You are on page 1of 33
DESCRIBING ENGLISH LANGUAGE SERIES EDITORS John Sinclair: Ronald Carter Patterns of Lexis in Text ie Michael Hoey 1 Questions about cohesion 1.1 The aim of this book Discourse analysis has become over the past ten years a major growth industry in modern language study, and several hundred books are published on the subject every year. This being so, any new publication in the field has perhaps to justify its existence. Earlier works needed to defend their discipline; current works need to defend themselves against the charge of superfluity. This book attempts to describe a new system of analysis based on the study of cohesion, particularly lexical cohesion. What distin- | guishes it from previous studies of cohesion ‘is that the attention is ~ not primarily on itemizing cohesive features but on observing how they combine to organize text. The analysis allows us to say some significant things about the way lexis organizes text and offers a new perspective on the way we store and process language that is of some: relevance to language teachers and applied linguists, as well as to theoretical and descriptive linguists. Out of the analysis is developed a methodology for the production of readable abridgements of text that is capable of some degree of automation. In short, I have tried to offer in this book fresh insights into text, and that, I hope, will be found sufficient reason for reading it. text analysis was still in its early stages jus joned was still little more than nt works were published that Sohesion may be crudely defined ical features of a sentence can \ ors (and successors) in a text. in part organized, in part —— eed answering fons that 1 Questi tence of thy ey nvenchisen Of ‘these elem, Bearerencon rroundi ents created, bY thE PITT 'Igok 10 titad resisted genoa a create? iy read sna that ‘sfactory j eie requ Phenome ars, such as pronominaliy,.® "andl interpretation” i ound Gee a irel in aan student clings sntence- yere ‘0 within sente nctions, W. : i € nor snce conju ‘actors were taken j Onk and sented once textual facto he nature of text ise] uN bat well handled ‘casting light on t itself, capable in turn oer studies referred to (or rather, the fitst to The earliest ©} as that of Quirk, Greenbaum, Leech, and Svar wide SS later chapters of this mammoth 8rammar w, (1972) Sees features that ground a sentence in its Context, buildin Soir 3 y the authors (particularly Greenbaum 1965, i devoted superseded by a much expanded dees ee br on previous works Seale cently 5 this a oy al, (1985). Another of these studies, Guwinsy tion cohesion in a stratificational fr, insane ect cont a its fost coe isan for some research studies in stylistics, Tye a ae paren most widely known, Halliday and Hasan (1976), i built heavily upon on two earlier works by Hasan, one published (1968), one unpublished (1971), which listed and classified the devices available in English for linking sentences to each other, and (less directly) on an unpublished, but widely circulated, report on. research into scientific text by Huddleston, Hudson, ‘Winter, and Henrici (1968) that used Halliday’s model of grammar. Unsurprisingly, given the reassuring connection with grammar and the insights provided into the ways sentences connect, these studi of cohesion proved influential both within the field of text anal and beyond it. Thus, in linguistics cohesive studies have proliferated, while in English language teaching it has become commonplace for materials to have units devoted (implicitly or explicitly) to encourag: ing the recognition and use of cohesive devices, How do these devices work and what is their place in a descriptio of text? Clearly, the answer will be slightly different according whose description one adopts. Si bove, Halliday and Hasan (1 976) Pts. Since, as noted above, : ae i TYe Probably had the widest audience, : set sealed description, the ans\ : Ronni vide. : Halliday and Hasan, the Organization of text (which th texture) i i / @) is made up (in large part) of relationships amongst | Questions about cohesion Swe tee ERmENES Some grammatical, which they refer to as eeadiclaencoets eee in kind and are divided into five | a eee >a heterogeneous group of criteria, though is made of the fuzziness of the boundaries. between classes of ties. The five classes are: conjunction, referenc s ‘ a ' e substitution, ellipsis, and lexical cohesion. ‘ Conjunction stands on its own as a category, and covers the use of adjunct-like elements by writers or speakers to mark the semantic relationships they perceive as holding between the sentences they produce. Thus, items such as however, alternatively, and on the other hand may all serve to mark a perceived semantic relation. We shall have little to say about these in this work; while they undoubtedly contribute to the semantic organization of text, they are in our view better treated as part of a larger system of semantic relations between clauses. Reference does not mark semantic relations; it is a semantic relation and occurs whenever an item indicates that the identity of what is being talked about can be retrieved from the immediate context. Thus, pronouns and determiners are both, in these terms, reference items. An example of each can be seen in the following (edited) extract from Masters of Political Thought: Vol. 1 From Plato to Machiavelli by Michael Foster, part of a passage about which we shall have much to say in later chapters. The pronoun is in bold type-face, the determiner in italics: 1.1 What is attempted in the following volume is to present to the reader a series of actual excerpts from the writings of the greatest political theorists of the past; selected and arranged so as to show the mutual coherence of various parts of an author’s thought and his historical relation to his predecessors or successors. . . - 1he book does not purport to be a history of yess iti ry, | ions i illustrate the litical theory quotations interspersed to illu ie Tt is lection of texts, to which I have nentary. ng to Halliday and Hasan, ‘occurs whenever one of a small i ical item in the text, the js nothing at all! As lines among these danswerms ‘ t stions that ne Que: indistinct. Thus, in answer to ves are indisti ; dl \ and reference a endear ea agrees yo catego eo epoes Agatha 8! c. of reference: unlikely question POF tlipsis, and in ‘e stitution, 9 : instance of substi Ido. a. No, but b > does. : a es Joos it to annoy us, I th : : c. Yes, she de do in a. replaces sing (No, uit T sing) acl ys re rgume oat dp in OC SC ae \ a an instance 0! by sing (Yes, she does sing) and it is therefore the absence a followed by sin 5,5 ane d that does in c. in combinatio, sing that is cohen ee ed singing in the bath describeg a Sar cere eae mee Tied ‘show, there are some the aussie! ieee to these fine distinctions, but the impor. gram them for textual analysis is not readily apparent. Further eo snaie conversation (and possibly insideicas wale ete iaa these categories, substitution, is rare, which is why the compari