Linguistic Dynamics: Discourses, Procedures and Evolution 9783110850949, 9783110101157

Citation preview

Linguistic Dynamics

Research in Text Theory Untersuchungen zur Texttheorie Editor Jänos S. Petöfl, Bielefeld Advisory Board Irena Bellert, Montreal Maria-Elisabeth Conte, Pavia Teun A. van Dijk, Amsterdam Wolfgang U. Dressler, Wien Peter Hartmann f , Konstant Robert E. Longacre, Dallas Roland Posner, Berlin Hannes Rieser, Bielefeld Volume 9

w DE

G

Walter de Gruyter • Berlin • New York 1985

Linguistic Dynamics Discourses, Procedures and Evolution Edited by Thomas T Ballmer

w DE

G

Walter de Gruyter • Berlin • New York 1985

Library of Congress Cataloging in Publication Data Main entry under title: Linguistic dynamics. (Research in text theory = Untersuchungen zur Texttheorie ; v. 9) Includes indexes. 1. Discourse analysis—Addresses, essays, lectures. 2. Mathematical linguistics—Addresses, essays, lectures. 3. Linguistic change—Addresses, essays, lectures. I. Ballmer, Thomas T II. Series: Research in text theory ; v. 9. P302.L54 1985 412 85-13164 ISBN 0-89925-099-8 (New York) ISBN 3-11-010115-7 (Berlin)

CIP-Kur^titelaufnähme der Deutschen Bibliothek

Linguistic dynamics : discourses, procedures and evolution / ed. by Thomas T Ballmer. — Berlin ; New York : de Gruyter, 1985. (Research in text theory ; Vol. 9) ISBN 3-11-010115-7 (Berlin) ISBN 0-89925-099-8 (New York) NE: Ballmer, Thomas [Hrsg.]; G T

Printed on acid free paper (pH 7, neutral) © Copyright 1985 by Walter de Gruyter & Co., Berlin 30. Printed in Germany Alle Rechte des Nachdrucks, der photomechanischen Wiedergabe, der Herstellung von Photokopien — auch auszugsweise — vorbehalten. Satz und Druck: Arthur Collignon GmbH, Berlin Bindearbeiten: Lüderitz & Bauer, Berlin

Contents Th. T Ballmer Introduction

Part I

1

Procedural Linguistics

T. A. van Dijk Strategic Discourse Comprehension

29

C. U. Habel Referential Nets as Knowledge Structures

62

H.-J. Eikmeyer, H. Rieser Procedural Grammar for a Fragment of Black English Discourse

Part II

...

85

Quantitative and Algebraic Dynamics

G. Altmann On the Dynamic Approach to Language

181

L. Hfebicek Text as a Unit and Co-References

190

M. Nowakowska An Algebraic Approach to Discourses and their Goals

199

Part III

Language Change

R. Keller Towards a Theory of Linguistic Change

211

B. Strecker Rules and the Dynamics of Language

238

H. J. Heringer Not by Nature nor by Intention: The Normative Power of Language Signs 251

VI

Part IV

Contents

Morphogenetic Linguistics

W. A. Koch Tension and Suspense: On the Biogenesis and the Semiogenesis of the Detective Novel, Soccer and Art 279 Th. T Ballmer The Psychology of Context Change

322

Name Index

357

Subject Index

361

Thomas T Ballmer, the editor of the present volume, was not able to complete his editorial work. He died under tragic circumstances. He was young, but his scholarly career within linguistic and on its borders was marked by creative power and productivity. Both these qualities are displayed characteristically in the present volume where new approaches and connections in the study of natural languages are presented. We are grateful for the impetus he gave to linguistic research and we will miss him as a colleague and a friend. J.S.P. Editor of the series

THOMAS T BALLMER

Introduction Is human language something static or something dynamic? At least during the first three quarters of this century this question has been decided in favor of a static view of natural language. This time span has been the playground of structuralism. It has been the structure of language which counted as the relevant object of investigation for a linguist and, similarly, for a philosopher of language. The challenge of linguistic sciences was the discovery, description and explanation of certain invariants, the rules of language. The regular and rule governed facts of language, the linguistic categories, and also the essence of language, namely the universals, were considered to be of major importance, in contrast to the accidentals which may change from time to time, place to place, situation to situation, and individual to individual. It is certainly a delicate matter to decide whether this textbook view of modern linguistics is entirely adequate and, if so, what the reasons are that the static conception of language got such a strong impact during the first seventy five years of this century. Lateron we shall have the opportunity to draw a more differentiated picture of the situation. For the time being we should, however, pose the question: is it tenable to stick to an exclusively static and structural point of view? The intention of editing the present book, Linguistic Dynamics, is based upon the conception that a static view of human language cannot really be sufficient. To be sure, this stand does not contradict the assumption that a static view is an even necessary and highly relevant pillar of the linguistic enterprise. On the contrary. It should be no doubt about the fact that a structural backbone of linguistics is categorially inevitable. But it should be maintained that this is only one pillar and not the only pillar. Static linguistics is a skeleton supporting and holding together the linguistic edifice. However, a skeleton alone without flesh and blood would represent a poor, dead and sterile linguistics. The vitality of linguistics rests upon a due account of the interplay of the static and dynamic aspects of language. By closer inspection it is seen that taking into account the dynamic aspects of language opens a wide field of linguistic research. The dynamic aspects of language are so manifold that one wonders why these have not been more vividly investigated recently from the point of view of theoretical linguistics. There are, on the one hand, the elementary temporal variations of the linguistic phenomena, like the diachronic changes of linguistic sounds, forms

