http://beaugrande.bizland.com/register.htm
In Mohsen Ghadessy (ed.), Register Analysis: Theory and Practice. London: Pinter, 1993, 7-25.
'Register' in Discourse Studies: A Concept in Search of a Theory
Robert de Beaugrande
1. The early heritage
Throughout much of linguistic theory and method, the concept of 'register' has led a rather shadowy existence. The term itself is not used at all in early foundational works, such as those of linguists like Saussure (1916), Sapir (1921), and Bloomfield (1933), nor do we find there any term we might classify as roughly equivalent. In such works, the lack is not too surprising. When linguistic theory is declared to be mainly concerned with abstract systems, as envisioned by Saussure, or with taxonomies of minimal units, as envisioned by Bloomfield, 'register' would be likely to seem a troublesome or even disruptive concept. It implies that the valence of systems or minimal units might not be established in the language as a whole but in some subdomain or constellation of contexts. A 'register' is certainly not a language unit, and is hard to define as a system of such units comparable, say, to the 'system' of 'phonemes' of a language, or to its 'system' of noun declensions or verb conjugations, and so on. Thus, a concept like 'register' would have contravened the early aspirations of linguistic theorists to make statements and set up schemes of the highest possible generality and abstraction.
However, we would expect to find the term, or some rough equivalent for it, in foundational linguistic works where the interest in discourse was quite pronounced, such as the collection-volumes by Pike (1967 [1954-60]) and J.R. Firth (1930, 1937, 1957 [1934-51], 1968 [1952-59]). Such linguists emphasized that they did not share the 'theoretical' commitments of their more conventional (or 'mainstream') colleagues, and that the major motive for this 'heresy' was a vital concern for actual speech and discourse and hence a mistrust of the drive toward abstraction.
In Pike's work, a possible equivalent for 'register' is 'the universe of discourse', which he considered able to 'condition' the 'meaning' even of his fundamental unit, 'the morpheme' (1967: 599). Such a thesis followed from his characteristic argument that 'the meaning of one unit in part constitutes' and 'is constituted of the meaning of a neighboring unit'; and that 'meaning' is 'one contrastive component of the entire complex' (1967: 609, 148f, 430).
Thus, Pike's interest in discourse domains or 'universes' reflects his awareness of the dependence of meaning on context. He suggested that units have a 'central meaning' with 'greater frequency' among 'the community' than 'marginal meanings', but 'special universes of discourse' can alter this proportion (1967: 601). Therefore, we might try to 'find a statistically' measurable 'set of common contexts', or set up 'a hierarchy of universes of discourse with progressive degrees of centrality' (1967: 600, 602). However, we might find 'no specific number of distributional orbits, or degree of remoteness from the central' (1967: 604). Pike's criteria for central versus marginal actually give prominence to the latter: 'the outer' 'orbits carry the greater communication energy' for 'hearer impact' (1967: 604). Major examples of the outer 'dependent or derived meanings' are 'idiomatic' meaning not 'predictable' from 'the meanings of its parts', and 'metaphorical meaning', along with 'poetry', 'puns', and 'slang' (1967: 601ff). Notice here that the marginal surpasses the central in ways reminiscent of the 'foregrounding' described by the Prague structuralists. Hence, what makes a domain of meaning or discourse special is the kind and degree of response and attention it receives -- in my view, an outlook we should still keep in mind in our search for a conception of 'register' today.
In Firth's work, a possible equivalent of 'register' might be the 'restricted language', which he defined as 'serving a circumscribed field of experience or action' and having 'its own grammar and dictionary' (1957: 124, 87, 98, 105f, 112). The emphasis here was on practical method. Such a domain is easier to manage than 'when the linguist' must draw 'abstractions' from 'a whole linguistic universe' comprising 'many specialized languages' and 'different styles' (1968: 30, 97, 118). 'The material is clearly defined: the linguist knows what is on his agenda, and can 'set up ad hoc structures and systems' for 'the field of application' (1968, 106, 116). In fact, once 'the statement of structures and systems provides' 'the anatomy and physiology of the texts', it is unnecessary' 'to attempt a structural and systemic account of a language as a whole' (1968: 200).
As domains of 'restricted languages', Firth looked to 'science, technology, politics, commerce', 'industry', 'sport', 'mathematics', and 'meteorology', or to 'a particular form or genre', or to a 'type of work associated with a single author or a type of speech function with its appropriate style' or 'tempo' (1968: 106, 98, 112, 118f). To counter the possibly divisive effects of such an outlook, Firth seemed to favor a compromise of sorts: 'linguistics' can regard each 'person' 'as being in command of a constellation of restricted languages, satellite languages' (compare Pike's 'orbits'), but these are 'governed' by 'the general language of the community' (1968: 207f).
Also possibly relevant for the concept of register is Firth's prominent notion of 'collocations': he suggested 'studying key words, pivotal words, leading words, by presenting them in the company they usually keep' (1968: 106ff, 113, 182). This 'study' may range between 'general or usual collocations and more restricted technical or personal' ones, between 'normal' and 'idiosyncratic' ones (1957: 195; 1968: 18). At times, the specific end seems to herald a profusion of varieties: 'characteristic distributions in collocability' can constitute 'a level of meaning in describing the English' of a 'social group or even one person' (1968: 195).
'The study of the usual collocations' resembles that of 'restricted languages' by making 'a precisely stated contribution' to 'the spectrum of descriptive linguistics' and by 'circumscribing the field for further research', e.g., by 'indicating problems in grammar' or aiding 'descriptive lexicography 'with 'citations' for 'dictionary definitions' (1957: 195; 1968: 180f, 196). We should state 'first the structure of appropriate contexts of situation', 'then the syntactical structure of the texts', and 'then' 'the criteria of distribution and collocation' (1968:19). Yet Firth repeatedly warned that 'collocation is not to be interpreted as context' (1968: 180, 1957: 195); apparently, he wanted 'collocation' to remain at a more abstract systemic level than that of text and discourse.
