Cohesion Paradigms in Paragraphs Author(s): Robin Bell Markels Reviewed work(s): Source: College English, Vol. 45, No. 5 (Sep., 1983), pp. 450-464 Published by: National Council of Teachers of English Stable URL: http://www.jstor.org/stable/376843 . Accessed: 07/05/2012 20:54 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. National Council of Teachers of English is collaborating with JSTOR to digitize, preserve and extend access to College English. http://www.jstor.org Robin Bell Markels Cohesion in Paradigms Paragraphs Whether reading The Great Gatsby or gradingfreshman compositions, readers expect texts to make sense. Beyond the happenchanceof contiguity, they expect words, sentences, and paragraphsto establish mutual dependencies and grow into cohesive wholes. The assumptionof cohesion is one of the basic rules of the languagegame in which we all participate.From a linguisticallyspare but situationally rich fragmentof dialogue like I talked to the man about the car. I've never seen a doctor drink like that. to a linguistically rich but situationally spare Shakespeareansonnet, cohesion can appearas a pragmaticsufficiency or an artfultapestry. Authorscan exploit it in novels, simply fulfill it in essays, or struggle for it in freshman themes. Implicit in all these variations,and in our sense that some texts never manageto be cohesive at all, lies the assumption of rudimentary paradigms of cohesive wholes. Whenever we judge texts to lack, fulfill, or exploit cohesion, our judgments arise from a determiningparadigm. The purpose of this analysis is to describe four possible cohesion paradigms for expository paragraphs.As the analysis develops it, cohesion embraces both the traditionaltextbook notion of coherence as connections among adjacentsentences and the notion that these local connections must produce a sensible whole. One impetus for presumingcoherence to be a function of cohesion comes from admonitions in composition textbooks to create coherence by repeating words or substitutingpronouns-which is really advice to create unity. The second impetus comes from the fact that coherence as traditionallyproduced by repetitions and transitionwords is not always commensuratewith a unified, sensible whole. To develop this concept, in this analysis I will combine several disciplines in a theoreticallysensitive and empiricallyverifiableaccount of cohesion patterns that should satisfy the formal criteria of linguists, the comprehension criteria of psychologists, and the esthetic criteriaof rhetoricians. On an ordinary,day-to-daylevel readers equate cohesion with the simple and sustaining fact that some sentence sequences make sense and others do not; cohesion distinguishes a sentence sequence from a mere agglomerate, a text Robin Bell Markels received a PhD from Ohio State University, where she currently teaches business and technical writing. This essay, based on her dissertation directed by Arnold Zwicky and Christian Zacher, recently won her an NCTE Promising Researcher Award. College English, Volume 45, Number 5, September 1983 450 Cohesion Paradigms in Paragraphs 451 from a nontext. As a result researchers have treated cohesion as an all but exclusively semantic relation. M. A. K. Halliday and Ruqaiya Hasan, in their enormously influential Cohesion in English (London: Longman, 1976), write: "we can interpret cohesion . . . as the set of semantic resources for linking a SENTENCE with what has gone before" (p. 10). The linkage is accomplished by the sharing of a referent-which Halliday and Hasan call a "tie"-by two or more sentences (p. 3). The criterion of the shared referent appears over and over in cohesion research: in van Dijk as shared semantic features, and in the work of Bellert and Kintsch as shared propositions.' Robert de Beaugrande's list of cohesion markers consists almost entirely of refined variations of the shared referent: recurrence, parallelism, paraphrase, co-reference, anaphora, cataphora, ellipsis, and conjunction ("Text and Discourse in European Research," Discourse Processes, 3 [1980], 289). Of these eight markers, only parallelism identifies something besides a semantic component. The semantic tie obviously contributes to cohesion, and studies by Lester Faigley and Stephen P. Witte report a high correlation between well-written student texts and the number of ties employed ("Coherence, Cohesion, and Writing Quality," College Composition and Communication, 32 [1981], 189-204). But as Faigley and Witte also report, the theory of ties does not explain sentence sequences that contain many ties but nevertheless lack cohesion, such as the following: John likes oranges. Oranges grow in California and Florida. My parents visited California last year. The example shows that the categories of ties formulated by Halliday and Hasan, and elaborated by other researchers, identify necessary but not sufficient conditions for cohesion, and it raises the question whether semantic conditions alone can ever be sufficient. Instead of treating cohesion as an exclusively semantic phenomenon, I want to argue that it is both semantic and syntactic. Even a general description of cohesion as the relationship between parts suggests that two elements are involved. While ties and the noun chains they create impart semantic consistency and thus create one element, only the interaction of the shared items with the syntactic information that thematizes them incorporates a second element and creates textual cohesion. Ties and chains considered without reference to their syntactic position or context turn out to be little more than glorified word lists. Moreover, only a definition of cohesion that embraces both semantics and syntax can explain the macrostructure of a text-that is, identify a structure that is not content specific. A merely semantic definition of cohesion limits macrostructure to theme or content, and it cannot bring us to the philosopher's understanding of cohesion as form. Ross Winterowd, in "The Grammar of Coherence," makes the two terms virtually synonymous: "If one perceives form in 1. Teun van Dijk, "Semantique Structurale et Analyse Thematique," Lingua, 23 (1969), 28-53; Irena Bellert, "On a Condition of the Coherence of Texts," Semiotica, 2 (1970), 335-363; Walter Kintsch, The Representation of Meaning in Memory (New York: Wiley, 1974). 