therefore, j | les. | por a pear eae oe what these classes of cohesive _ tie Eire is that they are all ways of repeating, and this is al§o true of much of the category that Halliday and Hasan loosely label lexical cohesion. Under this heading they include a variety of kinds of semantic relationship that can exist between lexical items, clustering them into two broad sub-classes: reiteration and collocation, As its name suggests, it is the reiteration sub-class that is concerned with repetition. Reiteration covers a range of ways in which one lexical item may be understood to conjure up the sense of an earlier item. Most obviously (but also, as we shall see, most importantly), a lexical item may exactly repeat an earlier item, Alternatively, an item may be ina relation of synonymy or near-synonymy with an earli item; in text 1.1 above, book is a near-synonym of volume. A thitt possibility is for a writer or speaker to refer back to an earlier item| Raa ese Hee has spoken of a spaniel, one may later neue coe ns ee to be talking of the same creature, isan pee 3 e end of the Previous sentence, is ava par ind of Feiteration—by general word. : reitersti a eee eea end Hasan note, the boundary bet ‘ation by Bubctordinatesan jee wee , ) tremely fuzzy, treet = ee Teiteration by general fe between the two types Questions about cohesion Jation and it i ienifi F ee it is of less significance for text analysis than for lexical analysis to distinguish them. Therein lies a c I ial point, While conjunction, reference, substi- tution, and ellipsis are markers of textual relation, the various types of lexical reiteration are in the first place types of lexical relation and only secondarily markers of textual relation. This is still more true of the other sub-class of lexical cohesion—collocation. Under this heading, Halliday and Hasan include a ragbag of lexical relations, many of which have no readily available name. Typical pairs of lexical items whose relationship would be included within this sub- category are boy ~ girl, wet — dry, order — obey, laugh — joke, ill ~ doctor, try — succeed, king — crown, and sky — sunshine (all examples, from Halliday and Hasan 1976: 285-6). It would be easy to criticize the heterogeneity of this motley assortment of pairings, but to do so would be unfair. At least Halliday and Hasan’s discussion of lexical cohesion acknowledges the existence of the important text-forming properties of lexis, even though the apparatus was not available for the precise classification of the kinds of lexical relation that perform this role. (To some extent Martin, 1989, has now supplied this deficiency.) Hasan (1984) acknowledges the weakness of the collocation category and concludes: While I firmly believe that behind the notion of collocation is an intuitive reality, I have come to accept the fact that unless we can unpack the details of the relations involved in collocation in the Firthian sense, it is best to avoid the category in research. The problems of inter-subjective reliability cannot be ignored. (p. 195) It will be noted that Hasan refers to ‘collocation in the Firthian sense’, But it is doubtful whether collocation in Halliday and Hasan’s work can safely be regarded as the same concept as Firth’s. Indeed, while they cannot be criticized for including a ragbag category of lexical relations, they an more reasonably be criticized for labelling these relations tio} ollocation has long been the name given to the al item has with items that appear with greater obability in its (textual) context (for example, Fi 66). Halliday himself uses the term this way in e yr example, Halliday 1966). Such a CQuestvr- ~ «in principle statistically ee (2S long as relationship is 1 P! xt), the pait of item: Me es cing refeng s enough Ce an reason to suppose that all the Paitings jy There is no Fauld necessarily be true Collocay by a intuition is © pairings as collocation ta TaResorlexic |relationship. reiteration and collocation R e key point rem: 3 SE ay and Hasan use the term) are lexical r way that Halliday a lations, es stance textual ones. In so far as they contribut in the first instance text! ation © creation and organizatio! at eC CeRtOnee f text, the organization is lexical, g i d textual relations ig. i ip between lexical an Ons is n course, the relationship Bees ; Saanich ‘As Halliday and Hasan (ibid.: 289) note: 19 1 unreliable guide—and the labeltin, 8 of has to some extent muddled the study ainly ar a Of bo (even j Without our being aware of it, each occurrence ot a coca carries with it its own textual history, a Pas locational environment [Author’s note: here collocation is being used in original sense] that has been built up in the course of dl creation of the text and that will provide the context within which the item will be incarnated on this particular occa This environment determines the ‘instantial meaning’, or tex meaning, of the item, a meaning which is unique to each sp instance. Thus, the text provides the context for the creation and interpre 7 tion of lexical relations, just as th e lexical relations help create t texture of the text, 4 To compensate for the loss of the category of collocation in later work, Hasan (1984), tightens up the categorization of les cohesion, incorporating within her sub-categories some | relations that would Previously have been handled unc ee } teas of collocation, The sub-categories of ies inset i ia i recognizes are listed in Table 1, ta Categories of lexical Cohesion A General leave, leaving, left leave, depart Questions about cohesion B Instantial i. equivalence the sallor was theit daddy, you be the patient, Ml be the doctor Hi naming the dog was called Toto; the : dog Fluffy Ni semblance — the deck was like a pool; all my pleasures are like yesterdays Table 1.1 ned the Some of these categories we shall have occasion to return to in Chapter 3. : We have devoted considerable space to discussing lexical cohesion in Halliday and Hasan’s work (and Hasan’s later modifications). The reason for dwelling on this aspect of cohesion is simple. Despite the fact that lexical cohesion is covered in Halliday and Hasan’s book in less than twenty pages (compared with over fifty for substitution), it is the single most important form of cohesive tie, even in terms of Halliday and Hasan’s own sample analyses at the end of the 1976 book. It is instructive to compare the frequency of the different types of ties in the analyses they provide. They analyse seven texts of a variety of kinds; Table 1.2 shows the frequency of each class of tie for each type of text: Lexical Reference Conjunction Ellipsis Substitution cohesion Children’s fiction Oral narrative Sonnet Autobiography Dramatic dialogue Reported interview Transcribed interview , 26 10 (including Conjunction) Bor, 4% (excluding Conjuncti 4 ee fo d answering .s that nee Question: lace, there j jon. In the first place, there is noyp. importance of lexical cohesion lationship with morelaa ‘cal item formin 1 re than an Eo ente etnh te al e ener eorall jay