocabulary. This means medicar students who know cardio-vascurar and.antcyrosing spondylitis may not know collocational items such as straighten your arm, easethepairz.They may alsoneedto be wamed of impossible collocates such as *trecttthepain. This sub-technicalvocabulary lies between generalEnglish and the technical vocabulary of a particular specialis-, Jrd is of great importance to ESP learners,as it is precisery this ranguage which they need to communicateabout their speciarismto non-specialists, such as patients, supplieror customers. Building a corpus for your learners The solution to the ESp teachers'problem is nowadays rerativeryeasy.A comparatively small corpus consisting of research articles in the same discipline can be gatheredby asking departmentalmembers to supply a recent

5 A u .S

ably I1On, itory i use ltrng 'tain ;'ealS need uage h for .rally erial :r. A

tltat )uite


Materials and resources

paper on disk. As little as a dozenarticles will probably reveal key words, and some of their most common collocates in that genre' and thus provide an excellent basisfor expansion.This can be done by the teacherproviding other collocates, with the help of a lafger, mole general, corpus, a collocation dictionary, or conscious scanning of other subject-basedmaterial. Later, studentscan add further examplesfrom their subject-specificreading. While searchingthe intemet for material on the Lexical Approach, I found a site listing prepositional phrasesfrom a colpus of material on fine art. Again' a brief sample demonstrateshow useful a relatively small subject-specific coryus can be in identifying languagethat studentsof that subject will need: IN in common,in comparisonto, in contrastto, in ffict, in private, in progress, in search of, in stYle ON on, attentionon, debateon, ffict on, emphasis eyeon,focus on, influenceon, relianceon, sectionon, serieson, impact on, on display, on exhibition, on show, on loan Such languagewill probably not be noticed by learnersin their reading unless they have been trained to recogniseits importance for their own wdting. It is also clear that many of the items would have benefited by being quoted with one or two more words of the original context. It is a good maxim to rememberthat there are more chunks than you think, and the chunks are often bigger than you first think. If learners of general English have a particular interest, it is comparatively easyto download a reasonablequantity of text relating to that interestto form a small corpus, and then use the sametechniquesto provide a core lexicon of relevantto both the particular subject and wofds, collocations and expressions the particular leatner. In time, learners can be trained to do this themselves. 'vocabularyteaching'which equipsleamersto expandtheir Here is a kind of individual mental lexicons in a way which is relevant, personal and a skill which can be taken away as a tool for life. Caution is needed with raw data A corpus provides incontrovertible evidenceof language which has actually been used but such data needs to be interpreted and used with caution, particularly when used for language teaching. If teachers are going to encoulageleamersto use corpora themselves,training and evenmore caution are both needed.One principle of enormousimportance is that teachersmust become so familiar with authenticexamplesthat they can selectout unhelpful examples - however interesting or amusing these might be to the teachers themselves- and direct their learners' attention to a selection of authentic examples chosen with a particular pedagogic purpose in mind. By all means introduce learners to authentic examples, but select such examples


rh qr H





to 1o


Fa pal c7a of are





up cor finr shc cor selr ofi pro

Materials and resources 197

and ean )ther rtlon ater, nda Ialn, cific eed:

first, so leamerscan explore without being overwhelmed.Avoid 'cleaning up' the examplesyou do select and, above all, do not present learners with vast quantities of unedited data without careful preparationover a period of time. Sorting a corpus Having assembleda co{pus, it needsto be sorted to provide evidenceof how each word is actually used. This posestwo problems - what do we mean by 'each word' of the language,and how do we carry out the immense task of sorting a corpus? The second task only became feasible with the advent of computerswith sufficiently large memoriesand sufficient processingcapacity to sort a huge amount of data. The computer still needsto be programmed to locate and list together every occurrenceof a particular word, which involves deciding what we mean by 'a parl'icular word'. Is hit 'a word'? What about close?Someexamplesmight help: The Beatles'firsthitwas . . . He was hit by a motorist who failed to stop. If you hit ittoo hard, you'll break it. If he hits it too hard, he'll break it. There'sno point in hitting it like that - it's alreadybroken. Only close friends of the couple . . . We close at 8pm on Sundays. Bull's-eye?- No wayl You weren't evenclose. He's impossible- he's got a completelyclosed mind. Far from helping, the examplessimply make you less sure what the phrase 'a particular word' means. corpus linguistics introduced two new terms to clarify the situation: the term word form, is used to mean every occurence of exactly the same group of letters, so the first three examplesof hit above are examplesof the sameword form. A lemma is different word forms which are all related to the same 'underlying' word, so examples 1-5 above are all included in the lemma 'hit' and all of save, saves, saved, saving, savings, saver,sayersare includedin the lemma 'save'. Teachersneed to be aware that it is quite possible that different parts of the samelemma may show different pattems, so it may be necessaryto look up more than one word form in exploring a single 'grammar point'. If you look up feel you will inevitably find examples of the present simple, not continuous;to find the latter you will need to lookup feeling, which will also find examples of the word as a noun. corpus evidence of the word form should could be very helpful for leamers, but it is also potentially very confusing unlessthe teachershas selectedthe examples,and in particular has selectedout ('zapped') any exampleswhich will only confuse.The advantage of a corpus is the range and naturalness the examples,but that doespresent of problems if used without due care. 1. 2. 3. 4. 5. 6. 7. 8. 9.

nless It is with mto rften ively form cn of t and :lves. their skill

ually rtlon, lg to ut10n

must dpful chers Lentic ly all nples


Materials and resources

9.5 Concordances
Nowadays computer concordanceprograms which find and display all the examplesin a corpus of a particular word are widely available and relatively cheap. The most common use the so-called KWIC (Key Word In Context) systemto display results. Each line displays an example of the selectedword, with the text arrangedso that all examplesof this word are printed under each other in the centre of the screen.Text is displayed on either side of the key word, and often arranged alphabetically according to either the word immediately before or the word immediately following the key word. [There are exampleson pp. 40142.1 Using concordances with learners This format allows the user to explore the collocates and colligations of a word with comparative ease. It is usually most helpful to draw learners' attention to the words which immediately follow a word they have selectedas of particular interest.If the key word is a noun, howevet, it is also likely to be helpful to note any adjectivesor verbs which occur within a few words to the left of the key word. With adjectivesor verbs, words to the right of the key word often show grammatical collocations such as a typical preposition or adverb. It may be helpful to note that a particular verb is always, or nearly always, followed by a time phraseor some other grammatical feature. The two main benefits of concordances for the classroom are that the examplesare always contextualisedand, most importantly of all, Ieamerscan see a large number of examples of the same item very quickly. Research suggestthat we need to meet a new item severaltimes - somewherearound sevento ten times - with at least parlial understandingbefore it is acquired. In normal reading or EFL textbook vocabulary recycling, it may take weeks, even years, before a leamer meets a particular lexical item, particularly a multi-word phrasal item, seventimes. With wisely-used concordancing this in lessons. can be achievedin a couple of two-minute sessions consecutive in order. Some caveatsare, however, Someprograms show words as collocatesif they occur within a small number of words (usually about five) of the chosenword. While this can be revealing it to a researcher, can confuse learnersunnecessarily.Consider,for example, these collocates (in the broad, descriptive linguist's senseof that word) of

unl, indi edit

wor part sirni muc than usrn mos that Bria I t(
S( \l ql -' ol




th Whilr probl



This ,


Usin-e . The

He said he wasjust netvoLts,in case things did not turn out as he expected. Just in casethings don't turn out as we expect. . . I'm just doubtful that, in this case, it is the best way to . . . I'm doubtful,and just in casethings do go wrong . . . There are severaluseful patternshere:just + (adjective), in case. . . , the fixed jzsr in case,in this case andthe clusterin casethings.It seems most phrases

follc . The, notr( fluer . The


. Obse exan

Materials and resources 199

the rely )rd, ach key ord Iere

unhelpful, however, to present all of theseto learners at the sametime in an indiscriminate, albeit authentic, list of real examples.Teacherswill need to edit concordances all except advancedleamers. for Many leamers still find it difficult to skim text without stopping at every word, and such learnersneed to be taught dictionary skills so they can find a particular meaning of a word. working with computer concordances is

that learnersare not overwhelmed. rfa
;I S

Brian Pooledescribes using concordances with a group of universitystudents: I find that the use of computer concordances key verbs, in addition of to improving their knowledge of subject and object noun collocates, seemsto improve their 'feel' for the finely-differentiated sensesof a verb, and hencethe range of nouns with which it can co-occur....The great virtue of concordancesis that they provide leamers with the opportunity to seelots of examplesof a particular word all at once- not something available in day-to-day target language exposure- and to derive from this not only an awareness frequent collocatesbut also of the kind of lexical word with which it has the potential to combine. while cautioning that somelearnerswill find concordancing,too dry and the problem-solvingaspectnot to their taste,,he concludes that:

las rbe the key tor rly the can Lrch und


)'a this )ns. Lber ,mg
n ll:e ' r

iof 'd.


. - ^ l


This view endorsesthe suggestion in chapter g that noticing examples in context without formal practice helps turn input into intake. Using concordances this way has severalbenefits: in ' They often answer straightforward questions such as .what word usually follows dffirent? Is it to,from or than?, ' They sensitiselearnersto the fact that text does consist of chunks and that noticing theseaccuratelydoes help them produce languagemore easily, fluently and accuratelythemselves. ' The processboth confirms your intuitions and, more disconcertingly, causes you to revise statements such as I've neverheard/seen it. ' observing real data should quickly persuadeyou of the danger of taking examplesaway from the natural contexts in which they occur and, even more, of the danger of inventing examplesin the classroom. [seep 167]


Materials and resources

Teacherswho are interestedin using coryora and concordancingwill find an extensivediscussionof the issuesin Biber's Corpus Linguistics,which also has an extensive list of commercially available co{pora, analytical software and on-line senrices. There are three major websites at the time of writing which are relevant to teachersinterestedin concordancingin the classroom: . An extensive bibliography on classroom concordancing on Tim John's website:htp:/isunl.bham.ac.uk/johnstflbiblio.htm. . Free sample concordances either of: on The British National Corpus (BNC) at http:II info.ox.ac.uk/bnc/ COBUILDd ir ect at http ://www.cobuild.collins.co.uk/direct info. html

Sev .Tl

w ofu

lat he pa fin hei

9.6 Referencematerials
Teachers interested in teaching collocation may like to look again at the dictionary they recommendto their leamers, asking themselveshow, if at all, it deals with collocation. Dictionaries are usually used for decoding - finding the meaning of unknown words - but there is no reason why such books cannot be designedto be useful as encoding tools too. Most of the standard EFL dictionaries provide information about collocation, sometimesexplicitly aspafi of the definition, more frequently aspart of the examples.This is often of limited help becauseonly a few collocates are given, and these are often the most frequent, and thereforethe ones learnersare most likely to have met in their languageinput. 1. Conventional EFL dictionaries We testedthe main EFL dictionaries available from Cobuild, CUP, Longman and OUP using the following questions: 1. What do you call the paper you need before you can drive a car? 2. What verb goes with exam to mean you did well and the opposite? 3. What verb goes with time to say you did somethingin a good, quick way? 4. Can you say very magnificent?If not, what is the corect expression? 5. Is it say the truth or tell the truth? And what is the opposite? 6. Is there anotherverb which means about the sameas admit (a crime)? 7. How do I use the verb confess? 8. What is the usual verb before a confession? All of the dictionaries provided some help and quite often the answersto our specific questions were contained in the dictionary entry. Unfortunately, however, the information was often difficult to find, as the entries could be very long and the specific information was often buried late in the entry in one of the examples. None of the dictionaries answered question four, highlighting again the importance of the teacherbeing proactive in providing nesative evidenceof this kind.

