1976. - BOOK  REVIEW:

M.A.K. Halliday and Ruqaiya Hasan 1976, Cohesion in English,
Longman, London, 374 pp. paperback £5.

Reviewed by C. George SANDULESCU, Stockholm.

(This book review was published in 1976 in the October issue of Working Papers in Discourse Analysis, Department of General Linguistics, University of Stockholm, as part of research conducted within the Discourse Analysis Project.)

1.            Cohesion is English can best be described as a bunch of compromises, which intelligently avoid the pitfalls of contradictions: it is a book putting across major points of theoretical and general linguistics through exclusive reference to the English language,in a way remotely reminiscent of what Chomsky and Halle were attempting to do in The Sound Pattern of English, -- a book which more or less started the newfangled modern tradition of making valuable linguistic generalizations via one language only.

            The other compromise is that it presents a new approach to a new issue, but at the same time aiming at a public which is meant to be as wide as possible (as Longman is very much in the habit of doing). In other words, the book is meant to be of use to the general linguist, to the applied linguist, to the average teacher of English as a first language, second language, or foreign language as well as to anybody dabbling in language theory at both amateur and professional levels. Accessibility is further increased by easily contextualizable discourse data largely extracted from one single book (Alice !); the authors themselves comment on their selection of data as follows:

                    (page 300) Consider the examples that have been cited throughout this
                    book. The vast majority of them have been drawn from either Alice 
                    in Wonderland
or made up. Why ? This is the only way to ensure that 
                    attention will be focused on the point at issue: either to use a text that is
                     so familiar that the reader will not pause over its interpretation, or to
                     construct examples that are so artificial that they avoid the problem.

            The last part of the last sentence may well be, at least in one of its possible interpretations, a well-disguised but devastating criticism of the way most adepts of a certain type of grammar are obtaining their data (for a discussion of the nature of linguistic data, cf. also SANDULESCU Stuttgart/August 1975).

            The book as a whole is placed within the clear theoretical tradition and frame of reference of HALLIDAY 1961 on the categories of grammar, a programmatic article which, in point of detail, still remains imperfectly understood by linguists in many parts of Europe. It further builds on HALLIDAY 1967 on transitivity and theme, and most certainly is an expansion and follow-up of HASAN 1968 on cohesion in spoken and written English;  finally, it is, to an indefinite extent, the outcome of mild co-operation with sociologists like Basil Bernstein (cf. footnote on page 288 on HASAN, forthcoming, about language in the imaginative context).

2.            HALLIDAY & HASAN 1976 is divided into eight chapters, primarily based on the meaning of different kinds of cohesion. As part of an introduction devoted to the important operation of providing correlational  definitions of fundamental concepts, mainly by relating text to texture (cf. RANSOME 1940, and the New Criticism), and text to discourse -- "the discourse comes to life as text" (page 299), (very 'scholarly & scientific' formulation, that !) --, the authors attempt to define cohesion on the basis of grammatical structure and linguistic system. (It is interesting to note here that the authors completely disregard the work of Widdowson, also originally from Edinburgh, who clearly distinguishes between cohesion and coherence.) To put it briefly, the authors advance the well-known view that cohesion is essentially a function of a set of categories, but what is new, interesting and perhaps important for the further advancement of theoretical linguistics is the 'componential' analysis of the notion. Thus, in the subsequent chapters, the authors discuss cohesion under five headings of reference (Chapter 2), substitution (Chapter 3), ellipsis (Chapter 4), conjunction (Chapter 5), and finally lexical cohesion (Chapter 6). The remaining two chapters function as separate conclusions to the book: Chapter 7, on the meaning of cohesion, takes up the points already made towards a definition in the Introduction, and processes them again in the light of the discussion provided by the five intervening chapters; the emphasis is again on texture (pages 295 ff), and the correlation with discourse (pages 326 ff).  The analysis of cohesion -- the last chapter of the book, and the most didactic and down-to-earth of all -- is written in clear textbook form; it provides rules of the thumb, called General Principles, and a very transparent coding scheme (where, unfortunately, overtaxonomization emerges at its worst) followed by seven sample texts. In the description of these texts, the theoretical construct of presupposed item chearly plays an overwhelmingly significant role in building up text cohesion.

