CO-ASSOCIATIVE COHESION IN ENGLISH TEXTS A PROGRESS REPORT ON RESEARCH INTO THE SYSTEMS OF LEXICAL COHESION IN EVERYDAY ENGLISH USE

A first-year course entitled "Prose Structures in Everyday l::ngllsh Use" follows the text of a book shortly to be available by the instructor, Dr. Michael P. Jordan. The course enables students to study the structures and styles of many forms of English use including news reporting, professional writing, and both serious and entertaining articles. Although this course provides a sound basis for the students to improve their own writing abilities, the main aim is to develop in the student an articulable awareness of the structures and linguistic signalling of English texts.


NEW WRITING COURSES AT QUEEN'S
The previous academic year at Queen's University, Kingston saw the successful introduction of two new courses in technical writing mainly for engineering and science students.
A first-year course entitled "Prose Structures in Everyday l::ngllsh Use" follows the text of a book shortly to be available by the instructor, Dr. Michael P. Jordan.The course enables students to study the structures and styles of many forms of English use including news reporting, professional writing, and both serious and entertaining articles.Although this course provides a sound basis for the students to improve their own writing abilities, the main aim is to develop in the student an articulable awareness of the structures and linguistic signalling of English texts.
A second-year course in "Basic Communication", also given by Professor Jordan, is a series of formal lectures on the basic elements of efficient writing which, he explains, need little detailed explanation and even less intellectual thought.The topics dealt with are thus the generalities and folklore of technical communication which can be taught at little expense, and learned quickly by the students.
Such subjects are: planning and outlining, definition, general communication concepts, word use, punctuation, grammatical errors, illustrations, conciseness and concepts of style.Students on this course are achieving a level of understanding of these concepts comparable to students who have taken a course costing 15 times as much.
Michael Jordan also established the principal course at Queen's several years ago, "Effective Technical Writing", which teaches the more advanced concepts explained in his research publications.The essence of the following article was presented at the Applied Linguistics Research Working Group Spring Colloquium at York University last April.

BACKGROUND AND AIMS
"The moment two sentences are placed together as members of the same paragraph, they enter into a semantic relation with each other whether we like i t or not."(Winter 1978, P69/90) This statement can be seen as the philosophical starting point of detailed research into "clause relations", which studies the many relations that exist between clauses and sentences.Winter's major work in clause relations (1978) not only discusses the more obvious relations conveyed by sentence adjuncts and subordinators, but also explains in detail relations dealing with compatibility and comparison.
Of relevance to this paper is his analysis of comparative denial (what is true of X is not true of Y) and comparative affirmation (what is true of X is also true of Y) as major features of the matching relation.This paper will show the compatibility of this analysis with more recent work dealing with inter-clause connection created by lexical repetition and related devices.
Another branch of study centered around Winter's work (1976) is known as "information structures" or "prose structures".This study deals with the types of high-priority information found in texts, typical patterns for such information, and the linguistic signals that indicate the types of information and show transition between them.
Although this work has received considerable attention (e.g.Hoey 1979Hoey , 1981Hoey , and 1983;;Fries 1982, and Jordan 1980, 19Bla, 198lb, 1982band 1983), much more analysis needs to be done.The introduction of old or obvious solutions to a problem into a text is a communicative need that is shown to fit the general pattern of continuity described in this paper.

Systematic study of lexical cohesion between clauses and sentences
of a text has received little detailed attention until quite recently.Halliday and Hasan' s (1976) chapter on lexical cohesion and particularly their sections on collocation and general concepts of lexical cohesion provide a useful starting point.Their categories of "reiteration" have since been expanded and placed within a larger nominal cohesive system of continuity (Jordan 1982c), and the relevance of lexical cohesion between clauses and sentences has been connected with theme (Jordan 1982d), it often being impossible or stylistically undesirable to reiterate a previous topic as theme of the following clause.Of special importance to us here are Halliday and Hasan's discussion of hyponyms and antonyms as lexically cohesive devices, and also work by Christophersen (1939), Hawkins (1978) and Jordan (1981) to define and explain intra-clause lexical connection created by lexical items "associated" with lexical items previously included in the text.This paper further elaborates on the types of association possible between lexical items in different clauses, and will thus show some compatibility of the systems of lexical cohesion both with clause relations and with information structures.