and Hasan analy fe 1.2), then Wain) that Haliiday © a } ve six times; it * that appears i : re thing! that Se anal eis pe that the final instance of ‘the bad 4 onable to cor vould Ing is i lexical relationship with Je previous five occurrences. w® if a ey the ead would fail we last. It is ne the earlier instances, es indeed these earl. connections ve rovide much of the eae comer Necessary . cone f occurrence. Lexical cohesion is the only interpret the Find ‘larly forms multiple relationships (though cohesion that regwany ). If this is taken into account, e does so too, . lexic aa eiceeenr a dominant mode of creating texture, fq 0 cohesion beco study of the greater part of cohesion is the study of words, the stu f cohesion in text is to a considerable degree the stud, and the study of cohe' : of patterns of lexis in text. 1 1.3 Questions concerning the place of cohesion IT have taken Halliday and Hasan’s account of cohesion to representative of the range of pioneering studies to which T earlier referred. If this book were a synthesis of work on cohesion, I would, of course, have wanted to include discussion of improvements to the approach (for example, Stotsky 1983; Jordan 1984a). But d occasional challenges to the basic concept of cohesion (for ex. Morgan and Sellner 1980), there seems little point in denying t English language contains devices, and enforces lexical relatior that may connect an utterance with other utterances anc Presence of these devices or relationships encourages reader to interpret the combi ngi might quarrel with the d theory), it is not sufficient, Questions about cohesion ¢ low dow a , 1 eis es the presence of cohesion contribute to the coherence of 2 How does the presence of cohesion affect the ways in which sentences are perceived to be related to each other as complete propositions? 3. Does cohesion contribute to creating the larger organization of a text (if such exists)? Question 1: cohesion and coherence The first of these questions presupposes that coherence is not SOD yon with cohesion. Widdowson (1978) notes that it is quite possible to encounter snatches of dialogue that manifest no instance of cohesive tie but which are entirely coherent in spite of that. He draws the conclusion that cohesion is best defined as ‘the overt, linguistically-signalled relationship between propositions’ (p.31). Coherence, on the other hand, is defined as the relationship between illocutionary acts (ibid. p. 28). Such definitions make the question about the relationship of cohesion and coherence hard to formulate, if not downright irrelevant. A similar strategy of keeping the two concepts entirely separate is adopted by de Beaugrande and Dressler (1981). Cohesion and coherence, for them, are two of the seven standards that a text must meet if it is to be regarded as communicative. (The others need not concern us here.) They argue that cohesion ‘concerns the ways in which the components of the surface text, i.e. the actual words we hear or see, are mutually connected within a sequence’ (p.3). Coherence, however, ‘concerns the ways in which the components of the textual world, i.e. the configuration of concepts and relations which underlie the surface text, are mutually accessible and relevant’ (p.4). Again, such a stance might be interpreted as meaning that question 1 is capable of no interesting answer. Hasan (1984) defines coherence as ‘the property of “unity”, of “hanging together”? (p. 181), and she notes that normal speakers are sensitive to variation erence. She claims that ‘textual coherence isa relative, not yperty, so that it is possible to rank a group of tex 961) from most coherent to least coherent’ (p. ‘suggest that coherence is only assessment. This allows us to make mm, will allow us to tackle the a = ns that need answering Question xd, We will assume that cohesion ig g = jon we have false ence is a facet of the reader's eval d that cohesion is objective, capable in tong words, ite coherence is subjective and judy ; w rom reader to reader (though, oie dent composition, there isin ce ; and coherence are distinguished If cohesion rs a question about how the Presence gf a reader to find a text coherent, cist qu of the text, ane a text. In oth automatic recogmitio’s Ny concerning it may ee cS realm of literature 4 nition, variation), eco manner, question 1 Ce ; cohesive tie predisposes ionships i ntence relationship: stion 2: cohesion and sei Question 2: c s Jationship between cohesion and coherence, itis lg Fier aot by themselves criceTia SEC aaaeaa oe ey presence of ties to make a reader And faa ae ee in Pip conversation theit absenconee eee Se ee lag thacuiiacieee One reason is that, i addition to perceiving, ties between words in the Sentenges aa encounter, we also see relationships between the sentences as whole units. Two sentences may be understood as being in contrast with cach other; one sentence may be seen as providing a reason for an earlier one (See my previous sentence, for example); a whole group sentences or clauses may be interpreted as exemplifying an earl generalization (as these clauses exemplify the generalization I ma in my previous sentence); and so on, Although there is no universally agreed way of describing the wa sentences relate to each other, that they do relate is obvious enot otherwise texts would disintegrate into collections of sen loosely connected lexi i i i excepta shared eee nero cat ala does cohesion (t contribute to th, a5 wholes? ( round,) IPS We perceive between those sentenc ee question can also be asked the other | Questions about cohesion arguments for, or to use, the system of description outlined in subsequent chapters, All that is necessary is to accept that sentences have a meaning together that is more than the sum of their separate parts, a view that underpins all work on discourse. Question 2, then, is concerned with cohesion’s contribution to this property of discourse whereby sentences mean more together than apart. Question 3: cohesion and text organi There are three possible ways of viewing text organization, all of which have been held by some linguist at some time or other. The first is that text has no organization whatsoever; this view seems to be losing ground, mainly because it flies in the face of our everyday experience. The second is that text has some organization, but that this organization does not have the status of structure, a structural description being one that permits one to make predictive statements about the data under examination. Thus, according to this view, it is not possible to make predictive statements about text organization. This is a position I have argued for (Hoey 1983a, 1983b, 1986a). The third way of viewing text organization is that text does permit of full structural description; this is perhaps the most common view and is held by such linguists as van Dijk (1972 et seq.); Longacre (1983); Pike and Pike (1977); Graustein and Thiele (1987), and Grimes (1975). A rather different version of this third view focuses on structure from