. Se exl inf

all< loo adji

. Unt Hot dicr you dict whe whir

. The coll< abea collc lexic

In sho difficul - medi

It is wi above sample questlo than drt to comr negativt 2.Prod

Tbe Lon producti

Materials and resources ),el

dan also ruare rt to ,trn's

Severalfeaturesof the dictionary entries became apparent: ' The emphasiswas very clearly on helping the learner understand unknown wordsor expressions. ' rncreased awareness in modern dictionaries of the phrasal nature of language means entries contain long lists of phrases containing the headword.while this is helpful for decoding, it inevitably makes finding a particular bit of encoding information more difficult. Try, for example, to find whether you should sayto spare time or to save time by looking up the headword time. ' Searchingfor a specific piece of information to help you produce the corect expression involves patience and the confidencl to ignore a lot of information not related to your query. often, the user also ieeds to think in a rather abstract way about ringuistic relationships. All the dictionaries allowed us to find driving ricence, but onry if you-realised you neededto look up not drive but crriving, which was not always an immediately adjacententry. ' unsurprisingry, the dictionaries are no help in answering questionssuch as How do you say 'nycke| in Engrish? For this sort of qr-"riion, a bijingual dictionary is needed- you need to know the name in bnghsh of a concept you have in your own ranguage(nycker= key).In the ri-" *uy that the dictionaries do not help when you do not know the word, they do not help when you do not know the kind of medium-strength rexicai collocations which learnersare most likely to need. ' The examples,while natural and helpful, tend to give the most frequent collocations which are often those which rearners are most likely to know already.They are much less good at giving those medium-strengthlexicar collocations which are one of the keys to expanding learners, mental lexicons effectively. In shorl, the conventional dictionaries provide help, but it is frequentry difficult to accessthe information you want and the most usefur information - medium strengthlexicar collocations - is unlikely to be there. It is worth mentioning that the sampler on the cobuild website mentioned above provided answers to an the questions except number six, and the sample of exam did not provide an exampre of faii an exam. For the first question,as with the dictionaries,it was necessary to look up crrivingrather than drive. Most usefulry from a reamer's point of view, however, its failure to come up with very,magnificent or say the truth provides vefy sftong negativeevidencethat thesecollocations are not sanctionedby usage. 2. Production dictionaries The LongmanLanguage Activator,describedon its cover as ,the world,s first production dictionary', tries to solve this problem. But it proved no more

: the
t oll

ding ooks dard citly rften rften met


,*-1. t l L *



I trll-

tour. idins


Materials and resources

successfulin our tests than the conventional dictionaries. The main problem was accessingthe information, which involves a fairly sophisticatedview of semanticfields. The later Longman Essential Activator, which has a comprehensive alphabetical index, proved a great deal more successful and provided the answersto most of our questionsquickly and easily. It also begins to provide important negative evidence so that, for example, the entry fot true contains 'Don't sayshe'ssayingthe truth. Sayshe'stelling the clearly markedwarning the truth.' Once again, however, it is less successful in tackling the allimportant area of medium-strength lexical collocations. For these, the relatively new collocation dictionaries are much more useful. 3. Collocation dictionaries There are now two collocation dictionaries which provide evidenceof words which co-occur. They are of most use to leamers who wish to activate languagewhich they half-know as they fulfil a role similar to the way native speakersuse a thesaurus- the printed lists serve to remind you of potential altematives to your original word. The dictionaries are, therefore, similar to concordancesin that they need to be skimmed rather than studied, with leamerstrained to ignore unknown words and to searchfor words which they recognise but which have not yet passed into their active lexicons. This processis endorsedby Skehan(A CognitiveApproach to LanguageLearning) when he suggests: It is proposedthat very often the pedagogicchallengeis not to focus on the brand new, but insteadto make accessiblethe relatively new. Of the two collocation dictionaries currently available: The BBI Combinatory Dictionary (BBI) places mole emphasison the total lexical and grammatical environment of a word, which can make the entries look rather intimidating. It is thereforeprobably more useful for more academicleamers who are used to using referencematerials and for whom grammatical acculacy is a priority. The LTP Dictionary of SelectedCollocations(DOSC) focuseson precisely the noun + verb, adjective + noun, verb + adverb lexical collocations which havebeenreferred to so often in this book. Its layout resemblesa conventional thesaurusand is designed to help learners activate half-known items in the way Skehansuggests. [For a sampleentry seep 38'] Celia Shalom of the University of Livetpool, in a survey review of thesetwo collocation dictionaries (Modern English Teache4April 1999) writes: If we are in favour of a more coherent approach to the teaching and leaming of collocation, we need to go beyond incidental treatment in the language classroom and help the learner really become familiar with collocations. I think such familiarity develops best when the learner is consciously aware of this tendency of words to go together' Explicit teaching about collocation can help studentsto develop a feel




Inr Cob coll colli shor

10 7 1

freq lexir for c


If le lexic . Tel .Le

ust . Lei lex exa ele


What perce

Whar of the usetl Whar

If lea teach spoke

Materials and resources )e!

tem /of

for it. ...There is still a long way to go in the teaching and learning of collocation. [Some recent] researchon collocational awareness found that half of a sample of English teachersin Switzerland talked about collocation to their studentswhile only gvo taughtit explicitly. The publication of these collocation dictionaries marks an exciting moment in the teaching of vocabulary: one in which the company words keep is being put firmly on the agendafor teacher and student alike. In contrast, one colocation resource which I have found disappointing is cobuild's English collocations on CD-R)M. It lists the twenty most common collocatesof 10,000 headwords.The user can suppress some grammatical collocates such as the, is, in, bfi even having aone that, the program only shows the 20 highest frequency collocates,which often include words such as any, own, new. A furtherproblem arises if you look up a word with several different meanings such as order or right, as all uses of the word form [seep r97l are dealt with together.This means you may onry get one or two high_ frequency collocates for any particular use and very few medium-strength lexical collocates.Finally, the examples,although ,"u1, huu" not been chosen for classroomuse, so they are sometimesless than helpfur for learners.

the ,ide tlns
lin o


xds \,ate
tlve Ltial rto ,'ith


9.7 Summary
If learners are to move off the intermediate plateau, they need substantial lexical input. This win be predominantlycollocational. This means: ' Texts need to be chosenwith their collocational content in mind. ' Learnersneed to be trained to search authenticmateriar for key words, usually nouns,and to notice the collocationalfeatures of the .t-r"*,. ' Leamers need to be taught the particurar importance of medium-strength lexical collocations and shown how to use the texts they meet. the examplesin conventionaldictionaries, collocation dicti,onaries and electronic resourcesas sourcesto enrich their own productive language.

ot1; ical
*_D' ^ ^ l JCU



Lich )nai the
[\\ O

Discussion Questions
what percentageof the input in your crassesis written English and what percentageis spokenEnglish? Do you think the current balanceis right? what kinds of different listening activities do you use? what use do you make of the tapescriptswhich are often printed at the back of coursebooks? you Do use them as spokeninput, even though they are printed in the book? What kind of narrative texts would you consider using for input? If leamers need a rot of spoken input, do you think you shourd increase teachertalking time? If not, how can you ensurethey are exposedto enough spokeninput?


Materials and resources

Can you see a use for corpora and concordancingin your classes?If so, in what way do you think you would need to preparethe material? In what way would you need to prepareyour leamers? Once learners can use a resource such as the COBUILDdirect sampler,they can check many things for themselves which they have previously had to consult the teacher about. Can you see your own role changing so that you train students to use such resources even if it means you spend less time 'language teaching' in the narrow sense?


Pete Unir EFL rela{ impl scri! appr
devil 'unai how

coltro vocal page


Thos class empl langu of mi parti( wnte on thr

Afc Exam levels be ab expec series . Horr .Do

gIan . Dol conl abili . Dol lean

Collocation and resting 205
SO; 111

Collocation and testing
Peter Hargreaves Peter Hargreaves is head of the English as a Foreign Language part of the university of cambridge Local Examinations syndicate, responsible for the EFL exams administered by the syndicate. rn this paper he examines the relationship between a learner's vocabulary and a learnerts level, and its implications for testing. He explains how the Syndicate is using a corpus of scripts produced by learners to ensure that vocabulary testing in the exams is appropriate to the learner's level. He explains how extending chomsky's idea of deviance can give a clearer picture of what we mean by a language item being 'unacceptable'. The paper is more theoretical than most in this book and shows how the establishment of valid and reliable tests is a complex matter, in which collocation now plays an important role. Readers new to the theoretical study of vocabulary 'level'and testing may find it helpful to read peter's summary on page22I before reading his detailed discussion ofthe issues.

, they ad to t you tlme

10.1 Introduction
Those who come into contact with language learners, whether in the classroom - their teachers- or outside the classroom - for example, their employers, need to have as accurate a picture as possible of the learner's languageability or proficiency. This language 'proficiency' actually consists of many different but related abilities, such as the ability to read texts for particular purposes,the ability to listen for factual information, the ability to write letters of a formal or informal nature, the ability to respondto enquiries on the telephone,etc. A formal examination system such as the cambridge Main Suite Examinations attempts, through a series of question papers at five different levels, to gather enough information about a candidate'slanguageabilities to be able to relate them (through the award of graded certificates) to what is expectedof learnersar eachof those five levels. This immediately suggestsa seriesof important problems: . How do we assess learrer's level? a ' Do items which test particular items of vocabulary or particurar points of grammar give a representative picture of leamers' generallanguagelevel? ' Do more 'global' forms of testingsuch as testsof spokeninteractionor continuous writing provide us with enough information about learrers' abilities? ' Do specific tests of vocabulary give an accurateand precisepicture of a learner's vocabulary resource,and, equally importantly, how do such


Collocation and testing

measuresrelate to what a learner can actually do if required to use languagein real situationsoutside the classroom,for example, at work? Such questionsare part of the everydayconcernsof my staff at the University of Cambridge Local Examinations Syndicate (UCLES) who are responsible for the five Cambridge Main Suite Examinations ie the Key English Test (KET), the Preliminary English Test (PET), the First Certificate in English (FCE), the Certificate in Advanced English (CAE) and the Certificate of Proficiency in English (CPE). In this article I want to discuss some of the factors which influence how levels of language ability can be differentiated, particularly in the area of vocabulary,and questionssuch as what it meansto 'knows' a word. In the process,I will demonstratethe increasing say a learner importance which collocation plays in differentiating language ability at the upper levels, not just in relation to formal examinationssuch as the FCE, CAE and CPE, but also for teachers who work with students in the area of vocabulary, prepare studentsfor examinations,and may themselvesneed to construct tests of vocabulary.


Wht leve brol perfr 'unit

pun( '1a)

L0.2 How do we define different levels?
A leamer's level of language proficiency can perhaps be described most usefully for 'lay' users such as employers, sponsorsor parents, in terms of what the learner can do with the language at different levels, for various purposesand acrossthe different languageskills - listening, reading, writing, speaking.A great deal of work has been done on the general definition of language levels under the auspices of the Council of Europe (see, for A example, Modern Languages: Learning, Teaching,Assessment. Common - Council of Europe, Strasbourg 1998), European Framework of reference and someof theselevels are already well-defined by detailed specificationsof languageobjectives,ie the Waystage,Threshold,Vantagelevels. As a tool for the comparisonof levels of proficiency acrosslanguages,and as an aid to the 'lay' user, a series of CAN-do statements are being developed by the Association of LanguageTestersin Europe (ALIE). These CAN-do statements cover typical language functions in three categories: Social and tourist, Work and Study. They encompassthe four languageskills - listening, speaking,reading and writing, and spanfive levels of languageproficiency from ALTE level 1, correspondingto KEI to ALIE to level5, corresponding CPE. Below is an example of a typical set of CAN-do statementsdeveloped by ALTE (still in the processof being validated) covering a specific level for a particular domain and function. This set is for level 5 in the domain of Study and the function of essay-writing: Can write an essay that shows an ability to communicate with few difficulties for the reader. The essay shows a good organisational

refer grafl leanr defu partl

In th the ii the c throu impli base

What lan-e psycl you i langu

In in UCL] projer purpa EFL Amer FCE, over I by le,

Collocation and tesring )eJ

sffucture, which enables the messageto be followed without much effort.


Can presentand support argumentswell. rs unlikely to make more than occasionar errors of grammar, vocabulary or punctuation. can write with understandingof the style and content appropriate to the task. can produce text which is proof-read and laid-out in accordancewith the relevant conventions. when it comes to assessing formally whether a leamer has achieveda target level of cAN-do in the language, statementslike the ones above have to be broken down into standardised tasks, so that it is possible to generalisefrom performance on such tasks the extent to which, for example, the learner is 'unlikely to make more than occasional errors of grammar, vocabulary and punctuation'. From a linguistic analysispoint of view, rather than that of the 'lay' user, the learner's language proficiency can be characterised by reference to a set of interrelated competences: communicative, lexico_ grammatical, socio-cultural, strategic, etc. How ,proficient' the language learner is can be translatedinto: where does the leamer come in a range of defined levels for thesecompetences appropriateto the use of the languagefor particular purposes? In the cambridge Main Suite EFL examinations ,knowledge and control of the languagesystem'are viewed as underpinning the learne-r's proficiency in the different language skills and are tested explicitly - at the upper levels through a separateuse of English or English in use paper - in addition to any implicit testing of knowledge and control of the language system in skills_ basedpapers such as Reading and Writing. what exactly (or even inexactly) is involved in 'knowledge and control of the language system' at different levels is not simpty an interesting psycholinguistic question; it is a crucial one if, u. un E*u-inations Board, you are attempting to make generalisationsabout the extent or level of a language learner's 'knowledge and control' on the basis of what words that user either recognisesor producesin carrying out certain test tasks. In investigating this whole area of vocabulary testing we are fortunate at UCLES to have access to the cambridge Leamer corpus (cLC), a joint project with cambridge University press (cup), foi internal research purposes.The cLC is a corpus of written material produced by cambridge EFL candidates of several different nationalities irom Europe and Latin America in responseto tasks in the composition or writing papers in either FCE, cAE or cPE from 1993 onwards.At the time of *.ltl.rg,it consistsof over 8 million words with additions planned. The corpus subanalysed "u'L" by level of examination (FCE, cAE, cpE), and also by candidates, first

h )f e
I o

E )f


iS I.