2. Propagation by Halliday: dilemmas of linguistics and semantics
It was a pupil of Firth's, Michael Halliday, who, along with his associates, eventually gave currency to the term 'register' as such. According to Halliday, 'the term' 'was first used' for 'text variety by [Thomas Bertram] Reid (1956); the concept was taken up and developed by Jean Ure (Ure and Ellis 1972)', and by Halliday et al. (1964)' (Halliday 1978: 110). Another source was the work of Basil Bernstein, who used the term 'variant' instead (cf. Bernstein [ed] 1973).
In Halliday's view, 'the notion of register is at once very simple and very powerful' and 'provides a means of investigating the linguistic foundations of everyday social interaction from an angle that is complementary to the ethnomethodological one' (1978: 31, 62). 'The theory of register' 'attempts to uncover the general principles which govern' the ways 'the language we speak or write varies according to the type of situation' (1978: 32). 'But surprisingly little is yet known about the nature of the variation involved, largely because of the difficulty of identifying the controlling factors' (ibid.).
Though he sees a possible parallel between the notion of what 'the member of a culture typically associates' and Dell Hymes' notion of '"communicative competence"', Halliday evades the latter notion as an 'artificial concept' which 'merely adds an extra level of psychological interpretation to what can be explained more simply in direct sociolinguistic or functional terms' (1978: 32). Halliday's uneasiness about 'psychological interpretation' (as stressed also in his discussion with Parret, cf. Halliday 1978: 38f), presumably influenced by Firth's similar attitude, creates predictable problems for any conception as complex as 'register', where the 'psychological interpretation' of language users is so essential and where agreement is harder to obtain than about other aspects of a language (see sections 3 and 4).
The central problem is how 'the "register" concept' can 'take account of the processes which link the features of the text' 'to the abstract categories of the speech situation' (1978: 62). The 'original' approach was to define 'register' directly in 'lexicogrammatical terms' (Halliday 1978: 111). For example, Jean Ure (1971) proposed a connection between 'lexical density and register differentiation', where the 'density' was measured by 'the proportion of lexical items (content words) to words as a whole' (1978: 32). Such work was typical of the classify-and-count methods that understandably dominated much of linguistics during the absence of more elaborate theories and methods of discourse.
Halliday warns against 'posing the question the wrong way': '"what features of language are determined by register?"' (1978: 32). Nor would it be fully adequate to ask 'what peculiarities of vocabulary, grammar, and pronunciation can be accounted for by reference to situation?' (ibid.). Instead, the really pressing question is 'which kinds of situational factor determine which kinds of selection in the linguistic system?' (ibid.). Stating the question this way is a major step forward, but makes answering it particularly difficult in that we now have, as it were, unknowns on both side of the equation, i.e., both for the situation and for the language.
Another approach to 'register' was to circumscribe it by comparing and contrasting it with 'dialect'. 'Dialect' was defined 'according to user', and 'register' 'according to the use' (Halliday 1978: 110). Also, 'dialects' 'differ in phonetics, phonology, and lexicogrammar, but not in semantics'; 'registers' 'differ in semantics and hence in lexicogrammar and sometimes phonology, as a realization of this' (1978: 35, cf. p .67). These definitions signal an important dualism in Halliday's work: 'lexicogrammar' differs both for 'dialect' and 'register', but for 'register', 'semantics' is interposed as the controlling factor. We thus encounter such formulations as these: 'register' is 'the clustering of semantic features according to situation type'; or 'a register can be defined as a configuration of semantic resources that the member of a culture typically associates with a situation type' (1978: 68, 111, 123). These formulations do not, I think, necessarily 'point in the direction of a functional semantics' following 'the theories of the Prague school', though Halliday says they do (1978: 63). The role of cultural situations was not prominently worked out by the Prague group, but is only indirectly implicit in their idea of 'functional sentence perspective' (cf. Beaugrande 1991).
Hallidayan argument, on the other hand, insists that 'the features of the text' should be 'considered as the realization of semantic patterns', and so this holds for 'register' as well (cf. 1978: 62). In his Introduction to Functional Grammar, he says: 'the relation of grammar to semantics' is 'natural, not arbitrary, and both are purely abstract systems of coding'; so 'there is no clear line between' them, and 'functional grammar' is 'pushed in the direction of semantics' (1985: xix, xvii). In Explorations in the Function of Language, he says: 'in principle, a grammatical system is as abstract (is as "semantic") as possible given only that it can generate integrated structures', i.e., 'its output can be expressed in terms of functions mapped directly onto others' to yield 'a single structural "shape"' that is 'multiply labeled' (Halliday 1973: 95). The equation of 'semantic' with 'abstract' occurred again when Halliday called for 'register' to be given 'a more abstract definition in semantic terms', rather than 'lexicogrammatical' ones (1978: 111).
Halliday's equation has noteworthy implications. Just as the history of 'general' linguistics has been marked by disputes over how 'semantic' the approach should be, the history of semantics has been riddled with controversies over how 'abstract' the approach should be. Semantics was often dominated by positivist and behaviorist proposals to make it more concrete by tying it to 'real states of affairs', 'features' of 'objects' (like chairs), or 'observed behaviors' (cf. survey in Beaugrande 1988).
Halliday's views are more complex and subtle. He suggests that 'semantic systems' 'relate' to 'grammatical systems' through the 'pre-selection' of 'options' (1973: 98). 'In some instances' can we go from 'semantics' 'directly to the "formal items": to the actual words, phrases, and clauses of the language', with 'no need' for 'grammatical systems and structures' (1973: 83ff). But this 'happens only' with 'a closed set of options in a clearly circumscribed social context', e.g., 'a greeting system in middle-class British English' or in a 'closed transaction such as buying a train or bus ticket' (1973: 83f). In genuine 'language, such systems are marginal', 'a small fraction of the total phenomena' among 'much more open' and 'general settings'. Due to 'indeterminacy between the strata', we find not 'one-to-one correspondences' between 'grammar', 'semantics' and 'phonology', but rather 'neutralization and diversification' -- 'many-to-many' (1973: 82, 93, 56f).