the perspective of genre (for example, Halliday and Hasan 1985; Martin 1985; Ventola 1987). Obviously the question of the relationship between cohesion and text organization will mean something quite different depending on which view of text one takes. In Chapter 2, I offer some reasons why the third view seems to have taken such a firm hold and suggest an alternative way of looking at text, Indeed, much of the book might be seen as an argument against a rigid view of the structure of text (though what I have to say may not be incompatible with the str uctural view associated with genre). To believe that s no nable to structural description (in ever, is not to assign it to chaos. g that text is patterned; much of devoted to describing this ohesion contributes to or book indicates that I think 3 Types of repetition 3.1 The aim of this chapter In the previous chapter, we analysed informally the patterns of repetition in a short popular-science text. Before we can do anything useful with the kinds of analysis we presented there, we must first identify more precisely what is going to count as repetition in our description. This chapter is, therefore, devoted to categorizing the various types of repetition, so that judgements as to the presence or absence of repetition can be properly replicable. 3.2 Repetition as links We argued in Chapter 1 that cohesive studies are, in part, studies of Jexical relation. In this chapter, we consider the kinds of lexical relation that we are interested in investigating, namely those that ~ permit repetition, To these are added a small set of cohesive features that are not lexical in nature but, likewise, enable repetition to take place. The latter are necessary to complete the picture, but it is the former, lexical, relations that are our main concern. Accordingly, ater attention is here given to the lexical relations than to the er kind. Many cohesive devices are not discussed here for no - reason than that we intend to make no use of them in our s, and what we do describe follows, jn the main, well-trodden is our not our topic, and we will have achieved we consider to be repetition ical method to be replicated. The ies and excluded is to be found in the wide and not in the discussion to be pro. lexical cohesive relations links. ablished term tie (as used by Answers from text analysis P iday and Hasan 1976) is that Halliday and Hasan include Halley on ding of tie than we do under the heading of li More the head a Cn Sea we do not include sentence conjunctions, collocat rain Soe ‘of reference item. A secondary factor is that tie seem” imply directionality more than link. 1h suppor ate Notice tha ‘one can talk more naturally of a an cing linked than of 1 on the other hand, how natural it is to Say that aaa ions, being tied. Compare, ais tied to b. i aie reno Our exemplification will be drawn mainly from text 2.1, analyseg in the previous chapter; for convenience of reference the text i, therefore, repeated here: 3.1 DRUG-CRAZED GRIZZLIES 1 A drug known to produce violent reactions in humans has been used for sedating grizzly bears Ursus arctos in Montana, USA, according to a report in The New York Times. 2 After one bear, known to be a peaceable animal, killed and ate a camper in an unprovoked attack, scientists discovered it had been tranquillized 11 times with phencyclidine, or ‘angel dust’, which causes hallucinations and sometimes gives thé user an irrational feeling of destructive power. 3 Many wild bears have become ‘garbage junkies’, feeding from dumps around human developments. 4 To avoid potentially dangerous clashes between them and humans, sc; mind-altering drug was responsible fon i haviour of this particular bear, no ese, Particular bear, chh the effets of giving grizzly beats or other mamman ens doses of phencyclidine. mals repeate 3.3 Simple lexical repetition What all types of repetition link have ; connection with our discussion of Wi Gee See they allow a speaker or writer to say somethin, Reyer 2) that something new may be added. The simplest fare cine the simplest kind of lexical relation, namely the hing pee is algo between two tokens of a type. at may exise in common is inter’s work in 52 coe — Types of repetition Simple lexical repetition oc ‘curs when already occurred ina text is re when a lexical item that has pee ea ea ne eaeaeraed with no geeater alteration than 8 of a closed grammatical paradi r igm. For example, in 3.1, bears in sentence § j pears in sentence 3, Benes nue simple lexical repetition of pépcticion of bearslinveeticenceia the only! varitiontkie keane eine enturely@icablmermea ie ree dgeeaicale ee ae ee this ENCE epee tte insular ox plical paradign. Bese Hee a ie most basic, and is what most eiorteee hey think of repetition, we need not henceforth refer to it as lexical—it will be sufficient to ref simple repetition, the lexicality being soreeane c ality being taken for granted. Only open-set lexical items can enter into such a link. We will not treat as simple repetition connections between such grammatical items as determiners, prepositions, auxiliaries, negatives, co-ordinators, subordinators, Sentence conjunctions (or conjuncts), sub-modifiers, or particles. This is an analytical decision, not a statement about the impossibility of repetition between grammatical items, still less a claim that such items may not materially contribute to the relatedness of sentences. Winter (1974) notes the contribution grammatical items may play in repetition-replacement patterns and there is tacit recognition of their importance in Chapter 7 when we look at the parallelism holding between (parts of) sentences connected by lexical repetition. Our discussion so far migh suggest that the notion of simple repetition is entirely unproblematic, but this is unfortunately not so. To begin with, a difficulty resides in the assumption that words retain the same meaning when they are repeated. As Rimmon-Kenan shen the whole sign is repeated, difference is (1980) points out, ‘even w g e introduced through the very fact of repetition, the accumulation of significance it entails, and the change effected by the different context in which it is placed.’ McCarthy (1987) draws attention to the fact that the meaning of a lexical item may be the subject of renegotiation ici in a conversation. ‘A speaker may choose to between two partic his or her co-conversationalist, but with a shetorical ploy this presumably can combines apparent acceptance of the e wi redefinition of them in ae toa len may be the subject of i ted, it semua where a word is not renegotia! 