Coherence, Cohesion in Writing Quality

8/6/2019 Coherence, Cohesion in Writing Quality

1/18

Coherence, Cohesion, and Writing Quality

Stephen P. Witte; Lester Faigley

College Composition and Communication, Vol. 32, No. 2, Language Studies and Composing.(May, 1981), pp. 189-204.

Stable URL:

http://links.jstor.org/sici?sici=0010-096X%28198105%2932%3A2%3C189%3ACCAWQ%3E2.0.CO%3B2-E

College Composition and Communication is currently published by National Council of Teachers of English.

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available athttp://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtainedprior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content inthe JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained athttp://www.jstor.org/journals/ncte.html.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.

The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academicjournals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers,and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community takeadvantage of advances in technology. For more information regarding JSTOR, please contact [email protected].

http://www.jstor.orgSat Jan 26 11:13:05 2008
http://links.jstor.org/sici?sici=0010-096X%28198105%2932%3A2%3C189%3ACCAWQ%3E2.0.CO%3B2-Ehttp://www.jstor.org/about/terms.htmlhttp://www.jstor.org/journals/ncte.htmlhttp://www.jstor.org/journals/ncte.htmlhttp://www.jstor.org/about/terms.htmlhttp://links.jstor.org/sici?sici=0010-096X%28198105%2932%3A2%3C189%3ACCAWQ%3E2.0.CO%3B2-E


2/18

Coherence, Cohesion, andWriting Quality

Stephen P. Witte and Lester Faigley

A quest ion of cont inuing interest to researchers in wri t ing is what internalcharacteristics distinguish essays ranked high and low in overall quality. Em-pirical research a t the col lege level has for the most part taken two ap-proaches to this q uest ion , examining err ors1 and syntact ic features2 whilegeneral ly ignoring the features of texts that extend across sentence bound-a r i e ~ . ~e i the r t he e r r o r a pproa c h nor t he syntac ti c a pproa c h has be e n e n -t i r e ly s a t i s fa c to ry . For e xa mple , E l a ine Ma imon a nd Ba rba ra N od ine ' ssentence-combining exp erim ent suggests that , as is t rue w hen ot he r skil ls andprocesses are learned, certa in kinds of errors accompany certa in s tages inlearning to wri te .4 Because th e sources of er ro r in wri t ten discourse are oftencomp lex and difficul t to t race , researchers can conclude l i t tle mo re than w hatis obvious: low-rated papers usual ly contain far more errors than high-ratedpapers . With regard to syntax, Ann Gebhard found that with few except ionsthe syntactic features of high- and low-rated essays written by college stu-dents are not clearly differentiated. Indeed, research in writing quality basedon conventions of wri t ten English and on theories of syntax, part icularlytransformational grammar, has not provided specific direct ions for the teach-ing o f wri t ing.

Such resul ts com e as no surprise in l ight of m uch c urren t research in writ -ten discou rse. Th is research-published in such fields as linguistics, cyb ern e-tics, anth rop olo gy , psych ology , and artificial intelligence-addresses qu es -t i o n s , c o n c e r n e d w i t h e x t e n d e d d i s c o u r s e r a t h e r t h a n w i t h i n d i v i d u a lsentences , ques t ions about how humans produce and unders tand discourseunit s of ten re fe rred to as te x ts .W n e such e ffor t tha t has a t t rac ted the a t ten-tion of resear chers in writing is M. A. K. Halliday and Ruqaiya Hasan 's Cohe-sion in E n g l i ~ h . ~l though Hall iday and Hasan d o not propose a theory oftext s t ruc ture o r examine how humans produce texts , they d o a t tempt todefine the concept o f text. T o the m a text is a semantic uni t , the parts o f

Step hen Witte and Lester Faigley teach at the University o f Texas at Austin. Bo th have pub-lished essays on research in the teaching of writing. They serve as co-directors of a researchproject on the evaluation of programs and courses in writing, funded by the Fund for the Im -provement of Pos t-Secondary Educat ion .


3/18

19 0 College Composition and Communicationwhich are l inked toge ther by explic it cohesive ties . C ohes ion, therefo re , de-fines a text as text. A cohesive tie "is a s em antic relation betw een an e lem en t ina text and some other e lement that is crucia l to the interpreta t ion of i t" (p.8). The two semantical ly connected e lements can l ie within the text or onee lem ent can l ie outs ide the text. H a l l i d a ~ nd Hasan call wi th in- text cohes iveties endophoric and re fere nce s to item s ou tside the text exophoric. An exam pleof an exophoric reference is the edi toria l "we" in a newspaper. Such refer-ences are exophoric because no antecedent is recoverable within the text .Exophoric references often help l ink a text to i ts s i tuat ional context ; but , asfar as Halliday and Hasan are con cerned , exopho ric references d o not con-tribute to the cohesion of a text . For Hall iday and Hasan, cohesion dependsupon lexical and grammatical re la t ionships that a l low sentence sequences tobe understood as connected discourse ra ther than as autonomous sentences.Even thoug h within-sentence cohesive t ies d o occur, the cohesive t ies across"sentence boundaries" are those which a l low sequences of sentences to beunderstood as a text .Halliday and Hasan's concept o f textuali ty, defined w ith referen ce t o re la-t ionships that obtain across "sentence boundaries ," suggests a nu m ber o f pos-sibil it ies for extend ing composi t ion research be yon d i ts frequen t m oorings insentence- leve l opera t ions and fea tures . The major purpose of the presentstudy is to apply tw o taxonom ies of cohesive t ies dev eloped by Hall iday andHasan to an analysis of essays of col lege freshmen ra ted high and low inquality. Because Cohesion in English is a pioneering effort to describe rela-t ionships between and among sentences in text , we ant ic ipate that cohesionwill be s tudied in futu re research add ressing the l inguistic features o f wri t tentexts . W e are part icularly interested in ident ifying wh at purp oses Hallidayand Hasan's taxon omies can serve in composi t ion research and what purp osesthey cannot serve.Halliday and Hasan's System for Analyzing and Classifying CohesiveTies

Cohesion in English specifies five major classes of cohesive ties, nineteensubclasses, and numerous sub-subclasses. In the analysis of cohesion whichfollows, we will b e co nce rne d with on ly th e five ma jo r classes-reference, sub-stitution, ellipsis, conjunction, and lexical reiteration a n d collocation-and th ei rresp ectiv e subclasses. T w o of the maj or classes-substitution and ellipsis-arem or e freq uen t in conversat ion than in wri t ten discourse . Subst itution replacesone e lement with another which is not a personal pronoun, and el l ipsis in-volves a dele t ion of a word, phrase , o r c lause. T h e effect of both subst i tut ionand el l ipsis is to extend the textual or semantic domain of one sentence to asubs eque nt sentence. T he wo rd one in sentence (2 ) i l lustrates cohesion basedo n subs t i tu t ion and the w ord do in sentence (4) i l lustrates cohesion based onellipsis.