2

Th. T Ballmer

and meanings, and there are, on the other hand, the highly complex linguistic actions and interactions like the speech processes, the speech acts and the ongoing discourse. Thus there are purely unconscious social changes on one side and consciously controlled interferences on the other. This span already opens a nearly unimaginable realm of facts and interrelations. It is thus not mandatory to add the actualgenetic, the ontogenetic, historiogenetic and phylogenetic dimensions of Linguistic Dynamics simply for the reason of legitimizing that there is enough substance to be dealt with in the field. On the contrary, it seems rather that there is too much material around. The average linguist is certainly overwhelmed by the quantity and quality of problems arising when opening the doors of Linguistic Dynamics. In fact it appears on these grounds that the reasons why linguistics so firmly sticks to the static aspects of language are not so much to be sought in a lack of phenomena and material. Quite the opposite is true. Linguistic Dynamics did not get the attention it should have obtained because of the excessive wealth of its phenomena and facts which must have stunned and even paralysed the mainstream linguist. The notorious anxiety to be overrun by the real world including its constantly changing appearance certainly must have been a constant source of motivation of the modern linguist to concentrate upon the linguistically palpable and stable: the static structure of language. We should not forget that it was the major merit of Ferdinand de Saussure at the beginning of this century to provide linguistics an existence independent of the proper sciences, the task of which is indeed to investigate the facts and laws of the constantly moving and fluctuating real world. De Saussure did this by his famous distinction of language and parole. Thus it became possible for him to inaugurate the proper and independent field of modern linguistics. Chomsky made a quite similar move by his distinction of competence and performance allowing him to stick to the autonomy of linguistics or, more adequately with respect to his concerns, to the autonomy of syntax. The viscid development from phonology to morphology, syntax and, to a much less extended degree, to semantics reveals the deeply entrenched conservation of the static structural dogma in modern linguistics (Static Structural Linguistics, SSL). In consideration of these facts we should adduce very good reasons for allowing us to deviate from the structuralist catechism. What is it that warrants that we are not getting overrun by the real world? How can we guarantee that linguistics does not become an all-comprehensive science? How do we prevent linguistics from being dissolved into biology or physics? The danger of losing our linguistic identity should render us extremely cautious when opening the doors for the real world. For quite a number of linguists this opening is already seen to take place when allowing for semantic and, more so, pragmatic issues. A similar situation occurs at the borderline of lexicology when encyclopedic knowledge becomes an issue. Transgressing the margin to Linguistic Dynamics is an even more radical step towards incorporating

Introduction

3

the real world into linguistic analysis. Thus we should have a vivid interest ourselves for a preferably conservative strategy, if we do not want to endanger the existence and identity of our very own intellectual residence: linguistics. Of course this issue is not a facile one. It is even loaded with considerable controversies of not exclusively rational nature, this is for certain. A case in question is the dispute between artificial intelligence and "the scientific study of language" which became famous as the Dresher-Hornstein debate against Winograd. The issue of the debate is, briefly stated, the procedural-descriptive controversy. Thus it is at issue whether a certain kind of dynamic, namely the procedural methodology, is a better one for linguistic theorizing than the more canonical Chomskyan paradigm of structural generative linguistics. But the turn to Linguistic Dynamics would be of an even more radical nature. Procedural Linguistics (PL) aims at modelling the psychologic reality underlying natural languages. I.e. Procedural Linguistics deals with the cognitive and to some degree also the emotional processes going along with the production and analysis of natural language expressions. Linguistic Dynamics is considerably more comprehensive. It is not tied to a special neighboring science of linguistics such as psychology. It is open for many allies. But whether this is a philology, a historical, a mathematical or a pure science is not determinded in advance. Besides philosophical and psychological matters sociology, anthropology, biology (including at least neurophysiology, ethology, ecology, evolutionary genetics) and even physics may be of concern. As abstract tool providers mathematics, statistics and computer science may be of great help. Linguistic Dynamics is not defined by the theoretical and empirical tools which are put to use, but by the kind and range of phenomena which are dealt with. The tools are selected according to the specific problems at hand. This being so special care is needed to approach the field of Linguistic Dynamics. In order to procede in an conceivable way let us consider the following three questions: (0.1) (0.2) (0.3)

What is the topic of Linguistic Dynamics? What are the relations of Linguistic Dynamics and Structural Linguistics? What licenses us to transgress Structural Linguistics towards a Linguistic Dynamics?

What is Linguistic Dynamics (LD) then? In spite of the seemingly immediate interpretability of this term we should try to not be mislead. The term may be given a number of different readings which should be carefully kept distinct. Some of these readings are really too primitive and simple-minded for serving as a viable candidate of interpretations, however, thus we should start out by stating what Linguistic Dynamics (LD) is not. LD does not mean that all there is to linguistics is dynamic. It does neither negate the existence of static phenomena nor of structural phenomena in

4

Th. T Ballmer

linguistics. Thus LD does not right away stand in opposition to Static Structural Linguistics (SSL). Nor does LD mean that all there is to dynamics is linguistic. There definitely exist nonlinguistic dynamic phenomena, e. g. the bursting of a vulcan. More relevantly LD does not imply that all there is to dynamics is the taking into account of temporal phenomena. Dynamics presupposes but does not exhaust temporality. Dynamics is thus more than the taking into account of temporal change. We shall inquire now what it is that is more. A very first way of conceiving of Linguistic Dynamics would be to say that it refers to more than statical phenomena of language. This primitive explication, although in consistency with the meaning of the term "dynamical", is nevertheless seriously misleading. It suggests easily that LD is concerned only with the temporal aspects of natural language and this, as we have been saying, is certainly entirely mistaken. Although LD deals with the temporal phenomena, it does not deal with them exclusively. LD is based upon and historically and systematically entrenched in the structuralism of modern linguistics. Thus LD explicitly assumes and refers to the existence of linguistic structures as they are postulated, described and structurally explained by the modern strands of linguistic research. The existence of linguistic categories and rules is thus explicitly not denied. What the status and kind of categories and rules are in Linguistic Dynamics need not, however, exactly and in every detail conform to these notions within Static Structural Linguistic. Thus we accept as a first but not yet fully adequate explication for LD: (1)

Linguistic Dynamics is a Structural Linguistics taking into account the temporal phenomena of Natural Language.