Nonetheless, Halliday retains a conventional provision when he stipulates the 'principle' 'that all categories employed must be clearly "there" in the grammar of the language', 'not set up simply to label differences in meaning' (1985: xx). Without some 'lexicogrammatical reflex', such 'differences' are not 'systemically distinct in the grammar' (ibid.). However firmly 'based on meaning', 'a functional grammar' is 'an interpretation of linguistic forms': 'every distinction' -- 'every set of options, or "system"' -- must 'make some contribution to the form of the wording', i.e. of the 'sequence' or 'syntagm' of lexical' and 'grammatical items' (1985: xx, xvii). Such provisions suggest that Halliday too is a bit worried about semantics getting overly 'abstract'. But instead of looking to reality or real objects and behaviors, he sees 'grammar' as the anchor to hold semantics down.
At stake here is a crucial issue in the emergence of 'modern linguistics'. Traditional grammarians had drawn their distinctions on the basis of the formal organization of their own language, or their own dialect of it. Moreover, as Bloomfield complained, 'a good deal of what passes for "logic" or "metaphysics is merely an incompetent restating of the chief categories of the philosopher's language' (1933: 270) (a practice which the philosophers associated with Chomsky's 'grammar' were later to demonstrate all too clearly).
What makes modern linguistics distinctive, I think, is the willingness to recognize distinctions as long as they are formally made in any language'. This factor suggests why the study of Amerindian and Afro-Asian languages provided such a major impetus: they made it possible and necessary to recognize whole new types of forms and of formal organization. For example, Bloomfield (1933: 175f) contrasts the 'sentence types' of English with those of Menomini and names 'surprise' and 'disappointment' as 'types of sentence' (1933: 175f). Here, a virulent 'anti-mentalist' felt justified in introducing psychological states into his 'grammar' because he saw formal markers for them.
Chomsky and his group, on the other hand, were avowedly mentalist, but were far less inclined than Bloomfield toward the kind of form-function compromise we just saw in describing Menomini. Instead, they wanted the formal aspects to be foregrounded in principled independence from semantic ones, as well as from psychological states. Since Chomsky's original approach was designed only from English, a formally sparse and frugal language, he was obliged to invent an explosion of 'underlying' forms and structures to nail down the kind of distinctions needed to attain 'formality'. In consequence, he became entrained in a steadily narrowing spiral of underlying formality until language as such disappeared, taking with it much of what Bloomfieldian and Sapirian linguistics would have been willing to admit under 'grammar', but what now appeared to be at best 'surface structure' -- and of course 'register' (or any similar concept) could have no place either.
As we can readily see, it is not Halliday's predicament that is new, but the solution he favours: to have semantics and grammar linked at every step. The problem of how to keep the two domains linked must of course be raised specifically for every language. Significantly, Halliday's own Introduction is entirely and explicitly constructed on English, as were many of his earlier works, de facto at least, though he started out as a Chinese linguist. However, he left open the prospect that his 'functional grammar' might be a 'general' one, for which he happened to be 'using English as the language of illustration' (1985: xxxiv). A similar approach might work for other languages, and appropriately enough, a 'functional grammar' for Chinese has recently been devised (Li and Thompson 1981), much in Halliday's spirit even if he is not cited there. The otherwise mysterious systems of particles in Chinese become much more tractable when we consider their functions for indicating the status of the message in context (cf. also Beaugrande and Mingliang 1989).
Halliday breaks down the 'register' by saying it is 'predicted' or even 'determined' by 'the categories of field, tenor, and mode' (1978: 62, 125). According to one formulation citing John Pearce (in Doughty et al. 1972: 185f), 'field refers to the institutional setting in which a piece of language occurs'; tenor' 'refers to the relationship between participants'; and 'mode refers to the channel of communication adopted' (1978: 33). In Halliday's own diagram, however, 'field' is the 'type of social action', 'tenor' is the 'role relationships', and 'mode' is the 'symbolic organization' (1978: 35). Later, we read that the three concepts are 'related respectively to the ideational, interpersonal and textual components of the semantic system' (1978: 125). 'Mode' gets special consideration: it is 'the rhetorical channel with its associated strategies'; it is 'reflected in linguistic patterns', but 'has its origin in the social structure'; and 'the social structure' 'generates the semiotic tensions and rhetorical styles and genres that express them' (cf. Barthes 1970) (1978: 113). Also, 'mode covers roughly Hymes' channel, key, and genre' (1978: 62).
Once more, we have difficulties determining exactly what the status and designations of Halliday's terms may be. 'Field, tenor, and mode' were evidently proposed as categories for describing situations rather than language per se, but the inconsistencies, especially in regard to 'mode', reflect the usual perplexities of making 'social' categories correspond language forms ('linguistic patterns'). Notice also that 'stylistic' and 'rhetorical' parameters are introduced, but their valence with respect to 'register' is not clarified. The 'social' categories are naturally far broader, and I doubt that we can insist, as Halliday did for 'semantics' (see above), that 'every distinction' must 'make some contribution to the form of the wording' in terms of lexical' and 'grammatical items'. In fact, Halliday's claim is a bit weaker, but still quite demanding: 'field, tenor, and mode' can 'make explicit the means whereby the observer can derive' the 'systematic norms governing the particulars of the text' (1978: 62). Thus, while 'deriving the situation from the text', 'the participant' or 'the observer' 'can 'supply the relevant information that is lacking' (ibid.).