4/18

Coherence and CohesionSubstitution

(1) Did you ever f ind a lawnmower?(2) Yes , I borrowed one f rom my neighborEllipsis

(3 ) D o you w an t t o go w it h me t o t he s t o r e?(4) Yes , I do.The remaining three categories include the bulk of expl ici t cohesive t ies inwri t ten Engl ish . The categories of refirenre and conjunction contain t ies thatare b oth gramm atical and lexical. Lexic'zl reiteration and collocation is restrictedto t ies which are presumably only lexical .Reference cohesion occurs when on e i t em in a t ex t po in t s to ano ther e l em entfo r i t s in te rp re ta t ion . Reference t i es a re o f th ree types : pronominals, de-monstratives and definite articles, and comparatives. Each o f the sen tence pa i rsbelow i l lust rates a d i fferen t type o f referen ce cohesion.

Reference Cohesion (Pronominal)(5) At home, my father is himself.(6) He relaxes and acts in his normal manner .

Reference Cohesion (Demonstratives)( 7 ) W e quest ion why they tel l us to do things.(8 ) T h i s is part o f growing up.

Refwence Cohesion (D efin ite Article)(9) Humans have many needs, both physical and intangible.(10) It is easy to see the physical needs such as food and shelter.

Reference Cohesion (Comparatives)(11) T h e older generat ion is of ten quick to cond emn col lege students forbeing carefree and i rresponsible.( 1 2 ) But those who remember the i r own youth do so less quickly.

Th e in te rp re ta t ion o f the under lined e lements in sen tences (6 ) , (a), ( l o ) , a nd(12) dep en ds in each case upon presupposed informat ion contained in thesentences imm ediately above i t.

A fourth major class of cohesive t ies f requent in wri t ing is conjunction.Conjunct ive elem ents are not in themselves cohesive, but they d o "expresscertain meanings which presuppose the presence of o ther components in thediscourse" (p. 226). Hall iday and Hasan dist inguish five types of conjunctivecohesion--additiv e, adversative, ca w a l, temporal, and continuative. Examples ofthese subclasses of conjunctive cohesion appear below and i l lustrate howconjunc t ive cohes ion ex tends the mean ing o f one sen tence to a subsequen to n e .

Conjunctiz~eCohesion (A dditi z~e )(13) N o one wants to be re jec ted .(14) A nd to prevent reject ion we change our behavior of ten


5/18

19 2 College Composition and CommunicationConjunctiz~eCohesion (Adversative)(15 ) Small chi ldren usual ly change their beh avior be cause th ey w antsomething they don ' t have .(16) Carol , however, changed her behavior because she wanted to becomepar t of a new group.Conjunctiz~eCohesion (Causal)( 1 7 ) Today's society sets the stand ards.(18 ) T he p eople mo re o r l ess fo llow i t [s ic ).( 1 9 ) Consequently, there exists the right behavior for the specific situationat hand.Conjunctive C ohesion (Temporal)(20 ) A fr iend o f mine we nt to an out-of-state college.(21) Before she lef t , she expressed her feel ings abou t playing roles t o win

new friends.Conjunctive Cohesion (Continuative)(22) Different social situations call for different behaviors.(2 3) This is something we all learn as chi ldren and we, of course, also learnwhich behaviors are right for which situations.

Coordinating conju nction s (such as and, but , and so), conjunctive adverbs(su ch as however, consequently, and moreover), and cer tain te m po ra l adv erb s an dsub ord inatin g conju ncti on s (su ch as before, after, and now) suppl y cohes iveties across sentence boundaries.

T h e last ma jor class of cohesive ties includes th ose based o n lexical rela-tionships. Lexical cohesion differs from reference cohesion and conjunctivecohesion because every lexical item is potentially cohesive and because noth-ing in the o ccurrenc e of a given lexical item necessarily ma kes it cohesive. Ifwe were to encounter the word this in a text , we would ei ther supply areferent f rom our working memo ry of the text o r re read the text to f ind arefer ent. Similarly, wh en w e enc ou nter a conjunctive adv erb such as however,we at tempt to establish an adversat ive relat ionship between two text e le-ments. In contrast , lexical cohesion depends on some "patterned occurrenceof lexical items" (p. 288). Consider the fol lowing sentences adapted from amounta ineer ing guidebook :(24) T he ascent up th e Emm ons Glac ier on Mt . Rainier i s long but re la-tively easy.(25) The only usual problem in the cl imb is f inding a route through thenum erous crevasses above Steamboat Prow.(26) In late season a bergschrund may develop at the 13,000-foot level ,which is customarily bypassed to the right.Three cohesive chains bind together this short text. The first chain (ascent,climb, finding a route, bypassed t o the righ t) carries th e topic-the way up th emountain. The second and third chains give the setting (Glacier, crevasses,bergschrund) (Mt. Rainie r, Steamboat Prow, 13,000-foot level). T h es e ch ains giveclues to the interpretation of unfamiliar items. For most readers, Steamboat


6/18

19 3oherence and CohesionProw is unknow n, bu t on e can infer that i t is a feature on M t. Rainier. Simi-larly, bwgschrund is a technical term referring to a crevasse at the head of aglacier where the moving ice breaks apart from the stationary ice clinging tothe mountain. In this text, a reader can infer that bergschrunds are associatedwith glaciers and that they present some type of obstacle to climbers, evenwitho ut th e final clause in ( 26).Lexical cohesion is the p redom inant means of co nnecting sentences in dis-course. Halliday and Hasan identify two major subclasses of lexical cohesion:reiteration and collocation. Reiteration is in turn divided into four subclasses,ranging from repetit ion of the same item to repetit ion through the use of asynonym or near-synonym, a superordinate item, or a general item.Lexical reiteration is usually easy to identify. An example of synonomyoccurs in (2 5) and (26 ) with the pairing of ascent and climb. Th e th ree o the rsubclasses are illustrated in the following student example:

Lexical Reiteration (Sam e Item ), (Superordina te), an d (Gene ral Item)(27) So me professional tennis players , fo r exam ple, grandstand , usingobscene gestures and language to call at tent ion to themselves.( 2 8 ) O t h e r professional athletes do similar things, such as sp iking a footbal lin the end zone, to at t ract attention.