This formulation is certainly not unambiguously interpretable. It is not meant, for instance, that LD deals only and exclusively with the succession of different structural states or patterns of (parts and aspects of) natural language. LD is more than maintaining that a categorization, a meaning assignment etc. is at one time so and at a later time different from that. LD is also more than enumerating and keeping track of sequences or sets of (temporally successive) structural patterns as it is done for instance in socio-linguistics (cf. e. g. New Wave theory, Goebl dialectography) or in frame-analysis (cf. pragmemes Rehbein/Ehlich 1972, Schank 1975, Minsky 1975). LD has to refer to such phenomena, of course. In order to be linguistically adequate, LD has to refer to linguistic structures and to the rules of their description. In addition it is necessary that LD refers to the temporal (linear or non-linear) sequencing of structures, as well as to the rules which describe such a sequencing. But, again, it must be emphasized that this is by far not sufficient! This level of elaboration of temporality would not properly transcend Statical Structural Linguistics (SSL). What concerns structures and relations, including the record of temporal successions can quite easily be made to fit into the realm of SSL (Static Structural Linguistics). To include rules of temporal succession,

Introduction

5

as a descriptive means or as psychological device, does neither transcend SSL. In fact, the rules of SSL can often be given an interpretation of temporal succession, although this was not intended or even denied originally. LD being a discipline accumulative upon SSL has to take these phenomena into account but is by far not exhausted by them. We should now eventually be getting in the position of grasping what LD is not. We have come to understand that it is not exclusively structural, even in case these structures include a reference to temporal relations. What LD is, that is what makes it go beyond SSL, is explicated a certain step more adequately by the following characterization. (2)

Linguistic Dynamics refers to the structures of language in the sense that they are a result of underlying processes.

LD thus refers to both the invariants (structures) and to the variants (processes) of natural language. LD considers these to be highly interactive. The crucial fact is that the invariants are not taken as simply given and fixed as it is the case in structural linguistics. The invariants, i.e. the structures and rules, of language are not self-sufficient entities. They cannot, hence, be explained by reference to themselves. Accordingly there cannot exist purely structural explanations in LD. This stands in absolute opposition to the paradigm of static structuralism. Structures, i.e. invariants, can be described, their relations can be observed, described, collected, compared, assembled to rules. Whereas such activities may prepare explanations they are not explanations. Structures, in LD considered as a product of processes, are explained by reference to the processes by which they come about. The processes, in certain situations, may lead to temporally stable phenomena of some sort, to relations remaining the same over some period, in other words they may lead to structural stabilities (this term is very much René Thorn's cf. Wildgen for the elaboration of Thorn's theory for linguistics). By reference to the notion of phenomenon, as used in science (cf. Hacking 1981) we may formulate now: (3)

Linguistic Dynamics considers linguistics as the science of linguistic phenomena, i.e. of linguistic invariants (structures including rules) being built up from linguistic variants (processes).

A crucial notion for LD is the notion of process. Thus we should ask first of all what a process is, how a process is to be described, how processes realize in linguistics and how they produce linguistic phenomena. The universal trait of a process is its temporal morphology: every physically or mentally real process begins and ends. More explicitely, a real process has a phase of beginning, a phase of going on ( = running o f f , duration, elapsing), and a phase of termination. Even more explicitly, we may take a typical, non-deficient real process to consist of five phases: beginning, acceleration, duration, deceleration and termination. Optionally there can be distinguished a number of further phase types: interruption,

6

Th. T Ballmer

resumption, intensivation, fading, iteration etc. One task of interest would be to enumerate all relevant phase types and to provide a method of justifying these (cf. Ballmer 1983). This would amount to a substantiated classification of aktionsarten, linguistically speaking. For Linguistic Dynamics a highly relevant question is what kind of processes may matter for building up and stabilizing linguistic structures. An overview of some such processes may be in order. The simplest processes are those for which nothing more takes place than the universal three or five phases mentioned. Examples of such processes are the weather processes like raining, snowing, sunshine and other ambient processes. In the course of more complex processes things get an existence, objects of specific contours get realized. Typical processes of this sort are the genesis of mountains, lakes, clouds. The coming about of objects give these processes the name of existence processes. These processes are inherently much simpler than fabrication processes, i.e. processes in the course of which hand made objects like vases, tools or even machines are coming about. The latter presuppose a living human agent whereas the former do not. Thus there must be processes between the pure existence processes and the fabrication processes which create in a stepwise evolution the necessary presuppositions for the latter. A highly relevant process of that kind is life itself (being created — grow up — live — age — die) which enables other process type to develop: spontaneous and controlled action, locomotion, grasping, transport and, later, modelling. Thus there seems to exist a dense succession of process types beginning with the simplest unstructured ones, mere processes, and leading to the human induced processes of manufacturing objects. This is an unfolding process of processes, a permanent complexivation. The question is where this evolutionary complexivation leads. What is the complete phenomenology of process types starting from the simplest and ending with the most complex and elaborated ones? What are the most complex processes? What we require from such a phenomenology is that it is complete. It should contain all relevant process types. No one should be missing. A complete phenomenology of process types would represent a map of all processes. Thus it would exhibit the proper topography of processes and provide a guide for orientation. According to Heege (1984) each attempt of a man to gain knowledge about the world is based upon the expectation that the world be organized (a "cosmos") and that this order can be discovered (!). This he calls the basic orientation hypothesis. It does not seem to be wrong to assume that the orientation hypothesis is correct also for the realm of processes. To be in possession of such a map would considerably facilitate the enterprise of Linguistic Dynamics. It would enable us to know which process types we are being investigating, which process types are to be considered because of their relevance, and especially their neighboring relation with respect to a given situation, and, which processes are (to be) neglected

Introduction

7

as farfetched with respect to the given situation. Every study in dynamics should be conducted on the basis of the knowledge of the complete phenomenology of process types. The lack of such a phenomenology would be rather like travelling without map and compass in a jungle. There is a high risk to get lost.