3. Recent trends
In terms of prospects for further work, we have two opposed options. The first option is to widen the scope by examining a variety of languages in terms of 'those features' of 'functional grammar' that are 'explicitly claimed as universal', notably the 'hypothesis' that the three 'metafunctions' -- 'the textual', 'the ideational', and 'the interpersonal' -- 'organize' 'the content systems' 'in all languages' (Halliday 1985: xxxiv). This option would be helpful if these 'metafunctions' are indeed 'related' to the three categories of 'register' (see section 2), but as far as I can discover, this option has not formed the major part of Hallidayan research. Perhaps one reason for this hesitancy is that earlier work on 'universals' was typically naive and premature. In fact, if we read Chomsky's Aspects closely, we may suspect the real attraction of 'universals' lay in the argument that they 'need not be stated in the grammar' of individual 'languages' but 'only in general linguistic theory as part of definition of the notion "human language"' (compare Chomsky 1965: 6, 35f, 112, 117, 144, 168, 225). So they were in effect one more dumping grounds, alongside 'surface structure', 'performance', and so on, for putting aside messy or intractable issues, notably, -- again, Aspects reveals this clearly -- 'semantics', which was handed over to the 'universals' pretty much wholesale (see Chomsky 1965: 160).
Moreover, the first option leads away from the concreteness that characterizes the Hallidayan approach and that motivated the notion of 'register' in the first place. To maintain that any one particular 'register' is 'universal' would strain the audacity of even the most hand-waving linguists, since 'register' is by its very definition firmly embedded in cultural situations. The overall fact that registers differ might be a universal, but to say so is simply to argue that 'register' is a generally justified concept. That cultural situations fall into different types is hardly open to serious dispute: these types are what culture is all about. Nor would many people deny that language and culture influence each other, though some might say (and have said) this influence is not the concern of linguistics.
Hence, it is the second option which Hallidayan linguistics would be likely to pursue and has in fact done. This is to retain a narrower scope by developing 'a grammar' for the 'analysis' and 'interpretation of texts of a broad variety of registers in modern English' (Halliday 1985: x, xv, xx). By his own reckoning, Halliday's 'account' has already served both 'practical' and 'theoretical' purposes', such as probing 'the relation between language and 'culture'; 'comparing registers or functional varieties of English'; 'studying socialization' and 'functional variation'; and 'analysing text, spoken and written', notably 'spontaneous conversation' (1985: xv, xviii, xxx).
My impression is that in most of this work the practical has run well ahead of the theoretical -- just the converse of formal (Chomskyan) grammar. A decisive case here is precisely the concept of 'register'. Its practical value is beyond dispute, but Halliday's own central theoretical book so far backs away from it in a somewhat disappointing way: the Introduction to Functional Grammar does not 'go into questions of register structure', which 'we are only beginning to be able to characterize' (1985: 290, xxxv). Halliday is content to remark in passing that 'register' is a key domain for examining how 'elements', 'configurations', 'collocations', and 'the patterning of clause themes throughout a text' may 'vary'; how 'a text' might 'deploy the resource of cohesion'; and how to give an 'account of English semantics' (1985: 318, 313, 315, 372, ix). He assumes that 'a speaker of the language 'knows' 'how likely a particular word or group or phrase is' 'in any given register'; but the 'treatment of probabilities' is also 'outside the scope' of the 'grammar' (1985: xxii; cf. Halliday 1973: 114). We thus cannot evaluate his view that 'registers select and foreground different options, but do not normally have a special grammar'; yet 'some registers do', such as 'newspaper headlines' (1985: 372; cf. 1985: 373-77). The reference to 'narrative, transactional, expository', and so on (1985: 372, ix, 318, i.a.) is not very illuminating, since these are not 'registers', but modes which may vary widely within one register as well as from one register to another.
The 'restricted languages' proposed by Firth (section 1) were construed by Halliday as 'extreme cases' of 'register' (1978: 35). Even so, Halliday acknowledges that they make up 'much of the speech' of 'daily life' in 'contexts where the options are limited and the meaning potential' 'closely specifiable' and 'explainable' (1973: 25ff). Exploring them might 'throw light on certain features in the internal organization of language' (1973: 27). Halliday lists 'games', 'greetings', 'musical scores', 'weather reports', 'recipes', 'cabled messages', and so on, along with 'routines of the working day' like 'buying and selling' (1973: 25f, 63). 'The language is not restricted as a whole' in such domains and 'the transactional meanings are not closed', but 'definable patterns' and 'options' do 'come into play', e.g., for 'beginning and ending' a 'conversation on the telephone' (1973: 26).
In discourse analysis and text linguistics, such issues have been typically treated as a matter of 'types' of discourse or text. The discourse analysts like Longacre and Grimes were chiefly interested in conducting fieldwork and realized that, however strong the allegiance of 'mainstream' linguistics might be to the abstract, 'isolated' sentences, the data on otherwise little-studied languages had to be extracted from discourse, and there were no a priori grounds for telling how general or specific any body of data might be unless its relation to discourse types was taken into consideration. In the work on 24 Philippine languages of Luzon, Mindanao, and Palawan, Longacre and 32 colleagues from the Summer Institute of Linguistics identified 'discourse structures' in types they called 'narrative', 'procedural', 'expository', 'hortatory', and 'explanatory' (Longacre et al. 1970). These notions were set up because the group found evidence that language structures correlated with discourse types, and that the correlation was clearly significant, not merely for the data analysis, but also for the discourse participants themselves. The point was therefore not so much to offer a 'universal' or complete typology of discourse, but to show that at least some types can be reliably identified in groups of languages and cultures. Much the same point can be made for the studies by the Grimes group (see Grimes, 1975, 1978), where the main focus fell on 'narratives'.
In its early stages at least, 'text linguistics' hardly engaged in the kind of fieldwork the discourse analysts were doing. Most work was done closer to home, mainly on English, German, French, Czech, and Russian, and proceeded by the usual methods of grammatical analysis originally developed for sentences, with minor modifications. Since linguistics were accustomed to setting up schemes of types, the same principle was readily extended to texts, especially when the text was seen merely as a 'unit' or 'level above the sentence' (e.g. Heger 1976). Traditionally, a main attraction of the sentence was the ease grammarians had sorting it into clear-cut types --- declarative, interrogative, imperative, and exclamatory. The recognition that such classifying might be quite a different matter for texts had to wait upon the hesitant realization that texts were fundamentally different entities from sentences (Beaugrande 1980, 1997)
A conference on text types ('Textsorten') had been held in Germany in 1972 to find 'differentiating criteria from a linguistic point of view' (proceedings in Guelich and Raible [eds.] 1972). On the whole, the results of conference were meagre: mainly just a general realization that the question was far more complex than prevailing 'linguistic points of view' were equipped to handle. If we follow through the arguments brought forward by Pike, Longacre, and Grimes, then linguistic abstractions, such as units, features, and structures, are at least as much a product of discourse or text types as the other way around. Linguistics -- especially the 'generative' kind -- has been a bit glib and naive in jumping from data sets to the language as a whole and skipping over the types as theoretically unimportant (or, what is more to the point, unmanageable).