452 College English discourse, he also perceives coherence, for form is the internal set of consistent relationships in any stretch of discourse, whether poem, play, essay, oration or whatever" (College English, 31 [1970], 828). Geoffrey Leech in "'This Bread I Break'-Language and Interpretation" writes: "Cohesion ... is the way in which independent choices in different points of a text correspond with or presuppose one another, forming a network of sequential relations" (A Review of English Literature, 6 [1965], 66). Both writers are using "form" in its philosophic sense "to signify the principle of unity which determines the nature of a concrete whole. ... . Form does not arise out of the mere combination of parts; it is, rather, that prior principle, which imposed on them, determines their relationships."2 An operational definition of cohesion, then, should account for: a) both semantic and syntactic elements; and b) the interaction of these elements to produce a formal, content-free macrostructure for basic paragraphs. Irena Bellert offers the most useful definition of the semantic component of cohesion: a necessary (though obviously not sufficient) condition of the coherence of a text consists, roughly speaking, in repetitions. The structureof each utterance S1 of a text S1, . .. Sn is such that at least one lexical item which occurs in it, or one propositionwhich can be inferredfrom it, can also be found withinthe utterancesof the sequence S1, . . Si-1, or within the propositionswhich can be inferredfrom those. (pp. 336-37) Ordinary experience confirms that Bellert's focus on item repetition captures the semantic constancy necessary for cohesion. But like many general concepts "item repetition" is more easily described than it is defined. One possibility is of course the straightforward recurrence of a term, as in The boys climbed the trees. The trees were oaks. Another is synonymy-which philosophers tell us doesn't exist-as in The boy climbed the fence. The youth is a gang member. But more complex and interesting relationships are also possible, for example the class-member relationship in I wanted some apples. But they were all sold out. Since the apples in both sentences cannot be the same, the reader must infer a class-member relationship between generic apples and those the speaker wanted to buy (Van Dijk, "Text Grammar and Text Logic," in J. S. Petdifi and H. Reiser, eds., Studies in Text Grammar [Boston: Reidel, 1973], p. 53). And this relationship can be complicated still further: The girls played house. The boys spit on them. All in all, the children were feisty that day. 2. Walter Davis, The Act of Interpretation: A Critique of Literary Reason (Chicago: University of Chicago Press, 1978), p. 1. Cohesion Paradigms in Paragraphs 453 Here "girls" and "boys" exclude each other semantically and cannot be construed either as item repetition or as two items sharinga common referent. But the two words designate membersof a single class and share the textual antecedent "children." But even when such refinementsare incorporatedinto the theory of ties, we have taken only the first step toward explainingcohesion. The second is to explain specifically how recurrencesare incorporatedinto sentences. Since cohesion obviously involves relationshipsamong parts, researchershave naively assigned the role of "part" to the sentence, without defining that critical component. Hallidayand Hasan are againtypical: "we can interpretcohesion as the set of semantic resources for linking a SENTENCE with what has gone before." But as we know, the sentence is a very elastic unit, with no syntactic limits to its length or complexity once the minimal requirementsof subject and predicate have been met, so that a sentence can itself lack cohesion when relationships among constituents are unclear and meaning becomes ambiguous. And a unit that is not certainto be cohesive in itself cannot become the basis for a theory of cohesion. In fact, any sentence provides two distinct structuralarrangementsof semantic information:at one level it presents the informationas a complete unit (the arrangementthat Halliday and Hasan assume); but at a second level-the level at which a sentence can lack cohesion-it presents that informationas a structuredarray of internal elements. The unit that includes both levels and is always cohesive-assuming semantic concord-is not the sentence but the subject-verb-objectsentence/clause. The kernel SVO sentence/clause reflects current linguistic interpretationsof syntax, and psychologists have accumulatedevidence to show that this kernel is the primaryperceptualunit as well, since it is the smallest unit exhibitingsemantic determinacy. As E. D. Hirsch explains, "The constituent words of a clause become determinateonly when they are perceived as functionalelements within the clause and these definite functions are fully determinedonly after the whole clause is perceived" (The Philosophy of Composition [Chicago: University of Chicago Press, 1977], p. 109). I suggest, then, that the kernel SVO sentence/clause controls the function of ties, creates the relationshipsamong internallycohesive parts, and provides the basis