            The bibliography at the end of the book is useful, but there are suspiciously too many relevant items which are not quoted, and many of the entries are only of doubtful relevance to the topic; but  this is quite another matter equally related to the philosphy of science, the sociology of linguists, and KUHN's "who's quoting who, and when". From among the entries that are either too conventional, too 'elementary', or downright irrelevant to a modern analysis of discourse, we mention CURME 1931, FRIES 1940, HARRIS 1963, JESPERSEN 1909-1949, KRUISINGA 1931, STRANG 1968, ULLMANN 1964, ZANDVOORT 3rd ed. 1965, etc.

3.            The great merit of HALLIDAY & HASAN 1976 is that it provides the first 350-page full-length discussion of a discourse phenomenon -- that of cohesion -- which has so far (with the possible exceptions of HASAN 1968 and GUTOWSKI 1974) received only very restricted attention (and that only in articles ten or twenty pages long).

            As a general notion, text is the basic semantic unit of linguistic interaction; the type of presupposition which provides texture in text (cf. also SANDULESCU Åbo/Turku November 1975), and which the authors are calling  cohesion, can extend over very long sequences of language, achieveing cohesive ties across very long stretches of text. Texts are, paradoxically, relatively free from constraints of time and depend much more on contextual relevance. Unfortunately, the authors occasionally resort to vague, fuzzy, and circular statements characterizing much semiological research in France:

                    (page 299)    What creates text is the TEXTUAL, or text-forming,
                     component of the linguistic system, of which cohesion is one part.

4.            From one  point of view, reference is either endophoric or exophoric, with the endophoric variety further subdivided into anaphoric and cataphoric. From another point of view,  there are three types of reference -- personal (through the category of person), demonstrative (on a scale of proximity, e. g. this 'near me', that 'not near me', yon 'not near either of us' (page 305)), and, finally, comparative reference,  by means of identity or similarity. This taxonomy leads to the treatment of most pronouns under reference, and not under substitution or ellipsis. Certain functions of it are discussed in terms of either extended endophoric reference (page 52), or generalized exophoric reference, as is the case with it is snowing, which is treated as a universal meteorological operator, essentially reflecting institutionalized exophora. Definite article the is, too traditionally perhaps, part of demonstrative reference. An interesting type of reference, however, which may be considered as one of the major new and important contributions of the book, is the treatment of the grammatical category of comparison, which is a great source of trouble  in all text processing, within a frame of identical vs. similar reference. It is at this point perhaps that Halliday's distinction between discourse structure and information structure becomes most obvious. Both general and particular comparison (e. g. (2.75a)(I was expecting someone different.) vs. (2.83)(We are demanding higher living standards.) provide an intelligent blending of a new approach with a highly traditional one, which turns the issue into an item of immediate and great interest to many, but of doubtful theoretical permanence.

5.            Passing on to substitution, and invariably operating on the surface level only (Halliday & Hasan never come anywhere near a deep structure or a consistently viewed semantic interpretation), the important point is made that whereas reference operates exclusively on the semantic level, substitution (including ellipsis) operates on the grammatical level alone. Reference is basically interpreted as a non-verbal relation (though including most pronominal systems !), whereas substitution is viewed as a purely verbal relation, essentially confined to the text. 'Exophoric substitution is fairly rare' (page 90). There are three types of substitution -- nominal, verbal, and clausal (one, ones, same;    do;    so, not). But there is a borderline where substitution shades into lexical cohesion, involving the use of general nouns. In discussing substitution, the authors bring into the focus of attention the important theoretical construct of  repudiation, or repudiated information:

                    (page 93)    In any anaphoric text, something is carried over from a
                    previous instance. What is carried over may be the whole of what there
                    was, or it may be only a part of it; and if it is only a part of it, then the
                    remainder, that which is not carried over, has to be repudiated.