One text linguist, Siegfried Schmidt (1978), contemplated two methods. Either we start with the intuitively given types and try to build a theory or model for them; or else we set up the theory or model and the try to deduce, construct, or reconstruct the types from there. Today, when the weakness and diminishing returns of deductive linguistics have been generally realized, the consensus runs clearly in favor of the first approach, the more inductive one. We pick one or more types that seem to be given in social practice and attempt to systematize some salient characteristics. The most active field here today is of course 'language for special purposes' (LSP), which has become one of the most conspicuous and successful areas of text linguistic and discourse analysis, as well as of applied linguistics (cf. Hoffmann 1987).
Even so, neither text linguistics nor LSP claims to have more than a very rough and ready classification of text or discourse types. Few people in those fields today hold high hopes that a rigorous typology will appear soon, and fewer still would insist that research cannot proceed without one. On the contrary, we have more than enough work to do if we are to systematize readily accessible text types. And in general, linguists today are less convinced that the goal of linguistics is the construction of abstract typologies of any kind, nor the creation of abstract systems of formal features or rules divorced from social or practical application. The surviving, unregenerate formalists have turned to computer programming, but even there, special-purpose domains have become a center of attention, witness the fact that expert knowledge systems are by far the most dominant concern in 'artificial intelligence' in recent years.
A further trend of decisive importance has emerged in 'discourse processing'. This research undertakes to show not merely that certain types can be found to correlate with language structure, but that these structures are relevant for the cognitive and communicative processes people actually perform in discourse. An eminent case is the 'strategic' model of van Dijk and Kintsch, which cites 'register' alongside 'style, 'text type, and communicative context' as factors affecting the 'selection of appropriate lexical items to express the concepts of the propositions' (1983: 292). As a central notion, van Dijk and Kintsch postulate 'superstructures': 'typical schemata' for conventional text forms', which 'consist of conventional categories, often hierarchically organized', 'assign further structures' and 'overall organization to discourse', and 'facilitate generating, remembering, and reproducing' (1983: 16, 54, 57, 92, 104f, 189, 222, 236f, 242, 245, 275, 308, 336, 343). These 'superstructures are not merely theoretical constructs of linguistic or rhetorical models' but also 'feature in cognitive models' as 'relevant' 'units' (1983: 237). 'During comprehension', they are 'strategically' 'assigned on basis of textual' 'information, i.e. bottom-up', yet also create 'assumptions about the canonical structure' and applicable 'schema', i.e. 'top-down' (1983: 237, 105). The 'superstructures provide the overall form of a discourse and may be made explicit' as 'categories defining' the 'type' (1983: 189, 235f). They are 'acquired during socialization' with 'discourse types'; 'language users know' the 'categories' and 'schemata' 'implicitly' or even 'explicitly' and 'make hypotheses' about them 'when we read' (1983: 57, 92).
Although here too, the details are not worked out for any large spectrum of types, the findings do indicate that we should enrich the mainly sociological approach proposed by Halliday and his associates with a psychological one, which (as we saw in section 2) he rejected. The compromise will be unavoidable because the social manifestations and situation types by themselves cannot provide all the criteria we need, and because the criteria they do provide require interpretation. This interpretation cannot be left up to the linguists or analyst alone, but must be traced in terms of the operations people in general carry out when processing discourse types.
Admittedly, the cognitive and psychological methods are not likely to produce any comprehensive formal typology. Experimenters are content if they can isolate at least some crucial differences between pairs of fairly uncontested types, such as narrative and expository (see also Freedle and Hale 1979). The project of isolating all the types and stipulating all the processes that may and may not apply to each is certainly quite remote and may remain impractable for the foreseeable future.
4. Future prospects
The trends I have outlined indicate that the concept of 'register' still needs to be reassessed. To begin with, we should grant what has generally been conceded for the types addressed discourse analysis, text linguistics, and discourse processing: that we are dealing with phenomena that cut across the usual schemes of 'levels' or components' and that involves far more than 'purely linguistic' factors. A register can at most be an open system, not a closed one or even a tidy one in which (to paraphrase a Saussurian formulation about language) 'everything holds everything else in place' (un systeme ou tout se tient). When we select a register to investigate, we must not expect or demand that we should list all the aspects its must have, and still less all the aspects it must not have. Instead, we must be content to postulate a register when a representative group of language users agrees that certain aspects are typical and predictable. The occurrence of a non-typical aspect does not undermine the register or suddenly transpose the discourse into a different register, but it will be likely to attract notice and to elicit some response.
By this line of argument, a 'register' is essentially a set of beliefs. attitudes, or expectations about what is or is not likely to seem appropriate and be selected in certain kinds of contexts. This explication may seem vague, but the phenomenon itself is inherently fuzzy. Like the 'superstructures' posited by van Dijk and Kintsch, registers are 'acquired during socialization' with 'discourse types' (cf. van Dijk and Kintsch 1983: 57). Registers 'may be made explicit' as 'categories defining' the 'type' (cf. 1983: 189, 235f, 92), but they usually are not; and in practice people often command registers that they could not describe very well, let alone justify by means of a theory. For example, virtually everyone knows that certain registers they use with family and close friends will not do in situations calling for written examinations or official documents; but how people know this and exactly how they carry it into practice is a predominantly indeterminate and intuitive matter.