In (28), professional athletes is, in this case, a superordinate term for profes-sional tenn is players. Professional athletes in other sports are encom passed bythe term. Th ings , in contrast, is a general term. Here things is used to referanaphorically to two behaviors, "using obscene gestures and language." Whilesuperordinates are names of specific classes of objects, general terms areeven m ore inclusive, not restricted to a specific set of objects. Th e oth er typeof lexical reiteration, illustrated by se nten ces (2 7) and (2 8), is same-item re -petition: attention is simply repeated.All the lexical cohesive relationships which cannot be properly subsumedunder lexical reiteration are included in a "miscellaneous" class called colloca-tion. Collocation refers to lexical cohesion "that is achieved through the as-sociation of lexical items that regularly co-occur" (p. 284). Lexical cohesionthrough collocation is the most difficult type of cohesion to analyze becauseitems said to collocate involve neither repetit ion, synonomy, superordina-tion, no r me ntion of general i tems. Wh at is impo rtant is that the items said tocollocate "share the same lexical environment" (p. 286). The following stu-dent example illustrates this principle:

Lexical Cohesion (Collocation)(29 ) On a camping t r ip with their parents, teenagers will ingly d o th ehousehold chores that they resist at home.(30) T h e y g a t h er woodfor af ire , he l p pu t up t he tent, and carry water froma creek or lake.

Although the underlined items in (30 ) are presented as the "camping trip"equivalents of household chores, the cohesion between sentences (29 ) and (3 0)


7/18

194 College Composition and Communicatzonresults more directly from the associations of the underlined items with camp-ing t r i p The underlined items in sentence ( 3 0 )collocate with camping trip insentence ( 2 9 ) . The mountaineering guidebook passage, however, is muchmore difficult to analyze. For one of the authors of the present article, an-tecedent knowledge of mountaineering allows Steamboat Prow to collocatewith M t. Rainier and bergschrund to collocate with glacier. For the other au-thor, neither pair is lexically related by collocation apart from the text wherethey are connected by inference. We will return to this problem later in thisessay.In addition to the taxonomy that allows cohesive ties to be classified ac-cording to function, Halliday and Hasan introduce a second taxonomy. Thissecond taxonomy allows cohesive ties to be classified according to theamount of text spanned by the presupposed and presapposing elements of agiven tie. Halliday and Hasan posit four such "text-span" classes. Member-ship in a class is determined by the number of T-units a given cohesive tiespans.' Taken together, the two taxonomies Halliday and Hasan presentallow any given cohesive tie to be classified in two different ways, one ac-cording to function and one according to distance. The four "text-span"classes contained in Halliday and Hasan's second taxonomy are illustrated inthe following paragraph from a student paper:

Test-Span Classes ilmmediate, Mediated, Remote, Mediated-Remote)( 3 1 ) Respect is one reason people change their behavior.( 3 2 ) For example, one does not speak with his boss as he would talk to afriend or co-worker.( 3 3 ) One might use four-letter words in talking to a co-worker, butprobably not in talking to his boss.( 3 4 ) In talking to teachers or doctors, people also use bigger words thannormal.( 3 5 ) Although the situation is different than when one speaks with a bossor a doctor, one often talks with a minister or priest different [sic]than he talks wi th friends or family.( 3 6 ) With t h e f a m i l y , most people use a different language when theytalk to parents or grandparents t han when they talk to youngerbrothers and sisters.( 3 7 ) People's ability to use language in different ways allows them toshow the respect they should toward different people, whether theyare professionals, amily members, clergy, friends and co-workers,or bosses.

Immediate cohesive ties semantically linked adjacent T-units. The repetitionof doctor in sentences ( 3 4 ) and ( 3 5 ) creates an immediate tie, forcing thereader to assimilate the content of ( 3 4 ) into the content of ( 3 5 ) . In contrast,the repetition of famil? in sentences (351 , (3 6) ,and ( 3 7 )forms a mediated tie.The semantic bridge established by the occurrence of amily in ( 3 5 )and ( 3 7 )is channelled through or mediated by the repetition of family in ( 3 6 ) . Thecohesive tie involving the repetition of famil? is not simply a series of im-mediate ties, because once a lexical item appears in a text all subsequent uses


8/18

195oherence and Cohesionof that i tem presuppo se th e f irst appearance, Immediate and mediated ties joinitems in adjacent T-units. Such ties enable writers to introduce a concept inone T-unit and to extend, m odify, o r clarify that concept in sub sequent andsuccessive T-units.

Remote t ies , on the other hand, result when the two elements of a t ie areseparated by one or mo re intervening T-units . Th e t ie between respect in ( 31 )and ( 3 7 ) is remote; here the repetit ion of the word signals to the reader thatthe semantic unit represented by the paragraph is now complete. Finally, tieswhich are both mediated and remote are called mediated-remote. An exampleof this type of cohesive tie appears in the repetition of bosses in sentences( 3 2 ) , ( 3 3 ) , ( 3 5 ) , and ( 3 7 ) . Here the presuppos ing bosses in ( 3 7 ) is separatedfrom the presupposed boss in ( 3 2 )by intervening T-units ( 3 4 )and ( 3 6 )whichcontain no element relevant to the particular cohesive t ie. Thus the t ie isremote. However , the presuppos ing bosses is also mediated through repetit ionsof boss in ( 3 3 ) and ( 3 5 ) . Hence the term mediated-remote. Skilled writers usemediated -remote t ies to interweave key "them es" within the tex t.Analysis of Stud ent Essays