)f )r
t1 i.

)f )r e e tr : E



Collocation and testing

language.Becausethe CLC cunently has more scripts from CPE than either CAE or FCE, and more from FCE than CAE, frequencies quoted in this article have been weighted to facilitate comparison. Invaluable though the CLC already is - and we expect it to become increasingly more useful in the future - it is important to bear in mind at all times that the data has obvious limitations. Firstly, it is entirely basedon what studentshave produced, not what they can recognise. Secondly, even as a sample of what studentshave produced, it is constrainedby the fact that the scripts are a responseto specific composition tasks, which inevitably have a strong influence on the languageof the topic areas.

Test infor reso
fhc o

CLC knon learn Mair learn relat' Hou knou

10.3 Testing vocabulary knowledge
At the lower levels of proficiency, there may be somejustification for a simple quantitative approach to defining levels of 'knowledge and control' of the language; that is, you could simply count the number of grammatical constructionsand headwordsthat a leamer is expectedto be familiar with in terms of recognition or production at a given level. So an elementary level leamer might be expectedto be familiar with, say, 1300 words such as a good vocabulary book might present and practise, while an intermediate learner might need these 1300,plus perhapsanother 1200 or so 'more difficult' (ie less frequent or more complex) words. The simple quantitativeapproachmay be studiedby comparing the frequency of occurrenceof headwords acrossCPE and FCE in the CLC and also with large corpora based on native-speakeruse. These comparisonsreveal some obviousfacts: . CPE studentsuse significantly more different words than FCE students. . CPE studentsuse more low frequency words than FCE studentswhere 'low frequency' relatesto citations in native-speaker basedcorpora. 1. Counting headwords Perhapsmore interestingly,however,the conclusionsthat can be drawn about learners' lexical knowledge at different levels from simple frequency of occurrence of headwords are distinctly limited. A search of the CLC for a number of randomly chosenwords which occur at either CPE or FCE level in the CLC and have a low frequency in native-speakercorpora (designation, unfavourable, snag, circulate, whine, stretcher, earthy, lik(e)able, puny, barmaid, self-supporting, pin-stripe) reveals the following: three occur at both FCE and CPE level (unfavourable, stretcher and lik(e)able) and seven occur only at CPE level (designation, snag, whine, earthy, puny, selfsuppofting and pin-stripe) bttt two occur only at FCE level (circulate, bannaid) - which perhaps tells us more about the social habits of FCE studentsthan about their vocabulary level!


Even u,het gam as de CPE (with cavea learnr to dal

The I the g occur claint refere

Or se A Hr rel

Ifale in bor Bearir respo the lei

Analv follou and nr

-Collocarion and testing )e)


ne all rat
i a he
, 4

Testing familiarity with a number of different headwordscan provide us with information about the 'quantity' of headwords in a l"ur.r".,s vocabulary resource,and even somelimited information about the learner,sknowledge of the grammatical significance of those words. At ucLES we will be using the cLC to develop the quantitative aspectof the characterisationof vocabulary knowledge further as part of the definition of what is typicalry expected of learnersat a given lever. we will also continue to include testsin the ucLES Main Suite examinations which are designed to probe how extensive the learner's knowledge is in terms of knowing cumulatively more headwords related to the appropriatelanguagefunctions. However, there is crearry much more to the jigsaw of a learner,svocabulary knowledge than simply familiarity with more or fewer headworos. 2. Parts of speech in vocabulary knowledge Even on the lever of simple familiarity with headwords, it is possible to probe whether the leamer has acquired some vocaburary knowLdge which has grammatical significance. Take, for example, words in th" ,a-'" family such as deny anddenial.In the cLC, the weighiedfrequencies of denyat FCE and cPE levels respectiveryarc 46 and,720;deniat(s) occurs onty at cpE revel (with a weighted frequency of 2). From this we might deduce,with the usual caveatsabout the sampling limitations of the CfC, that both FCE and CpE learnersare familiar with what is meant by ,denying,but there is no evidence to date that FCE level learnersare able to produce the nominal form deniar. The leamer'sknowredgeof what 'part of speech, a word is - whetherit has the grammatical property of being a verb (deny) or a noun (deniar), or occunence as either a noun or a verb depending on the co-text (for example, claim) - can in fact be tested directly without, of course, involving any referenceto terminology such as noun or verb, as in this example: only one of the folrowing words can occur in both branks in the sentencebelow. please circle the appropriate letter:

r1e he :al
^l cl

od reI


ruI of ' a 1n

A assert

B claim

C insist

D presume

He had the nerve to repudiated t h a t . . .

. . . that we all agreedwith him, but I totally

1'r: at



If a learner's answerdemonstrates that (s)heknows that only craim can occur in both slots, what does this tel us about her/his vocaburary knowledge? Bearing in mind that we would need to have the evidence of a number of responses such items before making any generalisation, to we can deducethat the learner's knowredgegoes beyond simple familiarity with the headword. Analysis of citations in the FCE and cpE subcorpora in the cLC indicatesthe following weighted frequenciesfor verbal (craim, craims, craimed,craiming), and nominal forms (claim, claims)of the ,word, claim:


Collocation and testing

Weighted CLC Verbal forms of claim Nominal fonns of claim

FCE 38 2





On the evidence of the large native-speakerbased corpora, verbal uses of claim appearto be about one and a half times as frequent as nominal usesin As the writing of native-speakers. can be seen from the above table, in the of claim arc generally much less frequent than the CLC the nominal uses data,but are also much rarer verbal useswhen comparedto the native-speaker at FCE level than CPE. The evidence suggeststhat we might expect leamers at CPE level, but not FCE level, to be able to use and recogniseclaim both as a verb and as a noun. One of the many caveatsthat needsto be mentioned here in connection with derivationally related words like deny and denial is that there can be no assumption of priority of one particular grammatical form of a word over another grammatical form. For example, because the noun form denial is found in CPE citations of CLC, one could not automatically deduce that forms of the verb deny world necessarilybe found or be more frequent - or 'less difficult'- than the noun form deniaL Looking through the CLC for forms of the verb insinuate, and the noln insinuation, which I will be referring to later on, I noted that only the form insinuating is found in the CLC, and at CPE level. This minors the fact that in large native-speaker corpora insinuating is more frequent than any of the other verb forms of insinuateor of the noun form(s) insinuation(s). 3. Dependent grammar patterns in vocabulary knowledge The grammar of words clearly extendsfar beyond the basic level of whether a word occurs as a verb or noun or both; it also involves the word's dependent pattems and constructions, and this is one of the most significant areas in differentiating a leamer's knowledge of vocabulary at various levels. At the risk of stating the obvious, I am distinguishing here, on the one hand, the knowledge of a grammatical pattem oI construction (eg how to form a thatclause) from, on the other hand, the knowledge that a particular vocabulary (eg item occurs with that pattemor construction thatclaim occulswithathatclause).The latter is what I mean when I refer to knowledge of the grammar of words. We already know that familiarity wtth claim as a noun but not as a verb is a likely distinguishing feature between FCE and CPE leamers in the area of production at least. We might now wish to probe whether there are grammatical pattems which occur with claim as a verb which might provide more subtle distinguishing featuresof the knowledge of the word claim. CLC evidenceindicates thaL26of the 38 weighted occuffencesof the FCE citations of claim as a verb have a following that-clartse, (They claim that road conditions aren't safe) b:ut only 2 have a following ro-infinitive construction

Kno item



S Simi


bl A

SN SI S1 The


by re prope accep This i
15 not

items betu'e as har SuchI profic which includ item r, is like patten with b dread

Collocation and testing lll

(she claimed to have afish bone in her throat). These figures contrastwith 6g occuffencesof claim + that-clauseand l7 of claim + ra_infinitive in the cpE citations. Knowledge of these two grammatical patterns of claim courd be tested in items such as the following: only one of the following words can occur in both of the blanks in the two sentences below. please circle the appropriateletter: A believes She . . She . . D claims . that she is more accuratethan her sister in her work. to be more accurate than her sister in her work. B tends C boasts

Similarly, words with more comprexpatterning, such as the three patternsfor remember,which all occur at cpE level in cLC, could be tested with items such as: only one of the words in A, B, c, D is appropriatein all three of the blanks in the three sentences below. please circle the appropriateletter: B agree C suggest D admit She didnot. . . . . . . . posting rhe letter. Shedid not. . . . . . . . topost the letter. She did not. . . . . . . . rhat shehadposredthe lerter. The difference between multiple co-text test items such as those in the example above and traditional test items based on choosing a word to fit a single sentence,is that the implicit assumptionthat knowlJge of a word is cumulative as the leamer reachesa higher level is made explicit. This is done by requiring the leamer to demonstrate greater knowledge of a word,s propefties and patterns, signalled by selecting or producing a word which is acceptablein a range of different co-texts. This is not to underestimatethe fact that testing pattems in depth in this way is not entirely straightforward. Take, for example, the question of level. Test rtems constructedfor a particular level of examination need to discriminate betweenlearnersclustered around that level with a view to classifying them as having adequate,good, exceptional, etc ability in relation to that rever. such testsare not designedto discriminate acrosswidely separated levels of proficiency. Hence, sinceclaimfollowed by a that clause is a feature of claim which is expected to be known at FCE level, it is probably redundant to include this pattern as a cpE lever test item. For a cpElevel test, a vocabulary item which occurs in cLC only at cpE and cAE levels and not at all at FCE is likely to discriminate better. For example, knowledge of the grammatical patterns of the verb dread, which is found in cLC at cpE and cAE levels with both to-infinitive (old people dread to go) and with verb + ing (I even dread thinking about the winter), might be testedinstead of claim: A remember


Collocation and testing

Only one of the following words can occur in both of the blanks in the below. Pleasecircle the appropriateletter: two sentences D hesitate C dread B refuse A avoid the I. . . . . . . . to contemplate future. I . . . . . . . . contemplating the future. The beneficial effect on vocabularylearning of suchitems is the way in which they illustrate the layers of grammatical patterns which make up knowledge of the grammar of words. An equally important teaching point that should be made is that it is comparativelyrare for apparentlyparallel constructionsto be freely interchangeablewithout some change in meaning or some restriction on use.An obvious example,repeatedfrom above,is the difference between posting the She did not rememberto post tke lettef Ske did not remember that shehad postedthe letter,wherethe first letter, andShedid not remember entails that she did not post the letter, in the second it is an open question whether she postedthe letter or not, and the third entails that she did post the letter. There are also more subtle differences associated with grammatical properties, to which I shall return when discussing collocation below. For example, when dread is followed by ro-infinitive the verb in the infinitive is usually in the semantic field of imagination such as contemplate,see, think and under the control of the subject of dread. For something out of the control of the subject such asfall ill, the verb + lrzg construction with dread seemsmuch more natural. For example: ? I dread to fall ill while I am travelling. I dread falling ill while I am travelling. There are practical problems associatedwith such test items with regard to sustainability,by which I mean the difficulty of constructing such items over a sustainedperiod, so as to ensure consistent and adequatesampling of the learner's knowledge of the grammatical properties of words, and not just testing what is relatively easy to test. 4. Collocations and vocabulary knowledge So far I have dealt with three important pieces in the jigsaw of the leamer's lexicon - simple familiarity with a quantity of headwordsused in connection with the functions and topics appropriate to a particular level of language proficiency, and two aspectsof familiarity with the grammatical propertiesof words: their word category and dependentpattems and constructions.That still leavesa lot of unchartedor unpredictableterritory in the jigsaw, and this is where collocation entersthe frame. 'collocation' has been well established in the description of The term 'colligation' as languagesince the days of Firth. It is usually contrastedwith in the definition in Robins (1964:234):



To rl to-rn colli

Alth xnpc The spec a gra we el end r

At th gene refer 1965 'gran


Chon borde
\ !