for an operational, reproducible definition of cohesion. Cohesion is attained when a dominant term, explicit or implicit, occupies concurrently the most importantsemantic position in the paragraphand also the most important syntactic position in each sentence of the paragraph.The following simplistic paragraphillustratesthe concept of term dominance: The Char-Baris a bar on High Street. The Char-Barswings. It permits dancing. The bar specializes in foreign beers. It seats 198 people. Dominance is attained not simply by repetitions of "Char-Bar" and its pronoun substitutes, but by the consistent appearance of those repeated terms in the subject position. When a term is repeated in each sentence but not in the subject position, dominance is not attained, and the paragraph lacks cohesion: 454 College English Alfred likes peaches. Oregon doesn't grow many peaches. We have a peach tree in our back yard. No one throws rotten peaches at politicians or ballplayers. As these two samples indicate, cohesion requiresthe meshing of both semantic and syntactic informationand, at least for some paragraphtypes, can be defined operationally. Psychologists have shown that semantic dominance is achieved most readily throughitem repetitionthat establishes an equivalence chain, in accordancewith the tie theory. They also report that the syntactic position of the repeated items directly affects reader perception of dominance: thematized items appearingin the subject position are better rememberedthan thematized items in the object position,3 which merely reminds us that English is a language of synthesis and position. Meaning,theme, and emphasis all depend on syntactic position, and in simple sentences the most importantsyntactic position is the subject position, the S in the SVO sentence/clause. In the first sample, the appearanceof "CharBar" in each sentence produces a thematized item and semantic dominance. Its recurrence specifically in the subject position gives it syntactic dominance as well, and this convergence of semantic and syntactic dominancein the individual sentences makes "Char-Bar"the dominantterm of the paragraph.In the second sample, "peaches" does not attain paragraphdominance, despite its semantic dominancein individualsentences, because it is not consistently in the dominant syntactic position. This account of cohesion as the convergence of semantics with syntax is also an account of paragraphform. Only when the repeatedterm keeps appearingin the subject position does the repetition create the rudimentaryunified structure that Kenneth Burke and Jonathan Culler independently identify as the series united by a common term.4 As I will show, paragraphstraditionallyclassified as analytic, descriptive, or exemplary can all exhibit this basic structure, which is relatively content free and incorporatesthe philosophicunderstandingof form as the unifying principle underlyinga text. The series united by a common term gives us a conception of form that begins to answer Francis Christensen's call for different kinds of "yardsticks" to measure the cohesion of differentkinds of paragraphs("A Generative Rhetoric of the Paragraph,"in Gary Tate and Edward P. J. Corbett, eds., Teaching Freshman Composition [New York: Oxford University Press, 1967], pp. 200-16). Indeed, it exemplifies the conception of textual form called for by H. B. Lathrop in 1918 when he coined the phrase "unity of connection": In a word,the unityof theme,directiveunity,the unityof a line, is not moretruly unitythanunityof connection,the unityof a chainor the unityof a wovenfabin a systematicway, as by ric. ... Coherence,orderliness,is simplyarrangement 3. Charles Perfetti and Susan R. Goldman, "Thematization and Sentence Retrieval," Journal of Verbal Learning and Language Behavior, 13 (1974), 70-79. 4. Kenneth Burke, "The Nature of Form," in W. Ross Winterowd, ed., Contemporary Rhetoric: A Conceptual Background with Readings (New York: Harcourt Brace Jovanovich, 1975), p. 183; Jonathan Culler, Structuralist Poetics (Ithaca, N.Y.: Cornell University Press, 1975), p. 91. Cohesion Paradigms in Paragraphs 455 deductionor induction,from cause to effect, from top to bottom or from bottom to top. There is no one plan that is preeminentlycoherent; a coherent result is produced by any systematic principleof guidance appropriateto the case.5 The Char-Bar paragraph illustrates the crudest instantiation of the series united by a common term. The following paragraph from Truman Capote's Other Voices, Other Rooms (New York: Vintage, 1948) illustrates a subtler instantiation in which the dominant term is implied rather than stated in a paragraph of description: The truck was of the Ford pick-uptype. Its interiorsmelled stronglyof sun warmed leather and gasoline. The broken speedometer registered a petrified twenty. Rainstreaksand crushed insects blurredthe windshield,of which one section was shatteredin a burstingstar pattern. A toy skull bump-bumpedover the rising, dipping, curving ParadiseChapelRoad. (p. 9) The various ellipses by which this paragraph avoids the tedious repetitions of the Char-Bar paragraph require a series of reader inferences to supply the predictably recoverable repetitions. A sentence like "The broken speedometer registered a petrified twenty" forces the reader to make what Susan E. Haviland and Herbert H. Clark call a "bridging assumption," through which an additional sentence is inferred: "The truck had a speedometer."6 Readers obviously make such inferences all the time, and Walter Kintsch has shown that inferred propositions attain the same status in memory as stated propositions ("Storage of Complex Information in Memory: Some Implications of the Speed with which Inferences Can Be Made," Journal of Experimental