Nonetheless, future work must include having everyday speakers describe the registers they know and the ways they use them. This kind of fieldwork will not provide all the data we need to understand how registers arise or how they are put to work. But we urgently need a more general perspective than any one linguist or school of linguists can bring to bear by attacking the problem among themselves.
Obviously, we need to clarify the social implications of registers. Solidarity is one important factor: the usage of a particular register preferred among a group with which one wishes to be identified. Conversely, dominance is important when a register is deployed to signal that one speaker or group has the right to assert priority over those who do not command the register, or at least do not command it as well.
We can therefore postulate a gradient between insider and outsider functions of registers in use. In this sense, command over a wide range of registers is a major implement of social power, and command of only a few is a typical drawback among the disadvantaged. In view of this fact, the lack of explicit attention to matters of register, especially in the educational system, is a effectual contributor to the maintenance of social inequality.
In schooling, the issue of register is usually treated on a purely negative basis. Learners are alerted when they have committed a violation of register, but are given fairly little systematic assistance in developing or diversifying their range of registers. This neglect is all the more grievous in that the entry to specialized fields of knowledge, particularly to prestigious ones like science and technology, depends materially on commanding the appropriate register. Yet even institutions that explicitly recognize the importance of the issue by building up programs like 'writing across the curriculum' seldom offer courses of study in, say, 'the register of physics', or 'the discourse of computer science'. Only the learners whose social background has already provided them with a wide command of registers are well-equipped to succeed, while the others tend to fall further behind than ever.
Science and technology also provide useful illustrations of domains wherein each register is associated with a corpus of prestigious or authoritative texts. The acquisition and skilled use of the strategies for the register decides who will be admitted to the domain in terms of who is authorized to contribute to that corpus as a profession. Yet authorities are understandably reluctant to acknowledge how far their status depends not just on 'knowing the facts' or carrying out research or design, but on producing and using texts about facts and on reporting or discussing research or design (a consideration Firth was fond of raising). Perhaps too, the authorities lack an explicit awareness of their own textual strategies and fail to appreciate (or to sympathize with) the problems confronting the outsider or the initiate.
On the other hand, we also need to clarify the linguistic implications of registers. It remains to be seen whether a theoretical framework can be found and developed that could subsume and situate already established practices, and if so, along what lines. This question can be broken down into several, and even then, answers are difficult to come by.
One major question is whether every instance of language in discourse belongs to some register. Few linguists would want to commit themselves to such a strong assertion, because the generality and 'abstractness' they claim for their theories and models would be endangered. Besides, it would no longer be admissible to present samples, say of 'English sentences' unless we also identified what register they belong to. In this regard, Halliday would be probably no different from anyone else working in 'general linguistics'. If the Introduction is a reliable indicator, he wants his 'functional grammar' to extend across all kinds of registers.
The alternative question would be whether only certain instances of language should be considered specific to some register. Here, we immediately confront the formidable problem of finding the criteria for telling which instances are and are not of such a nature. Halliday's statements are not utterly clear regarding this problem. If he tells us a 'register' is 'a configuration of semantic resources that the member of a culture typically associates with a situation type' (1978: 123), he seems to invoking some 'psychological interpretation' of the type he elsewhere repudiated (see section 2). If, on the other hand, he depicts a 'register' as 'a cluster of associated features having a greater-than-random (or rather, greater than predicted by their unconditioned probabilities) tendency to co-occur' (1988: 162), he seems to raise the prospect of deciding the matter by means of statistics. But he says the 'treatment of probabilities' is 'outside the scope' of his 'grammar', and he points out that 'the probability of such terms occurring in the discourse is also dependent on what' the speakers 'are doing at the time' (1985: xxii, 1978: 33). So we cannot expect to get the issue under control merely by statistical counts, such as Ure's 'lexical density' method (see section 2), because total frequencies may not be relevant for what seems likely or expected in concrete discourse situations, and may not yield the criteria we need to identify a register.
A more viable approach to 'the notion of register' would be to correlate the issue of 'probability' with some 'form of prediction' and to ask: 'what exactly do we need to know about the social context in order to make such predictions?' (Halliday 1978: 32). So far, however, Halliday's illustrations -- again quite typical of linguistics -- are commonsensical and intuitive, or merely assertive, and the 'we' in his 'we need to know' does not include the general public or does so only by implication. For example, he says: 'by and large, "scientific English" is a recognizable category, and any speaker of English for whom it falls within the domain of experience knows it when he sees or hears it' (1988: 162). Yet this 'recognition' is precisely what we still need to establish: not merely that 'any speaker' with 'experience' can do so -- already a stronger claim than we have empirical evidence for at the moment --- but by what standards and criteria. Halliday bypasses the problem by taking samples he considers clear instances of the register and relating the development of clausal strategies to the trends in science itself, for instance, the trend from speculation to experimentation, when 'doing and thinking' were 'brought together' (1988: 175).
Halliday's warning, cited in section 2, stresses that we should make 'situational factors' our point of orientation and work from there toward 'selections in the linguistic system'. But the speakers of the language might well be working just the other way: by picking out incidental features, such as the 'peculiarities of vocabulary, grammar, and pronunciation' Halliday does not consider really decisive, and using them to set up hypotheses or conclusions about what the situation (or the register) is likely to be.
We thus need to make more allowance for distinctions among speakers in terms of their specialization. The way a trained analyst like Halliday recognizes a register is doubtless different from the way naive language users do, especially non-scientists or quasi-scientists, to say nothing of the decorative use of 'scientific language' for special purposes such as advertising. As a member of editorial boards, I frequently have to review submitted manuscripts for which the handling of the applicable register is a conspicuous problem. The most frequent tendency is to salute the register without really using it, that is, to introduce specialized terms in order to show that one is or aspires to be an insider, but then to either apply the terms in vague and obscure ways, or to leave them aside in the subsequent discussion. In addition, many writers seem to believe that the 'scientific register' requires things to be stated in the most complicated and difficult manner. So when I approve a manuscript, I often edit it heavily to enhance clarity and readability and to insist that special terms either be used appropriately or replaced with ordinary language.
