Collocability in Languages for Special Purposes ( LSPs ) : Some Preliminaries

This paper is concerned with the language professional discourse communities use for their internal communication. The characteristics of these languages for special pwposes (LSPs) are many and varied, as well as being underresearched. The focus adopted here is to examine the phenomenon of multiword units, many of which are orthographic pluralities designating conceptual singularities. It is important to recognise at the outset that colocation is not the same as collocation. Analysis and systematisation of these textual "c1usterings" is intended to separate them into two radically different types of entity: multiword segments possessing terminological status; and collocative material. The methods used to achieve the above objective are both qualitative, i.e. micro-environmental analysis, and quantitative, i.e. statistical patterning exhibiting a certain level of frequency and constancy. Collocational material quoted here also shows by its configuration that discourse communities use collocations to which the general public are not inured and with which they may not necessarily be familiar at all.


Preamble
The above quotation -which, interestingly, chooses English examples to support its thesis -sets the scene admirably for what is to follow in due course on what is ~cknowledged to be a very significant aim, within text linguistics, discourse analysis and, of.course, lexicography and terminology studies: firstly, the location within running text and the subsequent analysis -either by hand or by computer -of units of meaning which comprise more than one (ortho)graphic word; and secondly and much more importantly, the over arching phenomenon of "chunking".However, before we can sensibly discuss these matters, the major focus of this paper, we must tread a path to the promontory from which we will best be able to view the scene.