Psychology, 94 [1972], 25-32). Bridging assumptions do not themselves provide a shared term for all sentences in the sequence; they do provide deep structure SVO subsentences that are incorporated in reduced form in the total surface structure. A transformational analysis of "The broken speedometer registered. . . ." identifies "the" as a reduced possessive generated by the bridging assumption, "The truck had a speedometer," and assigns that sentence to the dominant node in a deep structure tree diagram. Each subsequent sentence is subject to the same analysis, so that "truck" becomes the dominant term by virtue of its implied recurrence in each sentence, and by its occurrence in the highest syntactic position. The Capote paragraph exhibits the series united by a common term as the form underlying a paragraph of description. The following example reveals the same form underlying a paragraph of exemplification: Admen and packagers,of course, are not the only euphemizers.Almost any way of earning a salary above the level of ditch-diggingis known as a profession rather than a job. Janitorsfor several years have been elevated by image conscious unions to the status of "custodians";nowadays, a teen-rockguitaristwith three chords to his credit can class himself with Horowitz as a "recordingartist." Cadillacdealers refer to autos as "preowned" ratherthan "secondhand." Governmentresearchers 5. "Unity, Coherence, and Emphasis," University of Wisconsin Studies in Language and Literature, 2 (1918), 82. 6. "What's New? AcquiringNew Informationas a Process in Comprehension,"Journalof Ver- bal Learning and Verbal Behavior, 13 (1974), 512-21. 456 College English concerned with old people call them "senior citizens." Ads for credit and department stores refer to "convenient terms"-meaning 18%annualinterest payable at the convenience of the creditor. ("The Euphemism," Time, Sept. 19, 1969; reprintedin Paul Escholz et al., eds., Language Awareness [New York: St. Martin's Press, 1974],p. 18) Again the repeated term is implied rather than stated, and readers apparently infer as a bridging assumption some variation of "Other euphemizers are ... " which they mentally insert before sentences two through six. This inference sequence yields an implied common term for those sentences, while a transformational glance at the syntactic position of the inferred sentence shows it to dominate the actual text sentences-just as a whole sentence modifier like "For would dominate Thus the repeated bridging assumpthose sentences. example" on the based of tions, logic exemplification, introduce an implied "euphemizer" into each sentence, and that term attains syntactic dominance by inferentially functioning as a whole sentence modifier, the highest syntactic position of a sentence. Happily, neither Capote nor the author of the euphemizer paragraph produced tedious counterparts of the Char-Bar paragraph in which the paradigm of the series united by a common term is fully reconstructed. Instead they produced abbreviated yet self-sufficient versions of the paradigm, reminding us that discourse is elliptic and relies on form to guide interpretation. In neither paragraph do the sentences contain direct links with their immediate predecessors. The sentences are not locked into position, and in fact could easily be reordered. Nor does either paragraph fulfill Halliday and Hasan's prediction that sentence ties will cluster at the center of a paragraph (Halliday and Hasan, p. 97). Instead both fulfill the philosopher's prediction that unity of connection emerges from and is shaped by the total structure. There is no prescription for cohesion; different structures have different informing principles. And the preceding analysis begins to equip us with the different kinds of "yardsticks" that Christensen called for. A final example, again on the subject of euphemisms, represents the most structurally interesting and complex instantiation of the shared term paradigm: From a Greek word meaning "to use words of good omen," "euphemism" is the substitutionof a pleasant term for a blunt one-telling it like it isn't. Euphemism has probablyexisted since the beginningof language. As long as there have been things of which man thoughtthe less said the better, there have been better ways of saying less. In everyday conversation,the euphemismis, at worst, a necessary evil; at its best, it is a handy verbaltool to avoid makingenemies needlessly, or shocking friends. Languagepurists and the blunt spoken may wince when a young woman at a party coyly asks for directions to the "powder room," but to most people this kind of familiareuphemismis probablyno more harmfulor annoyingthan, say, the split infinitive. ("The Euphemism,"p. 17) Instead of a text whose bare bones rattle for our attention, here we find the structure fleshed out, more robust but also less conspicuous. Even a cursory examination reveals that the repeated term is "euphemism" (although sentence three substitutes a synonym), and that this term occupies the subject position in each sentence except sentence three, which substitutes the empty "there" in the subject position. Cohesion Paradigms in Paragraphs 457 Two qualities distinguish this paragraph from its predecessors: it conveys more information through more heavily modified sentences, and it uses the additional information to link each sentence to its immediate neighbor. The syntactic complexity of individual sentences, and the orchestration of that complexity into a textual whole, carry this paragraph into a realm of cohesion and craft that semantically based cohesion theories cannot describe. Theories based primarily on recurrence tallies, such as Walter Kintsch's propositional analysis (an elaborated version of Halliday and Hasan's