1976. Austin - Mapping Discourse Structure.

By C. George SANDULESCU, Stockholm.

(Paper given by C. George SANDULESCU at the Third International Conference of Nordic and General Linguistics, held at the University of Texas at Austin between 5 and 9 April 1976, and published in The Nordic Languages and Modern Linguistics 3,  Proceedings of the Third International Conference of Nordic and General Linguistics, edited by John Weinstock, pages 497-502.)

                                            [Full Text]

1. Introduction. The present research into discourse arises from the following set of dissatisfactions: widespread discontent regarding the descriptive adequacy of the initial symbol S of any transformational-generative grammar; substantial objections to the basic competence/performance dichotomy (cf. also Linell 1976); inherent inconsistency about the theoretical status of the constructs of context and style (which should ultimately be treated as features of discourse structure); excessive fuzziness of the notion of linguistic intuition. Summarizing all this one begins to wonder what is left of the major ideological premises and fundamental assumptions of present-day linguistic theory.

2. Angle of Approach. There are three options open in mapping discourse structure: (a) the model here proposed for the description of discourse should be ascribed no psychological reality; (b) the model should indeed evince psychological reality (in the sense adopted by Teleman 1976); (c) the model proposed for describing discourse should necessarily evince communicative reality, which is clearly superior to the more restricted (b) condition. A model such as the one triggered by the (c) condition must essentially be endowed with the following four components: syntactic, semantic, pragmatic, and axiological. Furthermore, the model must cope with two fundamental types of information -- homogeneous as well as heterogeneous (e.g., verbal versus non-verbal).

3. Incipient Outline. The fundamental constructs of the model are (a) the hypermorpheme, and (b) the abstract connector. The underlying cardinal assumption is that the sentence has a greater perceptual validity than the word. As such, it is a sentence-oriented, rather than a word-oriented model. In other words, it takes the clause-as-sentence -- i.e., the hypermorpheme -- to be the minimal rather than the maximal unit of description: hence, the 'hypermorpheme' label appended to this category (cf. Sandulescu 1976a). Sustained emphasis on the communicative reality of the model requires constant interplay between linearization and non-linearization: information conveyed via non-linearized structure of discourse may also carry assertions. The fundamental paradigmatic relation between hypermorphemes is the paraphrase relation: paraphrasing is conditioned by discourse structure via appropriateness in discourse. The discourse/text relation is a type/token distinction -- hence it becomes quite natural to speak of 'the text of a discourse'. The abstract connector is the major device in the process of linking hypermorphemes  across any types of boundaries.

4. Binary Structure of Abstract Connector.  The overt matrix Ko of the abstract connector K forms a closed system in sharp contradistinction to the covert matrix Kc of the same connector, which forms an open set. In simpler words, the overt part of it is only 'the tip of the iceberg'. But the Ko/Kc correlation in prospective structures (it must be emphasized that the picture in retrospective structures is quite different) functions as a 'data bank' for the actual linearization of the S-S' axis of manifestation. The open-ended Sz propositional content of Kc is contrasted with the conventionally featurizable content features of discourse structure'). There are two fundamental relations between the abstract connector and the linear manifestation:  (a) one-to-many; (b) many-to-one. The former is usually and conventionally handled under ambiguity, whereas the latter is essentially reducible to the paraphrase relation. The balance between Ko and Kc is the actual trigger of alternative realizations of one and the same semantic interpretation. The overt matrices contain, among others, indicators of discourse head compatibility, markers of deixis (time, place, and person), and diagrammed accounts of the semantic substance of conventional and non-conventional conjoiners.

        Any hypermorpheme,  taken in isolation, evinces (a) asserted meaning, and (b) presupposed meaning. Any given set of linearized hypermorphemes evinces in addition 'connective' meaning, derived from hypermorphemic juxtaposition, and reflected in either or both of the twin matrices of abstract connectors. However, it must be pointed out that these abstract connectors are no mere inventories of realizations; they should rather be viewed as ranges of potentials -- some being realized overtly, others covertly, and still others not being realized at all. But the model is pragmatics-oriented: the direct outcome of this is that production matrices never coincide completely with reception matrices. This is indeed one of the major sources of misreading, deliberate or unintentional miscomprehension, semi-intelligibility, etc.

9. Towards a Theory of Discourse Heads. It is here suggested that we do not speak in terms of words and morphemes at all (or whatever learned terms you may use to call them !), but rather in terms of discourse heads (which ultimately are highly sophisticated key-words-and-phrases if you so wish...). A celebrated illustration of this is the so typical Mr Jingle speech, which upon very close linguistic scrutiny is not so jingling as it may seem at first sight. By collocating ideas + sleep one manages, for instance, to foreground the acute incompatibility of discourse heads, though the tagging of green and furiously largely blurs the endocentric / exocentric distinction, and reduces communicative reality to near zero. Colourless (sic!) is only intended to increase the oxymoronic mess... (Quite reminiscent all  that of Rudolf Carnap's Piroten karulieren elatisch ! -- in fact, as close as ever Chomsky's 'performance' can  hope to come to Saussure's 'parole' !). 

        A more relevant instance of discourse head incompatibility leading to deviant discourse (cf. Sandulescu 1975 : 302) is provided by the following:

John opened the door and organic chemistry was required of all students, because if seventeen were not a prime, then the dairy association would have to back down. Thus, we can conclude that television will have a beneficial effect upon the nation's young.