To explore the usefulness of Halliday and Hasan's theory of cohesion inwriting research, we used their two taxonomies in an analysis of ten studentessays. The se essays we re written by beginning University of Texas freshm eno n the "changes in behavior" topic used in the Miami University sentence-com bining e ~ p e r i m e n t . ~rom 90 essays which had been rated holistically bytwo readers on a four-point scale, we selected five essays given the lowestscore by both raters and five essays given the highest score. We analyzedthese ten essays according to categories of error and according to syntacticfeatures, as well as according to the number and types of cohesive ties. Ouranalyses of error and content variables yielded results similar to those otherresea rche rs have reported-that high-rated essays are long er and con tainlarger T-units and clauses, more nonrestrictive modifiers, and fewer error^.^We anticipated that an analysis of cohesive ties in the high- and low-ratedessays would revea l similar gross differences. Th e results of o u r analysis con-firmed this expectation. At the most general level of analysis, the high ratedessays are much more dense in cohesion than the low-rated essays. In thelow-rated essays, a cohesive tie of so me typ e occurs onc e every 4.9 words; inthe high-rated essays, a tie occurs once every 3.2 words, a difference in me anfrequency of 1.7 words. Likewise, a large difference in the mean number ofcohesive ties per T-unit appears, with 2.4 t ies per T-unit in the low-ratedessays and 5.2 ties per T-unit in the high-rated essays. The figures for thisand the preceding index, however, are not precisely comparable because theT-units in the high-rated essays are, on the average, 1.64 words longer thanthose in the low-rated essays. By dividing the number of cohesive ties in anessay set by the num ber of words in that set, we arrived at anot he r gen eral


9/18

19 6 College Co~nposition nd Communicationindex of cohesive density. In the high-rated essays, 31.7% of all words con-tr ibute to explicit cohesive t ies while only 20.4% of the words in the low-rated essays contribute to such ties.T he ways in which writers of the high- and low-rated essays form cohesiveties also distinguish the two gro ups o f five essays from e ach oth er. W riters ofthe high-rated essays use a substantially higher relative percentage of im -mediate (High: 41 .6%/Low: 32 .8% ) and mediated (H ig h: 7.6Y:/Low: 0 .8 % )cohesive t ies than d o the writers of the low-rated essays. O n the o ther hand,writers of the low-rated essays use more mediated-remote (High: 25.9%/Low:36 .7% ) an'd remote t ies (H igh: 26.9?6/Low: 2 9.7 % ). Th ese percentages allowus to focus on some crucial differences between the two essay sets . Thelarger relative percentage of immediate cohesive ties in the high-rated essayssugges t s , among o the r th ings , tha t the be t t e r wr i t e r s t end to es tab l i shstronger cohesive bonds between individual T-units than do the writers ofthe low-rated essays. Analyses of reference and conjllnctive cohesion supp ort thisobservation. Writers of high-rated essays employ reference cohesion abouttwice as often, 84 .1 t imes to 47 .8 t imes per 10 0 T-units , as the writers oflow-rated papers . The largest difference in the occurrence of referentialcohesion is reflected in the higher frequency of third-person pronouns in thehigh-rated essays (H igh: 25.1 per 10 0 T-units/Low: 5 .1 per 1 0 0 T-units) . Thislower frequency of third-person pronouns in the low-rated essays may be adirect result of the less skilled writers ' attempts to avoid errors such as am-biguous pronoun reference. Because third-person pronouns usually referback to the T-unit immediately preceding, we can infer that the writers ofhigh-rated essays more often elaborate, in subsequent and adjacent T-units,topics introduced in a given T-unit.Also contr ibuting importantly to th e greater use of immediate cohesive tiesis the frequency with which the more skillful writers use conjzlnction to linkindividual T-units . Conjunctive t ies most often result in immediate cohesiveties between T-units. It is not surprising, then, to find that the writers ofhigh-rated essays employ over three times as many conjunctive ties (High:65.4 per 10 0 T-units /Low: 20.4 per 10 0 T-units) as the writers of low-ratedessays. Neither is i t surpris ing to discover that the more skil lful writersemploy all five types of conjunction while the less skillful writers use onlythree. As is the case with pronom inal references that cross T-unit boundaries ,conjunct ives are mos t of ten used to extend concepts in t roduced in oneT-unit to o ther T-units which follow immediately in the text. Th us the m oreskil lful writers appear to extend the concept introduced in a given T-unitconsiderably m or e often than do the less skillful writers . O n e ma jor effect ofsuch semantic exte nsio ns is, of course, essay length ; and this finding helps toexplain why the high-rated essays are, on the average, 375 words longer thanthe low-rated essays.The relative frequency of lexical cohesion gives another indication that thewriters of high-rated essays are better able to expand and connect their ideas


10/18

197oherence and Cohesionthan the writers of the low-rated essays. By far the largest number of cohe-sive ties, about two-thirds of the total ties for both the high and low samples,fall into the general category of lexical cohesion. Writers of the high-ratedessays create some type of lexical tie 340 times per 100 T-units or every 4.8words. Writers of the low-rated essays, however, manage a lexical tie just 161times per 100 T-units or every 7 .4 words. The majority of lexical ties ( 6 5 % )in the low essays are repetitions of the same item. This distribution is re-flected to a smaller degree in the high essays, where 52% of the total lexicalties fall into the same item subcategory. Writers of high-rated essays, however,form many more lexical collocations. Lexical collocations appear 9 4 times per100 T-units in the high-rated essays in contrast to 28.8 times per 100 T-unitsin the low-rated essays.

Cohesion and Invention

These cohesion profiles suggest to us an important difference between theinvention skills of the two groups of writers. Th e better writers seem to havea better command of invention skills that allow them to elaborate and extendthe concepts they introduce. The poorer writers, in contrast, appear deficientin these skills. Their essays display a much higher degree of lexical and con-ceptual redundancy. The high percentage of lexical redundancy and the lowfrequency of lexical collocation in the low-rated essays are indications of thisdifference. The text-span categories also point to this difference. In the low-rated essays two-thirds of the cohesive ties are interrupted ties-mediated-remote or remote ties-which reach back across one or more T-units, indicatingthat the writers of the low-rated essays generally fail to elaborate and extendconcepts through successive T-units.The larger proportion of interrupted ties in the low-rated papers stronglysuggests that substantially less new information or semantic content is intro-duced during the course of a low-rated essay than during the course of ahigh-rated essay. If more new information had been introduced in the low-rated essays, the writers would have had to rely more heavily than they didon immediate and mediated cohesive ties in order to integrate, to weave, thenew information into the text. The writers of the low-rated papers tend moretoward reiteration of previously introduced information than do the writersof the high-rated papers. Indeed, in reading the low-rated essays one can nothelp noting a good deal of what might be called conceptual and lexical re-dundancy. The following example illustrates this characteristic:

Some people have to change their behavior around different acquaint-ances. One reason is that they want to make a good impression on others.You have to act different in front of a person who is giving you a jobinterview because you want to make a good impression. You, most of thetime, act differently to fit in a crowd. You will change your behavior to


11/18

198 College Composition and Comml*nirationget people to l ike you. You change your behavior to agree with peoples[sic} in the crowd.