Point de depart
Chunking is a primarily mental phenomenon, the symptoms of which are to be found in linguistic formulations.Put in other terms, chunking is a psychological cause, associativity, which has linguistic effects, juxtaposition or blending.It follows that any macro-investigation of chunking as a phenomenon is best pursued by the micro-investigation of "chunks", stretches of agglutinating linguiStic material which are felt to represent segments of thought rather than just fragments.Chunks are also entities about which and about the use of which there is some sort of social consensus.However, chunks hardly ever seem to be static in developmental terms; rather, they appear to be on a trajectory towards "explicitness" and involved in a systemic combat driven by language users and aimed at a prevaricating homeostatic trade-off between lexis and syntax.Once they have reached it, whatever that means in orthographic terms, they are no longer inchoate and they are no longer ambivalent: they have, in fact, merely become "normal", if somewhat lengthier, units of language -in common perspective, at the very least.
It is a well-known but regrettable fact that very, very few language communities possess satisfactory collocations dictionaries, the global mission of which -in addition to their strictly utilitarian function -is to demonstrate, reactively and proactively, that colocation is not collocation.The former, that is, mere juxtaposition, is either volatile or an example of a "fixed" multiword unit, whereas collocation manifests associative regularities -sometimes based on assonance, sometimes on prosodic effect, sometimes on less easily definable criteria -that can be shown to be statistically significant rather than "binary" in the sense of present versus absent.The normal unavailability of collocations dictionaries is a great pity because that is exactly what advanced learners need and, indeed, what many native speakers hanker after too.In fact, it is not stretching things too far to say that first-class collocational control is the hallmark of the true L2 expert; collocational control is, of course, normally the last linguistic subsystem to be mastered by L2 learners who proceed to an advanced level.Correct deployment of collocations is particularly important for anyone striving for authenticity of performance within a particular professional sociolect, such as the language of medicine or economics.
Lexicographic tools are urgently needed to help those who have not had the advantage, from their early years, of "statistical exposure" to lexical patterning which, although often analytically idiosyncratic, is habitual.It follows that investigation of the phenomenon of collocability relates to the hidden, subliminal patterns and rhythms of language which need to acquire a description having its basis rather in the statistics of occurrence and co-occurrence.Such lexicographical tools need to be primed -in the sense of acquiring the data they treat -by computational tools, the purpose of which is to identify, excerpt and prioritise relevant lexical material in a form suitable for subsequent lexicographical treatment.However constant this general aim may be, the actual means of fulfilling it will differ from one language to another.In all cases, of course, it is a question of locating stable and authentic multiword material-this is probably the only initial common denominator.
The "intrusion" of terminology -in a double sense!
In languages for special purposes (LSPs), a particular problem is posed by the terminological usages which proliferate in such modes of discourse.It so happens that technical terms are themselves often composed of more than one Reproduced by Sabinet Gateway under licence granted by the Publisher ( dated 2011) http://lexikos.journals.ac.za orthographic word.Some criterion is therefore needed to discriminate between terms and collocations.The former are cognitivei entities inserted as such into flowing text, the latter are, predominantly, inserted separately into texl, yet their associativity helps' that text to flow and to cohere.However, the position is given a further twist by the fact that genuine collocability also manifests itself in LSP discourse.
Let us commence by suggesting that Bejoint and Thoiron's formulation "lexical units" can be the cause of some confusion as this piece of nomenclature focuses on the messenger rather than the message, so to speak.Is it not preferable to refer to cognitive units or cognitive entities -which, naturally, require linguistic form, at least for the purposes of interpersonal communication?Quibbles apart, all of these terms attempt to capture the "truth" that it is not only words which retain their referents outside text; word-groups and phrases may also do so.This is entirely in accordance with the customs, habits and social compact established within societies for structuring individual and group experience.These lexical pointers along with whatever they point to constitute the mental lexicon of individuals and, by aggregation, the working consensus of society about how to interact linguistically with the world.Items in mental lexicons are mappings, complex rather than simple, between mental images and the names which point to them.In some ways the names are secondary and subservient: they may alter, leaving their referent unchanged, for instance.Yet they have power: names tame!The "downside" of this, via a different analogy, is that the natural dynamics of mental images are reduced to slow-motion film or even to still photographs.Yet names are also the major instrument of ef£ability and they may acquire or discard referents as a result.Most of all, names, normally named "words", are at the very basis of social intercourse.Of course, much social intercourse takes place on a distinctly professional level, invoking/evoking -and instantiating by words -established concepts (Latin: ideas "taken together", i.e. holistically merged), phenomena or artefacts current within particular discourse communities (Des) for the simple (!) reason that they have been professionally introduced, negotiated and subsequently validated by them.This process is necessary so that DC members can communicate with each other on the same terms, in a double sense, understanding without either external or internal mediation and not merely comprehending each other's thinking and reasoning.In this way a DC's sophistication increases, first of all encyclopaedically (at least for the pioneers) and then terminologically.The route from term to concept is immediate for DC members; for those outside the DC it is most likely to be mediate.It can, obviously, also be a cul-de-sac.The people of a DC and their purposes always have precedence over their technolect, so to speak.I Putting this point in the parlance of sociology, the DCeven represented by its neophytes!-is the independent variable, the linguistic resources that they call on to express themselves individually and corporately are the dependent variable(s).A less obvious corollary, even to those directly involved, of this system of socio-professional "information exchange" is the continual growth -not necessarily linear!-in the given paradigm's cohesion.It should in fact be noted, in the margins, that this "growth" can sometimes lead to the sclerosis of orthodoxy, hopefully then followed by genuine renewal based on new relationships and a new dialectic, plus a new or at least revitalised discourse to go with them.In such extreme circumstances a DC's members may well be prisoners of their thought or "mind-set"; they are not normally prisoners of their terminology because they can realign and resemanticise it by agreement.
How and where to establish a "base" for investigative purposes?
Those who observe and analyse Des because of an interest, even a utilitarian interest, in their technolects are almost exclusively never members of them.This, of course, puts them in the empirically best possible but still suboptimal position to carry out their work as dispassionate investigators.The position is still less than perfect because little or no regard is given to what is encyclopaedically expressible (Le."encyclopaedic competence"?),only to what is actually expressed by linguistic means (i.e."linguistic performance"?).However, that is an easy statement to make.After all, there are constraints.What is more difficult is making a judicious and also successful choice of methods for the investigation.It seems that linguists are -very sensibly -increasingly resolving the problem of linguistically modelling discourse, not least LSP discourse, by primary recourse to quantitative rather than qualitative methods.Qualitative analysis of LSP discourse by "frontal attack" requires an exceedingly well-stocked arsenal of tools and techniques, many of them still evolving and maturing.The yield tends to be on a micro level, the analogy of intensive "case-study" approaches which always seem to studiously eschew any mention of "comparators" when the real questions are always the following.First and foremost, just how representative of wider practice and habits are the "findings"?Secondly, has the "gulf' between encyclopaedics and linguistics been successfully bridged?Quantitative methods offer a way round this dilemma -at a cost!These methods can summon from an arbitrarily large corpus exhaustive lists of linguistic "segments" along with their frequency of occurrence.Although the frequencies are numbers, this information is of high qualitative value because the individual items in the associated lexis can then be described in terms of their actual functional load in the running source text and also in terms of their putative "market share" in analogous "still-to-be-written" text.Such material is of prime value for pedagogical purposes, principally because it is direct and powerful evidence of authentic usage.This then is the nature and purpose of quantitative modelling in linguistics -specifically lexis in this case -and it is arguably the best route open to investigators, particularly those with strong pedagogical interests.