(4)

processes

inanimate existence processes

life processes

non-actions

actions

movements social interactions

operational actions countertrade

symbolic exchanges speech

abstract

context change processes anthropologic context change processes symbolic context change ( = procedures) unconscious controlled by nature causal behavioristic/direct guided natural

conscious self-controlled teleological symbolic spontaneous cultural

8

Th. T Ballmer

This justifies the presentation of an overview over the complete phenomenology of processes. It is obtained from a classification of the set of all anthropologically relevant processes, which in turn can be tapped by the linguistically realized process designators, the verbs. On the basis of the complete phenomenology of processes we are able to arrive at a hierarchy of process types which starts out from rather coarse but essential distinctions and produces an even finer and more practical grid of differentiated notions (cf. (4)). Every existing process fits somewhere uniquely in this frame or, if need occurres, in a refined version of this frame. Accordingly we may say that the majority of anthropologically relevant processes are context changes in the following sense: (5.1)

(5.2)

A human being is able, some canonical preconditions fulfilled, to bring about context change processes in a spontaneous and consciously controlled manner and to value their effects as to coordinate them with respect to other occuring processes. Especially actions, graspings, social interactions, symbolic exchanges (speech and abstract operations) are context change processes.

Context change processes, viewed form the psychological, neurophysiological and computational perspective, when stressing the fact of (digitalized, symbolic) programming, are also called procedures. It is, however, necessary to allow for a number of less-programmed and less-ritualized and, last not least, continuous processes of anthropological, social, biological nature, for which the term procedure is not obviously adequate. The term context change process, also for these considerably more general cases, seems still to be suitable. We summarize this: (6.1) (6.2)

Procedures are digitalized, symbolical, programmed context change processes. There is more to Linguistic Dynamics than merely Procedural Linguistics (for there are context change processes not being procedures but more general (physical and non-physical) processes).

(6.2) justifies as synonyms for LD: Process Linguistics or Processual Linguistics. We shall now try to answer the fundamental question about LD: which process type is the most central and characteristic for LD? In order to arrive at an informative and at the same time justified answer we have to enter into a discussion of Static Structural Linguistics with respect to some of its central and characteristic difficulties. It will turn out that SSL is not stable against two kinds of perturbations: formal and temporal. The chances that SSL hypotheses collapse are thus highly probable. There exist structures of varying degrees of depth: surface invariants and depth invariants

Introduction

9

so to speak. The varying levels of invariants have to be distinguished carefully. The lexicon of a language, e. g., is a surfacial invariant. It lies overt before our eyes. The lexicon is thus surfacial in the sense that it is open to immediate inspection. For a given time the lexical items of a language remain approximately the same with respect to their form, their meaning and their total number. But the lexicon is not an absolute invariant. It is stable at a time, but may undergo gradual changes as time goes on. It is thus merely quasi invariant. The set of idiomatic expressions, the set of sentence patterns, the set of textual patterns are surface (quasi-)invariants in a quite similar way. On the other hand there exist invariants of a deeper kind. These cannot be inspected as directly and easily. These invariants reveal themselves only by means of globally applied procedures such as classifications and statistical analyses. The semantic space of a language, language universals of various kinds, syntactic, semantic, pragmatic and statistical laws are depth invariants. Linguistic surface invariants, though stable against temporal and other variations (perturbations) are considerably irregular, thus are not formally stable. I. e. temporal and other invariance does not imply structural homogeneity at the surface. This fact, for a long time, made language resist a reasonable mathematical and computational analysis. Its irregularities, more precisely the irregularities of surface invariants, were adduced to argue against the possibility of mathematization. Depth invariants seem to show considerably more regularities, although often on normative rather than empirical grounds. Thus mathematization is easier. As long as the problem of relating depth and surface in a fully regular way is not solved, the regularities on the depth level are only of virtual use, this has to be kept in mind however. Static Structuralism attempted to cope with the problems of irregularity by assuming major regularities and entire cascades of subregularities. There are, however, at least two problems arising for Static Structuralism of such a kind: f i r s t l y , the task of classifying a specific item of data and of subsuming it under the appropriate type of (sub-)regularity. This task of rule-subsumption is soon seen to be highly context dependent. It cannot be solved on the basis of the regularities and subregularities already stated alone. The context types have to be classified and investigated with respect to their regularities themselves. There is absolutely no garantee, however, that the new level of regularities is in any way better off than the former, the contrary is true. The approach of SSL based on regularities and subregularities thus does not seem to be able to provide a convergent procedure for any viable broad coverage analysis. Secondly, there exist two temporal dimensions which are not accounted for: there may be new kinds of data coming in, firstly, epistemically novel data already existing but not yet analysed with respect of their rule type and, secondly, historically novel data, non-existent before, or also substituting obsolete data, thus reflecting historical change. Static Structuralism is neither robust (i.e. able to incorporate epistemically novel data) nor historically

10

Th. T Ballmer

adequate (i.e. able to incorporate historically novel data). The inventory of rules being fixed, SSL cannot learn nor forget. We may summarize the two deficiencies of the static structural approach as follows. (7.1) (7.2)

Static Structural Linguistics cannot cope with irregularities. For the task of rule subsumption cannot be solved. Static Structural Linguistics is not stable against data extension (epistemically) nor data change (historically).

This shows that SSL is not immune with respect to two kinds of perturbations: perturbations concerning form variation and perturbations concerning temporal change. SSL is easily falsified for these kinds of variation. Can Linguistic Dynamics overcome these difficulties? We shall argue that it can, indeed, and discuss the way it does it. Firstly, it should not be overly risky to accept the view that LD should be better off than SSL when temporal questions enter the field. Thus we may quite safely assume that L D should be able, to cope with epistemic and historic data change. Changes of any sort belong properly to the topic of LD. We shall argue, secondly, that LD, operating on the surface level but thereby using means and mechanisms on a depth level, is in a considerably more advantageous position than SSL with respect to irregularities. The solution of the serious problems mentioned (of SSL) reads as follows: (8)

At the heart of LD is the Linguistic Learning Situation, probably the most elaborated process of the bio-cultural evolution on this earth.