This paragraph from a low-rated paper has a fairly strong beginning: i t statesa topic in the first sentence, modifies that topic in the second sentence, i l lus-t rates the topic in the thi rd sentenc e, and gives ano the r exam ple in the f ou rthsentence. The next two sentences , however, s imply rei terate what is said inthe fourth sentence. The principal lexical i tems in the las t two sentences-change, behavior, people, and crowd-are repet i t ions of i tems int rodu ced earl ierin the paqagraph and offer l i t t le new informat ion. Al though for purposes ofat taining cohesion in a text some redundancy is a vi r tue, the redundancy inthe low-rated essays seems to be a flaw because these writers failed to supplyaddi t ional informat ion at the point where i t would be expected to appear .H ad this addi tional informat ion b een suppl ied, the wri ters would have had touse immediate and mediated t ies in o rder to connec t i t to the res t o f the t ex t .Com pare the p rev ious exam ple paragraph f rom a low-ra ted paper wi th th efol lowing paragraph from a high-rated paper.

I t is a job that really changes our behavior . A m on g othe r changes, wechange the way we dress . In many jobs college graduates want to lookresponsib le and mature , p ro ject ing an image of com petence . Th e co l legestud ent who w ore faded blue jeans is now in three-piece suits . H e feelsthe need to be approved of and accepted by his boss and associates.While he ta lked of socialism in college, he now reaps the prof its ofcapitalism. While in college he demanded honesty in the words and ac-tions of others , on the job he is will ing to "kiss ass" to make fr iends orget a promotion. Indeed, working can change behavior .

Notice that in the paragraph from the high-rated paper, behaz'ior is repeatedonly one t ime. Yet the reader never quest ions that the paragraph is aboutchanges in behavior . T h e wri ter repeatedly supplies examples of types ofbeha vior, which are l inked t o the topic by a series of lexical collocations (e.g.,behacior, dress, look responsible, blae jeans, three-piece sa i t s ). Clearly, the para-graph from the high-rated paper extends the semant ic domain of the conceptbehacior to include a number of differentiated lexical i tems. Low-rated papersrarely show such extended series of collocations.

Analyses of cohesion thus measure some aspects of invent ion ski l ls . Thelow-rated essays stall frequently, repeating ideas instead of elaborating them.O u r analyses also suggest that the wri ters of the low-rated pape rs d o no t havewor king vocabularies capable of extend ing, in ways prereq uisite for goo dwriting, the concepts and ideas they introduce in their essays. Indeed, skil l ininvention, in discovering what to say about a particular topic, may depend inways ye t unexplored on the p r io r development o f adequate work ing vocab-ularies. If stud ents d o no t have in the ir w ork ing vocabularies the lexical i tem srequi red to ex tend , exp lore , o r e labora te the concep t s they in t roduce , p rac-tice in invention can have only a l imited effect on overall writ ing quali ty.O u r analyses furth er point to the un derde velop me nt of certain cogni tiveski l ls among the wri ters of the low-rated papers . The low-rated papers not


12/18

199oherence and Cohesiononly exhibit a great deal of redundancy, but (as noted ear l ier) a lso includerelatively fewer conjancthe and reference ties and immediate and mediated ties.Besides lacking adequate vocabularies, writers of the low-rated essays seemto lack in part the ability to perceive and articulate abstract concepts withreference to particular instances, to perceive relationships among ideas, andto reach beyond the worlds of their immed iate experience.All this is to suggest that analyses of cohesion may be potentially useful indist inguishing between stages of wri t ing development. Clear ly, cohesionanalyses measure more sophist icated aspects of language development thand o erro r analyses and syntactic analyses. Co hesion analyses also give us som econcrete ways of addressing some of the differences between good and poorwrit ing, dif ferences which heretofore could not be explained ei ther to our-selves or to our students in any but the most abstract ways. We thus antici-pate that Halliday and Hasan's taxonomies can be usefully applied in devel-opm ental s tudies as well as in studies such as the p resent on e.

Cohesion, Coherence, and Writing QualityHo we ver prom ising cohesion analysis appears as a research tool and how -ever encouraging the resul t s of the present s tudy seem, we fee l tha t anumber of important quest ions cannot be answered by analyzing cohesion.T h e f irst of these quest ions concerns wri t ing quality. Th e quality or "success"

of a text , we would argue, depends a great deal on factors outside the textitself , factors which lie beyond the scope of cohesion analyses. Recall thatHall iday and Hasan exclude exophoric, or outside- text , references f rom theirtaxonomy of explicit cohesive ties. We think that writing quality is in partdefined as the "fit" of a particular text to its context, which includes suchfactors as the wri ter 's purpose, the discourse medium, and the audience 'skno wled ge of an interest in th e subject-the factors which are the cor-nerstones of discourse theory and, mutatis matandis, should be the cor-ners tones of research in wr i tten c o m p ~ s i t i o n . ' ~ e a re not a lone in th is v iew.Several s tude nts of wri t ten discourse-among the m Joseph G rimes," T eu nvan Dijk, l 2 Nils Enkvist ,13 and Robert de Beaugrande14-distinguish cohe-sion and coherence. They limit cohesion to explicit mechanisms in the text,both the types of cohesive t ies that Hall iday and Hasan descr ibe and otherelements that bind texts such as parallelism, consistency of verb tense, andwha t literary scholars have called "point of view."15 Co here nce cond itions, onthe o th er hand, allow a text to be u nd ersto od in a real-world setting . Hallidayand Hasan's theory does not accommodate real-world set t ings for wri t tendiscourse or , consequently, the condit ions through which texts become co-herent . We agree with Charles Fil lmore 's contention that

the scenes . . . [audiences} construct for texts are partly justified by thelexical and grammatical materials in the text and partly by the interpre-ter's own contributions, the latter being based on what he knows about


13/18

College Composition and Communicationthe current context, what he knows about the world in general, and whathe assumes the speaker's intentions might be.16