                As has already been pointed out, the book excels in the simplicity and clarity with which difficult fundamental concepts are explained. Illustrating the above definition, we get

                    (e. g. (3.10) (We have no coal fires; only wood ones.))

where fires is carried over anaphorically, but coal is repudiated. No research to date has managed to list all the far reaching consequences of repudiation in monitoring discourse in the two distinct situations of reception and production.

6.            Discussing substitution, Halliday deviates from language description as provided by, say, modern symbolic logic in an interesting way: not is a clausal substitute, nut a unary propositional connective (at least in one of its functions). This is again an important statement, as unary logical connectives have, in our opinion, an extremely precarious linguistic status in their capacity of 'connectives'. And it is quite refreshing to see Halliday going against all trends in modern symbolic logic simply in order to achieve perfect adequacy of linguistic description. A typical illustration is

                    ((3.100) ( - Has everyone gone home ?        - I hope not.))

In such examples, the authors seem to be acutely aware of the category of participant boundary (as discussed in Sandulescu, Oslo / April 1975), always graphemically representing it by a dash, though never feeling the need of substantiating it theoretically to any extent. Now and then, one comes across vague and insufficiently accurate definitions; here is one of them:

                    (page 133)    The word not can be interpreted as the 'portmanteau'
                    realization of the substitute and negative polarity.

(Any Barbara Hall-Partee would consider such a formulation as pertaining to the Age of Edward Lear during his days in San Remo . . . ) But the main question is: does it cover the whole domain of not, as a definition ?

            Concluding the chapter on substitution, the authors distinguish three types or contexts of clausal substitution: report, condition, and modality. But the final impression, particularly at this point, is that of an artificial attempt at incorporating all systemic data in an exhaustive way into a discourse model. It is perhaps in this section that Halliday's confusion between language and discourse reaches its highest point, and grammar is thus assigned far more onus, or load,  than it can ever carry. The consideration of discourse as a completely separate level of investigation is not given a single chance.

7.            Ellipsis is closely related to substitution in that it is one particular case of it -- "ellipsis is simply 'substitution by zero' " (page 142). This treatment of ellipsis, very much along the traditional-conventional lines, is highly favoured by an exclusively surface approach to discourse phenomena, with all the disadvantages deriving from it as in, e. g.

                    ((4:2) (Joan brought some carnations, and Catherine some sweet peas.))


                    ((4:3) (Would you like to hear another verse ? I know twelve more.))

There are instances of nominal ellipsis as is the case with ellipsis within the nominal group (never ever called 'Noun Phrase' !) or even presupposition of nominal elements; thre is also verbal ellipsis, i. e. ellipsis within the verbal group, and lexical ellipsis, as in --

                    ((4:57) (It may or it may not.))

A very interesting type is that of operator ellipsis, e. g.

                    ((4:64) (Some were laughing and others crying.)

or even better --

                    ((4:67)( - What have you been doing ?    - Being chased by a bull.))

            Presupposition of verbal group systems is discussed in terms of (1) Polarity, (2) Finiteness and Modality, (3) Voice, (4) Tense, where the attempt is again obvious to include the whole of grammar into an allegedly comprehensive and consistent discourse model.

            Somewhat similar taxonomies are then being applied to clausal ellipsis, which may be modal and propositional. An interesting discussion is provided by ellipsis in question-and-answer sequences (pages 206 ff), and indirect responses (pages 212 ff). But it is indeed surprising that modern research into pragmatics as well as recent ethnomethodology-based studies of conversation and dialogue find no place in this book (apart from a passing mention on page 327). Lack of emphasis on genuinely substantiated pragmatics (sic!) factors, and a somewhat slipshod treatment of the issue of semantic interpretation, simply under 'meaning', give a more than traditional touch, and a purely 'amateur' imprint to what is in the last analysis just another set of taxonomies (in a situation where more 'originality of theory' was being expected ...). But the book is indeed in the mainstream of the so very British 'teach the teachers of English-as-a-world-language' attitude (rather than furthering the debate on theory).