How my interventions are received by the authors I usually don't find out, but the issue is doubtless sensitive, because the tendencies I rebuke may form part of the author's own self-image as 'a scientist'. The editors, however, seem to agree with me, since they keep sending me more manuscripts, and frequently the ones they think 'need some rewriting'. Sometimes I have completely revised a manuscript which editors had already accepted on the basis of the author's professional standing, but which they couldn't bear to print in such clumsy or opaque styles.
Though this evidence is merely anecdotal, it does suggest two points. The first point is that 'register' is evidently not just an issue beset by 'mixed or borderline cases' (Halliday 1988: 162), but a fundamentally indeterminate domain, directly subject to the current motives and aspirations of both potential and actual users, of both outsiders and insiders. I find a wide variation in what is evidently judged to belong to the 'scientific register' and its subdomains ('psychology', 'educational research', 'linguistics', etc), and in my view, a good deal of it is inappropriate for the goal of effective communication, as opposed to the goal, say, of laying claim to prestige or 'insidership' for the author.
The second point is that even established insiders do not agree about their own register, or do so at best in holistic and intuitive ways. Odd biases get carried over not from science, but from quirky handbooks on grammar and style, such as that the 'scientific register' disallows the use of the first person singular, and perhaps the second person as well. Imaginary 'rules' no professional linguist or grammarian seriously upholds appear suddenly in the judgments of editors and copy-editors for scientific publications. In my role as author, I have had many skirmishes over this.
And yet the 'scientific register' is typically picked out as a prize example and taken at face value much as we saw Halliday doing. If even that register is so disputatious, what about others? Is there such a register as 'unscientific English', as the prescriptive responses of copy-editors seem to imply? Or, is the 'scientific' actually a very loose agglomerate of registers, divided not merely for the various fields, such as those for physics, psychology, and so on, but in still finer detail for particular professional organizations, proceedings, journals, roundtables, and so on?
It is plainly time to re-open the case for 'register' in the broadest possible terms. This time, we should be keenly aware that demands for generality and abstractness involve strategic trade-offs, of which 'normal science' tends to focus only on the more favourable side. The abstraction sought across the board -- in psychology, linguistics, and philosophy, especially logic -- has always been a two-edged sword. The same contextual factors which control everyday processing and communication and thus make things simpler are typically viewed by researchers as additive and unmanageable complexities that make things more complicated. Therefore (to stay with the metaphor) the 'sword' that is wielded to clear the field cuts away the vital supports needed to keep the issue securely under control.
Halliday's 'functional grammar' thus already marks a great step forward by showing a richer range of factors whereby the organization of discourses, and of clauses in particular, is affected by the current status of the knowledge involved: its relative degrees of importance, newness, and topicality (or 'thematicity'). However, Halliday has hesitated to state whether or how these factors apply across different registers and whether or how far they are modified. The really thorny task of determining how this might be the case is only just beginning to be undertaken. As my illustrations for 'scientific English' indicated, the handling is still a bit too facile and speculative, and even the seemingly 'obvious' registers can be disturbingly indeterminate.
An elaborate combination of strategies will be needed to get a handle on the issues of 'register'. Undoubtedly, we need to consult in great depth and detail with people who, by some reasonably secure measure, are recognized as skilled users of a given register. These experts should act in several roles. First, they can be observed giving advice to their students or other initiates on register use. Second, they can report their own prior experiences as students or initiates trying to enter the domain, and especially on any problems encountered along the way. Third, they could respond to or rate the appropriateness of sample texts which are systematically varied along whatever lines are believed to be involved in the register, including lines they may themselves suggest. These three uses of register experts would do much to fill in the gap which now exists between the intuitive appeal or plausibility of the notion of register and the body of evidence needed to give the notion some socially documented substance.
Another tactic would be to gather data which might help to indicate what registers can be identified in the first place, and if so, by what means. I have been applying this tactic for some time by directing my own students on numerous small projects of this kind. In some cases, the students were themselves accredited users in the register, as for the projects on the discourse in pool-hall gambling or in a certain 'fraternity' house, or the social 'small-talk' among users of university VAX computer system ('VAXers'). Here, the students had the advantage of being familiar with using the register, but in no case did they have a very clear idea of the factors involved before the actual data-gathering began.
The controlling influence of the social situation was certainly as powerful as Halliday or Firth could wish. Pool-hall betting is illegal in Florida and therefore requires a register outsiders will not understand. The fraternity house had made a point of developing a register not even known to other fraternities, though more for motives of upholding exclusivity than shielding illegality. The dialogue among VAXers, who cannot see or hear each other, reflects a shift in social pressures, including the freedom to either reveal or conceal oneself without the usual worries about possible consequences.
The findings showed, as might be expected, a characteristic mix of special lexical items with specialized strategies covering whole discourses. In pool-betting, the turns of the betting negotiation -- to establish the type of game, the rules, the handicaps, and the amount of money at stake -- were found to be carried out in almost every recorded case, though specific words were not usually prescribed except for the game names, e.g. 'nineball' versus 'snooker', which themselves said a good deal about the kind of player and the amount of money likely to be involved. In the 'frat', special terms clustered around activities which in the American middle-class environment are not so much talked about, such as drinking alcohol (e.g. 'turbo-slam', a way to drink beer upside down), having casual sex (e.g. 'to bust'), and stealing (e.g. to 'schwartz' or to 'ninja'). The VAXers were most clearly characterized by their choices of special names and descriptions for themselves, borrowing heavily on science fiction and popular movies and television.
In other projects, the students were outsiders, as in the study of prison discourse by a student whose husband happened to be a sports director in a Florida prison. This method has the advantage that special factors of the register stand out by virtue of their seeming unfamiliarity. My student was 'shocked' at the extent of what she called 'obscene language', though how far this aspect was due to the hostility the inmates sensed in the situation and how far it might have reflected the dialects of the predominantly lower-class inmates could not be determined from the sample, and we did not use direct interviews to gather the data. Aside from special terms like 'e.o.s.' ('ee-oh-ess') instead of 'end of sentence', higher-level strategies included thematic attempts to impress other inmates with claims about what one was or did in the outside world, matched up with numerous blunt formula showing disbelief; and formulations for maintaining that one went to jail for some unjust accusation or mistake in the legal system, rather than for some real fault or crime of one's own.