Sounds are heard, fragmented, parsed, interpreted, meanings assigned, inferences drawn, understanding enhanced, actions (including speech actions) undertaken, effects (results, consequences) of all sorts induced. Nothing, really nothing, hinders human beings from understanding fallaciously produced expressions, novel expressions, expressions neither fitting a specific paradigm nor a set of rules. Chomsky's ideal speaker condition is an idealization. But it is neither theoretically needed nor true to the facts. It is not a necessary prerequisite for a viable conception based upon LD. Non-ideal communication works for many quite common situations. For instance, epistemically and historically new words may be learnt item by item. N o lexical rules are necessary, certainly not. Moreover, for every new rule coming about the initial situation is such that only one expression is representing it. This is the one-item state of a rule. Nothing prevents a human being, particularly a child, to learn a single highly novel expression and add it to his memory. Any such (pre-) analysed expression could then, in principle, serve as a pattern for establishing a novel rule. Thus we may normatively expand the expression cafeteria to teeteria, milkteria, sirupteria, waterteria, oilteria, ga^teria and thus establish a rule starting from the one expression cafeteria. An expression can

Introduction

11

be added to somebody's memory and correspondingly also a subregularity of limited or general range. No overall regularity is required, neither for the production nor for the understanding of linguistic items, in principle. There may, and in fact do, exist superimposed regularities for a considerable number of communicative reasons, mostly of global character involving large numbers of linguistic events. This topic is not to be investigated here immediately, because it leads right in the heart of LD: to Procedural Linguistics, Quantitative Dynamics, Language Change and Morphogenetic analysis. The Linguistic Learning Situation is central for LD, because being based upon this situation means to be adaptive for the cases problematic for SSL (cf. 7). The Learning Situation is a remedy for at least a majority of mutual misunderstandings, be it because of inattentiveness, mispronouncing, misapprehension, novel expressions, not following an expected rule, new notions, novel semantic relations etc. But the Linguistic Learning Situation is also central because of its being the pivot of (historic) linguistic change. Minimal errors of production and reproduction, intentional alterations, creativity and innovation, in other words socio-biological mutations, may have the effect to propagate through entire populations. If they do, they become linguistically relevant on the level of a linguistic (sub-)society, which amounts to historic change. This is true for sounds, words, grammatical rules, meanings, modes of expression etc. Thus there exists, from the point of view of Linguistic Dynamics, a close and intimate relationship between Learning Situation and Language Development. The Linguistic Learning Situation is thus characteristic for LD mainly because of its properties. These are: processual, interactive, incremental. This is true not only on the more surfacial level of sounds, words and form-pattern, but likewise on the semantic and pragmatic levels of meaning and context. Information is transferred, knowledge enhanced, inferences drawn, conclusions obtained, strategies conceived, and tactics realized. The (mutual) Learning Situation is, as it seems, not far from being the most evolved and complex dynamic process type existing in the world of nature and culture. The following analysis scheme of a promise demonstrates that it is possible to analyse the very same phenomena from a very surfacial and static structural perspective, but also more and more deeply and dynamical. At the lower end there is a indicated (9.7), how a promise script may ontogenetically develop via implicit learning stages. (9) (9.1) (9.2) (9.3)

I promise you to come. Description of the Sound Shape: phonemes used, rules used, assimilations, dissmilations, ... Grammatical Structure: syntactic categories, subcategorization, valencies of the verb, transformational relations, ... Semantic Interpretation: Meaning relations, Denotations, Meaning projection rules, ...

12 (9.4)

(9.5)

(9.6)

(9.7)

Th. T Ballmer

Pragmatic Rules: — Propositional Content — Preparatory Rule — Sincerity Condition — Essential Rule

Conversational Maxims — Quality — Quantity — Relation — Manner

(Illocutionary Point) Frame Patterns: Temporally/Spatially recurrent patterns: Schemata, Scripts, Speech Act Classification Schemes, ... Context Dynamics: — unconscious/conscious Context Changes — uncontrolled/controlled Context Changes — consequences, results, influences — context change programs Learning Situation as an Ontogenetic Deploiment of Frame Levels (for reasons of illustration!) immediate

Maxim eats an orange ( = P (Maxim))

action:

action mediated by force: (brut nature, emotive)

Maxim does not want to P.

Thomas operates upon Maxim in order to make him P.

Maxim does P.

action mediated by command or suggestion: (enactive)

Maxim does not want to P.

Thomas tells Maxim (in a high/low voice) to do P.

Maxim does P.

action mediated by escalating iteration: (interactive)

Maxim does not want to P.

Thomas tells Maxim (in a increasingly N. higher voice) to \ do P.

action mediated by request and promise

Maxim does —• Thomas tells —• Maxim says —• Maxim does not want Maxim (...) he does P. N^ P. to do P. N. xMaxim does to P. — not P. Maxim does Maxim does • Thomas tells • Maxim says he does P. P. not want Maxim (...) to P. Maxim does to do P not P.

promise and threatening

-* Maxim does P. Maxim does — not P.

Thomas tells Maxim that Q (good or bad)

This analysis provides us with various informations. First of all, there is an indication that a promise viewed in its own genetic deploiment helps to judge what belongs to its essentials and what not. The similarity between promisses