        Discourse heads like John and the door are said to be compatible, whereas organic chemistry, a prime, the dairy association, and television are not compatible within this particular discourse structure (though each clause taken separately is perfectly well-formed -- in the sense given the term by TG grammars). The overall impact is otherwise quite similar to the sleep/green/furious/idea/colourless kind of argument. Casual spoken discourse is characterized by a completely different type of discourse head compatibility than non-casual written discourse: the latter evinces far tighter zonal constraints than the former. It is this very principle that accounts for the rather jerky and random nature of one, and the more sustained and relatively smoother nature of the other. Casual communication does jump from pets to politics, non-casual communication does not. 

        In point of linguistic materialization, the discourse heads are very often Noun Phrases, but in narrative discourse of the 'telegraphic' kind in particular (e.g., Veni, vidi, vici), it is the Verb Phrases that take up the discourse head function, with severe sequentiality constraints imposed upon them. A fast reader does take in text in terms of discourse heads; the major operation that he performs is that of pertinent discourse-head identification on the basis of (a) textual assertive structure, (b) textual presuppositional patterning,and (c) previous exposure to similar discourse. A well-reconstructed retrospective structure will tell the experienced reader a great deal about the corresponding prospective structure in point of discourse heads. 

        Discourse heads are communicative units which are hierarchically organized, and consequently they do indeed structure discourse. It is the discourse-head sequentialization  that ultimately determines syntactic structure, particularly from a communicative point of view, rather than vice versa. A theory of discourse heads must take into account  both conventional paradigms of the type of 'referential hierarchies', which are in the process of being described by Pike (1975), and non-conventional paradigms, which are for the most part the nonce juxtapositions normally occurring in casual spoken discourse, where a process of idiosyncratic value ascription is taking place. 

10. Conclusions. 

        1. In addition to laying the foundations for a possible explicit description and formalization of discourse which should go far beyond the highly restrictive boundaries of sentential linguistics, the model cursorily sketched here also provides an explanation of language function via discourse structure.

        2. The model leaves open the question as to what the actual relation is between discourse structure and information structure: the nearest point of contact and basis for an analogy between them are provided by a theory of discourse heads, coupled with a network of referential hierarchies, perhaps as derived from Pike.

        3. It has become a commonplace to state that text as a construct has no theoretical status: the present resarch is aimed at proving, however, that discourse is a fundamental descriptive category, on a par -- in point of importance -- with the initial symbol S of modern conventional linguistics. The whole field of research is in fact triggered into existence by the unsatisfactory descriptive adequacy of the initial symbol S.

        4. Sentential word order is determined by discourse structure and not vice versa. The reason is simple: sentential word order is ultimately triggered from within the abstract connector. In consequence, it is not at all a mere feature of sentence structure, as conventional linguistics vehemently propounds.

        5. The construction of an explicit model for the description of language segments larger than the sentence with data derived from a great number of widely different languages is evidence not only in support of universals, but also -- and more significantly -- in support of differentials of discourse.

        6. The communicative bias emphasizes the inter-personal (i.e., micro-socio) rather than the intra-personal (i.e., psychological) parameters; hence, the stress on prospective and retrospective structuring as well as on the discourse heads.

        7. Based on a close interrelationship between hypermorphemes and abstract connectors, the model provides a globalistic and wholistic  approach to language description, rather than an atomistic one. Assigning the hypermorpheme the status of minimal descriptive unit, the point is forcefully being made that the 'clause-as-sentence' has greater perceptual validity than the lexical item.



Eliasson, Stig.     1976. 'Theoretical problems in Scandinavian contrastive phonology.' (in this volume).

Grimes, Joseph.    1975. The thread of discourse. (Janua linguarum, minor 207). The Hague: Mouton.

Hale, Austin.    1973. Clause, sentence, and discourse patterns. 4 vols. Norman: Summer Institute of Linguistics. University of Oklahoma.

Linell, Per.    1976. 'Current perspectives: the contribution of Adolf Noreen'. (in this volume).

Longacre, Robert.    1972. Hierarchy and universality of discourse constituents. Washington, D.C.: Georgetown University Press, 2 vols.

Longacre, Robert.    1976. Discourse. (mimeo; forthcoming in print.)

Lowe, Ivan.    1969. 'An algebraic theory of English pronominal reference.' Semiotica I (4) : 397 - 421.

Pike, Kenneth.    1967. Language in relation to a unified theory of the structure of human behavior. The Hague: Mouton. 

Pike, Kenneth.    1975. Multidimensional feature arrays from affix to discourse. (mimeo). 

Rommetveit, Ragnar.    1974. On message structure. London: John Wiley & Sons.

Sandulescu, C. George.    1975. 'Displacement constraints on discourse.' In: Papers from the Second Scandinavian Conference of Linguistics, ed. by Even Hovdhaugen. Oslo: Lysebu, pages 282 - 312.

Sandulescu, C. George.    1976. 'Presupposition, Assertion, and Discourse Structure.' In: Reports on Text Linguistics: approaches to word order. ed. by Nils Erik Enkvist and Viljo Kohonen. Meddelanden från Stiftelsens för Åbo Akademi Forskningsinstitut. No. 8, pages 197 - 214.

Sandulescu, C. George.    1976. 'Structuring discourse connectors.' (forthcoming). 

Sinclair,  John.    1975. Discourse in relation to language structure and semiotics. Wenner-Gren Foundation for Anthropological Research (mimeo).

Teleman, Ulf.    1976. 'How to be concrete in linguistics.' (in this volume).

Trail, Ronald.    1973. Patterns in clause, sentence, and discourse. 4 vols. Norman: Summer Institute of Linguistics, University of Oklahoma.

Wise, Mary Ruth & Ivan Lowe.    1972. 'Permutation groups in discourse.' In: Language and Linguistics: Working Papers 4. pages 12 - 34. Washington, D.C.: Georgetown University.