TOWARDS AN INTEGRATED THEORY OF COHESION
Some progress has already been made in integrating the cohesion of clause relations with that created by lexical, pronominal and grammatical techniques.Host of the ways that educated users of the language adopt to "re-enter" co-referentially a topic already introduced into the text (e.g.substitution, synonymy, full and partial repetition, naming, and generic nouns) have been identified; and the "re-entry" devices within the sentence have been identified as ellipsis, listing, and the use of relative, non-finite and verbless clauses (including apposition).These technqiues provide connection within and between clauses by virtue of the same topic being involved in two or more statements in the text.The relations that exist between the clauses (which are often in addition to those created by lexical cohesion) are part of a clause relational study, and an introductory analysis has been made (Jordan 1978) of short extracts of text which exhibit both types of cohesion simultaneously.This work has been expanded slightly in Jordan 1982d.

6
The major point of similarity between the system of lexical and related cohesion patterns on the one hand, and relations between clauses on the other is that of "re-entry" of part of the text back into the text.
For the lexical branch of cohesion, a noun or nominal group is re-entered into the text to allow the speaker/writer to add something about that topic, or about something associated with it.For clause relations, a clause (or sentence or paragraph) is re-entered into the clause so that more can be said or written about it.As an example, pronominal this usually has at least a clause as its referent or antecedent, and thus it re-enters that clause into the text for elaboration.
The previous sentence is an example of the occurrence of both types of coherence working together to provide cohesion between the topics of discussion of a text and the clauses of the text.Here, in simplified form, is its structure: Cohesion in this sentence is achieved not just through the use of the substitute it to show extension of the domain (Hudson 1968) of the topic, but also by thus which re-enters the first clause to create a "CAUSE-EFFECT" relation between the two clauses.(Reiteration by nonthematic its referent or antecedent and that referent are further nominal re-entries, but we will not be considering such complications here.)The sentence we are studying also re-enters the preceding sentence, giving au example; this is more clearly seen by the expansion "As an example of this", where thiswould refer to and thus re-enter the previous sentence to create a relation of "EXAMPLE" between the two sentences.It seems highly probable that the notion of any coherent stretch of text being re-entered into text will provide a powerful mechanism for integrating our understanding of these two branches of cohesion.
Integrating well established concepts of information structure with these two branches of cohesion is much more difficult, but perhaps not insoluble.The clause relation of "EVALUATION" fits perfectly with clause relations as evaluative information answers questions about a previous part of the text, but the boundaries between "Situation" and "Problem" and between "Problem''and'!>olution" are not so easily resolved; at present perhaps we have to regard them as special cases of clause relations, but that is a far from satisfactory explanation.A particularly difficult aspect of information structure to integrate with other concepts of cohesion has been the introduction of competitive or old solutions into the text as these are not reiterations 01 teentries of something previously mentioned in the text.The difficulty is explained by comparing synonyms and antonyms as cohesive devices in a text.A synonym enters the previously included lexical item coreferentially, whereas an antonym enters something that is opposite or complementary to that item.What is more significant is that synonyms provide a means of continuing the discussion about a given topic, whereas antonyms introduce something new (but related to the topic) in order to provide some comparison between the previous and newly introduced topics.
The detailed work on lexical cohesion and other re-entry techniques has so far dealt with the means by which a topic in the text or something associated with it is included again in the text, and the explanations provided have excluded the entering of antonyms, old solutions and co-hyponyms as these are not means of re-entering the given topic into the text.The explanation now needs to be expanded to accept these techniques as a separate branch of lexical cohesion, and this will be achieved in the following section of this paper.The next three sections deal with the implications of this extension in the use of co-hyponyms with comparison in everyday English use, and there follows a discussion of how co-hyponymy enables us to explain more fully some aspects of information structure.The final section examines implications of this work and directions for further research.