Introduction

13

and warnings could, on this basis, be discussed in a more substantiated manner than without insight into the developmental stages. Secondly, (9) seems to represent a historical trend following the levels of analysis indicated from (9.1) to (9.7), say. It starts from a surface analysis of expressions (phonetic) and proceeds deeper and deeper (phonemes in a system, morphemes, syntax, etc.). The historical trend seems to free itself from the static bounds and is turning recently towards the dynamic. This raises the question what the historical influences are upon which the dynamization of linguistics is based. First of all we should state that this novel tendency does not stem from linguistics itself nor from the philologies, although there exist dynamical streams in linguistics and the philologies in the 19th century. We may cite here Schleicher (1873) who stated "Languages are natural organisms which grow according to specific laws, develop and then grow older and die." Paul (1909) and Steinthal (1881) point out the dynamic character of the speech act and related psychologic processes. The missing link from these conceptions to the more recent ones is the structuralism which, today, enables us to cope with the complexities of language. As I can judge the contemporary situation, there are two major and a third minor source for the rise of dynamic issues in linguistics. The first, catastrophe theory (Thom 1975), is mathematical originating from physical settings (mechanics, electrodynamics) and making use of differential topology. The second source is the computer and the technology related with it. It is gaining impact since the beginning of the fifties, but since a few years takes influence upon the procedural theorizing in the information sciences themselves (via artificial intelligence), psychology (cognitive science) and linguistics. Even in the philosophy of language and mind there is felt the influence of the procedural conceptions. As third source of dynamic reasoning, until now however of yet minor influence, should be mentioned G. K. Zipfs PsychoBiology of Language and Human Behavior and the Principle of Least Effort. The developments in computer science (CS) and applications including computer games, text processing, personal computers, view data (videotext, btx, selex, mupid), computer aided instruction (CAI), design (CAD) and manipulation (CAM), and even robots (robotics) certainly enhance the contact and working interaction with highly interactive systems. In a very short period a great many people will get used to systems which may serve themselves as models for rather complex dynamic systems. This will train many theorists to observe, control and get insights into dynamic processes which were beyond their horizon until now. The complexity of structures treated by these interactive systems will increase steadily. Data structures and dynamic structures get repeated, iterated, nested and recursively intertwined. Parallel processing, synchronization, dynamic processor generation will become useful tools of dynamic computer architecture and dramaturgy. This entire development will be the major determinant of a consolidation of procedural thinking and hence, in the end of LD.

14

T h . T Ballmer

Catastrophe Theory, in Thorn's sense, is being generalized, by developments like Haken's Synergetics, Bifurcation Theory and the like, and thereby integrated into the canon of physics. These approaches are adequately called processual. There will be a noticeable tendency of reconstructing biological, social, cultural and even linguistic phenomena within this processual Physical Approach (PA). Whether the Procedural Computer Science Approach (CS) and the Processual Physical Approach (PA) will have the opportunity to get closer to each other or even to merge remains an open question for the time being. Quantitative Linguistics in the sense of Zipf could, however, join the realm of the Processual Approach. The problems and the tools are certainly cognate enough. On the basis of a phenomenological analysis of the structural and processual morphology of processes one could hypothetize that there must be a common denominator of the procedural CS and the processual PA. There may even be several bridges between the two. One, as the author of these lines has suggested, could be Blastematics (cf. Ballmer 1981). From what has been said so far we are now in the position of summarizing the topic and task of LD in the following manner: (10.1)

(10.2) (10.3)

Linguistic Dynamics deals with all processes, especially context change processes, which lead up to, and hence are a presupposition for linguistic processes including the linguistic learning processes and the processes of historical language change. The task of Linguistic Dynamics is to explain on processual, or specifically procedural grounds, the surfacial and deep invariants of natural language. Linguistic Dynamics comprises purely temporal changes, unfoldings, motions (kinematics), actions, interactions by forces (dynamics in the classical sense), briefly processes and procedures in rising complexity. At the top stands, as mentioned, the linguistic learning process.

We are now going to describe what the contributions to this volume are. We intend to point out the relations among the contributions themselves as well as to the field of Linguistic Dynamics. The papers of this book can be divided into four groups. The first group, comprising the papers of T. v. Dijk, of Ch. Habel, and of H.-J. Eikmeyer/ H. Rieser deal with what belongs to the field of Procedural Linguistics. These papers, accordingly, are concerned with questions of discourse grammar, including questions of strategies concerning discourse comprehension, dynamical properties of knowledge structures, guiding principles of procedural utterance analysis. The topic of these papers is confined to the level of specific linguistic acts. Thus neither the entire language nor the historical course of events of entire languages are in focus, here. The second group of papers

Introduction

15

including those of G. Altmann, L. Hrebicek and M. Nowakowska belong to the field of Quantitative and Algebraic Linguistics. Their intent is to analyse natural language processes from a numerical, specifically statistical point of view. This conforms to the view that taking into account dynamics implies the necessity of considering the continuum which enforces a quantitative approach. Nowakowka's paper is included here because, although being algebraic in nature, it bridges the gap between the logico-algebraic approaches and the purely quantitative ones. The third group of contributions deals with a classical topic of linguistics, Language Change, but from a novel perspective. The questions raised and answers given by R. Keller, B. Strecker and H. Heringer belong properly to linguistic dynamics or pave the way towards such theories. There is a tendency to make use of biological reasoning, based upon evolution theory. This trend is even more explicit for the forth group of papers, Morphogenetic Linguistics, written by W. A. Koch and Th. Ballmer. This last group deals with the question of how the ontogenetic and phylogenetic dynamics of linguistic and non-linguistic forms are related. How do forms on different levels of the bio-cultural development arise, persist and decay, how are these processes controlled and, before all, are there morphogenetic laws governing the substantial evolution of the wealth of forms? The first group of papers deals with Procedural Linguistics. As such the human procedures of dealing with linguistic utterances, in the production and comprehension mode, are of central interest. The point of view of Procedural Linguistics draws the attention from the pure form of linguistic utterances away and focuses upon the algorithmic, strategie and psychologic processes of fabricating and understanding language in the speech act situation. The very first paper of that group, and hence of the book, starts out with the topic of "Strategic Discourse Comprehension". T. v. Dijk states that the discourse comprehension models on the market are structural rather than strategic. They are constructing representations and based on rules. But actual processes have a strategic nature. The strategic aspect of discourse comprehension is crucial for a cognitive model. What is a strategy then? Stemming, together with the notion of tactics, from a military context it took its way through economics to (the theory of) games. Language can be construed, according to Wittgenstein and others, as games. The general goal is of course to win the game, that is to have better final results than the opponents. A strategy, a method of winning of a game, has, according to v. Dijk, typically a certain number of steps. Strategies correspond to problem solving processes. Text understanding is a complex task although not immediately experienced like that. It is, hence, not in an obvious sense rule governed but rather individual. Nevertheless language comprehension strategies are to be optimal with respect to the processing resources, time, attention, memory space etc. v. Dijk takes discourse strategies to be, relative to a goal of semantic, pragmatic, and interactional understanding, as a (pre-)