1 ence “a Mn ‘drift’. Without speakers ate ac Ine aedae chaciciment tae slow shift in the course of a conversati oat beginning. is slightly different from one ne ae fe oe Pe It is difficult to assess the implica : a : Drift can only be described in terms: of something h constant, communication would soon break down i ‘shifting. Oaveeuend was that it was unreliable and On am ee ie Tippee othe; hand, polysemy isa further indica Carthy describes seene™t inconstancy of lexis. The phenomenon MCAT 1 © ST WS Seems ty throw doubt on the role of simple repetition across oe boundaries; it must be a matter of future investigatio a deft whether, and if so to what extent, the phenome ot drift renegotiation exists within single-speaker utterances or in written texts. We can assume, however, that repetition oto sS Speech boundaries at least betokens a desire to make connection with the previous speaker’s utterance and that drift likewise can only occur when something is being held constant. In other words, simple repetition may not be total repetition—but it is undeniably partial repetition. A second difficulty relates to the polysemy alluded to above. The neat compartmentalizing of senses in a dictionary disguises the protean nature of much of our vocabulary. Lexicographers find that for many of the most common words there are myriad subtleties of difference in meaning. How, it might reasonably be asked, can one claim in the light of this that repeated lexical items mean the same? How, indeed, can one be sure one is dealing with the same lexical item? There are two answers to this. To be; gin with, we note as Types of repetition third difficulty i per eae that, once we have identified all clear instance Corer aey ith a residue of doubtful cases which fall halfway along the cline between grammatical and lexical Examples from text 3.1 ording to (sentence 1) and particular (eentence 5); an adverbial such as sometimes might be added to this jist. Although it is apparent that according to might be treated as a compound preposition, the possibility of (near)paraphrase ee Jexically based phrase such as in the opinion of edges us towards greating it as lexical. For the analytical method we describe, however the decision is not crucial. As will become apparent as we proceed it ig not the presence of isolated links that has significance, but the clustering of links. In consequence, the stray doubtful case will normally cease to matter once these clusters are identified. This suggests why it is that hearers or readers are not thrown by stray links and means that, for the purpose of our analysis, it is sufficient to ensure that the main grammatical items are excluded. re 3.4 Complex lexical repetition Closely related to simple lexical repetition is complex lexical repetition. This occurs either when two lexical items share a lexical norpheme, but are not formally identical (as defined in our discussion of simple repetition), or when they are formally identical, but have different grammatical functions. In 3.1, an example of the first type is drug (sentence 1) and drugging (sentence 4), and of the second, humans (sentence 1) and human (sentence 3). The second example is, in fact, a marginal case; if human is assumed to be an adjective, then it is indeed in complex lexical repetition with humans. If, however, it is deemed to be a noun- modifier, then it is in simple repetition with humans, since nouns are capable of appearing in ae. position and this is not, i jcal function. thereon, ead Paes Jationship of complex lexical in the context of the text in "gs to ensure that the paraphrase of ping may be (roughly) paraphrased administering @ drug to’; human “of humans’. It will be noticed yne to accept the possibility of teal or j, ‘ a nd imagina form a res ir no; poe "ePetition. On ihe sei adjectiy, Tepetit, tring to a es, On. Lexi.) the sa, fe iy “ANNOY, after ail al items with we Obiect that On) it simply assumes close approxi i ee eae Se yee TePetition i ae ee : to complex lexical repetition ane efer henceforth re a as be Ae ee nee 7eO ee Taba Waly OW VO tare 1983) We should ng aie with knobs on. Stotsky ( 3), is just simp Pn in some detail, treats it as quit ru aces and terms pairs of items so conne simple ted deriva he collocations that a particular word forms with For a start, the » different from those that a closely related words may be quite ea 6); economist and economy, for example, may form (Renouf 1 a oe dine oes of orcs aa en w that complex repetition is not » as eee sent expect items in true repetition toh ees el profiles. Since, novel oN are only rega repetition as significant when it forms eas any com : repetition that occurs in these clusters must—by de: inition—share common context, even if, in the language as a whole, the items hay quite different collocational profiles. es, exact paraphras Plex > Who di. € distin, 3.5 Text-forming versus ‘chance’ lexical repetition It will have been noticed that we cen mentione, Ff e “object? ( ned. If a pair of i the grammar of ve ve : ve “niently said to reeneeeia Types of repetition second criteri me iB iterion for distinguishing text-forming from chance repetition is contextual. Of each pai al sstions 5 ach pair a i : : pair of lexical items the question: é: a nes Searen or related contexts? or . Do the items share comr i i J 5 mon relationships with neighb lexical items? or aie Is there w ' c. Is there whole or partial parallelism between the contexts of the items? What these questions seek to discover is whether the contexts of the items provide any evidence for regarding the lexical repetition as serving some function. In a long text, the contextual criterion can be difficult to operate, particularly if certain words are recurring with great frequency. It is, therefore, necessary to add an important rider to this criterion, namely, that if a lexical item appears for the third (or more) time ina text, it is only necessary to establish a contextual connection with ' one of the previous occurrences for the item to be treated as forming a repetition link with all previous occurrences. This can be represented quite simply diagrammatically as in Figure 3.1: fa repetition established Figure 3.1 rrence of the ‘link triangle’, a useful analytical y casion to employ again below. n only be answered as matters of ful scope for dispute over the ways ed, Also, they may be valuable in no use for automatic analysis. all the evidence points to chance ficant proportion of the lexical the analysis that follows, as _ Answers from text analysis \99 « not necessary to distinguish tex: come apparents 1 1 cor normall being elimi t form) walt become APPA ye latter NOME g climinated from 8 from chance Ecol stage. However i is SE to be abiet® ne analyse Fink between 8 pair of feria items, even ied oY justify the Toe needed for the practical purposes of the anal a pe representing the decision processes that are being aa iv when we identify @ lexical item as being a simple or com lade -yious item, is in the form of a group of once repetition of a pre flow charts. The first flow chart (Figure 3.2) n worth investigating. further flow charts: Figur ouMCALTENS | simply establishes whether there, ‘As can be seen, Figure 32 a ¢ 3.3 is primarily (though ng a phenomeno entrance into Y Ener | SIMPLE | YES the items LEXICAL graphologically ae > REPETITION or phonologically, low chart identc Figure 33) | ce the items graphologicall or phonologically, similiar : Ne similariy ele items sole ' OT the