The crux of the matter
We are now at a point where we can begin to draw together the two main strands of this paper.The first strand is obvious enough: how do we isolate within running text, firstly, DC cognitive units expressed by single orthographic words and how do we isolate those cognitive entities which are by definition conceptual "singularities" even though they are at the same time orthographic "pluralities"?Obviously, such matters of procedure and the variety of contingent cruces will depend on the particular natural language involved in any such investigation. 2In some languages, of which German may be taken as a suitable representative, the linguistic designation/configuration of many concepts, both within everyday life and within specialist DCs, often occurs via the compounding of separate lexical items into one orthographic unit.The linguistic process involved is one of holisticisationwho can say whether this is a true reflection and model of the Gestalt formed by the fusion of the latent (!) mental constructs?Just as the concept is "taken together" and becomes molecular, so the component [Latin: "putting together"] names are similarly batched and merged:
English achieves the same objective of terminologically nOminating a concept by juxtaposition and/or hyphenation.Phrasal formulation is also often an option in English: guidelines for price-fixing / guidelines for fixing prices. 3Ii: should be apparent, even on this slender basis, that delineating technical terms in running text is not an easy business as far as English is concerned.It is, of course, not unimportant which of the three available methods a language uses, and in what "mix", for structuring syntactic meaning: element order, function words, or inflection.The characteristics of English, with its "residual" inflection and consequent reliance on "neighbourhood" are particularly impervious to analysis by traditional methods.The chunking phenomena• of interest yield only to subtle environmental analysis, often supported by statistical profiles.An instance of this is the English "phrase term" stability augmentor pitch axis actuator housing support which brackets as [[[[[stability augmentor] [

pitch axis]] actuator] housing] support].5
See also Appendix I for a brief excerpt of text on navigation 6 with terminological usages "coded up" -in an intuitively simple manner -as single ortho-graphic words.The paragraph shown represents only about one tenth of the entire text but the artificial graphic system used cuts the token count by 70 items, while -because of the batching effect -reducing the number of types by only a mere half dozen!

A fundamental distinction and its implications
It is vitally important to understand that the formation/usage of such multiword units in English -and in many other languages -has nothing whatsoever to do with collocability, the chronologically second but thematically and substantively first strand in our discussion.It is a fact of life that, alongside everyday collocations of greater or lesser currency, professional collocations also exist.It is to this feature of language that we now tum our attention, encouraged and helped by the following dichotomising principle: If multiword units are terms they are not collocations; if they are collocations they cannot by definition be terms.
Operationally and computationally, this principle reduces to the need to distinguish in English between those spaces between orthographic words which separate and those which glue and bond.Fortunately, this problem normally solves itself in the course of computer-driven analyses by dint of the statistical force of grouping.

Collocation: a macro phenomenon with micro differentials
Collocation is a linguistic phenomenon which manifests itself via the habitual but often seemingly idiosyncratic or inexplicable 7 association by co-occurrence in a microcontext, in parole, of one word with another.The strength of such associations may vary, both within a community and over time.There is some basis for stating that collocation is a phenomenon which can partly be measured in terms of its surface systematicity.Collocation is a linguistic subsystem which individual L1 users acquire, as noted above, not by study but by sheer exposure to the phenomenon from their early years on.Each new stimulus adds incrementally to the speaker's ability to control his/her own linguistic behaviour and to make certain judgements about other people's linguistic behaviour.In many ways, collocational control -which also embraces the ability not to produce incorrect collocations!-is not merely a hallmark of linguistic maturity, it is also a kind of shibboleth.That is why it is so important -it is also why collocational control is difficult to master, seriously so if the person attempting to establish control is not a native speaker of the language concerned.Such a perSon has to adopt or devise a chronologically optimised regimen for acquiring collocational control: the problem is that emulation of native speakers is too slow and simulation is too risky.The simulation has to be con-

Reproduced by Sabinet Gateway under licence granted by the Publisher ( dated 2011)
http://lexikos.journals.ac.za nectionist anyway, developing its own neural network, to use two terms in a literal rather-than a metaphorical sense.
Given that collocations occur freely and naturally in all those sociolects which are "everyone's property", it would be somewhat strange if they did not also occur in special languages, the LSPs associated with and "owned" and regulated by DCs which group professional persons by gradations of various types.It is a very pertinent -and underresearched -question as to whether exposure to the collocational habits of technolect speakers instils collocational orthodoxy of linguistic behaviour in the same way as in general circumstances.One hypothesis must be that there are differences between the two situations alluded to.People joining obvious professional groupings do so at a time in their lives when their adult native language habits are more or less established.Nonnative speakers have a more obvious problem.Either way, there is a learning and adaptation process which has to be much more purposive and "supraliminal"; this, in its tum, requires deliberate reflection, learning and greater awareness about linguistic, sociolinguistic and metalinguistic matters.The urge to adopt "group-speak" is more keenly felt even if "group-thought" lags behind somewhat!Yet there is an ongoing tension between the demands of the logical and ontological framework of the professional DC and the requirements of illocutionarily successful linguistic formulations and terminological codifications.The "genre" problem within a DC's discourse is an added subtlety, not least in terms of the collocational practice of DC members vis-a.-visother DC members or "layfolk", to name only one such communication axis.