16

Th. T Ballmer

programmed plan for the execution of steps, such that the end-result (understanding) is optimal given the (other) goals of the language users and cognitive resources available, v. Dijk, then, sets out to fill the notion of strategy with linguistic content. There are investigated different kinds of strategies (coherence, predicate coherence, completion of coherence, global coherence). There is distinguished between micro- und macrostrategies. The latter concern the architecture of text and discourses. The major strategy is to integrate text information into world representation, v. Dijk deals on that basis with a great number of specific problems: difficulty of texts, readability of texts, acquisition of frames and scripts, macrostructures, microstructures, subjective strategies, motivation. Moreover he deals with speech act understanding, social interaction and cognitive strategies, v. Dijk comes to the conclusion that the strategic approach exhibits a considerable number of advantages. Strategies provide meanings where apparently there is none, they establish coherence before enough data have been obtained, they derive global themes that can be used to process further information top-down. In general, he states, strategies provide easy, fast and flexible means to handle vast and complex amounts of textual and communicative data. Ch. Habel's "Referential Nets as Knowledge Structures" deals with structural and dynamic properties of the knowledge of facts, rules and objects. Processing language is seen as a knowledge-dependent process. This processing is based on large amounts of knowledge and belief. It is seen to be, moreover, as context dependent. In the paper, a formal logic-oriented approach for representing and processing different subtypes of knowledge is put forward. The knowledge structures are interpreted as representations of the speakerlistener's knowledge about the language, the world and the participants in the communication situation. The specific subtype 'knowledge about objects' is studied in more detail. The corresponding knowledge structure, called Referential Net (RefN), is introduced in a formal manner and exemplified with some anaphoric problems. Operators cp is assigned an effect e, times z.j are introduced, the synchronization of times is considered. The processing of referential relations is executed in a particular component of the discourse processing system, the referential processes (ReP). The dynamics of referential nets is focused upon. The thesis is that humans operate as information processing machines. A list of basic information processes is used (symbol creation and destruction, manipulation of symbol structures, storing and retrieving symbol structures, tests and comparison of symbol structures). The aim is then to provide, firstly, the searching and choosing of the best Referential Objects RefO in the Referential Net RefN and, secondly, manipulating the RefN by Referential Processes (RefPs): (for RefOs) creation, deletion, fusion, splitting; (for descriptions of RefOs) attaching, detaching, changing; (for attributes) expansion, erasing, modification, transformation, integration. The procedures for changing knowledge structures (i.e. context or discourse models etc.) were described in a detailed algorithmic manner.

Introduction

17

The third paper dealing with Procedural Linguistics is "Procedural Grammar for a Fragment of Black English Discourse" by H.-J. Eikmeyer and H. Rieser. This paper adheres to a procedural approach of grammar in the sense that the interpretation of expressions is made dependent upon the change of context parameters. Thus, contexts and context changes are needed to provide and intuitively satisfactory semantics for linguistic utterances such as Here it is warmer than here. A grammar is conceived as a (description of the) device of the planning of speech production and reception. Eikmeyer and Rieser abandon the venerable idea that the system of language exists independently of language users and contexts. The modelling of the techniques which speakers apply in analysing (and producing) spoken language is taken to be the core of a new paradigm: procedural grammar. The spoken language is in the center of interest. Utterances are perceived as going on in the flow of time. These utterances are related to what is called grammatical backgrounds. This enables the language user to cope with incomplete information. The forming and relating of linguistic chunks is thereby of vital importance. And, in order to control the understanding of the utterances, the interaction between speaker and recipient is taken into account. Grammar is based essentially on understanding, i.e. it is conceived semantically. Eikmeyer/ Rieser oppose the assumptions of recent theories of grammar: the ideal speaker-listener, the homogeneity of speech communities, the competenceperformance dichotomy, and the type-token distinction. Also they deny the usefulness of the de-facto separation of linguistic levels such as phonetics and phonology of grammar proper. Levels should be intensively interacting. Eikmeyer/Rieser start out from certain Naturalness Conditions. In deviation to de Saussurean, structuralist and Chomskyan linguistic, regularities are set up with respect to a class of contexts, i.e. in a context-dependent manner. There seems to be a division of labor between contexts on the one hand and utterances on the other. Therefore one is not looking for the optimal syntax, semantics or pragmatics fully covering a given fragment as traditional theories of grammars are. The recipients are viewed to take their analytic tools according to the grammatical issue at hand. The technical means provided by Procedural Grammar are global principles (like the Long Sentence Principle) and a local analysis by a flexible syntax and semantics. Grammaticality does not lie at the centre of this approach, for there are procedures of re-analysis and the anticipation concerning the structure of events to be analysed. A fragment of Black English discourse is then analysed on this basis. The second group of articles in this book is concerned with Quantitative Dynamics. As has been mentioned, the dynamical doctrine implies to analyse a given phenomenon finer and finder. This leads, as far as the temporal dimension is concern, to the discovery and analysis of the temporal continuum. A consistent and systematic dealing with dynamical questions automatically leads to the task of coping with the continuum. One important way to manage