The 1 deviar illustr

(3r (31

Accor of wot becau have t the rig but hz subjec

The dr rules i involv Chomr

Collocation and testing 213

Groups of words considered as members of word-classesrelated to each other in syntactic structureshave been called colligations to be distinguished from collocations which refer to groups of words consideredas individual lexical items irrespectiveof their grammatical classes and relations.
'hich edge ld be to be ltron teen i the

To illustrate the distinction, consider the examples discussedearlier: verb + /o-infinitive is a colligation, dread + think a collocation which exemplifies the colligation. Although the colligation/collocation distinction is a valid and useful one, it is important to keep in mind that there is no sharp dichotomy between the two. The description of English involves very general rules at one end of the spectrum,which we call grammar, and from thesevery generalrules there is a gradual move through a continuum of more and more qualified rules until we end up with particular statements about words at the lexical or vocabulary end of the spectrum. At the grammatical end of this continuum, the gradual move from highly generalised rules to more qualified rules can perhapsbe usefully illustrated by reference to chomsky's discussion of deviant sentences (chomsky I965:l52tf). Having spent 40-50 pages of the densest analysis on the ' grammatical' sentence: (1) Sincerity may frighten the boy. chomsky then discusses sentences (2), which he describes havins ,a the in as borderline character': (2a) The boy may frighten sincerity. (2b) Sincerity may admire the boy. The fact that these sentenceshave a 'borderline'

stl0n it the tical For r,e is think f the tread

rd to

f the Just

rather than a ,clear-cut' deviancy is explained by chomsky in terms of hierarchy of ,violation' illustratedin (3): (3a) Sincerity may virtue the boy. (3b) Sincerity may elapsethe boy. According to chomsky, (3a) is most 'deviant'because involvesa violation it -virtue is a noun wherea verb is needed. of word category t3ul is lessdeviant becauseelapse is a verb but it has the wrong syntactic feature as it does not have the pattern verb + noun (object). (zb) is least deviant becauseadmire is the right word category (verb) with the right syntacticfeature (+ object noun) but has the wrong 'selectionalfeatures' becauseadmire only occurs with subject nouns which are human, such as boy, not abstractones like sincerie, The degreeof 'violation'becomes less as we move from the highly general rules about word categoriesor parts of speech to the more qualified rules involving what chomsky refers to as 'selectional features'. However, chomsky's 'hierarchy of violation' does not deal with the vast area of

net's ctron uage es of That I this nof n'as


Collocation and testing

'deviancy' resulting from using unacceptable collocations which ale not coveredby violations of'selectional featufes'. Just becausea verb has the 'selectional feature' of occurring with abstract subjects and human objects, this does not mean that all abstractsubjectsare appropriatewith such verbs. As can be seen from the examples below, both seize and consume,unlike admire, do occur with abstractsubjects,but not, it seemswith sincerity: Fear may seizethe boy. ? Sincerity may seize the boy. Envy may consumethe boy. ? Sincerity may consumethe boy. While vocabulary knowledge may involve a number of qualified rules of the 'selectionalfeatures',which excludesomecombinations kind Chomskycalls of words, most of that knowledge involves simply having acquiredor leamed specific combinations of words - collocations - which are acceptable,while others are not. A secondpoint to note in relation to the colligation/collocation distinction is that knowledge of a collocation, if it is to be used appropriately,necessarily involves knowledge of the pattenis or colligations in which that collocation can occur acceptably. 'individual lexical items When Robins (see above) refers to collocations as irrespective of their grammatical classesand relations', he is highlighting a key characteristic of collocations which is that they can cut across grammatical relations. So, for example, the collocation slow-walk is equally evident in the colligation: verb + adverb, walk slowly, as it is in the colligation: adjective+ noun, slow walker. However, it is important to keep in mind that the colligational options for a particular collocation are not unlimited, and vocabularyknowledge includes knowing what the options and limitations are. For example, the collocation short-walk as in he went for a short walk, doesnot occur acceptablyin the colligation verb + adverb- ?walk shortly or with the -er form of the noun as in ?he is a short walker. To suggest that there is a different walk here from the walk in slow walk merely begs the question as to what constitutes the meaning of a word and what role collocation has in defining that meaning. It is particularly evident that many noun collocations are consffained with respect to whether they occur in subject or object relation to a collocating of verb; many nouns collocate with particular verbs or subclasses verb either as subjectsor as objects,but not both. Hence,as they developtheir knowledge of collocations, leamers need to be aware that this knowledge must include the grammatical patterns or colligations in which the collocatescan occul Referring again to the verb claim, with few exceptions this only collocates 'human', reflecting the real-world situation that with subjectnouns which are 'people claim things', not the other way round. it is generallythe casethat



Bo col \\o an



COT n/n r""

cla coi

AI \\ 1t

- f 'oc


A " bot re\ ran co{ 'nu

uhl pro fror





Collocation and testing /lJ


There are, however, exceptionssuch as these based on examplestaken from the CLC: There were many things that claimed my attention. . . . plague and tuberculosis,which claimed many lives . . . Both of these examples, as it happens, illustrate the extended nature of collocations. In the first, we fail to record something significant about the word claim if we note simpry that it occurs with thing"asisubject noun; it is an important fact about the collocationar characieristics lr chim that inanimate subjectssuch as thing canoccur in subjectrelation wrthcraim only when claim occurs with the noun attention as object. In other words, thing occurs as subject with the extended collocation ctaim_attention (cf the comparabre extended collocations: merit-attention, deserve_attention, command-attention).Similarly 'diseases' and other ,ills, substitutablefor plague and tuberculosis collocate with craim as nou, ,rbi*s only when claim is in the extended collocation craim-rrfe (ct tne comiarable extended collocations: take-life, account for_ life).



ed ile
1S lv


10.4 Grammaticarpatternsand corlocations in testing
; able to produce a particular collocate rishing separatemeaningsof that word o be shorthand for somethins like:



llv he

cases conerare disrinc,,"".";T::;;in:["rTons' also with

whichin som"

rot nd 'a tlk

he r1e ith



IC \

hrar nd.

A word rlke correspond, for example, has (at least) two distinct meanings _ both attestedin the cLC - 'exchangeletters' and 'match,, a factwhich is revealed by the occurrenceof the word correspond withtwo very different ranges of colligations and associatedcollocaiions. with its first meaning correspond can occur in the progressive form and is typically restricted to 'human' collocates: The two frieirt, hor" been corresponding for some time, whereas with the second meaning correspond typicalry does not have a progressive form but has a greater variety of collocaies exemplified in: correspond to/with (the) truth/needs/advertisement/reality/expectations (all from the CLC). The distinctive features of a word rike correspond couldbe used to test the extent of a leamer's knowledge of this word at different levels: Fill the blanks by using each of the following once only: A correspondwith C correspondto B correspond D are corresponding l' I am happy to say that things have quietened down in the hostel now that Louisa and pablo . . regularly again. 2. It is interesting that your views and mine . . much more closely now than they ever did in the past.


Collocation and testing

3. You can hardly call yourselvespen friends when you eachother about once aYear. 4. It's handy that your expectationsof what you will get from a g o o dh o l i d a y . . . . . . . m i n e . . Apart from such examples where colligation and collocation features can distinguish basic meanings of a headword, it is the distinctive andlor overlapping ranges of the collocating items with each word which map the detailed contours of knowledge of that word. Above I referred to the continuum from grammatical to lexical information about words. At the lexical end of the continuum there is also a gradation of relationships that collocating items have with a word - a gradation which moves from loose to close or vice versa. and Second This is brought out in an articleby Howarth (1998):Phraseology table below, based on nativeLanguage Proficiency and illustrated in the speaker data. Using a variety of formal criteria, Howarth categorises collocations according to their degree of restrictiveness,thus identifying degrees of conventionality, in particular the use of a verb in figurative, technical or de-lexical sensesor according to the degree of limitation on the permitted substitutions.He stresses imporlance of seeingthe categoriesas forming a continuum rather than discrete classes'
combination COMPARE EMPHASISE INFLUENCE INTRODUCE PAY MAKE GIVE DRAW SET behaviourl levels, results, size autonomy, concept, link, rights content,culture.groups bill, amendment,motion attention, heed decision, imProvements credit to sb, Preferenceto sb line sroreby srh category free collocations

coff ame 1ir e


Nod leveI the r conit mto.

HouI collor
IJ. --



co hir


Level I collocations restricted restricted collocations Level 2

restricted collocations Level 3 figurative idiom pure idiom

One r colloc highel ar aila examF caullo purpol needs
+L^^^ tllc5g I


I have attempted in the table on the next page to illustrate how such a continuum might apply to a single vocabulaly item such as the word claim, already used extensively above in discussing glammatical pattems and extendedcollocations. Most of the examples in the table involve claim as a verb, so the words listed occur as nouns in object relation to claim; two of the examples involve the noun form of claim. All the examples apart from the figurative category are taken from the CLC'

one \\ { absolu standa any oc non-de name1 'some,

I also l verb+lr that wc

Collocation and testing )lJ

Combination with claim
compensation,benefit, allowance attention, credit lives stake (stake a claim) lay (lay claim to - not *lay a/the claim



Restricted collocations levet 1 Restricted collocations level 2 Restricted collocations level 3 Figurative idiom Pure idiom



td 3
)S lq

Howarth suggests that learners'lexicaldifficulties lie chiefly in the restricted collocations'since idioms and free collocations are largeryunproutematrc,. He concludes: A comparisonbetween fNative-speaker]performance and fNon-native_ speaker]errors suggestthat at an advanced level leamers are lexically competent and rrave successfully internalised the more restricted co'ocations and semi-idioms. There remains, however, the vast hinterland of less restricted combinations ...


10'5 sources - native-speakercorpora and dictionaries

'iL lij



one of the most fertile sources for sampring the grammar of words and collocations appropriate to different levels of proficiency, especially the higher levels, are the large native-speaker co{pora, part of which a'e now available in some limited form via th. Int"*"t. These provide a range of examplesand useful frequencydata but they must be usedwith considerable caution - it is always necessaryto bear in mind the differences between the pulposes for which colpora were produced, and the students, vocabulary needsin relation to the specific sylrabusor rearning gout. tn"f -e following. My searchesof the large native-speakercorpora to date have confirmed that thesecorpora provide only a parlial selection of all the pattems or collocates one would expect to find for a given rexical item - rro. do tt with an absolute guarantee that what you do find in them would "y".o-" Le accepted as standardEnglish. Searchingthrough the occurrences of sincerityto check on any occurrences ,rinceriQas subjectnoun of with frighten as in chomsky,s non-deviant sentenceabove,I was amused to find thatlhere was one citation, namely chomky's own sentence quoted from a book on languug" a, a 'somewhat peculiarexample'. I also found among the dependentpattems of sincerfiy an example of for+ verb+ingas in he had no forfincring sorutions- not, I think, a pattem -sinceriQ that would be judged to be standaidEnglish. And as a diversion,I stumbled


Collocation and testing

on a special kind of millennium bug in the large native-speakefcorpofa: it that about one sixth of the large number of citations of millennium are appears 'n'. Should we take the fact that this spelling with one spelled with only one 'n' occuls in significant numbers of native-speakercitations as sufficient proof that millenium is an acceptablealtemative spelling? If occurrenceof a dependentpattern oI collocation in the large native-speaker corpora does not necessarily gualantee that the pattem or collocation is standardEnglish, it is also true that non-occulrence of a particular dependent pattem or collocation in the colpora (or occurrencein less than a statistically that such a pattem or significant frequency) is by no means a gLrarantee collocation is non-standardor unacceptable. Retuming to an earlier example, here is the entry for insinuation in the LTP Dictionary of SelectedCollocations(DOSC): INSINUATION V: defend oneself against,deny, make, (dis)prove,reject A: horrible, nasty, serious,unfair Where V: indicates that the cited verbs or verb phrases collocate with insinuation as object, and A: indicates that the cited items collocate as adjectives wttt insinuation. Examining large corpola' however' I compiled the following: V: ooze.resent.wince at A: astronomical,clear, disparaging,filthy, foolish, heinous, hidden, indirect, silent Intuitively, the DOSC list of collocations looks to be a familiar selection of verbs and adjectives that one might expect to come acloss or use with insinuation, but the overlap with the comparablecollocations in the cotpora happensto be minimal. Native-speaker corpora are clearly very valuable as sources of authentic learning and testing material, and for checking on frequencies,and typical cotexts of lexical items, such as the fact that deny, referred to above,is almost invariablyusednegatively- I cannot deny . .. - or in questionform such as Who would deny . . . ? UCLES' approach to producing items for testing English in use at various levels, whether with a grammatical or lexical focus, 'corpus-informed" not corpusis what is sometimesreferred to nowadaysas based. When it comes to dependentpattems and collocations, the nativespeakercorpora need to be used in conjunction with other referencematerial such as advancedgrammarsand collocation dictionaries.