A LSP-based investigation of collocability
The (ongoing) research on which the above remarks are based is an investigation of LSP discourse in the areas of finance/banking and -partly -of medicine. 8The texts used in the investigation come from two types of source.Firstly, the holdings of the Aston Scientific and Technical English Corpus (ASTEC) provided the medical text.Secondly, and quantitatively more importantly, the finance and banking material was read from a commercially available CD-ROM which contains the entire text of the 1993 editions of the Financial Times, classified into sub corpora designated by theme and genre.'Once a prima facie satisfactory corpus of text has been identified, it can then be processed.For very large corpora (i.e. 2 million tokens) the machine processing is carried out in a UNIX environment, via a suite of programs expressly designed for the purposes described in some detail here.The method of initially selecting and then identifying multiword textual segments is simple enough.The objective is to find, inspect and count batches, no more than arbitrary fragments really, consisting of nine orthographic words.Once found, these fragments can be sorted and arrayed by their "middle", i.e. fifth, word.The scanning window then moves one word to the right, so to speak, and iterates the same basic process.Some further sophistication is available -at adm:.•tional computational expense -by the following method which has been shown to be operationally viable with groups of two words (dyads) and three words (triads).The algorithm works as follows: in the case of dyads, the words comprising them must be within a sentence boundary, with no intervening punctuation marks.Neither member of the dyad may be a function word.For triads the strategy is similar, with the "concession" that the middle word in a triad may be the function word of.The yield, unlemmatised, from the algorithm is sorted by frequency and then inspected with a view to identifying segments rather than fragments.Many of the segments are proper names, either institutional or personal, of encyclopaedic significance to the DC concerned.Those that are not are either terms representing cognitive units or they are potential collocations.A list of 553 triads -representing occurrence frequencies descending from 308 to 3 inclusive -was derived, by the above "unintelligent" methods, from a sample of the World Stock Markets corpus.Of the items in this list 62% represented cognitive units (interest rate cut), proper names (UBS Phillips Drew), jargonistic turns of phrase (dealers took profits), or collocational expressions (volume remained flat).Similarly, a list of 1268 dyads -representing occurrence frequencies descending from 431 to 5 inclusive -was then derived from the World Stock Markets corpus.Of the items in this second list 44% represented the same set of lexical categories: cognitive units (corporate earnings), proper names (Bill Clinton), jargonistic turns of phrase (depressed sentiment), or collocational expressions (Paris fell).An analogous analysis of a medical corpus drawn from the British Medical Journal yielded 121 terms from a list of 883 triads (see Appendix II).
Computational experience has shown, at this stage, that the main operational requirement is copious output which can then be refined by classification even if the discard rate is high.A surfeit of information is better than a dearth thereof!The steps needed to achieve copious output are easy enough to understand and appreciate.A large corpus will yield a rather long list of types, to which is appended the number of tokens for the relevant type, that is, its occurrence frequency.No attempt is made at all to group the types together into their "lemma set" and hence attempt to look at the behaviour of the canonical form.This can be done separately for English text, but even for inflected languages (such as Polish) it is more profitable in fact to look at the types independently of their lemma "allegiance".The point is that the various forms of the lemma all lead different lifestyles in text, not just in terms of frequency but also in terms of the lexical company they keep.The grouping action of lemmatisation actually obscures important information about type frequency and functional load within text and -above all -about type behaviour in terms of semantic differentiation.Be that as it may, it is probably the type frequency list (plus token counts) which is the main priming material for further investigation.Clearly, the analyst has choices to make, invidious choices, possibly.Attention tends to gravitate either towards types with a fairly high frequency or towards items of almost curiosity value.Having once selected an item for further scrutiny, the analyst goes back to the corpus an9 retrieves all the lines containing the item in question.This is the necessary preliminary to the study of the said item's distributional characteristics -that is, seeing how its meaning is "defined" by the constancy of its neighbours, its morphological variants and, if such applies, its -often Janus-like -polysemy (see Appendices ill and IV).This is, in fact, the only logistically feasible way of studying words as "chameleons".It can be noted in passing that any search profile can be entered as a string rather than as a fully-fledged word if interest is actually focused on a lemma or hyperlemma (as in the set: STRUCture, reSTRUCture, STRUCtural, STRUCturalism, deSTRUCtion, conSTRUCtive etc.).The result of the actual search is a set of one-line concordance citations which may supply information that is adequate for the analyst's purposes.If this is not the case, the entire sentence from which the keyword has been somewhat artificially extracted can be retrieved for inspection.As a further specific aid to collocational studies, a so-called synoptic chart (see Appendix V) can be produced of the node's left and right neighbours to a depth of four items.The individual items arrayed are "decorated" with their own occurrence frequencies.This is the basic method of attack for the purposes of collocational analysis but it is worth mentioning in this precise context that an overview of the valency pattern of the node words selected often also comes across very forcefully.Valency patterning is, of course, held by some to be an integral pattern of collocation studies.The techniques enumerated here have also worked, incidentally, to good effect on material in languages other than English.
It is clear from the initial stages of the above ongoing programme of research into LSP collocability that, as regards text generation, a marked degree of formulaic writing exists in the World Stock Markets financial press.This is true with respect to sentence structure and to the choice of words used for the purposes of qualification.This is either a tendency to avoid any disorientation in readers' minds that might be caused by novelty of usage or -just possibly -the result of summoning up, by reason of time pressures, certain prefabricated utterance structures of an algebraic kind and leaving the few variables mostly unchanged (see Cowie 1991 and 1992).
The techniques described immediately above achieve their greatest effect in the way they present for inspection the left-hand and right-hand neighbour(s) of the nodes chosen.Scrutiny of these environments leads naturally and fairly rapidly to the identification of any collocations present.Let the professional collocations used to describe the quality of stock market trading in Appendix VI speak for themselves and invite reflection, but let not the merits of the material in Appendices II, III and IV be overshadowed.in the process!