Hence lexical collocations within a text are understood through cues whichthe writer provides and through the reader's knowledge of general discoursecharacteristics and of the world to which the discourse refers.Thus lexical collocation is in all likelihood the subcategory of cohesion thatbest indicates overall writing ability, as well as disclosing distinctions amongwritten texts that represent different discourse modes and purposes. Anexamination of lexical cohesive ties shows how writers build ideas, how theyare able to take advantage of associations to weave together a text. But afundamental problem lies in the analysis of a writer's text. Whose colloca-tions do we analyze-the reader's or the writer 's? One simple proof that thetwo do not always coincide can be found in the unintentional sexual refer-ences that students occasionally produce-the kind that get passed aroundthe faculty coffee room.Consider again the mountaineering guidebook passage in sentences ( 2 4 ) ,(25), and (26). We have already established that for mountaineers andglaciologists, bergschrund probably collocates with glacier, but for many otherpersons the two items do not collocate. Yet a naive reader presented this textprobably would not stop to consult a dictionary for the lexical item,bergschrzlnd, but would infer from its context that it is some type of obstacleto climbers and continue reading. Herbert Clark theorizes that we com-prehend unknown items like bergschrzlnd by drawing inferences.17 We makeinferences on the basis of what we can gather from the explicit content andthe circumstances surrounding a text, through a tacit contract between thewriter and reader that the writer will provide only information relevant to thecurrent topic. In the case of the mountaineering passage, the circumstancesof the text greatly affect our understanding of it. The type of text-aguidebook-follows a predictable organization, what has been called a scriptin research o n artificial intelligence.18 Th e guidebook contains a series oftopics with a clear, yet implicit, goal: to inform the reader how to get to thetop of a mountain. We expect the author to give us only information relevantto the particular route. Accordingly, readers understand bergschrzlnd as anobstacle through a combination of cues--overt signals in the text such as theparallelism of the bergschrund sentence with the sentence about crevassesabove it and, for those readers familiar with the type of text, implicit signalssuch as the following of the guidebook "script." Although Halliday andHasan do not include parallelism in their taxonomy, parallelism often createsa cohesive tie.Cohesion and coherence interact to a great degree, but a cohesive text maybe only minimally coherent. Thus cohesion-based distinctions between textsrated high and low in quality can be misleading. Besides explicit links withina text, a text must conform to a reader's expectations for particular types oftexts and the reader's knowledge of the world. A simple example will illus-trate this point:


14/18

Coherence and Cohesion(38 ) T h e quar terback threw the ball toward the t ight end(39) Balls are used in many sports.(40) Most balls are spheres, but a football is an ellipsoid.(41) The t ight end leaped to catch the bal l .

Sentences (39) and ( 4 0 ) ,while cohesive, viola te a coherence condit ion thatthe wri te r provide only informat ion re levant to the topic. Th e m ajor problemwith this sho rt text is that a reader ca nn ot cons tru ct wh at Fillmore calls areal-world scene fo r i t ; that is, the text ne i ther se em s to have a c lear purpo senor appears to m ee t the needs of any given audience. Because i t has no c learpurpose, i t lacks coherence, in spite of the cohesive ties which bind it to-gether. In addi t ion to a cohesive uni ty, wri t ten texts must have a pragmaticuni ty, a uni ty o f a text and the world of the reader . A descript ion o f the fi t ofa text to its context, as well as descriptions of what composition teachers callwri ting quali ty, m ust specify a varie ty of co heren ce condit ions, many of th emoutsid e the text i tself.

Implications for the Teaching of CompositionOne impl ica t ion of the present s tudy i s tha t i f cohes ion i s be t te r un-derstood, i t can be better taught. At present, in most college writing classes,cohesion is taught, explicitly or implicitly, either through exercises, class-r o o m i n s t r u c t i o n , o r c o m m e n t s o n s t u d e n t p a p e r s . M a n y e x e r c i s e s n o texplic i t ly designed to teach cohesion do in fact demand that s tudents formcohes ive t ie s . Open sentence-combining exerc ises , for example , offe r asmuch practice in forming cohesive ties as they do in manipulating syntactics t ruc tu re s , a f a c t w h ic h ma y e xp la in t he suc c e s s o f c e r t a in s e n t e nc e -combining experiments as well as the failure of research to link syntacticmeasures such as T-u nit and clause length t o wri ting qual i ty. l9 An op ensentence-combining exercise abo ut Ch arl ie Chaplin might contain a series of

sentences beginning with the name Charlie Chap lin. Such an exercise would,a t the very least , demand that s tudents change most of the occurrences ofCharlie Chap lin to be in order to produce an acceptable text . Students work-ing e i ther from contextual cues or from their knowledge of Chaplin mightalso use phrases like the comic genizls o r the little tramp to subst i tute for theprope r na me Chaplin.If cohesion is often implicitly incorporated in writing curricula, coherenceis often ignored. A great port ion of the advice in composi t ion textbooksstops a t sentence boundaries . Numerous exercises teach c lause and sentencestructure in isolation, ignoring the textual, and the situational, considerationsfor using that structure. The passive is a classic example:(42 ) Th e pol ice apprehe nded the suspect as he left the bank.(4 3) H e is being held in the county jail. (43a ) T h e police a re holding the suspect in the co unty jail.