8.            One whole chapter is entirely devoted to the highly important question of conjunction in discourse. The positive point about it is that the authors discard both the 'traditional-conventional' and the 'symbolic-logic' approach to conjoining, and discuss the issue solely from the point of view of adequacy of linguistic description. (The critique of this chapter is made on the basis of Sandulescu, Helsinki / October 1976, where the matter was discussed in far greater detail.) Without specifying it, the authors distinguish clearly between conventional and non-conventional conjoiners, and attempt to assign conjoiner status to a restricted range of non-conventional items, such as next, actually, therefore, besides, in addition etc (page 231).

            The authors admit that the issue of conjoining is somewhat blurred --

                    (page 232) by the indeterminacy, or perhaps flexibility, of our
                    punctuation system [;] the sentence itself in a very indeterminate

            After making the important, but highly controversial, point that "cohesion is a relation between sentences, not a relation within the sentence", the authors go on to state that --

                    (page 233) a new sentence starts whenever thee is no structural (sic !)
                    connection with what has gone before . . . [ and ] . . . a conjunction
                    occurs in first position and has the whole sentence as its domain.

            But the actual correlation between hypermorphemes and conventional or non-conventional conjoiner is left unanalysed. It is simply not true that --

                    (page 238)  the conjunctive relations are not logical but textual.

            It is in fact much more correct to formulate it as a Principle and, expanding their domain, say that

                    conjunctive relations are not ONLY logical but ALSO textual.

Monitoring discourse, or as Halliday & Hasan are putting it "the time sequence in the speaker's organization of his discourse" (page 239) is a complex process, which has only been partly investigated by various disciplines such as psychology, communication theory, interactional analysis, and symbolic (or formal) logic. By way of digression it must be said that we tend to forget that the syllogism was the first type of discourse ever to be investigated properly (in point of Information Structure), and that happened as early as Aristotle. Well then, these widely disparate approaches have never been integrated into a unified and unitary model. It is this ambitious task that discourse research is undertaking in several countries at the present moment. But the authors of the book under review fall short of that goal. The table of conjunctive relations on pages 242-243 is just another traditional taxonomy, enlarged and modernized it is true, but still evincing all the pitfalls of overtaxonomization. After all, and to put it crudely, is there much point in updating Kruisinga, and Zandvoort, and Jespersen ?

9.            In discussing temporal relations (page 261), there are rudimentary attempts at providing a Feature Analysis, but how can one possibly ever distinguish between structural  relations and cohesive relations (page 264) in such a dogmatic way ? 'Continuatives' is an interesting category, though a 'ragbag' one, and the cohesive function of intonation is dismissed far too quickly. So, what we are left with in the end is just another ragbag -- that of the lexical cohesion of Chapter 6, which should make interesting reading both in point of data and of interpretation to all those having an interest in what is currently called 'semantic networks'. For it is at this stage that semantic networks, in the very narrow sense of the word, come into the picture, and are viewed both syntagmatically and paradigmatically. But it is unfortunate that the level of generalization is so low and many of the token data fringe the trivial.

10.            Reading the chapter devoted to collocation (pages 264 ff), one again gets the feeling that slight modifications in the definition of fundamental theoretical constructs can easily lead to substantial modifications in the overall angle of vision. This is very much the case with collocation, which, in spite of its strong syntagmatic suggestion, is here taken to mean potential (vs. actualized) semantic paradigms, which presuppose but never fully linearize, co-occurrence, e. g. series of synonyms (climb, ascend), series of antonyms (boy, girl; child), closed lists (days of the week, months of the year etc) etc. It is true that such items do form cohesive chains in a text (page 286), but in addition to that there are the paradigmatic 'chains' which evince considerable pragmatic variability, i. e. from  one individual to another, and largely account for the sometimes astounding discrepancy in the quantity of information per partial sequence of text when monitoring discourse in reception; we know full well that a doctor 'sees' more in the (same) medical text than a nurse does, and the nurse, if she is bright, may 'see' more than the patient indeed does.  The authors disregard such pragmatics [sic !] aspects completely, as do most linguists. The text selected for illustrating  collocation (pages 286 ff) for instance, has been chosen expressly in order to minimize such variability in potential range of internalized paradigms. Lexical cohesion, as the authors amply show, is indeed based on reiteration and collocation. That is, without doubt, a good beginning, but one is left wondering whether it is in any way half the bottle. 