CI ler wc col 1ev an(

an( col difi the

ata col Gol


acc a-crl ask cha con


g1\-t harlmp hllr offe stai( The DOI moa

The oprn


L0.6 Sources the learner corpus(CLC)
As I have suggestedabove,the CLC is already proving invaluable in helping us at UCLES to build up a picture of the vocabulary knowledge associated (albeit in productive medium) with different levels of proficiency (FCE, CAE,

justi obje resp shar valur Fron

Collocation and testing )l) it 1e


is nt ly




of rh

cPE), for example, in terms of number and frequency of headwordsat each level and evidenceof the range of grammaticarpatterns used with particular words at each lever. It is certain to prove equally valuable as a source of collocations which might be expected to be known by studentsat different levels, especially when used in conjunction with the native-speaker corpora and well-researchedcollocation dictionaries. A particularly interesting and fruitful area of collocations for both teaching and testing pulposes, emerging from searches of the cLC, are those collocations which are made up of words which individuany have a very different frequency from their frequency as collocations. Take, for example, the word opinion' As one might predict, opinion is very frequent in the cLC acrossall three levels FCE, cAE, cpE becausecomposition tasks frequentry require the candidateto expressopinions. The verb/o rm alsooccurs in cLC at all three levels, least often at FCE (10 weighted occurrences). However,the collocation/orm-opinion occursonly at CpE level (9 occurrences). Going through all the verbs listed in DoSC with opinion as object, the weighted frequenciesin CLC are as follows; CPE CAB FCE accept 5 L agreewith 6 z 47 ask (for) 6 6 8 /1 change 10 8 confirm z express 10 40 10 form 4





44 have 70 J+ 94 rmpose 1 influence 1 L offer 1 state 1 There are no entries in the cLC at any level for the following verbs listed in Dosc as collocates of opinion: air; convey,discount, dissenT from, encrorse, modify, mould, seekout, stick to, sway, trust, venture, voice. The verbsfrom DOSC arenot, of course,the onry verbs which collocatewith opinion as object. Here are some other examples from the cLC not in the DOSC entry for opinion:




a t




n_q _ D

ed F

justify 1 2 object to 1 _t respect 1 2 share 9 9 I2 value 2 From this kind of data we may be able to staft building up a picture of


Collocation and testing

incremental ranges of collocations which are significant in characterising lexical knowledge at different levels of proficiency, especiallythose involving vocabularyitems, which, looked at in isolation, have frequencyprofiles which would not automatically lead us to associatethem with different levels of proficiency.

tor cun One

10.7 Approaches to testing collocation
Testsinvolving recognition of appropriatecollocations are a standardpart of UCLES' examinations. Some typical examples of collocations from CPE Paper 1 are listed below together with the relevant test item. The test focus parl of the collocations is in bold: 1. breach-code (of ethics) the medical profession's code of Any doctor who . . rePrimanded. ethics is severely D breaches C ruPtures B cracks A fractures 2. pursue-point She obviously didn't want to discussthe matter so I didn't . . the Point. D chase C Pursue B follow A maintain 3. lend-weight (to) These documentslend . . B weight A depth to the reporter's accusations' D gravitY C volume

Ider that the I othe blan a pr{ sele



4. power-wane This is the author's tenth book and it is clear that her creative powerhas... B dissolved C suspended D dispelled A waned It should be noted here that in selecting the appropriate collocation, the leamer also decides that the so-called distractors,when inserted in the gaps, *to actually form deviant expressions,such as *to rupture a code of ethics, chasea point, *to lend gravity to something.It is worth noting, therefore,that successfulperformance on such test items involves knowing both what is possibleand what is not. Testing collocations in this way has worked well enoughover the yearsin that performance on such tests correlateswell with other measuresof high-level language ability used in the CPE. As part of the process of developing a revised version of cPE (due to be introduced in Decembet 2002), work is being carried out on colpora to check on frequencies,among other criteria, for establishingappropriatedifficulty levels of such collocations. Experimental test formats are also being trialled, which attempt to probe the cumulative nature of lexical knowledge as evidencedby the learner's ability

It shr candi not aq a ranl for er ler-els
irrrr t}


on the


So. hc atteml framr produl

1. Fan criterii in la1' For ex levelsr corpor

Collocation and testing /)l


ch of

to recognise not just one collocation (as in the examples above from the current cPE Paper i), but a range of collocations in which a word occurs. One of theseformats is illustrated below. Identifying the most appropriate answer for this item involves recognising thar opinion collocatesin suitable contexts with impose, with popular, and in the phrase in a matter of opinion, all of which are taken from Dosc. The other nouns (fashion, feeling, will) fit appropriately in one or two of the blanks, but not all three.A variant on this type of item is also being trialled as a productive test, where the candidatehas to produce,rather than recogniseor select,the appropriateword that fits two or three contexts. Circle the word which fits in alt three sentences: B opinion C feeline D will (a) You cannot simply come in to an existing situation and lmpose your . . on everyonelike that. (b) Though he may have good reasonsfor introducing such measures, popular . . is likely to prevent them from workins. (c) Shemay insist on such a dresscode in the office, but whether it's correctto do so is a matter of It should be noted that if this type of item is going to be used to test candidates'vocabulary knowledge at one particular level, such as cpE. and A fashion

of )E

he ps,

just the one that is at the appropriatelevel, given that it might reasonablybe assumedthat the two lower level collocations should be known to the leamer on the basis of correctly identifying or producing the higher level collocation.


10.8 Summary
So, how imporlant is collocationin testing the learner'sproficiency?I have attempted to answer this question by placing collocation within the framework of vocabulary knowledge - whether for recosnition or productionpulposes- summarised below: 1. Familiarity with increasing numbers of headwords which can on some criteria be classified as appropriateto incremental levels of proficiency (or, in lay terms, more difficult). For example : puny andsnag. These are describedas 'appropriate' to higher levels only on the criterion of low frequency in both native-speaker corpora and the CLC.

,^t i Et


.L^ .11C



Collocation and testing

2. Familiarity with distinct but related lexical items which are part of families of words. For example: the noun denial andverb deny. 3. Ability to distinguish different grammatical categories of the same word. For example: knowing claim occrttsas both a verb and a noun. 4. Ability to distinguish colligations of words' For example claim + ro infinitive and claim + that-clatse. 5. Ability to differentiate basic meanings of a word using knowledge of its colligationsand collocations. correspond(a = write letters;b = match) For example'. 6. Incremental knowledge of collocations of a word including different degreesof restrictiveness,and the ability to recogniseidiomatic uses. For example: claim-bag gage, claim- compensation, claim- credit, claim-lfe, stake a claim, lay claim to. I have also discussedin this article soulceswhich UCLES is making use of to ensure systematic and consistent sampling of vocabulary knowledge, especially at the upper levels of proficiency. The importance of collocation in the framework of vocabulary knowledge should be clear; the extent of the contribution which knowledge of collocations makes in differentiating incremental levels of proficiency is still an area for further research. At UCLES we will be continuing our investigationsinto the areaand how to test it validly and consistently,making use of a variety of sources,including collocation dictionaries and both nativespeaker and learner corpora. At the same time, validation projects are continuing under the auspicesof ALTE to link leamers' performance in the formal testing situation with their CAN-do abilities outside the classroom.


E 11 R

Discussion Questions
'vocabulary test' for your class, would it concentrate on If you make a anything other than words they have studied recently? 'new words'in the input they don't only think of learners How can you ensure meet? From a teaching perspective,Peter Hargreavespoints out how complex the idea of 'knowing a word' can be. Which of these aspectsof vocabulary in building do you emphasise class: a. new words b. word grammar c. collocations Do you think different vocabulary building strategies are important at different levels? If so, what strategiesare most important at theselevels: a. elementary b. intermediate (FCE) c. advanced(CAE) d. mastery (CPE)

Collocation and testing /)!

Chomsky A. N. (1965)Aspecrs the Theoryof Syntax, of MIT press HilI, J. & Lewis, M. (Eds) (1997)LTp Dicrionary of Selecred Collocations, LTp Howarth, P. (1998)phraseology secondLanguage and proficiency, AppliedLinguistics, vol. 19,No.1pp 2444 Robins,R. H. (1964)General Linguistics: introductory an suruey, Longman


)f I
Itr (r


t]3 te


te n'


A world beyond collocation

A world beyond collocation: new perspectiveson vocabulary teaching
Michael Hoey Michael Hoey is Professor of English at the University of Liverpool and author of the award winning Patterns of Lexis in Text, ouP 1991. He is a descriptive linguist rather than a language teacher. In this more theoretical paper, based on the talk he gave at IATEFL Manchester 1998, he discusses his own language learning in the light of his theoretical insights into lexical patterning. Teachers may find it interesting to see how research, which initially appears remote from the classroom, can have practical consequences,particularly for the content of teaching materials. He suggests that learning individual words is relatively inefficient; that learning collocations is more efficient, but as our understanding of text increases, it may be that colligation will play an increasing role in classroom materials and teaching.

pafl ma5 ofp bel pra( avel

The com (aB onl Hug thou cout sulte and u-ho mad voca

11.1 Learning new words
to unit dedicated Applied Linguistics,I Despitebeing the head of a language am not myself an experiencedlanguageteacher.The little experienceI have of languageteachingis well out of date and forms no basisfor giving practical advice to anyone. I am, it is true, surroundedby staff who daily renew the connectionbetweenlinguistic theory and chalk-faceteaching and I learn from them, but this does not alter the basic fact that I carry no authority as a practitioner of TEFL. Before my honesty leadsyou to skip this chapter,let me quickly add that I do, however, have another kind of experience that may partially compensate.I may have no language teaching experience,but I do have daily experienceof being a languagelearner. Every weekday,more or less without exception,I attempt to leam a language. I give a minimum of 20 minutes a day to the task and often more. I work set meticulouslythrough setsof materials,listen to tapes,read the passages (occasionallygoing beyond the textbook), tackle the exercisessuggested, memorise lists of words, and engage in painful conversation with anyone competentto listen and reply. My motives are a genuine desire to command another language,a wish to speak something of the languageof any country I may be visiting, and a determination to put myself through the hoops that languagelealners go through in order better to understandtheir problems. Sadly, I am not a successful learner - there is no language that I could truthfully say I have a command of, though I am an adequatereader of a couple. My first goal has therefore not yet been met and my second only

Suen them of sc



accon archi carpE Num one two three

slxt) comfl


Collo examl

it, tha imme and to the

A world beyond collocation


partially so. Paradoxically, my failure to become a super-leamer,capable of mastering languageswith consummateease,meansthat my third objective of putting myself in the shoesof the averagelanguagelearner - continues to be fruitfully met. The experience I bring to a book like this, aimed at the practising language teacher, is that of the typical well-motivated but only aver agely competentleamer. The materials I use vary greatly in style. Some are structural in style, some communicative, some eclectic. over the past year I have worked with Sueiios (a BBC courseon spanish) A vous la France (anotherBBC course,this time , on French), swedish in Three Months (an optimistically titled course from Hugo), Cantonese: completecourse A for beginners(a Teachyourself Book, though a far cry from the courses they used to publish) and a Routledge course entitled Colloquial czech. All of these have merits; none is perfectly suited to my needs.I come to them with researchexpertisein text linguistics and lexical studies.I find nothing that relatesto the former in any of them; the whole discourseperspectivemight never have existed for all the impact it has made on the authors of these materials. They cannot, however, avoid vocabulary,so the question arises:how do they chooseto teach vocabulary? suefios, A vous la France and to a lesser extent Colloquiat Czech favour themed word-lists, by which I mean that the lists are constructedon the basis of some common semanticproperty, such as colour, measurement sporting or activity. Abbreviated examples from Suefios(pp 254-5) are the followine: Occupations actor/actress accountant architect carpenter Numbers one two three

r e n e
S 0


v c

actor/actriz contable (Sp) ; contador/ora arquitecto/a carpintera eIc. uno/un; una dos tres
sesenta sesenta y cinco etc.

L, d

sixty sixty-five


suefiosalso usesunthemedlists (lists madeup of words that havenothing in common) to accompanytaped examples.Both reachyourself cantonese and colloquial czech use such lists as their main way of teaching vocabulary,for example,from page 16 of Colloquial Czech:
it, that, this to immediately, at once hned and a to the left vlevo

o a


very to the right then

velmi vpravo potum


A world beyond collocation

What holds such a list together is simply its relevance to a piece of taped dialogue that the leamer is expectedto be attempting to make senseof. The list's main virtue is that each item it contains is readily contextualisedby reference to the tape. On the other hand, as can be seen, the items leamed form no system. The lists tn Swedishin Three Months are of both the kinds mentioned,but in addition some are structurally oriented.An (abbreviated)example is: Here is a list of some of the most common deponentverbs: to breathe to be, to exist to hope to vomrt to succeed andqs finnas hoppas krrikas lyckas

Noh Ther

Hav Therr

I sus itisr to er

This of elr some undo leris l


In so far as Cantonese,Colloquial Czech and Swedish in Three Months tely on lists of this kind, they are clearly adopting a conservative strategy for vocabulary teaching,one tried and trustedfor centuriesbut one that takeslittle account of what this book has been about, of what we now know about the way words work. Suefios and A vous la France are more adventurous in this respect. Both employ shadedsectionsthat talk about the way one sayssomething.Theseare useful, for example: sometimes To ask for a ticket you can say: Quisiera (comprar) un billete para (ir a) Valencia. I'd like a ticket to (go to) Valencia. For the type of ticket you want, say: wn billete de ida (y vuelta) a single/returnticket un billete defumadores/nofumadores a smoker/non-smoker Sometimes,on the other hand, these sectionsare more contrived: say: To ask about seasons, Cudntas estacioneshay ? are How many seasons there? Cudndo es el invierno/verano/etc? When is winter/summer eIc? I And to answer, say: El invierno es de diciembre a marzo. Winter is from Decemberto March. Hay...estaciones. Thereare...seasons.