result of a shared prefix and / suffix 7 > Entrance to SIMPLE LEXICAL REPETITION Flow chart NO or NOT APPLICABLE NO or NOT APPLICABLE ‘of one of the items include the referent is the ‘meaning of| the items the same? ie. is a paraphrase of one item a paraphrase possible to paraphrase one of the items 4o that it includes the other, or so that it shares an item with the paraphrase cof the othe items. share a common context, or are their contexts related in, some way does one of the items share a context swith an item that is in a repetition relation with the other do the items share ‘a common context, or are their contexts related in some way. ? \ “Answers from text analysis Entrance to. COMPLEX. LEXICAL. REPETITION Flow chart are the items paraphrased in such a way that a paraphrase of one of the words may contain the other, YES do the items share a common context, for are their contexts related COMPLEX LEXICAL. REPETITION does one of the items share context with an item that is ina repetition relation with the other exclusively) With com, With paras “Parately belo rts wo; Ww. whi, rk, let us consi We treated j td we ra ir Jected, e tem appears as Bears W cha: o Types of repetition sentence 2. Entering flow graphologically or Hoa I ME oR eee logically identical, so we move to the righe Li © to the right- hand side of the diagra : yar mmomelostemy, lal Raa ‘Are the items graphologic- Be velco the question STeUERY siMilaMiGMuseveeeNtie main Gee ern eAR ERR between the items solely the Be ris EES USN EERE ee ees explicable solely in terms ee aT ag difference between the items paradigm, and is there no ae on choices from a grammatical that difference?’ It is clear that aay available in English to express difference is entirely accounted for in graphological or phonological d T Y accounted for in terms of the singul iT paradigm. he two items, therefo: gular or plural repetition flow chart (Figure 33) ore ad the simple lexical HG aterhe ame ee oiicn me career cE naa Pe aanneneinenn Py Serer See be ‘no’; on the Pe eee ont ancestor tes ae le 2 that of the other (the the pair as an instance of See rencdee am eae E iomress as EEE with a more doubtful pair. Gives in 1 ‘ ntence 5 look, on the face of it, as if they might be an instance of repetition. We start, again, in Figure 3.2. Although the items are not graphologically or phonologically identical, they are similar, their similarity not lying in a shared affix, and the difference between them is accountable for in terms of verb form. This then means that they enter the simple lexical repetition flow chart. The first questions of Figure 3.3 that ask whether they have the same referent or whether the referent of one is included in the other do not apply, so the issue becomes one of whether the meaning of the two items jis the same, that is, whether a paraphrase of one is a paraphrase of the other. In this instance, the two items do not mean the same: in sentence 2, gives can be paraphrased as in sentence 5, giving is better paraphrased is forces us to consider whether it is possible ms so that it includes the other (or shares e of the other). In this case, there is no | the criteria for repetition and are of the pair drugging and drug in j milar route through the charts in this case, the previous pair. I 5 | The significance of repetition nets 5.1 The aim of this chapter » We showed inthe previous chapter how recording lexical cohesig ‘ and other forms of repetition in matrix form allowed us to ident Za Sentences that were bonded, that is, connected by an above-aver, level of repetition. Such sentences, we saw, were sometimes at some distance from each other, and formed networks of connections which wwe termed, and represented diagrammatically as, nets. The purpow of this and the next three chapters is to explore the properties of these nets. In this chapter, we consider the nets as complete entities, We note how the nets permit us to identify marginal and central Sentences in the passage we have been investigating and show how we can use them to create summaries of the content of the Passage; we also note how the nets allow us to identify topic-opening and topic-closing sentences, In subsequent chapters, we examine closely the nature of the bonds themselves, 5.2 The removal of certain features of cohesion with the same Proviso) is it relevant t0 other, non-adjacent, packages. Like- njunct may Spell out (one of) the 100 7 : he significance of repetition nets es in the passage features, since these interfere with the connections bens mete ee ections being made with sentences re in the text. Justification for making such ie given a Chapter 2, section 2.6, when we dine e rem unwanted cohesion from ti t = grtvies. Most of the modifications are macs of eanntee a reference of the; it is assumed that these are not error in themselves, even though the means whereby readers accurately identify what is being referred to are still not adequately understood. Where a reference needs to be supplied, it can usually be recovered from the immediate context. If an automatic procedure is adopted this step must, at our present state of knowledge, be omitted; as will shortly be seen, in most cases the modifications are not required. To show what the passage looks like once the local non-lexical cohesion is removed, a sample paragraph from 4.3 is given below in its modified form; the whole text with such cohesion removed is given in Appendix II: 5.1 17 What, then, is the advantage which we may hope to derive from a study of the political writers of the past? 18 A view prevalent in earlier ages would have provided a simple answer to [the] question [of what advantage we may hope to derive from a study of the political writers of the past]. 19 A work of politics, it would have been said [in earlier ages}, is the handbook of an art, the art of pores 20 i E gue: a i dge or skill in the art of carpentry may Soil a we his knowledge is made available to ese who aspire to be good carpenters, So a man of superior a fom i £ politics may set down his knowledge ina book for the ; © business it is to found, govern, or is> [2 the art of governing] is what no difficulty in determining pha ed from the study of great political works] will be consulted ern states. ‘Answers from text analysis P lectively for purpose Te mode ee eens: the condineaes no, aoa a I He sete by quoting the original; it win, ae See eS euiniiaa treo gtne bonds themselves, py; oe ESC rgatl nelCaeSle ne ees words in the text thar nd Pave bbe removed while ajaarelbratheg indicate words been added, (Thus, has no ie che Sr works] indicates a modification to oe etc ine in square brackets follows a pronoun or a « be cae brackets, the material in square brackets i aes Se alternative wording, not as supplementing the cohesive device if Ces nn of 5.1 might give rise to og aes of concern, Firstly, in sentence 21 there isa slight doubt as to what this refer This is not a common state of affairs in written text, though its frequency may be higher for the spoken word. The problem, if