Implications and Renvoi
The implications of what precedes are numerous but can nonetheless be classified as either practical or theoretical.Let us deal with the practical implications first.Computationally, the impediments impinging on both principle and practice are minimal.Largely, the name of the game continues to be the game of the name!In terms of particular investigations a great role is played by pedagogical impetus: in fact, the agenda can safely be driven by this as no gulf is likely to open up between those who seek and find materiafO and those who place it in the lexicographer's/terminographer's crucible.
On the level of theoretical considerations, there is much more to speculate about.Let us ask some questions which are not merely rhetorical.What is the precise nature of the mismatch between orthographies and the cognitive units they must find designations for?What is chunking -with respect to term formation and collocability -and how can synchronically-and diachronicallyfocused studies be pursued with both descriptive and analytical purposes?How does the primeval soup of language congeal in different ways and in what contrasting and competing ways within particular languages?What is the logical, semasiological, ontological and analogical nature of, say, the sets (or classes?) of adjectives collocating with certain nouns or groups of nouns?Is this all sui generis or is it describable in generic terms?

Acknowledgement
I am very grateful indeed to Anthony P. Cowie for his expert and penetrating critique of this article and also for several suggestions about how it might be improved, all of which I have adopted with alacrity.

1.
For further discussion of the linguistic and sociological basis of Des, see Knowles 1997.

2.
Let us, for the purposes of this discussion, set aside the cross-cultural differentials and the occasional mismatches between the encyclopaediC networks of "national" Des.Such mismatches cause communication problems by cropping up in LSP discourse potentially addressed to all members of a supposedly international and "shared" specialist/intellectual culture.

3.
It may be that Gennan's/Gennans' tendency to generate complex/compound neologisms in the fonn of large numbers of one-word lexicalisations (their words are Komposita or Zusammensetzungen!) is explained not just by the linguistic resources available but also by holisticising perceptions.4.
Languages such as Arabic possess a mechanism known as iDaafa (annexion) in which two (occasionally more) words, while remaining orthographically separate, unite to fonn a separate cognitive unit.An instance of this is Arabic ra's maal, literally head of money, i.e. capital.These units are fully lexicalised and are atomic in the sense that no other linguistic material may be inteIposed between the two orthographic words involved.Hebrew, Farsi and Turkish also have very similar structures.Note that these compounds are exactly that, they are not collocations, although collocation might well have been their origin.5. See Varantola 1984: 42. 6.
From the article "Navigation" in Amerongen 1977.7.
Why do English speakers always enter a caveat?Enter, here, is a truly transitive verb, a usage very different from -entering-a building.mainly targets the that because there have been so many m committing one of the biggest attract night stay in a Marne la Vallee trippers or staying at nearby its us and Wellworth food retailiDg d standards and with its own 39 / Turkish fears grow on foreign I might place some of its small te and the postponement of many be said that many, if not most of he company along with other agri 9, has resigned to follow other production and flat February d yesterday showed the ratio of terday showed that the ratio of business rates impose on 'Our core Brothers of Belfast, and the US to power the new Gulfstream GV GV flight of the new It is true the and billed as the fastest-ever ion last year.However, the 186 o find an education pundit or a However, Mr Hata was responding to A timely combination of new prothough the company that once had a boomiDg een told that his parliamaDtary 'The goveramaDt'.