u ithc conil becau the ca usage e-\prel then t


In difl from tr

of rl hr that nl a u-0rl made empla

You nr list ir words emplo


each r reveal belou' profes disting yesteft collocl

A world beyond collocation


)ed lhe by red

No hay estacionesdefinidas. There aren't well-definedseasons. Hay una temporada lluviosa y una temporada seca. There is a rainy season and a dry season. I suspectthat many leamers want to know how to buy a train ticket, but that it is only ecologists and geographerswho have a burning need to know how to enumeratethe seasons. This doesnot exhaustthe strategies thesecoursebooks - thereis use also use of exercises encourage to guessing the basisof similarity of soundand also on some discussion of related words - but the strategies I have illustrated undoubtedlydominate.How successful, then, are thesemethodsof teaching lexis? I can only speak for myself as a very average reamer, but I am consciousof quickly forgetting themedlists. Becausethe words are learned without referenceto any context in which they might be used,they tend to get confused with each other. unthemed lists work somewhat better fbr me because the accompanying tape contextuarises words I am rearning,but the the contextualisation is apparently accidental and varies in usefulness. The usage notes likewise vary in usefulness- if an opportunity to use the expressions arisesreasonablysoon after they have been leamed, then these expressionstend to stick, but if on the other hand no such opportunity arises, then they drift out of memory.


for ttle the


11.2 Why word lists are dangerous
In differing degrees,all the methodsfor teachinglexis I have mentioned suffer from two weaknesses, of which I shall mention immediately and one the other of which I shall spring upon you later. The first and immediaie weakness is that none show any awareness the importance of collocation, the of company a word keeps, an important matter, as previous chapters in this book have made abundantly clear. with that in mind, look again at the list of types of employment quoted above from Swefios. You might reasonablysupposethat you would encounterall the words in this list in very similar contexts.you might reasonablypredict that employment words llke architect and accountont would share many collocates employ(ed), work(ed), good, trained as spring to mind as obvious possibilities. Yet, as you would now predict having read the rest of the book, each word also has its own collocates.sometimesexamining large corpora reveals some surprises:actor, for example,has these collocates: [see note belowl best, former, (perhaps an indication of the insecurity of the acting profession),director, and, most surprisingly, sir, in examplls sochas The distinguished actor sir lan McKellan was invited to Downing street yesterday-.. In contrast,accountanl does not have sir as one . of its major collocates;carpenter, which is considerablyless common than the other


A world beyond collocation

words in the list, has aged, father, son, among its main collocates. These differencescan be explainedin terms of the word's characteristicgrammatical pattems, of which more below. For the moment, we can note that the collocatesof carpenter do not refer to the job, unlike the collocatesof all the other words on the list.
[Notice that this paper uses a different, more technical, definition of collocation than most of 'words which occur within a few (six) words the others in this book. It defines collocation as on either side of the headword in naturally occurring spoken or written text': under this definition actor and Sit; carpenter mdfather are collocates.Although this technical definition seemsa long way from the classroom,it can reveal pattems which are helpful for teachersand materials writers, see Chapter 7. Ed.l

What all this suggestsis that themed lists hide great variety of use in a spuriousconformity. It is reasonableto assumethat the collocatesof all these words will similarly vary in Spanish(or any other language),though of course that remains to be proved. If they do, learning the words in a list will not guide more imporlantly and the learner into producing natwal-sounding sentences: less tendentiously, list-learning will not help recognition of the words in reading, nor will it help with guess-workfor accompanyingwords. So is this weaknessalso evident in EFL textbooks?The answer is a modified 'Yes'. If we look at Headway, it avoids lists as such and sometimesgoes well beyond them. One strategy used (New Headway English Course: guesswork: Bookp 30) is to encourage IntermediateStudent's Work in three groups.

Ii hl


sfi o0 sa

Group A Group B Group C

Read about the writer. Read about the painter. Read aboutthe musician.

Read your extract and answer the questionsabout your person. Try to guessthe words underlined from the context. Then use your dictionary to check the words. This activity is inherently good, encouragingas it does sensibleguessingin a natural reading context and inviting the learner to make use of collocational information. The dictionary is only brought in as a checking device. Much of the time, though, the strategiesfor teaching lexis are less natural. Often the effect of the chosenmethod is similar to that of using an unthemed list. Consider the following example, also from New Headway Intermediate: You will hear Bert Atkins, who was born in 1919, talking about his school days [Accompanying tape]. Check these words in your dictionary: chalk a slate a cloth to knit Apart from giving the leamer practice in using a dictionary, this activity serves no different purpose from the list from Colloquial Czech.A similar chargecan be levelled againstthe following activity from the samebook:

Th inl thr

cI( lea

A world beyond collocation


se al
le le

VOCABULARY Art, music and literature Use your dictionary to look up any new words. Look at the nouns below and write them in the correct column. composer instrwment chapter brush drawing ART poem band tune banjo novel author palette bugle portrait pianist MUSIC painter sketch biography fiction pop group oil painting orchestra detectivestory play

of ds
tls ln rd



se 1e rd m

This strategyencourages leamers to build up themed lists for themselves, the but while this gives the leamers ownership of the lists they create,the lists themselves have the sameweaknesses any other kind of list. Sometimesthe as strategy adopted is one that seeks to combine the themed list with collocational information. consider the following further example from the samebook: VOCABULARY AND LISTENING: SpoTt Make a list of as many sports and leisure activities as you can think of. Use the pictures to help you. [Pictures omitted] WiIe in play, go, or do. There are three of each. . tennis . exerctses . jogging Can you work out the rules? . athletics . volleyball . aerobics . football fishing skiing


A .I dl

choose some of the sporls or activities from your list and fill in the columns below. use your dictionary to look up any new words that you need. sport/activity football play, go, or do? play people place equipmenr ball, boots

[. :

goalkeeper stadium footballer footballpitch referee



The effect of this sfategy is to get the leamer to organisehis or her knowledge into themed lists and of course supplementthat knowledge with the help of the dictionary. If that was all it did, it would be no more than a fussy way of creating themed lists. But what of course it does in addition is invite the learnerto investigatethe common collocatesassociated with eachof the items


A world beyond collocation

in the list. Furlhermore,it recognisesthat eachof the items in the list will have different collocates; it reminds the learner, implicitly at least, that such lists cannot be treated as homogeneous.

11.3 The importance of context
So far, so good. But so far is not quite far enough.To begin with, vocabulary definitely takes a back seatto grammar in Headway and, as already noted, the vocabulary activities vary in value. But there is a more fundamentalproblem here, which is a direct product of the fact that vocabulary items need to be It taught with their most common collocates. has always been impossibleto exposelearners to mote than a fraction of the vocabulary that they might be felt to need. How much more impossible it must be to teach them the that everyitem has on average in necessary vocabulary context! Ifwe assume around ten important collocates,we have in effect multiplied the vocabulary leaming task by ten. Admittedly some of the words which make up the collocations will be items we might want the learner to acquire anyway, and leaming items in context may be easierthan learning them out of context, but for all that the problem will not go away. There is a greal deal of vocabulary to be learned, and a gteat deal to be learned about each item ofvocabulary. One strategy adopted increasingly in recent years is that of developing specialisedmaterial for the purposeof teaching vocabulary.One such work is English Vocabularyin Use by Michael McCarthy & Felicity O'Dell. This contains 100 sections, all but seven of which are directly concemed with teaching classesof words, mostly semanticallyorganised,though in some cases morphologicallyorganised. In an early section (p 2), they offer the learner advice that is so eminently sensiblethat it is worth quoting extensively: What does knowing a new word mean? . It is not enoughjust to know the meaning of a word. You also need to know: (a) what words it is usually associated with (b) whether it has any particular grammatical characteristics (c) how it is pronounced . Try to learn new words not in isolation but in phrases. . Write down adjectivestogether with nouns they are often associated with and vice vetsa,e.g. royalfamily; rich vocabulary. . Write down verbs with the structure and nouns associated with them, e.g. to add to our knowledgeof the subject; to expressan opinion. . Write down nounsin phrases, e.g. in contactwith; a train set; shadesof opinion.

Th Bu

the thr rb1 un


- .+,




1- __



noi A'in

kin narl


the witl
I O i

con s191 clil s1_s

4 world beyond collocation



Write down words with their prepositions, e.g. at a high level; thanksto your help. Note any grammatical characteristicsof the words you are studying. For example,note when a verb is irregular and when a noun is uncountableor is onlv used in the olural. . Make a note of any special pronunciation problems with words you're learning. (Quotedby permission the publisher,Cambridge of Universitypress.) This is excellent advice and deservesto be etchedinto any learner's memory. But there are two points that need to be made about it. First, it is very much leamer-advice,not teacher-advice.Indeed it might even be seenas advice to the learner on how to minimise the damaging effects of strategiessuch as those we were earlier considering. Second,it is advice that is much harder to follow than it seems,and from a materials writer's point of view it is virtually impossibleto provide what is needed.The very next sentence the Mccarthy in & o'Dell book unwittingly provides clear evidenceof this. They follow their advice with the following activity designed to make the reader think about what they havejust said: How would you record the following? b) dissuade c) king e) independent f) get married A perfectly worthwhile task - as long as the learner doesnot check the answer key at the back of the book! Here we are told that some possible answersare: a) b) c) d) e) f) a chilly day to dissuade someonefrom doing something a popular king / to crown a king up to the ears in work independent of someoneI an independentcotntry get married to someone a) chilly d) wp to the ears

ry :he

be to be
.L^ .tlc oe

ll.y :he nd lut

no :IS

fris irh


At first sight theseanswers seemreasonable. on closerinspectionthey do But not in most casesring true. In my corpus,for example,in lz00 instances of king, there is only one caseof popular king - as opposedto thirteen of porn king. Nearry 40vo of instancesof king actually occur in conjunction with a name often followed by a number, for example King charles II, and thar might well be the most natural form in which to record the word. If, though, the lower casek invites a non-namingresponse, most common adjectives the with king are new,former, last, .future. Likewise the most common collocate to the right of independent state, not country, though country is also a is common collocate. Even a chilly day, though an acceptablecombination, is significantly outnumbered in my corpus by a chilly wind, a chily night, a chilly eveningand, above all, a chilly receptiont Moreover, and perhapsmore significantly, a chil\, day is outnumberedby multiple-word phrasessuch as a


A world beyond collocation

chilly May day or a chilly damp overcastday. Notice also the significant but covert collocation in lt's pretty chilly this morning. But these are minor matters. The real problem lies in the fact that these answersdo not reflect serious generalisationsabout the way these words are used. I noted above that all the methods for teaching lexis we have been looking at suffer from two weaknesses. Firstly, they take insufficient or no account of collocation. Secondly,there is a world beyond collocations. We have got so fixated on collocations that we do not see that they sometimes group in generalisableways and these groupings then account for examples of word combinations that are not collocations.

prosody 11.4 Semantic
The frst important point 'beyond collocation' is that words don't just combine with chosenother words, they combine with chosenmeanings. The idea is associatedwith John Sinclair (1991) who commented:Many usesof words and phrases show a tendency to occur in a cefiain semantic environment For example, the verb happen is associatedwith unpleasant things - accidentsand the like. prosodyas This phenomenon beenlabelled 'semanticprosody'. Semantic has I am defining it occurs when a word associateswith a particular set of meanings.So, for example,in principle, almost anything can be chilly; people and food can, afler all, both be cold so why should they not be chilly? Yet we find in fact that the word occurs in the company of certain kinds of meaning rather than others. The table below shows the semantic prosodies of chilly. For the sake of contrast I have included the frequenciesof people Nrdfood, below the line in the table below, just to show up the significance of the true semanticprosodies.
Semantic prosody frequency (out of 352) 79 58 30

learne certaii proso( all gor beenp prosod words, reveal others occur I of textr chilbthey ctr These
15 tlue

prosod not col 'time'

Morec4 yet all point h for the as indi vocabul the wq prosodl



unit of time place weather

22Vo l6Vo 97o 7Vo 4Vo 37o 2Vo 27% 27o

temperarure ill people watery things [metaphorical] people food


a chilly overcastaftemoon chilly Comwall it's pretty chilly the chilly breeze a decidedly chilly -10C a chilly patient a chilly bathtub a chilly reception a chilly lavatory attendant chilly rolls with Iceberg lettuce

8 96 6 I

This lal the virl assocra or typic charactl from ler arbitra{ for lean of the I to ask d courseb

11.5 I
Even st words a namely list fron

Excepting the metaphorical use of chilly, which does not strictly involve semanticprosody, the prosodieslisted above the line in this table account for 86Voof the occurrencesof chillv in mv data. What this means is thal if a

A world beyond coilocation


leamer wants to learn chilly they would do best to learn that it occurs in cefiain kinds of context rather than all contexts. Seen like this, semantic prosodyis a kind of generalisation basedupon the collocates word has.Like a all good generalisations, covers word combinations it that might not have beenpicked up as collocations. For exampre a chilry decadefitsthe semantic , prosody of 'unit of time' but is not a particularly common combination of words' on the otherhand,as we will see,it is u purtiul generalisation, data as revealthat somecoilocates which we might expectto occur actuallydo while others do not. It is particularly interesting, foi example, that chilty does not occurmuch as a literal descriptorof people,sincepeopre are a commontopic of texts and talk. It would appear that ir you say of a person that they are chilly, you mean that they are ill or have an unattractive temperament,not that they could do with sitting near a fire. Thesesemanticprosodiesare more than the accumulation of collocations.It rs true, as you wourd expect, that there are lots of collocations in each prosody' Thus chilly coilocates with morning, night, evening,day; butit does not collocatewith minutesor decades, both yet of theseare examplesof the 'time' prosody. chilly also collocateswith mountain but not with tent or Morecambe(despitewhat visitors to the north-west of Engrandmight think), yet all three words exemplify the 'place'prosody. Don,t miss the important point here: such semanticprosodiesare potentially powerful generalisations for the languagelearner.It is no longer n"."ruu.y to leam endlesscollocations as individual combinations, which as I remarked earlier seemed to make vocabulary learning harder. Instead, what the leamer needs to do is to learn the word in combination with an absolutely typical representativeof the prosody as long as (s)he also knows that it IS typical. This last point is crucial. I commentedearlier that unthemed word-lists have the virtue for me of being contextualisedby the texts with which they are associated. But, I added, I was never sure whether the contexts were natural or typical. unless one knows that the collocation one is leaming is absolutely characteristic the way the word is used,more than of half the value one gets from learningthe word in its contextdisappears. [As we saw on pp r32r3, the arbitrarygapsin what we might expectby generalisation.un problems for leamers.Edl Mccarthy and o'Dell may not reflect "uu." the semantic prosodies of the words they seek to teach, but at reastthey are encouragingthe leamer to ask the right questions. you would scanthe pages of most if iy language coursebooks vain for the slightesthint that words in havesemantic prosodies.