A student following her teacher's advice to avoid the passive construction


15/18

202 College Composition and Commtlnicationmight revise senten ce (4 3) to (4 3a). If she did so , she would viola te the usualsequence of information in English, where the topic or "old" information ispresented fi rs t .20 n active sentences, such as (43a), wh ere th e o bje ct expres-ses the topic, a revision to the passive is often preferable. Avoiding the pas-sive with (4 3a) would a lso requ ire the unnecessary and un econom ical repet i -t ion of police and suspect. Consequently, maxims such as "Avoid passives"ignore the coherence condit ions that govern the information structure of atext .Other discourse considerations are similarly ignored in traditional adviceon how to achieve coherence . As E. K. Lybert and D. W. Cummings haveobserved, the handbook injunct ion "Repeat key words and phrases" oftenreduces c o h e r e n ~ e . ~ ~u r analysis o f cohesive ties in high- an d low-rated es-says substant ia tes Lybert and Curnmings ' point . While the low-rated paperswe examined contain fewer cohesive t ies than th e high-rated papers in equiv-alent spans of text, the low-rated papers rely more heavily on lexical repeti-t ion. Also contrary to a popular not ion, frequent repet i t ion of lexical i temsdoes not necessarily increase readability. Roger Shuy and Donald Larkin'srecent study shows lexical redundancy to be a principal reason why insurancepolicy language is difficult to re ad.2 2O u r analysis of cohesion suggests that cohesion is an im portan t prop erty ofwri t ing qual i ty. To some extent the types and frequencies of cohesive t iesseem to reflect the invent ion ski l ls of s tudent wri ters and to influence thestyl is t ic and organizat ional propert ies of the texts they wri te . However, ouranalysis also suggests that while cohesive relationships may ultimately affectwriting quality in some ways, there is no evidence to suggest that a largenumber (or a small number) of cohesive t ies of a part icular type wil l posi-tively affect writing quality. All discou rse is con text bound-to the dem and sof the subject matter, occasion, medium, and audience of the text . Cohesiondefines those mechanisms that hold a text together, while coherence definesthose underlying semantic re la t ions that a l low a text to be understood andused. Co nseq uent ly , coh erenc e condi tions-condi tions gov erne d by thewri ter ' s purpose, the audience 's knowledge and expectat ions, and the infor-matio n to be conveyed-militate against prescrip tive appro aches to the teach-ing of wri t ing. Indeed, our explorat ion of what cohesion analyses can andcannot measure in student writing points to the necessity of placing writingexercises in the context of complete written texts. Just as exclusive focus onsyntax and other formal surface features in writing instruction probably willnot better the overall quality of college students ' writing, neither will a nar-row emphasis on cohesion probably pro duce significant ly impro ved w ri t in gz 3Notes

1. Most notably, Mina Shaughnessy, Errors and Expectations, (New York: Oxford UniversityPress, 1977) .


16/18

2 0 3oherence an d C ohesion2. See Ann 0 . Geb hard, "Wri t ing Qual i ty and Syntax: A Transformational Analys is of Th reeProse Sam ples," Research in the Teaching of English, 12 (Oc t obe r , 1978) , 2 11-231 .3. Composit ion theorists , however, have not stopped at sentence boundaries. Several of the

early efforts to describe relat ionships across sentences are summarized in Richard L. Larson,"Structure and Form in Non-Fict ion Prose," in Teaching Comp osition: Te n Bibliogra phical Essays,ed. G ary Ta te (For t Worth, Texas : Texas Chri st ian Univers ity Press , 1976 ) , pp. 45-71. T hereader will no te certain similarit ies between Hall iday and Hasan's taxon omy o f cohesive t ies setout later in the present essay and the work of previous composi t ion theorists .4. "M easuring Syntactic Gro wth : Errors and Expectat ions in Sen tence-Co mbinin g Pract icewith College Freshmen," Research in the Teaching of English, 12 (Octob er , 19781, 233-244.5. N o co mpreh ensive overview of work in discourse in all of these fields exists at present .Extens ive bibl iographies , however , can be found in Current Trends in Textlinguistics, e d .Wolfgang U. Dress le r (Ber l in: de G ruyter , 19 78 ) and Rob er t de Beaugrande , Text, Dircourse,and Process (Norwood, N.J . : Ablex, 1980) .6 . (London: Longman, 1976) .7 . T he t e rm T-uni t , of course , comes f rom Kel logg Hunt ' s Grammatical Structures Written atThree Grade Lecels, NC TE Research R eport N o. 3 (Champaign, I ll .: Nat iona l C ouncil of Teach-ers of Engl ish, 1965 ) . H un t def ined a T - u n i t as an independent clause and al l subordinate ele-ments at tached to i t , whether clausal o r phrasal. Hall iday and H asan d o not use th e term T-uni t ,bu t they do define their four "text-span" classes according to th e num ber o f simple and complexsentences that th e presuppos ing elem ent of a cohesive t ie m ust reach across for the presupp osedelement (see pp . 34 0-3 55) . Th ere is good reason to de fine the four "text-span" classes in termsof T-units . T o examin e only cohesive t ies that span the bou ndaries of orthograp hic sentenceswould ignore the large number of conjunct ive relat ionships, such as addit ion and causal i ty, be-tween independent clauses.8 . Max Morenberg, Donald Daiker , and A ndrew K erek, "Sentence Combining at the Col legeLevel: An Experimental Study," Research in the Teaching of English, 12 (O c t obe r , 1978) , 245-256. Th is topic asks students to wri te ab out why w e act different ly in different si tuations, usingspecific illustrations from personal experience.9. Detai led analyses for data summarized in this sect ion are repo rted in Stephen P. Wit te andLester Faigley, A Cotnparison of A naly tic an d Syn thetic Approaches to the T eaching of College Wr it -ing, Unp ublished manuscript . Th e high-rated essays are, on th e average, mor e than twice as longa s t h e l ow- ra te d e s sa ys (64 7 words1270 word s ) . E r ro r s i n t h re e m a j or c a t e gor i e s we recounted-punctuat ion, spel l ing, and gramm ar. Grammatical erro rs include erro rs in verb tenses,subjec t -verb agreement , pro noun re fe rence , pronoun num ber agreement , and dangl ing or mis-placed m odifiers. Th e low-rated essays exhibi t an err or of so me type nearly th ree t imes as oftenas the high-rated essays--one every 29 w ords as oppo sed to o ne every 8 7 words . For example ,errors in end-stop punctuat ion, resul t ing in ei ther a comma spl ice or a fragment , occur nearlyeight times as often in the low-rated essays as in the high-rated essays. Misspelled word s are overfour t imes as frequent in th e low-rated essays, and gramm atical errors appe ar ove r twice as often.Syntactic comparison s were made according to the nu mb er of words per T -unit and clause,and according to the frequency and placement of nonrestrict ive or "free" modifiers. The low-

rated essays contain T-units and clauses considerably shorter than the high-rated essays (High:15.3 word s per T-u nit , 9.3 words pe r clauselLow: 13.7 word s per T-unit , 7.5 words per clause).No nres trictiv e mod ifiers in all positions-initial, medial, and final-appear in th e T-u nits of thehigh-rated essays nearly three t imes as frequently as in th e low-rated essays (High : 28.5 % of allT-units contain nonrestrict ive modifiers1Low: 10. 1% contain nonrestrict ive modifiers). Th ehigh-rated essays also have twice the percentage of total words in nonrestrictive modifiers thatthe low-rated essays have.10. Stephen P. Wit te , "Toward a Model for Research in Writ ten Composit ion," Research inthe Teaching of English, 14 (February, 1980 ) , 73-81.11. The Thread of Discourse ( T h e H a g u e : M o u t o n , 1 9 7 5 ) .12 . Tex t and Context (London: Longman, 197 7) .13. "Coherence , Pseudo-Co herence , and Non-Coherence ," in Reports on Text Linguistics:Semantics and Cohesion, ed. Jan-Ola O stman (Abo , Finland: Research Ins t i tute of the A boAkademi Fo undat ion, 19781, pp. 109-128.