GBRMAH
will help gOV8ramaDt'.tactics by banks in the small ple is commonplace in the small fares may benefit the asset.They are .teepediD the it will be the in monitoring the bu.iDe when customer retail decisions and would not affect stacIe in building constructive t the well-publicised series of business operations business operations recently put up for sal business operations would be about half the business opportunity.business park for smaller industrial and se business park in partnership with the Churc business park, industrial areas and housing business park.business parks are on the periphery.business parks in the north-east, are not p business parks in the north-east, are not p business parks, business parks, has asked National Westmins business parks, there are hundreds of acres business partner of Ms Lansing's.business partner reporting on the ripeness busines.partners.business people need is panic action to ref business people who business people, judges or politicians.business people, led by its deputy business performance confirmed the broad we business pessiBdsm, and the political hiatu business plaD business plaD and may decide to sell core 0 business plaD for the company, says that th business plaD once it was complete and had business plaD to lenders later this month a business plaD was complete.business plaD were business plaD which business plaD which it updates business plaD which, it claims, will 'retur business plaD written in spring last year h business plaD, in spite of business plaD, the airline plans to open at business plaDDing at Itochu, business plaDDing has been based on very co business plaDS business practices business practices are reminiscent of business problems, the company has plans to business product with a finite constituency business profits for the year business profits to Pounds 11.9m from business property costs such as rent levels business prospects is still holding business prospects prompted profit business rate in 1990, brought in an estima business rate is set by the government, but business rate revenue into the national non business rate was collected business rate.However, the Department of t business rate.This year, though, The Depar business rates -and council tax when that business rates -and council tax when that business rates could lead to unexpected inc business rates had been cut by Pounds 1.25b business rates impose on business invest business rates non-payment'.business rates were 'the tip of a nasty ice business rather than one which generates a business reached Pounds 1.154bn.business relationships between the two business relationships' .
or from the Freemans mail ord.r nearby to haD41. the volume of _11 pending is put the way of black that only gangsters are in the they are .ti11 in ers were informed.was open for kely to remain in charg. of the y no comprehension of ruDDing a this week.had we been in unning a town is like ruDDing a in turning round the 'OK beDking ince •they are part of our cor   yet another day of extremely active trading.Elsewhere, Peugeot was knocked bac ore than Dollars 3 1/2 1/2 in active trading after the company said it backed Pr s less than Dollars 22 1/4 in active trading.Merck also saw heavy turnover, clo dipped 0.29 Dollars 49 3/4 in active trading after Merrill Lynch cut its long-te isregarding down 1/8 at 44 in activa trading.Citicorp was up 7/8 at 21 5/8 in a icorp was up 7/8 at 21 5/8 in active trading.Canada ACTIVITY was limited, with ose 9.7 to TAIWAN advanced in active trading with the weighted index putting on g OSLO jumped 2.3 per cent in active trading as the economic outlook appeared na erage climbed 4.9 per cent in act iva trading on rumours that the last minutes of RONTO ended little changed in active trading boosted by a large block Nikkei in m A HELSINKI closed higher in active trading, as the HEX index closed 13.1 or SH ocks ended slightly higher in active trading.Based on preliminary Advances led t close off intraday highs in active trading.The STOCKHOLM gained in moderate t Toronto share prices lower in active trading.1,889.5 in thin trading.Dealers a continued its recent rally in active trading on the first day of the Other state 2.1 per cent and saw its most active trading this month on before closing FFr13 and arbitrageurs, In spite of active trading by domestic investors, traders said ion.The TSE-300 fell 15. ng demand for Astra made up for dull trading