ASSOCIATED RE-ENTRY EXPANDED
The early work of Hawkins (1978) and Jordan (1978) demonstrated the existence and use of associated nominals (newly introduced nominals associated in some way with a previously introduced nominal).Later developments (Jordan 1981(Jordan , 1982c) ) established several types of association: e.g.whole-part, cause-effect and other logical associations, 7 and intangible features or characteristics.Hore detailed study of lexical cohesion in non-thematic positions (Jordan 1982a) provides some introductory analysis of chains of associations including hyponyms, co-hyponyms, hyponyms of hyponyms, etc.It is now necessary to distinguish clearly between three major types of association: hyponymic, superordinate, and co-hyponymic.
The work on associated nominals has so far concentrated on hyponymic association, in which cohesion is created by a new nominal group which is a hyponym of the original nominal group (or "trigger").Because of the inherent compatibility of the "re-entry" systems of clause relations and lexical cohesion, it should be clear that the trigger does not have to be a nominal group, but can be a clause, a sentence, a paragraph, or indeed any coherent stretch of discourse.In hyponymic association, the new nominal can be one of many types of hyponym as shown by: Here is an example of an untriggered associated nominal (i.e.there is no mention of the trigger in the new nominal): 1 Thh pro1raa h touaJt, 'beceuae thigh• are hard to reduce, particularly for .......,, lut 1D juat one llODth, you can have thiDDer, fir.er thi1ha Partial superordinate association also occurs with the use of suah, Zike and of this type to indicate some generalisation of the trigger to a complete sub-set defined as being all items like or of the same type as the trigger.Context is vital in understanding some partial superordinate associations.For example, th~e of T would be a partitive hyponymic association for a plural T, but a partial superordinate association for a singular T. As with all associations, the new nominal has no definite meaning without reference to the trigger.
The various types of superordinate association deserve a great deal more attention, but they are not the main point of this present paper which centres around co-hyponymic associations.DimeMione of the traMfo1"fter and Soldsr piM on tli.tranefo1"fter.

CO-HYPONYMIC ASSOCIATIONS
The co-hyponyu are not directly connected, however, aa the di•ndone form one part of the phyaical characterietica of the tranaformr, vhereu the aolder pine are co.ponenu of the tranaformr.
In Example 3 both associated nominals can be seen to be associates of the implicit associated nominal "Parts of the drive belts".Further complications occur when a superordinate term is classified in mutually exclusive ways.If, for example, a group of people is classified first by sex, then by age group, and finally by education, there will be nominal groups such as "Those between 18 and 25" and "Those with a university degree", and such nominals will only be seen to be connected as overlapping co-hyponyms of the main superordinate term.

CO-HYPONYMY AND COMPARATIVE DENIAL
If we treat co-hyponymy in a wide sense, co-hyponymic association does not have to occur in such natural sets as we have just seen.It can occur whenever two or more topics are used as a basis for comparisonfor similarities (comparative affirmation) and/or differences (comparative denial).The common saying that apples cannot be compared with oranges does not always apply in natural language.Obviously a more meaningful comparison is achievable when comparing oranges with tangerines rather than with apples, but the higher superordinate fruit forms a valid basis for some comparison between apples and oranges.
Even oddly different topics such as tennis and motherhood can become co-hyponymic when they form part of a comparative denial: 5 A co11ple of yean ago it vu tnnia.llov it He• a lot of vo.en are reaching for 110therhood, before it'• too late.(Good Health, JVD/Jlll 81, Pl6) The comparative denial involve• tiaa, vith the aignala A ooupZa of H•IU"• ago and """• Th• act1v1tiaa for yo11ng VOOl8D (the a11parordinate concept) are identified a• t.nnia and 1110tharliood, and ve H• that vbat ia tr ... for tennia ia not true for 80tberhood -and that thia baa cbaqed vith the.
Comparative denial is even more clearly marked in the following example:  The two previous examples were opening statements from semi-formal articles in the entertainment/informative genre.Comparative denial is a common opening for such writing, and it is often signalled by But or However as sentence starters early in the text as denial is one of the conventional "surprises" signalled in this way.Here is an illustration:

CO•HYPONYMS AND SET DETAILING
Although comparison is an important motivation for co-hyponymic association, another is the need to introduce two or more members of a superordinate set of topics or things to enable each to be described in turn.
In the following example, there are two sets of co-hyponymic families within a complex "Problem-Solution" framework.
ro1t1r Killar are to delian • •Y•t•• that burn• die•el fuel llllod with eir to 1enerete ateu by direct contact with voter.Both the It•end the cOlll>uotion 111eo could be injected into the reaervoir, evoidina probl ... of 1t1101pberic pollution due to the ca.buotion.
The locbtdyue Divilion of Rockwell lntarnational are vorlting on 1 11-ilar idea in vbich the ateu i• 1ener1ted in "down-hole" heat uchanaera.rather than by direct coo tact, and the axhau1t a••e• are vented to lt9Dlphere.Tbio avoida the po11ibility of "pluaging" the reaervoir with particle• 11n1rated during coabuotion of the fuel.Another locltvell propooal which ii bein& otudied ii to uoe an electric heatar at the bottOll of the well to produce ota••• The Doll plan• to aelect the teclmique with the belt potentiol for field teattna in 1980.
llaanwhile, under another Do! contract, the General Crude Oil Coapany io inv11tig1tin1 another .. thod of recovering additional oil frOll underground reoervoira.The c1111pany ii to telt an in-1itu cOlll>uation proc111 that burn1 part of the oil in I reaervoir to heat the ra . .inina oil.Tb• reduced vi1co1ity enableo the c011bu1t1on 11111 to drive the previouoly unrecoverable oil to I producin1 wll.In practice co-hyponymy often involves comparative affirmation and comparative denial, as co-hyponyms have some aspects in coumon and some different.This was shown in Paragraph 3 of Example 9 and is further illustrated by:

CO•HYPONYMS AND INFORMATION STRUCTURES
The identification of co-hyponyms has already helped to explain the information structures of Example 9.In that example, two sets of evaluations were seen to be connected by co-associative nominals, and other applications can readily be understood.For example, "agreement" is comparative affirmation of two or more evaluations of a topic, and "disagreement" is comparative denial of different evaluations of a topic.
In addition, many problems arising from a given situation can also be recognised as co-hyponyms.
The most difficult aspect of information to explain in terms of a coherence pattern involving successive re-entry of a topic of description has been the instance where an "old" or previous solution to a problem is introduced and its deficiencies discussed as a basis for then introducing the "new" solution, which does not have these deficiencies.This is the whole basis for the concept of improvement, a subject discussed briefly in Jordan 1980 and dealt with in more detail in Jordan 1983.The information pattern is typically found in advertising, as shown in the following advertising extract.

REVIEW AND PREVIEW
Although several systems of lexical cohesion have been identified, their uses in combination need detailed analyses, snd related systems of implicit connection and perhaps allusion remain to be analysed.The whole area of lexical cohesion is an important study in its own right, which has many possibilities for practical application.Hore significantly, however, the integration of lexical cohesion with other branches of discourse study is now promising to help in the development of a more general theoretical approach which describes and explains many elements of cohesion and continuity in contemporary English use.
The types of relations that exist between clauses need detailed analysis, but it now seems likely that the general approach to an explanation of clause relations is compatible with that for lexical cohesion.Huch of the work in information structures can now also be related to the system of description of lexical cohesion.It is known that many predictable types of information occur together frequently in texts, and that one important relation -that of EVALUATION -readily fits into a generalized description of textual continuity.As future work in information structures develops, its points of contact with detailed descriptions of clause relations and lexical cohesion seem likely to increase and thus add to our general understanding of textual cohesion.
Whether an integrated theoretical explanatory framework for text will ever be fully compatible with a theory of grammar remains to be seen.
Our knowledge of both aspects of language is still such that any pronouncements regarding an eventual overall "granunar" for clauses, nominals, sentences and text would, I feel, be premature.