17/18

204 College Composition and Commz~nication14. "The Pragmatics of Discourse Planning,"Journal ofPragmatics, 4 (February, 19801, 15-42.15. Susumo Kuno describes some linguistic features controlled by point of view, which hecalls "empathy." See "Subject, Theme and the Speaker's Empathy-A Reexamination of Re-lativization Phenomena," in Subject and Topic, ed. Charles N. Li (New York: Academic Press,19761, pp. 417-444.16. 'Topics in Lexical Semantics," in Current Issues in Linguistic Theory, ed. Roger W . Cole(Bloomington: Indiana University Press, 1977), p. 92.17. "Inferences in Comprehension," in Basic Processes in Reading: Perception and Cotnprehen-sion, ed. David LaBerge and S. Jay Sarnuels (Hillsdale, N.J.: Erlbaum, 1977), pp. 243-263.18. R. C. Schank and R. P. Abelson, Scripts, Plans, Goals, and Understanding: A n Inquiry intoHu ma n Knowledge Structures (Hillsdale. N.J.: Erlbaum, 1977).19. See Lester Faigley, "Names in Search of a Concept: Maturity, Fluency, Complexity, andGrowth in Written Syntax," College Composition and Comm unication, 31 (October, 1980), 291-300.20. The theme-rheme distinction is the product of the Prague school of linguistics. Otherresearchers use the terms "topic-comment" or "given-new" information to refer to essentially the

same concept. See Vilem Mathesius, A Funct ion al Analy sis of Present Da.7 Englis h on a Gen era lLinguistic Basis, ed. Josef Vachek ( Prape: Academia, 1975). Also relevant are Wallace L. Chafe,Meaning and the Structure of Language (Chicago: University of Chicago Press, 1970); M. A. K.Halliday, "Notes on Transitivity and Theme in English: 11," Journal of Linguistics, 3 (October,19677, 199-244; Herber t Clark and Susan Haviland, "Comprehension and the Given-New Con-tract," in Discourse Production and Comprehension, ed. Roy Freedle (Norwood, N.J.: Ablex, 19771,PP. 1-40; and Liisa Lautamatti, "Observations on the Development of Topic in Simplified Dis-course," in Text Linguistics, Cognitive Learning, and Language Teaching, ed. Viljo Kohonen andNils Erik Enkvist (Turku, Finland: University of Turku, 19781, pp. 71-104.21. "On Repetition and Coherence," College Cotnposition and Cotnmunication, 20 (February,1969), 35-38.22. "Linguistic Consideration in the Simplification/Clarification of Insurance Policy Lan-guage," Discourse Processes, 1 (October-December, 1978), 305-321.23. John Mellon and James Kinneavy both make the point that while the Miami Universitysentence-combining students improved significantly in writing quality, they may not have doneso because they learned to manipulate syntactic structures better, but because they were taughtto put together complete texts. See "Issues in the Theory and Practice of Sentence Combining:A Twenty-Year Perspective," in Sentence Combining and the Teaching of Writing, ed. DonaldDaiker, Andrew Kerek, and Max Morenberg (Akron, Oh.: University of Akron, 1979), p. 10;and "Sentence Combining in a Comprehensive Language Framework" in the same volume, p. 66.

WPA: Writing Program AdministrationBeginnin g with i ts fa ll , 1 97 9 issue, th e W P A IC7ewsletter became a refereed jour-nal, W P A : Writ ing Program Adtninistvation, I t is published th ree t im es a year byt h e C o u n c i l o f W r i t i n g P r o g r a m A d m i n i s t r a t o r s a nd b y t h e S c h o o l o fHu ma nities and Office of Publications, Bro okly n College-City University ofNew York . Contr ibu t ions should be addressed to Kenneth A. Bruf fee , Edi torW P A , English D epar tmen t , B rooklyn Col lege , Brooklyn , N Y 112 10 . Cor re -spondence concerning subscr iptions ( included in membership in the Council ofWri t ing Program Adminis t ra tors ) should be addressed to Joseph Com prone ,WPA Treasurer , English Department, University of Louisville , Louisville , KY40208.


18/18

You have printed the following article:

Coherence, Cohesion, and Writing Quality

Stephen P. Witte; Lester Faigley

College Composition and Communication, Vol. 32, No. 2, Language Studies and Composing.(May, 1981), pp. 189-204.

Stable URL:

http://links.jstor.org/sici?sici=0010-096X%28198105%2932%3A2%3C189%3ACCAWQ%3E2.0.CO%3B2-E

This article references the following linked citations. If you are trying to access articles from anoff-campus location, you may be required to first logon via your library web site to access JSTOR. Pleasevisit your library's website or contact a librarian to learn about options for remote access to JSTOR.

Notes

19 Names in Search of a Concept: Maturity, Fluency, Complexity, and Growth in WrittenSyntax

Lester Faigley

College Composition and Communication, Vol. 31, No. 3. (Oct., 1980), pp. 291-300.

Stable URL:

http://links.jstor.org/sici?sici=0010-096X%28198010%2931%3A3%3C291%3ANISOAC%3E2.0.CO%3B2-Z

http://www.jstor.org

LINKED CITATIONS- Page 1 of 1 -

NOTE: The reference numbering from the original has been maintained in this citation list.
http://links.jstor.org/sici?sici=0010-096X%28198105%2932%3A2%3C189%3ACCAWQ%3E2.0.CO%3B2-E&origin=JSTOR-pdfhttp://links.jstor.org/sici?sici=0010-096X%28198010%2931%3A3%3C291%3ANISOAC%3E2.0.CO%3B2-Z&origin=JSTOR-pdfhttp://links.jstor.org/sici?sici=0010-096X%28198010%2931%3A3%3C291%3ANISOAC%3E2.0.CO%3B2-Z&origin=JSTOR-pdfhttp://links.jstor.org/sici?sici=0010-096X%28198105%2932%3A2%3C189%3ACCAWQ%3E2.0.CO%3B2-E&origin=JSTOR-pdf

Documents

Coherence, Cohesion in Writing Quality