analysis of ties), are self-contradictory in that they recognize the syntactic capacity of sentence surface structure to show relationships, but then use that recognition only to identify deep structure propositions. But since deep structure propositions consist essentially of a verb plus noun(s), they are susceptible to a wide variety of surface-structure transformations. Kintsch's deep structure proposition (Bake, Mary, Cake) may be represented by any of the following surface sentences: Mary bakes a cake. Mary is baking a cake. A cake is being baked by Mary. The baking of a cake by Mary ... Mary's baking of a cake. . . . (Kintsch, Representation, p. 14) But in a given context the requirements of cohesion and craft may make one of those surface sentences highly preferable. Kintsch's procedure reduces sentences to their primeval, pre-text stage in order to isolate recurrences into recoverable items and enable us to track recurrence chains. But in stopping with only the most limited syntactic information, the procedure also reduces whole texts to lists of propositions that obliterate the syntactic complexity reflected in an author's surface craftsmanship. As surely as maturity and grace in individual sentences are measured in part by the simple number of syntactic elements, one wants a theory capable of identifying the various degrees of cohesion reflected by the syntactic complexity of entire texts. Kintsch and others committed to propositional, semantic theories of cohesion justify their preoccupation with deep structure analysis by the fact that people do not remember texts verbatim. I. M. Schlesinger takes this view: "As for storage in memory, there is experimental evidence that the sentence content or deep structure tended to be retained and the surface form to be discarded. . . . Phrase structure boundaries may thus serve as cues to deep structure relationships. Note, however, that the surface structure is not accorded an independent status in the comprehension model" (Production and Comprehension of Utterances [Hillsdale, N.J.: Erlbaum, 1977], pp. 138-39). These statements typify an almost wholesale bypassing of surface structure on the tenuous assumption that the macrostructure of a text is analogous to the deep structure of a sentence. But there is no particular reason why the questions asked by transformational grammar in seeking to explain the structure of sentences should be the appropriate or the only questions for discourse analysis in seeking to understand the structure of texts. For example, transformational grammars are verb-based and treat the parts of sentences as instantiating verb frames. But in discourse analysis one can 458 College English just as easily imaginethe verbs being subordinatedto nouns or topics. Similarly, transformationalanalysis cannot by itself account for the discourse function of articles, which for good reason are given low priorityin sentence grammar,but which are of consummate importancein discourse structure. Despite the theoretical inconsistencies by which sentence maturity and discourse maturity are made equivalent, and despite previous evidence that the syntactic informationavailable in surface structures affects reader comprehension, the question of "How do readersget their semantic information?"has only recently been asked. Even if readers do not remembera text verbatim, syntax may still contributeto what they remember,or to how they rememberwhat they remember, and the theoretical objections I have raised to propositionalanalysis were recently confirmedby the experimentsof David Kieras. Asked to identify the topic of a passage, Kieras' subjects consistently chose what I have been calling the dominant term: the term semantically dominantby virtue of its recurrences, and syntactically dominant by virtue of its subject position. Kieras concludes: "Hence the macrostructurebuildingprocess can be viewed as being based primarilyon semantic context, but heavily guided by the surface form of the passage and passage sentences" ("The Role of Major Referents and Sentence Topics in the Constructionof Passage Macrostructures,"Discourse Processes, 4 [1981], 14). The surface structurecontributesnot only to the cohesion and comprehensibility of a paragraph;it also contributes to its craftsmanshipby providing immediate sentence links. In any of my earlier examples the author could have chosen to begin each sentence with an enumerative transition word such as "first," "second," or "next." But these would have been superfluousto comprehension and would have conveyed little information in themselves; they would have been mechanical insertions as if from a misunderstood freshman English lesson. But in the last "euphemizer" example, insertingsuch transition words (except for "for example" before the last sentence) would be downright illogical and semantically incongruous. Here the sentences are immediately linked by message-contributingadverbialssuch as "since the beginningof time," "as long as," "in everyday conversation," "at worst," and "at best." Besides conveying additionalinformation,these adverbials lock individual sentences in place and create mini-contexts for adjacent sentences within the larger series pattern, even though the adverbials do not participate directly in the "euphemism" recurrencepattern. Roland Barthes observed this phenomenonin largerdiscourse units, and he called the mini-patterns"kernels" and "satellites" (cited in Culler, p. 219). Barthes distinguishedbetween kernels, which link up with each other to form plot, and catalysts or satellites, which are attached to kernels but do not link into sequences. As JonathanCuller points out, "kernel and satellite are purely relational terms; what is a kernel at one level of plot structurewill