there be one, is, however, one that relates to an earlier Stage of analysis. namely, that of identifying the repetition devices of a Passage, |y creating a modified version of the text, we are doing no more than recording decisions of analysis taken at that earlier stage, Secondly, although for the most part the changes or additions are minor, occasionally treating a sentence as a self-contained Package of information produces surprises, Thus, it will be seen that sentence 1§ has all of sentence 17 inside it. Equally striking examples of this Phenomenon are sentences 15 (in 4.1) and 39 (in 4.3). Once the non- lexical cohesion is removed from sentence 15: 5-2 In this respect his position is like that of the critic of a work of art. it looks like this: 53° In Tespect [of bei iv matic in eing tentat; G suggesting lines of ng tentative and not dogmat €asoning and Conveying lights he has dy [of the writings which follow] and of i that have it Politica, Th he significance of repetition nets 54 Ifhe did he would certa up-to-date, and would n inly confine his the pages of Galen, With non-lexical cohesion removed, it ite ‘ads: 5.5) It AAI ‘ader of this book] w ne 2 a to provide the same kind political stu S ; peas nae ee his medical text-books provide for th mes ease newould certainly confine his attention to ie » and would no more think of learning his art from Aristotle than th i an the medics ? peetnceed ea ecametcal student thinks of learning his [expected the works of instruction for the In other words, the first cl . f F previous sentence with oe Siena ees Se eae All of this should nei j Sa ould neither surprise nor worry us. Such sentences only demonstrate the importance of repetition devices in making text cohere. They are, in any case, the exception rather than the rule in our chosen text, as a glance at Appendix II will confirm. It is worth ee that the adapted versions have played no part in establish- ing bonds. ‘A consequence of this is that the presentational choices a writer has made in producing the original version may become inappropri- ate in a shortened version, or when pairs of non-adjacent sentences are juxtaposed. We will, however, on occasion want to use these same kinds of cohesive devices to connect non-adjacent sentences; in particular, we shall want to make use of conjuncts such as however and nevertheless to highlight contrast relations between packages of information (in the manner described in Hoey 1983a). In so doing, we will be doing no more than making our own presentation of the packages of information for the purposes of illustrating clearly the relationship between the packages. Our objective will be to alter formation itself nor the relationship perceived bya ch, where there is more than one relation, the rably force one relation to the fore at the inserted conjunct indicates the loss o! ‘ot make reference t0 Its content. — e guid ny sah | Bg ns v “Answers from text analysis bondin 5.3 The net as record of oa a : 1 be able to investigate in so 1s we have created (which, c). It may be remembered ¢h combine to make a iin Chapter 2 we suggested that bibliographical mete? an academic’s @uvre might be gat importance of a work 10 ed to other works or was itself re Sa el ene refered to by other work that ei as marginal, while a work which was a}, seled to and which made plentiful reference to other referred to ane central. A work that subsequent works tefe characterize a have opened a topic, while a work that Ne eotee cae ded as having a summative back to earlier works was regar ic. It is productive to rounding off the development of a oo i‘ Pp. a the application of each of these aspects of the ma ky 2 to do this we need to interpret the nets as records of the onds sentence forms with the others in the passage. Using the two nets in combination, each sentence in the fo ; sentence passage can be assigned a two-figure co-ordinate, the { showing the number of previous sentences to which it is bonded, second showing the number of subsequent sentences to which bonded. Thus, sentence 20 in Figure 4.3 has a co-ordinate of since it is bonded to one earlier sentence (sentence 19) and t Subsequent sentences (sentences 21, 22, 23, 25, 26, 28, 34, 3 fu) If one or more of the bonds is doubtful in some respec figures may be provided, the i et We are now ina Po pee properties of the two (Figures 4.2 and 4 nce co-ordinates: 2 EDI) WY Bo pi a ae ae ay CG Ns aie Rea sre) So 4 ayaa 33 (50) 6 (10) JS (0,1) 25 (1°) aa ye OPT ope (64 8 Qi) WONA56)(A,5), 5°, (00) 2 (0,0) 18 (0,0) 28 (83) (7 10 (0.0) 19 (05) 39 {8:3) (7,3) 20 (1/9) 39 web) Table 5.1 0 (0,0) 104 The s ignificance of repetition nets 5.4 Marginal sentences The first thing one mii iB One might notice about Table 5.1 is th : 1 is that, as our bibliographical met: eee es ithe ‘aphor predicted, there is reaped tone lcted, there is grest variation between six or more other se bisrarerb RN eahae fences are i rences, while eight are bonded tc aaeaan a onded to none at all. The eight sentences th 1 s that manife jn that what they have st no bonds may be di i RINE ERIS Cine enna Bear are etter bull lexically upon chats Bane eine chins ches ary subsequent statements. This is, they ave frequently of rae aie Pere no purpose. As we shall see, in providing some necessary ancill in making a reader's task easier or Beeeacncaniait'be ose, a ary information without which the Meee esiacinatgict ona Nevertheless, we eet ee ey do not directly contribute to effect of their omission is on. ee ie els eee te Fee clysing e coherence of the passage we have The sentences P that form no .. ee ean all with other sentences, that 10, 13, 18, 27, 30, and 31. The beasties Cate eee cr entre 18 nd 27 aes , various. s 18 and 27 are ex: what have been termed signalling sentences (see, f oe 1983a) in that their main function is to sarees are t rlier sentet Jater sentences; in effect they function like sentence-length eee Bi: a expect such sentences to be marked out as aaa indeed, we would be i i : pect 1 b concerned if they were not, since they are effectively metalinguistic. The other sentences do not so readily identify themselves as _ marginal. With the benefit of hindsight, one can see that sentence 9 is exemplifying a point that is not further pursued and that sentence 10 rejecting a possible topic of discussion, but it is not clear that these sights could have been arrived at by other means. How then can we sure that we have identified anything real? One strategy that can be adopted is to consider how the text reads yut the ser we have singled out as marginal. With this in = from the extended passage We have been ive marginal sentences. Enough context o allow the reader to form his or her shortened text reflects the ‘Answers from text analysis 1: 10 is Sentences 5,9, — ingthe effect of omitting sentences 5,9 yy cons! er in by ; } We begin sthe book does not purport to Ps ay of Dali, SiO! degen cees The 0 notations interspersed to illustrate Mi theory, with q her a collection of texts, to which Tha tory. 3 It is rathe Jy a commentary. 