11.5 Colligation
Even semantic prosody, however, is insufficient to account fully fbr how words are used.There is a world beyond colocation and semanticprosody, namely that of colligation. Let us return for a moment to the ,employment, list from suefiosdiscussed earrier. first sight,as a setofcountable,concrete At


Alryorld beyond collocation

nouns sharing quite a lot of meaning, we might expect words lke architect, actor accountant,carpenter to sharequite a lot grammatically as well. So we might expect them all to take definite and indefinite articles (the Finnish architect, a church architect). We might expect them all to take classifiers (church architect) and possessives(the Academy's architect). We might expect them also to occur within possessiveconstructions (the whims of an architect; the architect's bombastic ego). We might expect them to occur in parentheses (WeUsCoates, the architect of Lawn Road Flats in Hampstead) and apposition (British architect Will Alsop). Any good grammar would encourageus to have all theseexpectations,and in so far as every item ofthe list is grammatically capable of appearing in all these constructions, such expectationsare all entirely reasonable. 'likely to appear'. Grammars But 'capable of appearing' is not the same as over the years have got obsessive about recording what the language is capable of doing, and there has been all too little attention given grammatically to what it actually does, or, more accurately, what we do with it. This is where the concept of colligation comes in. Colligation can be defined as 'the grammatical company a word keeps and the positions it prefers'; in other words, a word's colligations describewhat it typically does grammatically. The following table, basedon occulrencesin my corpus, showsthat the words accountant, actot actress,architect, carpenter differ grammatically amongst themselves,despite our expectationsthey should all behave alike; they have, in other words, different colligations. Notice the particularly high numbersin bold and the particularly low ones underlined in the table.
Grammatical construction
indefinite article classifier tpossessort construction ie's & of NP tpossessedt construction apposition parenthesis l4Vo I'lVo 27Vo 87a 3lVo l27o lSVo 13Va

A 1 urrl inr

Na likr likt ltet but har

tha wol con car, coa AId ah gar thal


Act, rne;

G c(| m

accountant (1045 instances)



actress (1710)

architect (2020)



Yor to(

. 26Vo l27a l0Vo 8Vo 4Vo











pr0 sefl odd sirl haq coll


Tol and Mc1


A world beyond collocation


'employment' A glance at this table shows that words do not behave in a the grammatical constructionsthey occur uniform fashion when it comes to in and with. The word carpenter has a much higher likelihood of occurring with an indefinite article or in a parenthesis(Mr Morland, a carpenterfrom Nofting Hill . ..) than does,say,architect.The word accountantis much more likely to occur with a classifier (a wages accountant) and actress is more likely to occur in apposition (actress Debra Winger) than any of the other items in the list. It is quite possibleto possess accountant(my accowntant) an but virtually impossible to possessan actress (as dreamy-eyedteenageboys (the mother of the have long known). An actressmay however be a possessor actress Fay Compton) as may a carpenter. What the table does not show is that even with regard to this similarity betweenactress and carpenter the two 's words actually differ. Whereas carpenter occurs in constructions and ofconstructions in roughly equal proportions (the son of a Lithuanian carpenter; the catpenter's apron), actressoccurs almost exclusively inthe ofconstructlon. Alone amongstall the employmentwords, architect is not distinguishedin the abovetable by its unusually high or unusually low associationwith one of the grammatical patterns mentioned - but it is distinguished from the others in that, as the supplement to the table below shows, it is alone in being frequently used as a metaphor (He was the main architect of the peace plan). Actor is the only other word in the list with any significant record of metaphorical use.
Grammatical construction metaphor actor accountant (1045instances) (3re4) 5Vo none actress architect carpenter




You might imagine that colligation and semanticprosody are applicable only to concrete nouns. One of the other lists quoted from Suefiosis that of the numbers.You might be forgiven for thinking that at least thesecould safely be learned out of context becausethese, surely, would not have collocations, prosodies. suchchance, No I'm afraid.It is possible colligationsand semantic to show thal sixty has a strong semanticprosody with time (e.9. sixty years), occurring with a unit of time 23Voof the time but that sixty-five has no such prosody, occurring with a unit of time only 8Voof the time. Slxfl also has semanticprosody with markers of imprecision (over sixty, almost sixty, sixtyodd), occuning with such markers 2l%oof the time, but I have no instanceof sixty-five occurring with a marker of imprecision in my corpus. On the other hand, while both words have a great preferencefor beginning a sentence, that colligation is much sffonger for sixQ-five (86Voas opposedto Tl%o). To show how it appliesto vocabulary leaming, let me againpick on McCarthy and O'Dell's sample answers. (I stress again lhat I am not picking on McCarthy & O'Dell becausethey are bad, but I am discussingtheir examples precisely becausethey are about as good as it currently gets.) The sample


A world beyond collocation

answersthey give for married and up to the ears both creak somewhat,and it is worth looking at why. For these words they offer in their answer key the followingsuggestions: a) up to the ears in work b) get married to someone of In my corpusthereareonly 14 examples wpto ... ears,usedmetaphorically, then,McCarthy and twelve of theseshow a collocationtttithin.In this respect, & O'Dell's answer is typical of the way of the word is used. But a still strongerpattern is overlooked.Thirteen out of fourteen of the examplesin my corpus colligate with possessivepronouns. rather than a definite article, for example: wp to his ears in exasperation,wp to my ears in debt. So McCarthy & O'Dell's answer is colligationally untypical. Furthermore, ten of the occurrencesin my corpus manifest semantic prosody with bad things (such as treason and narrow and sterile rwles) and five of these bad things are instances debt.McCarthy & O'Dell's work is a marginalcaseof of this and not one that reflects the prosody clearly. Thus a proper description of up to ... ears is the following: upto...ears


n1arril, ra,hich to ihe v'as p) are lc nlarnt some l
L 5 U - L I

oI }t (?t


uirh tir

(131 possessive
with (1)


good (3)

(5) miscellaneous

Obr io on uhl tu o frr phra-<

debt (5\



Put more straightforwardly, wp to his ears in debt has a one in five chanceof occuring when you useup to ... ears; up to the ears in workhas a one in 800 chanceof occurring. Guesswhich one I think a leamer should learn.
[A word of caution may be neededhere. If the corpus Michael is using is heavily skewed,for example towards news reporting, its most typical examples,although typical of that genre,may not be typical of the whole language or, for example, another genre such as informal basedon corpus evidence,while conversation.It can never be repeatedtoo often that statements undoubtedly evidence of what is in, that corpus, may not be as typical of the language as a whole as we are tempted to believe. From a pedagogicalpoint of view, in addition to raw colpus evidence,factors such as immediate usefulnessor relevanceto the needs of particular students may need to be taken into account when choosing the exampleswhich will be most helpful for a particular class. Ed.l

subie' r','ithir some Again more l

l c1


A similar picture applies to get married. Of 257 instances of clauses containing the verbal group get married, at least 92 (36Ea) have as their

Inau ongml conhn lin-eui apart prepar teache

A world beyond collocation


y 1


subjecta referenceto the couple ( "Why did AwntieElaine and Uncle Marc get married?" askedOlivia.) A further 26have an indefinite subject,ustally you, which may be referring to a couple.Another 97 GSqa)have one of the parties to the marriage as subject but no mention of the other pafty (Monica Zanotti was planning to get married this spring.) On the other hand, only five (ZVo) are followed by a prepositional phrase starting with to. The phrase ger married is far more likely to occur with a positive time or place expressionof somekind (though not usually both), such asplanned to get married one day or wanted us to get married in church - 28voof instancesof the expressiondo so - but this is twice as likely to happen when the couple is subject. Thus a fuller description of get married would be:
get married

l I f f


with time/place expression(29)


(42) positive (0) negative


poSitive (20)


\ \

negative (9)


obviously, the exact distribution of thesecharacteristics may differ depending on whether get, getting or got is selected,but informal inspection of the other two forms suggests great variation. For example,the past tenseform of the no phrase got married shows the same tendencies.Out of 172 instancesof gol married, 4l (27Ea)have the couple as subject; a further six have an indefinite subject,and only eight (5Vo)are followed by a prepositional phrasebeginning with to, three of which are separated from the verbal group by punctuation of somesofi. Again, what all this means is that sentences such as the following are much more likely than McCarthy & O'Del|s get married to someone: I cried when I watched Jill and Mark get married on Tuesday. They'replanning to get married at last, in Monaco. In a way, this analysis only confirms the wisdom of McCarlhy & O'Dell's original advice to record every new word in its grammatical context. It also confirms, however,that intuition, even the intuition of the best lexical applied linguists,is likely to be flawed.And if it is true of McCarthy & O'Dell, who, apart from being steeped in lexical knowledge, also had time aplenty to prepare their book, how much more true must it be of the poor language teacher, rushed off his or her feet, preparing materials with inadequate


A world beyond collocation

resourcesand asked about words in class without the chance to check the answersout!



So what can the languageteacherdo? How can (s)he teachvocabulary so that the naturally occurring colligations and semantic prosodies are picked up? Haven't I actually made the teacher'stask still worse? The answerto the last question is'Yes and no'. Certainly,I have redefined what it is to leam a word well. But the featuresof words that I have mentioned are naturally occurring features and there are strategiesthat can be used 1o ensure that words are learned wilh maximum usefulness.The first thing to note is that words acquire colligations and semantic prosodies, as well as collocations, by being repeated in similar contexts. As we leam our first language,we build up in our headsa profile of the words we are leaming. The so-called LanguageAcquisition Device in a baby's head is more likely to be a set of concordancing 'software' that enables us to find regularities and recuffent features in our linguistic experience, rather than any abstract grammar-makingdevice. So the first necessityis, unsurprisingly,exposure to as much naturally occurring language as a learner is capable of attending to. If learner or teacher has accessto computer concordances,so much the better; Tim Johns at the University of Birmingham has shown for years how much a learner can pick up from simple, small concordances. But what if learners or teachers have no access to a computer? In such circumstances,there are still useful strategiesthat can be adopted for the leaming and teaching of vocabulary.Most texts can be shown to be networks of repetitions,for the obvious reasonthat texts tend to be about somethingand whatever that 'thing' is, it is likely to be repeatedmany times in the courseof the text. In addition, becauseit is impossible to say everything at once, as Eugene Winter once noted, we often pick up something we said earlier in order to add something to it that we could not say on the first occasion. Take this chapteras an example.My computer tells me that the most common lexical items in the chapter so far ate words, language, learner and word, which is hardly going to come as a surpriseto you, The word words occurs 49 times up to the end of the previous paragraph;word occtrs 26 times, language 21 times and learner 28 times. This means that you have had more than 20 opportunitiesto seeeachof thesewords in action, to absorbtheir collocations, colligations and semantic prosodies. Without realising it, every time you encounteredone of thesewords, my text was subtly reinforcing or modifying your mental lexicon, which already contains collocational, colligational and semantic prosodic information about these words, though you may well not be particularly aware of having that inforrhation. So, my chapter is in fact a kind of linearly organisedconcordanceof the words words, word, language and learner (as well of course as many others).