become a satellite at another, and a sequence of kernels may itself be taken up by a thematic unit" (Culler, p. 220). The patternedrecombinatory features of the series united by a common term-transformations, if you wishprovide the possibility for its integrationinto more richly structuredand informative paragraphs,as we shall see shortly. Cohesion Paradigms in Paragraphs 459 None of the paragraphsthus far consideredcontains a topic sentence. Richard Braddockdemonstratedthat paragraphscan and do get along without topic sentences,7 and we are now in a position to ask the more interesting question whetherthey should have topic sentences. The foregoinganalysis suggests, quite simply, that it all depends. It depends on whether the topic sentence would require developmentby two or more recurrencechains, so that we would be dealing with a different structure from the single term totality analyzed thus far. Should a paragraphbe without a topic sentence? Yes, if it is a single-termparagraph, which, as the "euphemizer"examples show, can be cohesive and stylish and elegant without a topic sentence. But in turningnow from the paradigmof the series united by a common term to more complex paragraphstructuresin which equivalencechains are conjoined and embedded within one another, we must look at paragraphswith topic sentences. The following example again illustrates the principle that cohesion requires the meshing of syntactic with semantic informationin a simple combination of two equivalence chains: John likes Columbus. He likes its low inflationrate and its very low crime rate. He enjoys its many foreign food stores. He likes the city's extensive park system. Finally, he likes the culturaland sportingevents that the city provides. The "John" chain is easily identified:"John," or a pronoun substitute, appears in each sentence, and in each appearance occupies the dominant noun phrase position. The second chain, "Columbus," is instantiated through "its" and "city." Here as with the single-termseries, the mere presence of repeated terms does not insure cohesion: John likes Columbus. He likes its low crime rate and its low inflationrate. Someone told John that Columbus is located in the geographic middle of Ohio. While visiting friends in Columbus, John saw a Clippers' game last summer. Woody Hayes bumpedinto John at the ColumbusCivic Center. In the conjoined chain paradigm,cohesion is attained when 1) the two terms appear in or are inferrablefrom each sentence in the paragraph,and 2) one term appears consistently in the subject position and the other in the object position (i.e., the dominantnoun phrase constituent of the predicate). Here the definition of dominanceis extended to incorporatethe second and syntactically equal noun phrase in an SVO constructionwhen that phrase participatesin an equivalence chain. The patterncan be representedschematicallyas follows: A verb B A2 verb B2 A3 verb B3 An verb Bn 7. "The Frequency and Placement of Topic Sentences in Expository Prose," Research in the Teaching of English, 8 (1974), 287-302. 460 College English The schema illustrates the pure double chain paragraph,in which the "John" chain meets all the requirementsfor a single series paragraphfrom the subject position, and the "Columbus" chain meets those requirementsfrom the object position. Taken by itself the "Columbus" chain would form a single series paragraph. The double chain paragraphcontains a straightforwardtopic sentence of the kind Braddock called "simple"; that is, a complete, stated T-unit representing the main idea of the paragraph.But what is stated, complete, and simple for the reader is not always straightforwardfor the author, even when the topic sentence is as grammaticallysimple as "George Jones is a poor basketballcoach." Especially in academic prose, this set-relation sentence, when developed as a topic sentence, commits the author to showing how George Jones exhibits characteristicsthat locate him in the set of poor basketball coaches. The topic sentence might be developed by means of two equivalence chains-"Jones" and "reasons'-as follows: George Jones is a poor basketball coach. One reason is that he is consistently outcoached in importantgames. Another is that he does not use his players to their full potential. Moreover, he can't teach offense. It might also be developed appropriatelyby only one equivalence chain, as follows: George Jones is a poor basketballcoach. He is consistently outcoached in importantgames. He doesn't use his players to their full potential. And so on. But the mere use of equivalence chains can also produce an inappropriatedevelopment of the topic sentence, as follows: George Jones is a poor basketballcoach. He is consistently outcoached in importantgames. He doesn't use his players to their full potential. Finally, he is a Mormon. This version illustrates that regardless of the number of equivalence chains, a set-relationtopic sentence requiresthe authorto supply an acceptable inference chain if the paragraphis to attain cohesion. In that final sentence a reader either sees a new topic signaled or else sees red, rejecting the inference that being Mormon characterizes poor basketball coaches. In either case the cohesion of the paragraphis broken. While all paragraphsultimately require some amount of inference to create cohesion, in the set-relationparagraphthat requirementis primary.By contrast, the authorof the "John likes Columbus"paragraphcould keep plugging"John" and "Columbus" into subsequent sentences without buildingan inference chain to connect up predicate material. Thus, while a sentence like "John likes peanuts" would break the cohesion, a sentence like "John likes Columbus' flat and.ugly landscape" would maintaincohesion even for readers who