in other issues.The Affarsvarlden index slipped 1.09 to 279.36 in dull trading.Astra was the most in yet another Jax, JOHANNESBURG was firmer in dull trading with the overall index 14 higher at cross-shareholding, issued a glo~ trading US share prices were flat-to-lower ted the market.The company's glo~ trading trading days it has lost more than PARIS dropped 2.9 per cent in glo~ trading on the last day of the account L6,O of the per cent lower after a heavy trading session.The TSE-300 fell 15.3 to a stocks finished lower after a heavy trading session with most in busy trading T stocks finished lower after a heavy trading session marked by the afternoon tra rices ended mixed in continued heavy trading.Nikkei average gained marginally a months before the in extremely heavy trading.Although much of that rise was con d Wednesday's gains in further heavy trading as the JOHANNESBURG revived in late ram trading combined 24 3/4 in heavy trading on disappointing fourth-quarter ear ollars 3/8 to Dollars 9 5/8 in heavy trading after the A late selloff sent Toron 5 shares STOCKHOLM advanced in heavy trading as strong demand for Astra made up CKHOLM rallied 4.1 per cent in heavy trading on speculation, confirmed market.F e prices were flat-to-lower in heavy trading yesterday as the market trading.Lo sing FFr13 better at FFr506 in heavy trading of 129,325 shares STOCKHOLM advance -CHIP stocks jumped sharply in heavy trading on US stockmarkets Because of the ck markets rallied strongly in heavy trading on unexpectedly good trading patter f the last trading day for the Heavy trading in Westpac's new shares and options HANNES BURG ended mixed in lacklu.tretrading ahead of today's public US SHARE pr days.The NEW ZEALAND saw lacklu.tretrading as the NZSE-40 index moved up just omposite rose 2.79 to 637.16 in late trading.Shares in Tele-Communications led trading day before to Ll,500 in late trading.The telecommunications company had ORT saw a flurry of activity in late tra4i"g.having moved in a narrow activity er cent.HONG KONG fell back in late tra4i1lg after the Sino-British Joint Liaiso re, Peugeot was knocked back in late tra4i"g to close FFr15 or 2.8 being delayed 25 to AUSTRALIA lost ground in late tra4i"g as the local currency weakened agai ~rs.SINGAPORE closed higher in late tra4i"g as foreign institutions came into M ing by overseas institutions in late tra4i"g.BOMBAY fell to a low for the 1992/ as the JOHANNESBURG revived in late tra4i"g.helped by a better performance fro d the but recovered slightly in late tra4i"g to finish down 18.47 at l,476.0l. a NG ended lower in volatile but light tra4i"g.The Hang Seng Index of the day's t er the recent MANILA dipped in light tra4i"g as the composite index fell 7.35 to in 20m shares.MANILA slid in light tra4i"g but brokers still said that they ex ces remained in easier vein in light tra4i"g yesterday as losses, and took its d technical rebound led by Kepco light tra4i"g yesterday, writes Patrick Harverson 53, while the KUALA LUMPUR saw light tra4i"g continuing as the composite index d into MANILA closed lower in li.tle •• tra4i"g as the composite index shed 10.92 t in trading and the lower in li.tle •• tra4i"g: the all-share index shed 8 to 3,25 x dropped 7.52 to 348.75 in moderate tra4i"g PARIS dropped 2.9 per cent in gloom otest SEOUL ended firmer in moderate tra4i"g on a technical rebound led by Kepco ng.The STOCKHOLM gained in moderate tra4i"g as domestic interest rates fell bac ity.Declines SEOUL rose in moderate tra4i"g but sentiment remained nervous ahea us SHARE prices were mixed in mode.ttra4i"g yesterday as the Thanksgiving activ ars 5/8 to Dollars 63 1/4 in "ervou.tra4i"g ahead of retailer's trading and imp he SINGAPORE fell sharply in "ervou.tra4i"g on talk, confirmed after the tradin ing on AUSTRALIA was firmer in quiet tra4i"g.The All Ordinaries index rose 9.7 ices moved marginally lower in quiet tra4i"g ahead of the release of low of 17,1 xed after drifting sideways in quiet tra4illg.Gains in there was no discernible PORE ended marginally lower in .lacktra4i"g ahead of the holidays.The NEW ZEAL X contract MILAN ended lower ill .lowtra4i"g and the Comit index fell 4.17 to 44 Y56 to Y695 on increased .peculativetra4i"g, after rumours of at l42.9m shares luation of the crown 3,134 in .tea4¥tra4i"g.The gold index added 24 or 3 per c in the morning session on techllical tra4i"g related to the hands in the first h gen was also