83
Spur aaar cable drive belu are dad111ed to operate with "co ... rcial ft arooved apur a••r• 14i 0 or 20• pr•••ure angle and becauH of tbe ain&l• .. jor cable feature will twiat and tum ae deaired.JOl!ket and u:Untkd dl-ive piM are .. de of polyurethane for 8&X1-veer condition•.Cabl• ia •ultiple atranded atainlaH ateel for ta,..ile atren1th and flexibility.(DedlJl !n&ineering, Dec 75, P48) The triaa•r i• the topic of daecription Spur g..uo aable dl-ive b.lta, and there are tvo untrigered hypon,..tc aHociated nollinah Jacket and u:tended dl-ive pine and Cabl• which are thu• co-hyponyu.lo theee aHociatione the tri11er ia linguiatically recoverabh ae Jacket and a:tflnded dl-iv• piM weed in tli.b.lte and Cable incZM/Ud in t/,.b.Zte.Example 3 includes two associated nominals which are closely cohyponymic as they are both physical components of the trigger.Although all associated nominals of a given trigger must be cohyponymic to some degree, they are not always as closely related as in Example 3. If we think of the co-hyponyms in Example 3 as being "siblings", the ones in Example 4 should be regarded as "cousins": 4 The Tricoll liaht-vei&ht power tranaformr for printed circuit mountiaa on .1 etandard arid baa power rat101a frOll 3. 5 VA to 20 VA • • • Dimmeione are 1.9" JI 1. 5" " • 59" for the 3. 5 VA type, 3" JI 2.3" " .l"for the 20 VA type.Solder piM are 1old plated.(Canadian Electronic• ED&inaerina, Apr 74, P62) The tr111er ta 7'h4 f'Maoll Zight-wight power tranefu...er etc. and the tvo co-hyponymic aaaociatu are Dimeneione and Soldsr piM, for which the aeaociatione have linguietically recoverable trigger• by , no details are given of the co-hyponyms for "Parts of the drive belts" such as "Physical characteristics of the drive belts", "Patent applicability for the drive belts", etc., and thus the information given is closely related.In contrast, the information in Example 4 is less closely related because the co-hyponyms are hyponymic associated nominals from different branches of association from the main topic of description:The tricoll Uaht-1aht paver tranafo ...
n.e anvil and plunger •re cold-formed froa wire in one p••• by !lco lndu1trie1.Pl'c11ioue model• had acrew-uchined anvil• and plunaera.(Material& !naineerina, Apr.81, Pl38) The firot paragraph provides three denial• of what the cataphoric topic of deocription does not do, the clear implication being that that i• what earlier nutcracker• do.Thi• ii implicit comparative denial, with the co-hyponyu aloo being iilplicit.The second parearaph provide• detail• of the new 8Ddel and how it vorlto, and the final paragraph •antiono it• an11il and plungel' aa an untriggered hyponyaic eaooci•te.The find aentence introducea Pl'e11ioue rnodtlle11(a co-hyponym of the new 90del) in order to provide inforaation of difference between the old and the new.Again comparative deniel ie involved: vhat 11 true for the anvils and plunaero of the new 8Ddela ia not true for the anvile and plungero of previous 8Ddeb.
Winter's work in clause relations can now be seen to have significant points of contact with the description of lexical cohesive techniques, and combinations of clausal and nominal intra-clause connecting devices now deserve serious attention.
Clearly both chair and table have a mutual co-hyponymic relationship,• but in actual use we can also recognise that co-hyponyms are associated hyponymic nominals related to the trigger (furniture here) as the super- .,,.,.,, o.ZZa and ...aiZ "PA"" oaZl•, which are both partial aub-aeta of all apazw calla.111• comparative denial la apparent from the But and the no in the aacond aentenca: what ia true for 80at apazw cell• la not true for •nail aper.calla.
12 Zinc •tronaer than ironT That'• vhat Danair di•covered vhen they •tarted lookina for another aaterial to replace the Cl••• 30 iron for the cap of thi• air ~r.[Illu•tr•tion provided.]Iron juet couldn't do the job; •ir vent• broke frequently •nd porooity c1u1ed uchining probleu.They finally turned to • zinc foundry alloy • • • The zinc alloy eliainateo brealulge and poroaity; uchinina ii faoter (zinc foundry alloy• routinely uchine 3 to 5 Uaea fHter than CHt iron); and, H a bonua, llanair 1et1 & •uperior f inllh on the part.(Material• !naineerina, Apr 81, This is used effectively on the following example, where the comparison of effectiveness of the new solution with previous models ie implicit in the first paragraph and is not made overt until the final paragraph.13 It doeon't .... h, crush, or shatter the shell.It doeon't flip little fra.-nu into the potato Hlad, down into the carpet, or luve t..._ in the nut uat• where they can break a tooth. PS4)Althouah sine alloy 1a bein& c-ared vith en .i-ntd•tal (iron) in th11 &IWlple, they are co-hyponyu for the aimple reeeon that they •re being coapared -•n el-nt can be validly c-ered with an alloy ju•t •• orangea c•n be validly compared with •ppleo.The iron 11 introduced &1 the old aolution which had deficienciH of oouldn 't do tlw Job •• the uaH&Mnt, and ail' 11ent1 bl'Okl f'l'equentl11 end pol'Oeity •• the opecif ic baaia for tha aeoeaament (concept• of evaluation dia-cuHed in Jordan 1983); thaoe are identified •• p1'0bl4me in lftOOhining.