reject "flat and ugly landscape" from the class of likeable things. This difference in inference requirementssuggests that there are at least two kinds of simple topic sen- Cohesion Paradigms in Paragraphs 461 tences: the "argument" topic sentence ("John likes Columbus"), whose basic schema is two equivalence chains, and the set-relationtopic sentence ("George Jones is a poor basketballcoach"), whose basic schema is equivalence chain (or chains) plus inference chain. None of the paragraphtypes examined thus far-the single chain, the double chain, the conjoined equivalence and inference chain-has been particularly complex. The structuresare limiting,and neitherthe chains themselves nor their various combinationscan accommodatemuch information.Probablybecause of this limited informationcapacity, most paragraphsin fact combine these structures and consist of one primaryequivalence chain and two or three secondary chains that appearin many but not all sentences of the paragraph.In comparison with the others, the mixed form is more informationallyrich and characterized by a stronger sense of thematic progression-without, however, sacrificing its integrityas a unit. It produces a strong sense of backwardanaphora,or vertical structuring,along with its strong sense of advancing theme, or horizontal and linear structuring.The dynamicjuxtaposing and balancingof these two potentially conflicting conditions is characteristicof this paragraphtype, and this dynamic juxtaposing has been a principal impediment to describing paragraph structure.In a text characterizedby thematic progression, sentence subjects are constantly being replaced, and the inherentcontrol imposed by a repeated subject no longer exists: John likes cheese. Cheese is made in Wisconsin. Wisconsin is the home of my best friend. This example fits Osten DMines' discourse category of "simple linearprogression, in which the rheme of one sentence becomes the theme of the next,"8 but it obviously lacks cohesion, and furthertextual requirementsfor the linear pattern must be identified. One of several mixed patternsinvolving linearprogressionis that in which the single series pattern operates as the primarybut not dominant pattern in the paragraph.In Barthes' terminologythe primaryseries remainsas a reduced kernel while secondary equivalencechains arise from the primarychain and develop the series topic in relation to other topics. Where a single series paragraphdevelops a concept internally by talking about, say, the color, shape, size, and atomic structureof tomatoes, the mixed linear paragraphmight talk about tomatoes in relation to lettuce, celery, and cucumbers. The following paragraph illustratesthe mixed form based on the single series: Imagine,if you will, a miniscule"chip"shavedfroma crystalof silicon.On this chip are all the components of an entire informationstorage and programming system-a full-fledgedcomputer, in other words, that takes up less space than the first four letters of this paragraph.That is the microcomputerof tomorrowand its prototype already exists. The informationit stores could come from anywhereThe Libraryof Congress, The New YorkTimes, the personal bankingrecords of thousandsof taxpayingcitizens. It could come from the tape cassettes of psychiatrists, the daydreamsof novelists, the logbooks of birdwatchers.Soon information 8. Cited in Nils Erik Enkvist,LinguisticStylistics (The Hague:Mouton, 1973),p. 120. 462 College English of this sort will be availableto anyone and everyone at the flick of a switch. You'll be able to plug into it. So will the government.So will the guy next door. Your kid will be able to plug into that information-your doctor, your thesis advisor, your gtfru, your garbage man. And when everyone is all plugged in and accessing, in unison, this monumentalnew universe of data, the existential situation is going to change. (TheTechnolPeasant*SurvivalManual [New York: Bantam, 1980],p. 9) As an aid to identifying the underlying recurrences and chains, the following schema lists the subjects and objects of the individual sentences: SUBJECT (You) implied Chip That exists Its prototype Information It (information) Information You Government Guy Kid situation OBJECTOR PREDICATE NOMINATIVE chip componentsfor an entire informationstorage and programmingsystem microcomputer anywhere tape cassettes, daydreams,etc. anyone it (information) (information) (information) information going to change The list spotlights "information" as the primary equivalence chain, as it occurs in either the subject or object position in eight of the twelve sentences (counting the compound sentence as two). In its semantic and syntactic bulk, the "information" chain replicates the single series pattern, but now in reduced form. The list also reveals the secondary equivalence chains and indicates how they must be introduced to prevent their appearance from turning the paragraph into a wild rampage of new topics. The "people" chain is the most conspicuous secondary chain because of its length: You'll be able to plug into it. So will the government. So will the guy next door. Your kid will be able to plug into that information-your creditor, your thesis advisor, your guru, your garbageman. "People&' forms another series involving fewer recurrences than "information" in the primary series, and this secondary series is introduced without signalling a new topic or paragraph and without creating another Wisconsin cheese sequence. Here two criteria must be met: 1) the secondary chain term must be introduced as the object in a sentence whose subject is the primary chain term ("Soon information of this sort will be available to anyone and everyone .. ."); 2) when an instantiation of the secondary chain term appears in the subject