A world beyond collocation



hat rp? red red to to

So what will you subconsciouslyhave absorbedabout the word words in the course of reading this chapter?well, firstly you will have absorbedthat it occurs repeatedly with certain other words, putative collocates in fact: these are:use(d)(3 times - 67aof possibleoccasions), learn(ed/s)(7 times -l4Te, - l)Vo), and new (3 times - 6Vo).Examinarionof my 100list(s) (5 times million-word corpus shows that all these words are bona fide collocates of words and one of them, use(d), is the commonest lexical verb words associates with, occurringalmost 5Voof the time in my corpus. You will also have unconsciously absorbedthe fact thaLwords occurs in an idiom - in other words - twice in my chapter.of course,this idiom proves to be very common in the languageas a whole, occurring 1774 times in my 100million-word corpus. So much for the collocations. You will also have noticed, probably without noticing that you are noticing it, a significant colligation, namely that it is common for the word words to be immediately followed by some form of specification, example,the words 'actor' and 'actress';this happenssix for times in this chapter (l2vo of the time). This is a putative colligation of words - and examination of the large corpus confirms that this is so. Specification occurs after words 1731 times (lIEo) in the 100-million-word torpus, of which 1259 (over 8o/o)are individual words or phrasesor sentences less than four words long. So my chapter has inadvertently reinforced a very common colligation of words. The sameis true of word, only evenmore so in this chapter.In appositionwith someother item, for example,the word 'accowntant' the word 'carpenter',iI , occurs a fifth of the time in my chapter.This reflects its use in the language as a whole. Out of 12,839instances word in my lOO-million-wordcorpus, of 272I occw with apposition,that is 27Voor, as in this chapter,one occurrence in frve. Another colligational feature of word you will have unconsciously absorbed is that it functions as a classifier in my chapter three times: word combinations, multiple-wordphrases,word-lists;this happens almost ljTo of the time in the 12,839 examples.You will also have absorbedthe fact that word collocates in my chapter with use(d), occurs, learn, and - as this sentenceillustrates - collocates; thesefour collocates account for 58Voof all instancesof word in my chapter.As for words, the collocate use(d) is very commonin the language a whole; one of the forms of use occurs as withword 1051 times or 87oof the time. The collocateoccur(s) is much less frequent, barely registering ar O.lvo of the time. on the other hand, if one looks at the semanticprosodiesof occwrsin a generalcolpus we find that it occurs nearly 4vo of the time with a linguistic term as subject (or with a pronoun as subject which has as its reference a linguistic term) - a very high percentage,given that my corpus is taken predominantly from the Guardian newspaper,not noted for its extensivelinguistic discussions.Add to this the fact thar occurs

|rst he be
Lnd act

to ng


-1.^ R5

nd of

on rd. 49 ge l0 us. ou n,g nd
iot r 2



A v)orld beyond collocation

when is used as shorthandfor can be defined as what occurs when two and a half per cent of the time, and the fact that on another two and a half per cent of the time it takesa mathematical musical subject(quasi-linguistic, or both) you have strong evidence that rny chapter's use of the combination of and word and occurs is entirely natural. So, this chapter has subconsciously reinforced the semantic prosody that occwrs takes a linguistic or quasilinguistic subject. Notice that the last collocateof the four just mentionedis not one that would be picked up in a large generalcorpus. The writers of this book have all been working to changeyour perception of the way words work and part of what that has involved has been the adding of collocates to your mental concordance word. So the way a word is used and understood modified of is as well as reinforced by the texts we encounter. The point is clear: a short natural text is already creating the collocations, colligations and semantic prosodies of the words we encounter.For native speakers,most of the time what happens is simple reinforcement of collocations and other pattemsthat we already know and recognise,but as the exampleof the word collocatesshows,it is also possiblefor a text to modify or add to the collocationsand other languagefeaturesthat we recognise. For the leamer, of course, all encountersare like this. So what we need are ways to intensify the leamer's encounterswith words. As I warned at the beginning, I am a descriptive linguist, not a practising languageteacher,and it is likely that any advice I give here can be bettered by you. But two approachesmight be suggested.The first comes from Jane Willis who, in a fascinating paper on materials development, showed how she had got her learnersto use a text to produce a manual concordance.She took a group of words that occuned commonly in the text and got her studentsto go through the text looking for one or more of thesekeywords. In her case,the keywords were grammatical in nature, but the activity can be used with equal effect on lexical items, as long as you have verified in advancethat the words being searchedfor have occurred a sufficient number of times to prevent the search being boring or frustrating. The studentsthen had to copy the keywords down in the contexts in which they found them, lining the keywords up under each otherjust as is done automaticallyon a computerconcordance. She then got them to reflect on what pattems they were finding. Obviously, for full effect the studentsshould be linguistically sophisticated even if not linguistically advanced,though value could be gained by any group willing to go beyond being force-fed. Another way of using the vocabulary - the hidden concordances the text of is to have students look for words that group together and then search for other sentences with the same group of words in them. For example the first sentence this chapter:Despitebeing the head of a languagewnitdedicated of to Applied Linguistics,I am not myselfan experienced langwage teacher.

A dt ot





I1n raI Gi clt str_
]i l



rr],rti{ir llJllli

A world beyond collocation


A number of lexical items occur in this sentence: head, language(twice), wnit, dedicated, applied, linguistics, experienced,teacher; since it has a referent outside the text, the pronoun l could be addedto the list. The studentis then told to look for another sentencewith three of these words in it. Thev don't have far to search:the very next sentencefulfils the condition: The little experience I have of language teaching is well out of date andfomnsno basisfor giving practical advice to anyone. The studentcan then look for pattem and variation. (S)he will certainly spot: language teacher langwage teaching which point to the possibility of a colligation whereby language is noun modifier to a noun derived from a verb (cf language learning, langwage acqwisition). the sametime, (s)hemay note the variations: At
experienced experience I have of language tectcher language teaching

and, if (s)heis really sharpand linguistically on the ball: I am not myself I have little
experienced experrcnce

The student then looks for another sentencewith the same lexis, trying not to read the intervening sentences but simply searching for the cluster of words. Lo and behold, (s)he will find one at the end of the paragruph:I may haveno langwage teachingexperience I do havedaily experience but ofbeing a language learner I should say here that I have made no changesto my text whatsoeverin order to illustrate the point I am making, nor did I write it with the intention of using it as an illustration. Now the tentatively identified colligations are reinforced. We have: experienced I experience have of experience being a of
language language language language teacher teaching teaching learner


The learner now has encounteredone of the colligations of langwage fow times in a manner likely to reinforce the learning, but (s)he has encountereda range of ways in which experience (and its related adjective) can be used. Given the parallelism of learner wilh teacher, it could be added to the word cluster being sought. If 1 is included in the search,the next sentencethat the student will find is the very next sentence'. Every weekday, more or less without exception,I attempt to learn a language. The main point here is the confirmation of the fact that the previous noun headswere derived from verbs:
learn a

language languagelearner

! J1


A world beyond collocation

If however1is not included,the next sentence be found will be at the end to of the third paragraph:The experienceI bring to a book like this aimed at the practising language teacher is that of the Qpical well-motivated but only ge avera Iy compet nt I earner. e It will be seen that this sentencereinforces what the learner has picked up about experience, and by now the leamer will also be picking up that experience alwaysassociated is with someone, this caseL It alsoreinforces in the noun-modifying quality of langwage as well as showing that the noun headsit accompanies occur on their own (averagelycompetent can learner). And so on. There is one other point to be made about the sentences studentwill have the been looking it, namely that they will characteristically- if the text is nonnarrative - make sensetogether,as will be seenbelow: Despite being the head of a language unit dedicated to Applied Linguistics, I am not myself an experiencedlanguageteacher.The little experienceI have of languageteachingis well out of date and forms no basis for giving practical advice to anyone. I may have no language teaching experiencebut I do have daily experienceof being a language leamer. The experience I bring to a book like this aimed at the practising language teacher is that of the typical well-motivated but only averagelycompetentlearner. Thus the studentsget two benefits at the same time. They have a controlled task to perform that will result in the raising of their consciousness about the nature of the collocations, colligations and semanticprosodies of a group of words - though I hope I do not have to add that there is absolutely no need for them to ever hear whisper of such terminology.At the sametime they have got sensefrom authentic text without having had to read everything in the text. In this context I should emphasise that the text doeshave to be authentic; inauthentic text may distort the featuresI have been describing.

11.7 Summary
The featuresI have been describing,as well as thosedescribedelsewhere in this book, are the natural result of the way we encounterand acquire our first language.Word-lists, despite the promise they offer of allowing the learners to correlatetheir first languagewith the languagebeing learned,deprive them of much of the information they would naturally have if learning those same words in the language community. The strategiesI described in the last few pages are not the only, and probably not the best, ways of enhancing the leamers' vocabulary so that they learrrnot just the meaningsof the words but the environments they occur in. But they surebeat word-lists! No wonder I have never succeeded a languagelearner: the coursesI have as taken have not only denied me accessto the collocations of the words I have

A world beyond collocation


learned, but they have hidden from me the whole rich world beyond collocation that underlies what it r"atty -"u.r. to know a word.

Discussion euestions
#:iHf;"tr have noticed vou when learners lisrsof individual learn words
Michaer Hoey is ress happy with many themed word lists, such as lists of sporrs or fumiture, than with unthemei fisrs drawn from a;;;", dialogue. experience



reflecthis assessm"nt or-rn.." differenr

Do you anticipate any probrems implementing the suggestion, made by severar contributors to this uoot, oi"n"ouraging learners to record new language multi_worditems, in ,cleaning l" it up,first? "orrt"^t,^;ithout

244 Bibliography

Baigent, M. (1999) Teachingin Chunks: integratinga lexical approach,'in Modern English Teacher, Vol. 8, No. 2 Biber, D., Conrad, S. and Reppen, R. (1998) Corpus Linguistics, Cambridge University Press Brown, G. (1917) Listening to SpokenEnglish, Longman Bygate, M. (1996) Effects of Task Repetition: appraisingthe developing languageof learners, Willis and Willis in Coady, J. and Huckin, T. (1991) SecondLanguageVocabularyAcquisition, Cambridge University Press Fernando, C. (1996) Idioms and Idiomaticity,Oxford University Press Hill, J. and Lewis, M. (Eds) (1997)LTP Dictionary of Selected Collocations, UTP Hoey, M. (1991) Patterns Lexis in Text, Oxford University Press of Hoey, M. (Ed) (1993) Data, Description,Discourse, HarperCollins Howarth, P. (1998) Phraseology and Second Language Proficiency, in Applied LinguisticsVol. 19, No. 1 Hudson, J. (1998) Perspectives Fixedness, Lund University Press on of Johansson,S. (1993) "Sweetly oblivious": some aspects adverb-adjective combinations present-day in English, in Hoey, 1993 Krashen, S. and Terrell, T. (1983) The Natural Approach,Pergamon Larsen-Freeman, D. (1977) Chaos, Complexity, Science and Second Language Acquisition, in Applied Linguistics,Vol. 18, No. 2 Lewis, M. (1993) The Lexical Approach,LTP Lewis M. (1997) Implementingthe Lexical Approach,LTP Lewis, M. (1996) Implicationsof a Lexical View of Language,in Willis and Willis Lewis, M. (1991) PedagogicalImplications of the Lexical Approach, in Coady and Huckin Nattinger, J. and DeCarrico, J. (1992) Lexical Phrases in Language Teaching,Oxford University Press Poole, B. (1998) Corpus, Concordance, Combinability, in IATEFL Newsletter, Feb-March 1998 Rudzka, B., Channel, J., Putseys,Y. and Ostyn, P. (1981) The Words you Need,Macmillan Schmidt, R. W. (1990) The Role of Consciousness Second Language in Learning,in Applied Linguistics,Vol. 11, No. 2


S] L



Tr R

T EI T \t


T] 1, v!

\1 H




sinclair, J. (1991)corpus, concordance, colrocation,oxford university press Skehan, P. (1998) A Cognitive Approach to Language Learning, Oxford UniversityPress Stewart, I. (1990) Does God play Dice?, penguin 19 Ter-Minasova, S. (1996) Language, Linguistics and Life: A view from Russia,Moscow StateUniversityAssociation

Sunderland, P. (1998) Lexical Chunks,inArena, Issue

Timmis, r. (1999) Language corpora: what every teacher should know, in ELI News & Views, year 6, No. 1 Thornbury, S. (1998) The Lexical Approach: a joumey without maps. in Modern English Teacher,Vol. 7. No. 4 watts, P. (1999) using textual analysis to teach source use, paper, 33rd TESOL Convention Wilberg, P. (1987) One to One, LTp Williams, A. (1998) Negotiatingin Chunks,in Arena, Issue 19 willis, J. and winis, D. (1996) chailenge and change in LanguageTeaching, Heinemann


I S B N1 - 8 9 9 3 9 6 - 1 1 - X

Sign up to vote on this title
UsefulNot useful