11.            The last but one chapter, devoted to the meaning of cohesion, discusses variability of tension (tight / loose) at the level of semantic texture; the sketchy diagrams, however, are merely hypothetical, and would require a great deal more explanation before they begin to make real sense. The notion of Imaginary Texture is most attractive, and makes one think of Voltaire himself, misleadingly presenting the first edition of Candide to his readers as a text "traduit de l'allemand", meant to suggest certain imaginary expectations in point of texture. 

            Though solving but few theoretical problems in any way adequately, the authors go on and ask theoretical questions in the discussion of data such as --

                    ((7:13) (Spurs played Liverpool. They beat them.))

How do we actually know who 'actually' beat who ?

                    ((7:15) (The cops chased the robbers. They eluded them.)) 

                    ((7:18) (These ponies the children were given by their 
                                        grandparents. Have you seen them ?))

            Such instances come very close to language material of the type --

                                The box is in the pen.

extensively discussed by BAR-HILLEL (1960 : 158 - 163). On the basis of such material it is quite clear that any general procedure capable of achieving the context resolution of ambiguity would in such instances in particular have to have access to an unlimited amount of factual information. This open-endedness in text processing, which often accounts for the difficulty of the subdiscipline, is artificially restricted by Halliday & Hasan in the discussion of the above examples to the issue of co-interpretation.

12.            Finally, in addition to accepting the idea that a text is in one way or another structured on the basis of a set of factors, such as texture, cohesion, (un)marked focus,  etc (the absence of such structuring would lead to "a non-  or pseudo-  text" !), the also advance the concepts of information structure (page 317) and discourse (page 327).

            There is no attempt anywhere in the book of correlating these three major types of structuring beyond the sentence and of providing a differentiated description of them. It has been well known for several years, ever since the ETIC Symposium on Language for Special Purposes, Halliday has been promoting the idea of information structure, without ever taking the real trouble of minutely explaining what he really means, and what the far-reaching consequences of such a construct would be, if it were to be clearly dissociated from language.

13.            The conclusion of the book gives a set of general principles stating that a tie is a directional relation including both the linearized elements and what is presupposed by them; such ties may be either immediate, or mediated, or remote. And the distance between at least certain ties can be counted in number of sentences. But can it really ? Particularly if we are prone to abandon the narrowly 'surface' or 'linear-manifestation' approach that the authors themselves have adopted . . .

14.           As far as our conclusion regarding the present book review is concerned, the book constitutes an excellent presentation of problems, a set of lucid analyses of discourse data, accompanied by an overtaxonomized attempt at providing an explicit description. This the authors have in common with SINCLAIR & COULTHARD 1975. But then there is the more regrettable attempt -- which Sinclair was very much aware of, and carefully avoided -- namely, that of incorporating as much of the discourse data as possible into the lower levels of grammar and lexis, where Halliday certainly feels more at home. 

            But  the fact remains that discourse is a separate level of investigation requiring specific methodology and its own theoretical constructs, and the authors have in no way managed to disprove that principle. On the contary, they seem to have reinforced it.

15.            The very great merit of the book is that it takes up and discusses quite a few of the so far taboo problems of the linguistic discipline, and this is done in a very accessible way for a wide public, and in simple language, with clear definitions, constantly emphasizing the practical consequences. There are, it is true, quite a number of problems which are swept under the carpet -- such as, (a) the real correlation between linear manifestation and semantic interpretation at the level of what we have called hypermorphemes (which is indeed the real discourse level), (b) how conjunction relates to recent research in modern symbolic logic (cf. the treatment of not, as a substitute rather than unary connective), (c) the status of pragmatics (sic !) factors at least in a discourse of the type (This item of furniture can be bought in China but not abroad.),  which contributes considerably to placing the speaker in space, and last but not least, (d) the real significance, if any, of topic/comment research within a consistent model of  "semiotic interaction" (cf. page 320).

            The book does not solve many problems nor does it put across any definite model for the description of discourse, but when all is said and done in point of explanation, it still remains an impressive edifice to look at on the long and tortuous Jakobsonian path of adequate Fragestellung in linguistics.