position, an instantiation of the primary chain term must appear in the object position ("You'll be able to plug into it."). The semantic pattern is A verb B B verb A Cohesion Paradigms in Paragraphs 463 Both these criteria reflect the necessity for back anaphora to prevent the noncohesive linear progression of the Wisconsin cheese sequence. And from the perspective of information theory, these two criteria reveal how new information-the secondary chain-is introduced and developed through old, known information without signalling a new topic or paragraph. The following segment illustrates another kind of secondary chain: On this chip are the componentsfor an entire informationstorageand programming system-a full-fledgedcomputer, in other words, that takes up less space than the first four letters of this paragraph.That is the microcomputerof tomorrow,and its prototypealreadyexists. The recurrence pattern for this sequence can be represented as A verb B B verb C Since information about the computer (item B) is introduced through old information, the sequence meets the first criterion that the secondary term be introduced as the object in a sentence whose subject is the primary term. But because the new information about "microcomputer of tomorrow . . . prototype already exists" (item C) is not then incorporated into the other chains but is linked only once, the final sentence of the segment is of marginal importance both informationally and structurally. With only a single recurrence link a sentence like this can easily be pruned without disrupting either cohesion or sense. It is an allowable aside within the structural and cohesive patterns of the paragraph, but critical to neither. A final question arises in connection with the sentence, "On this chip are all the components for an entire information storage and programming system-a full-fledged computer," and that is the introduction of "information" in the primary chain through its emergence from an adjective rather than a subject or object position in the sentence. Informationally rich in authorial choices, this sentence sets up possible chains for "chip" and "computer" as well as "information." The richness stems from the surface combining of several deep structure kernel sentences, roughly: "On this chip are components. The components are for a system. The system stores information. The system programs information." Although "information" is initially subordinated in the adjective position in the surface structure, the deep structure definition of adjectives originating from SVO structures give it the potential to be actualized as an object. Thus introduced as one bit of information in an opening, framing sentence, its eventual participation in a primary recurrence chain is predictable, not an ad hoc exception. This sentence shows in fact how the author's handling of SVO sentence/ clauses as parts of sentences, as well as complete sentences, creates cohesion. In outlining how the current work of linguists and psychologists can be joined with rhetoric in the study of cohesion, I hope I have also shown how this total study also leads to a literary and philosophic sense of form (even in the mundane expository paragraph), and also to a practical pedagogy for English teachers. In showing how the particular series united by a common term anticipates para- 464 College English graph development by analysis, example, or detail, and in showing how even simple topic sentences like "John likes Columbus"and "George Jones is a poor basketball coach" have different formal and inferentialrequirements,I hope to have shown finally that paragraphsare indeed structures-crafted units that do not depend on mere authorialwhim, as Paul Rogers suggested in 1966,11 and as Arthur Stern repeated ten years later in "When is a Paragraph?"(CCC, 27 [1976], 253-57), which Tate and Corbett reprintedin 1981 in The WritingTeacher's Sourcebook. From the perspective of such writers, or even from the deeper perspective of Mary Louise Pratt's Towarda Speech Act Theoryof Discourse (Bloomington:IndianaUniversity Press, 1977), with its typical currentquestioning of any form in discourse beyond what the reader brings to it, my research must seem reactionary.I hope it is; and while it affirmssome old values, I hope it enables us to take another step toward understandingdiscourse not only through mathematicalrepresentationand recall experiments, but also through the structuralpatternsthat do in fact underliediscourse and make style possible. These patterns can be seen after all as distillations, not reductions-d istillations analogous to a person's perception of a shape on a page that allows her to say, "This is an 'a' not an 'o'," but also allows her to say, "This 'a' is more elegant than that 'a'." 9. "A Discourse-CenteredRhetoricof the Paragraph,"CCC, 17 (1966),2-11. SEARCH FOR NEW EDITOR OF COLLEGEENGLISH Donald Gray's term as editor of College English will end with the May 1985issue of the journal. The Search Committee for his successor will welcome letters of application, to be received no later than November 7, 1983. Letters should be accompanied by a vita and one sample of published writing by the applicant. Applicants are urged to act promptly to explore the feasibility of assuming the demanding responsibilities of the editorship, which will require some institutional support in addition to that provided by NCTE. The applicant selected to be the next editor will spend the 1984-85 academic year effecting a transition and preparing for her or his first issue in September 1985. The initial appointment is for four years and is renewable for three years. Applications should be addressed to: College English Search Committee, NCTE, 1111 Kenyon Road, Urbana, Illinois 61801.