Whether we have solid theoretical justification for maintaining that the data gathered in such fieldwork belong to or even constitute a 'register' can of course be debated. The same question must continually be raised for any research concerned with 'register', since society itself has no exact criteria for deciding what the necessary and sufficient conditions of a register must be. The vital criterion is how the participants in situation types view their discourse and how far they in fact adapt their general discourse strategies to fit the type, and both factors were evidently operative in the domains we studied.
This twofold strategy of working directly with actually accredited insiders on the one hand while doing indirect grassroots fieldwork among presumed insiders on the other hand can offset the trade-off I cited between abstraction versus control. The insiders can describe how they in practice fit their own discourse to a register and vice-versa. The fieldworkers can strive to notice any aspects that seem special or specific and hence potentially relevant for a register. I suspect that we will still have a good deal of indeterminacy in the data, somewhat along the lines of centers versus 'orbits', 'margins', or 'satellites' envisioned by linguists like Pike and Firth (section 1). And we will probably find more registers and more shadings within registers than we would like, especially variations which do not show up reliably in lexicon or grammar as Halliday stipulated for his approach (section 2).
In return, however, we stand to gain a firmer empirical base for treating linguistic data at large, and for deciding how wide our claims should be. We have, I fear, been much too eager to make wider claims than were really justified by the materials we had at hand. A more comprehensive engagement with the notion of register can materially brighten the future of language study not merely as a factory for turning out formalisms and sentence diagrams, but as a participant in broad social research and education programs.
Bibliography
Barthes, Roland (1970), 'L'ancienne rhétorique', Communications 16.
Beaugrande, Robert de (1980), Text, Discourse, and Process, Norwood, N.J., Ablex
______ (1988), 'Semantics and text meaning: Retrospects and prospects', Journal of Semantics 5, 89-121.
______ (1991). The heritage of functional sentence perspective from the standpoint of text linguistics. Linguistica Pragiensa 34/1-2, 1992, 2-26 and 55-86.
______ (1997) New Foundations for a Science of Text and Discourse. Stamford, CT: Ablex.
________ and Mingliang Hu (1989), 'Interference at the discourse level: Definiteness and topicality in English and Chinese', in Albrecht Neubert and Heide Schmidt (eds.), Interferenz in der Translation, Leipzig, Enzyklopädie-Verlag.
Bernstein, Basil (ed) (1973), Applied Studies Towards a Sociology of Language: Primary Socialization, Language, and Education, London, University College Research Centre.
Bloomfield, Leonard (1933), Language, New York, Holt.
Chomsky, Noam (1965), Aspects of the Theory of Syntax, Cambridge, MIT Press.
Dijk, Teun van, and Walter Kintsch (1983), Strategies of Discourse Comprehension, New York, Academic.
Doughty, Peter, Pearce, John, and Thorton, Geoffrey (1971), Language in Use, London, Arnold.
Firth, John Rupert (1930), Speech, London, Benn.
_______ (1937), Tongues of Men, London, Watts.
_______ (1957), Papers in Linguistics 1934-1951, London, Oxford.
_______ (1968), Selected Papers of J.R. Firth 1952-1959, London, Longmans.
Freedle, Roy, and Hale, Gordon (1979), 'Acquisition of comprehension schemata for expository prose by a transfer of a narrative schema', in Roy Freedle (ed) New Directions in Discourse Processing, Norwoood, Ablex, 121-136.
Ghadessy, Mohsen (ed) (1988), Registers of Written English, London, Pinter.
Grimes, Joseph (1975), The Thread of Discourse, The Hague, Mouton.
_______ (ed) (1978), Papers on Discourse, Arlington: Summer Institute of Linguistics.
Gülich, Elisabeth, and Raible, Wolfgang (eds) (1972), Textsorten, Frankfurt, Athenäum.
Halliday, Michael Alexander Kirkwood (1973), Explorations in the Function of Language, London, Arnold.
_______ (1978), Language as a Social Semiotic, London, Arnold.
_______ (1988), 'On the language of physical science', in Ghadessy (ed), 162-78.
________ et al. (1964), The Linguistic Sciences and Language Teaching, London, Longmans.
Hasan, R. (1973), 'Code, register, and social dialect', in Bernstein (ed).
Heger, Klaus (1976), Monem, Wort, Satz und Text, Tuebingen, Niemeyer.
Hoffmann, Lothar (1987), Kommunikationsmittel Fachsprache, Berlin, Akademie-Verlag.
Li, Charles, and Thompson, Sandra Annear (1981), Mandarin Chinese: A Functional Reference Grammar, Berkeley, University of California Press.
Longacre, Robert et al. (1970), Discourse, Paragraph, and Sentence Structure in Selected Philippine Languages, Santa Ana, Summer Institute of Linguistics.
Pike, Kenneth (1967), Language in Relation to a Unified Theory of the Structure of Human Behavior. The Hague: Mouton.
Reid, Thomas Bertram (1956), 'Linguistics, structuralism, philology', Archivum Linguisticum 8.
Sapir, Edward (1921), Language, New York, Harcourt Brace and World.
Saussure, Ferdinand de (1916), Cours de linguistique générale, Lausanne, Payot.
Schmidt, Siegfried (1978), 'Some problems of communicative text theories', in Wolfgang Dressler (ed), Trends in Text Linguistics, New York, De Gruyter, 47-60.
Ure Jean (1971), 'Lexical density and register discourse formation', in Perren, G.E., and Trim, J.L.M. (eds), Applications of Linguistics, Cambridge, CUP.
Ure, Jean, and Ellis, Jeffrey (1978), 'Register in descriptive linguistics and linguistic sociology', in Uribe-Villegas, Oscar (ed), Issues in Sociolinguistics, The Hague, Mouton.