boosted by a .hort-termtra4i"g buy The Tokyo Stock Exchange suspen ency index rose 0.6 to 103.2 in thi" tra4i"g.BUYING by public funds and dealers r in active trading.1,889.5 in thi" tra4i"g.Dealers again blamed weekend polls by some institutional buying in thi" tra4i"g and the lower in listless trading: PUR drifted to a lower close in thi" tra4i"g as investors remained market-makers ines to a fifth straight day in thi" tra4illg.The trading after a one-day halt, owed.PARIS continued to fall in thi" tra4i"g, and some sell orders from the UK V eased just 0.05 to OSLO fell in thi" tra4illg as speculation about a devaluation d by SINGAPORE closed firmer in thi" tra4i"g as the Straits Times Industrial KUA firm but off the day's highs in thi" tra4i"g.The thin trading in very thin trad to SKr33.5 MILAN ended mixed in thi" tra4i"g and the Comit index eased just 0.05 sted by a large block Nikkei in thi" tra4i"g HONG KONG closed sharply lower on p that China TAIWAN retreated in thi" tra4i"g as the weighted index lost 28.25 to ort policies.AMSTERDAM rOSe in thi" tra4i"g with the CBS Tendency index gaining e fourth consecutive session in thi" tra4i"g as many BOMBAY fell again in tradin the HONG KONG eased slightly in thi" tra4i"g, as modest early gains were erased rading, genuine NEW ZEALAND saw thi" tra4i"g as the NZSE-40 index eased 13.78 to ay's highs in thin trading.The thi" tra4i"g in very thin trading, Reuter report four weeks.SEOUL slid in very thi" tra4i"g in a technical adjustment after the ading.The thin trading in very thi" tra4i"g, Reuter reports from Tokyo. the New have been al ba.iDe •••• rvic •• have both recovered lost ba.iDe.. .PeD4!Dg foreea.tba.iDe ••• trat.gy for years to come -is no ba.iDe••• trat.gygoes further than a simil ba.iDe.. .trat.gy.' ba.iD •••• trat.gy.ba.iDe.. .tudi••• ba.iDe ••• upport in recent weeks.ba.iD... teDaDci.. to housing ba.iD ••• teDaDt.the right to ba.iD ••• t.~.' he reckons.Meanwhile he i ba.iD ••• t.~.' ba.iD ••• travel and tourism.ba.iD ••• travel caused by the Gulf war and ba.iD ••• travell.r.more choice.ba.iD ••• travell.r•. ba.iDe.. turned over about twice as much as ba.iDe.. unit consists of three product ope ba.iDe.. unit focused ba.iDe •• unit to sell its new Canadair Regi ba.iDe •• unit'.mission is to build, on top ba.iD ••• unit, GPT Cablecom, to tackle the ba.iDe •• UDit.beneath those groups in its ba.iD ••• UDit.' (SBUs) in Vevey for each Ne ba.iD ••• unit •. ba.iD ••• volume in the insurance industry, ba.iD ••• volume.are still depressed and ba.iDe •• volume.during the quarter, the ba.iD ••• volume.in the next three months.ba.iD ••• volume.remain well below ba.iD... volume.to ba.iD . . . . . .ho14iDg up, helped Great Unive ba.iD... was reported to have come from the ba.iD ••• was smokers' accessories.Today, p ba.iD ••• wer.har4 hit.They fell from ba.iD ••• will attract coapetitiOD, but is c ba.iDe •• will be lower.Last year S4 per ba.iD ••• will be threatened by the new law.ba.iD ••• will have access to female busi ba.iD ••• will start from ba.iD ••• with Hanoi, the sensitive issue rna ba.iD ••• with a view to ba.iD ••• with U., not ba.iD••• writ.4OWD.and closure ba.iD ••• -frieD4ly 10catiOD in ba.iD ••• -orieDtate4 research institute, say ba.iDe •• -r.lat.4 bankruptci.. to total indi ba Gateway under licence granted by the Publisher( dated 2011) 3 to active trading, writes Emiko Tera2ono in Tokyo.tr ding day, the longest winning active trading by public funds was prompted by fin se Dollars 7/8 to Dollars 10 in hu.y trading after the company TORONTO stocks fi ed Dollars 7/8 to Dollars 17 in hu.y trading after FRANKFURT saw a flurry of act lars 3 3/4 to Dollars 55 3/4 in hu.y trading after the TORONTO stocks finished 1 Dollars 1/2 to Dollars 8 3/4 in hu.y trading ACTIVITY surged in the morning sess lars 1 3/8 to Dollars 32 1/8 in hu.y trading after Mr PARIS eased on profit-taki ollars 1/2 to Dollars 65 3/8 in hu.y trading after Merrill Lynch, TORONTO share vy trading session with most in hu.y trading TORONTO stocks ended slightly highe US SHARE prices held steady in hu.y trading yesterday as financial markets Hone y at the end of a quiet and cautiou.trading adjusted positions and there has be December options chairman's cautiou.trading statement on Monday. of trading was g SINGAPORE closed lower in cautiou.trading as sentiment was dampened by Hong C ing TAIWAN remained weak in cautiou.trading ahead of Saturday's parliamentary B