4 T have treq! hae SE a of Aristotle, Aurel and th el ad ite a book abou ; ee sae the student, than to wr ee tea the Rccesible OLS eae eee iti the main oe for a further sendy of cise Worl ve subs 4 1 rtaKe it. nee - an but as an incentive to undert. these authors, assage has been quoted a gi A ae ae oo and sometimes Ee ee selected not because it was the most impo; eer peas one on which I had something to say, 71 re es tried to cover all the ground, and shal ec part if the reader is stimulated, by the samples w] ave offered, to complete a commentary of his own. 1, 8 The selection has been confined to a few authors, for reasons not only of space, or of limitations of my own knowledge (though either of these reasons would have been sufficient), but because it is part of the plan of the book to concentrate attention upon the most important works, 11 If the commentary Is secondary to the text, still more so must be any introductor y remarks which I make here . . 0 Itwill, [hope, be apparent th; has not disrupted the argui furnish Support for our ¢ A laim th; ; fo at these sentences are marginal. Po mite he though, is the fact that there is, at least f ae s : © coherence of ¢ i Z finds it likewise Coherent, then ° e eae aaa ur clai i a Sentences jg stil] more Strongly support ie Concerning thes Tmay seem Strange, co ment in this aes Mnsclentific, to, appeal rochereneh Posie As we commented i, S a n fact, no other form of the text, but Hee baD teri en eae 2 i esion is a pro} theteet Theewaeass ® aiprope ‘ 4 4 The evid, Nee for the ee phe teader's otra at a Particular Passage o 106 The significance of repetition nets of sentences makes sense can onl sen 4 ly be that readers i This is particularly the case because the proces Bee gbridgement are not well understood; there is no anes y ; there is no agreed asesning the adequacy of a summary on abridgment. woul S Bore eyes le Ke Present statistical support for alee a ee eae perce) tests, but if the reader does not find the Se era r coherent, it is doubtful whether he or she will be convinced by evidence that others did. The reader will be asked to = pee pence make sense and/or are representative of the much of this chapter and most of thi Chapter 9 we consider wha i ees t the implications are f i pee p re for the reading Sentence 18 Jf we omit sentence 18 from the paragraph hich it have the following: Se ee 5.7. ...-17 What, then, is the advantage which we may hope to derive from a study of the political writers of the past? 19 A work of politics, it would have been said, is the handbook of an art, the art of governing, 20 Just as a man of superior knowledge or skill in the art of carpentry may compile a work in which his knowledge is made available to those who aspire to be good carpenters, so a man of superior wisdom in the art of politics may set down his knowledge in a book for the instruction of those whose business it is to found, govern, or 21. If this is what political theory is there is no preserve states. difficulty in determining what advantage may be expected from the study of great political works. 22 They will be consulted for purposes of instruction by those who have to govern states . . - reaten the sense of the ‘o the question posed in in the missing time adjunct in earlier t clause of sentence d from the inse ; d as grammatically essential and ip for ellipsis to > the criterion we set UP ©” t 1 3.9: Other substitution Jinks). While 18, again, does not thi 19 providing an answer ¢ 4)9 hin sewers from text analysis i Ai me adjunct may Fender the conn, | ir does not stop them maki, 4 port for our iden omission of the t e a ces unnatur two senten ‘As before, then, Y unbonded sentences Sat ce 27 from the passage ise »f omitting senten , C18, re conception of Ee ics as an acher of it rests uy i political philosopher i be ee Dena vl it is imposst which it is im Shey would needitolbelaa masters in the art of governing, ands stacesmea apprentice themselves to them in order to learn ¢] [However] few political philosophers have themsel any mastery of the art of governing, and fe statesmen have owed their success to the study writings... The effect OSian25) This enti writers of polit As we noted above, sentence 27 is a signalling sentence, the f of which is to deny the truth of the previous claim and p ground for an alternative view of the truth. It is, the surprising that there is no disruption to the argument of when it is omitted. We have chosen to insert the conjunct howey mark the contrastive nature of the relationship between senten and 28 (a task originally performed by sentence 27), In dawes Ae doing no more than re-presenting the Packages of infor ti the light of the changed context in which they are appearing : Sentences 30 and 31 The effect of omit 4 following extract. ing sentences 30) and 31 can " se) j ati ‘ ; unique, 32 Of nit Which the Politician has t Tue that every cary ee ta ‘able is in a situation eve} ni iffero tet Carpenter y a The significance of repetition nets the statesman h. : atesman has to act— man I mean not only the profe on every citi 1 politici pees ty citizen. who has a share in the governmenteeinis cae Thy tiations ate unique ina more thoroughgoing See enchant Bot t0 be covered by a single, unchanging Beaanuiriecer i auiting only to be differently applied vould-be statesman were ceed in eli system from the w to succeed in eliciting a orks of a previous writer, i sora s it would inevitabl bea system applicable only to an age already past ... : If the comparison between the politician and the carpenter seems rather sudden, it should be remembered that it has been prepared for in sentence 20 (which can be found in 5.7 above), and is in part motivated by that sentence. 1 hope, however, that the majority of readers would concur that the paragraph makes sense without the marginal sentences as it did with them. Sentence 13 Irmay have been noticed that, in defiance of good rhetorical practice, Idid not discuss sentence 13 in its proper place. The reason for this is that its removal does not produce a satisfactory paragraph: 5.10 ... 12 In commending the writings which follow to the reader’s attention, I will indeed stake my credit on the assertion that the study of them will correct the judgment and enlighten the understanding upon matters in which it is important to be enlightened and correct. 14 The introducer may suggest lines of reasoning, he may try to convey certain lights which he has himself derived from the study, but in doing this he must be tentative and not dogmatic, and in the last resort he must say to the reader, “Go and read for yourself, and try whether this is confirmed by your experi- | ence.” 15 In this respect, his position is like that of the critic of a work of art. 16 However useful the critic’s remarks may be in preparing an approach to the work, they can never he reader | ¢ from the necessity of studying the work right, on the basis of this study, ‘standing in judgment on the jin the first place ..-

You might also like