18

T h . T Ballmer

this is to use quantitative and related means. In a first paper "On the Dynamic Approach to Language", G. Altmann enters upon the history and the more recent developments of quantitative dynamic linguistics. In the 19th century linguists were concerned with language change. But their rigorously deterministic view and the lack of deductive systems led to a stagnation. According to Altmann, de Saussure banned the discussion of dynamic phenomena. Mere descriptions and no explanations were the result of this act. Zipf, although unrecognized for a long time by linguists, is considered to be the founder of dynamic linguistics. He took a scientists point of view and introduced statistical analysis. Russian mathematicians took up Zipfs ideas of a dynamical philology. Whereas the static approach is qualitative, algebraic and rule oriented (Carnap speaks even of mythical), the dynamic approach is quantitative, uses differential equations (analysis) and laws yielding explanations. Language is seen as being a selfregulating system in a communicative equilibrium, (meta-) stable against disturbances, with an inherent evolution. Altmann discusses a number of laws, such as the Piotrowski's law which is intended to explain and describe the course of change of large class of individual linguistic entities. The change in language is conceived as proceeding like an epidemic. Other laws are mentioned concerning the multiple variability of language phenomena, the behavior of language in its use and even the structure of language. This demonstrates, once more, that structural results are well looked after even in a dynamic linguistics. L. Hfebicek in his article "Text as a Unit and Co-references" makes an attempt to approach a textlinguistic phenomena on a quantitative level. After a characterization of what a text in this framework could be, a quantitative variant of co-reference, a notion of text coherence, is introduced. There are formulated a number of hypotheses concerning the number of references and co-references in a text. For these hypotheses empirical verifications are designed. The sample of analysed texts is a random sample of the population of Turkish texts. The essential hypothesis (z = ak-v/n) is proven by the Ftest. There is also a stochastic treatment given of the problem. It is not claimed that a dynamic solution of the problem under consideration rules out other approaches, but a dynamic approach is considered to come closer to what language really is. M. Nowakowska, in her paper "An Algebraic Approach to Discourses and Their Goals" takes a point of view of language dynamics which considers the structural constraints on discourses as imposed by composite goals (of the language users). Considering questions of optimization and sequential decision making, this approach stands in the interdisciplinary field between quantitative and algebraic ways of dealing with natural language. Nowakowska shows that the situation of written and spoken discourse is essentially that of multiple criteria decision making in a fuzzy environment. Discourses are treated as admissible strings of sentences, i.e. the set D of discourses as a

Introduction

19

fuzzy subset of L*, whereby L £ V* is the language, a crisp subset of V* (V: the vocabulary). Admissible termination moments of discourses are definded in terms of degrees to which initial parts of a discourse constitute discourses. The process of developing initial parts of discourses is the main object of interest. The goals of a discourse are divided into three kinds: positive, negative, third type goals. Further kinds of goals are distinguished (context free, context dependent, multiple goals, prerequisite and independent goals). The major result of the paper is a theorem on the condition whether something is a discourse fulfilling a composite goal. Applications are envisaged: textbook discourses, salesman's pitches. The use of Bellman equation is advisable in these matters. The third group of papers deals with Language Change and the notion of Rule. It seems to be a longstanding problem to decide what the status of the processes are which go on when a language changes. Are these actions in the sense that human beings act in a controllable controlled manner, spontaneously and consciously or are these language change processes natural processes, lawful but inanimate and of physical nature? Are these processes teleological (teleonomical) or rather causally conditioned? Linguistic Dynamics should be concerned with such questions and provide an answer. The notion of linguistic rule will be of special relevance. As the following articles show, the dynamic point of view raises a number of new questions and provides perspectives of novel solutions. Even there seems to develop a change in the general perspective as to what counts as a problem and a corresponding solution. The first paper of this group, then, is R. Keller's "Towards a Theory of Linguistic Change". N o one, Keller states, knows exactly how languages change. Three questions are raised accordingly. First, what are the reasons that the development of a theory of linguistic change were hindered, secondly, how should language be conceived in order to entail a theory of its change, and thirdly, what would the structure of such a theory be? Is language an organism, in Schleicher's sense, which changes itself, or do, rather, people change it? Keller claims that we are faced here with Phenomena of a Third Kind. We are put in the middle of old linguistic questions. Is language natural (physei) or artificial (thesei, nomo)? Keller points out that there are phenomena which are the result of human actions but not the goal of their intentions, e. g. the inflation of the dollar, our population growth, the beaten path across the lawn, our language. Natural language is thus not (simply) natural nor (simply) artificial. Language is constitued by necessary effects of the actions producing them, so called collective consequences. The true definition of Phenomena of the Third Kind, as already W. v. Humboldt seems to have discovered, is such that it can only be a genetic one. We do not understand if we do not know how it develops what the process of its origin is. The development is part of the object, after all. Keller illustrates phenomena of the third kind by the (simplified, linear) analysis of a traffic jam. The security margin leads to an effect, the jam, which is not intended by the person who

20

Th. T Ballmer

actively slowed down at some moment in the motorcade. Explanations of Phenomena of the Third Kind are dubbed invisible-hand explanations. Such an explanation includes the depiction of personal motives, intentions etc., the depiction of the invisible hand process, and the depiction of the collective consequence, the structure produced. Thus there is involved a dynamic microlevel, a structural macrolevel, the collective consequence, and the structure produced. Thus, there is involved a dynamic microlevel, a structural macrolevel, the collective consequence, and the genetic process linking the individual microlevel (energeia) and the collective macrolevel (ergon). An invisible-hand explanation is intentional in its beginning and continues causally. The range of control for actions is limited, as we know. For various reasons invisible-hand explanations are not so much of prognostic than of diagnostic value, Keller states. This difficult methodological question must be considered carefully, before definite conclusions can be drawn. It is maintained, after all, that an invisible-hand explanation explains the explanandum, a phenomenon of the third, as a causal consequence of individual intentional actions, which, at least in part, realize similar intentions. On the basis of these conceptions the notions of static and dynamic maxims are introduced. Certain maxims, static maxims, are optimal and stabilized. Others, the dynamic maxims, are not. They may have the effect of "contradicting" themself and each other. They render the system of behaving and speaking instable. This is an inherent motor of destabilization. Examples of such dynamic maxims are discussed. One such example is that one should speak (behave) in a manner that people pay attention to you. Finally the relation between individual competence and socio-cultural evolution, in a quite sociobiological fashion, is investigated. Keller states, however, that whereas, in nature, the variations evolve according to chance, with regard to communicating we create variation already in anticipation of the selection to be expected. The second article in this group, written by B. Strecker deals with "Rules and the Dynamics of Language". After some introductory explications on the notion of rule, and especially normative and descriptive rules, Strecker criticizes the unreflected use of rules in linguistic theorizing. Especially the relation between the implicit rules underlying linguistic behaviour and their formulation seems to get little attention. Moreover the span between rule governed behaviour and libertinage (or even chaos?) seems to be a neglected area, too. The reasons for and against the existence of implicit rules underlying behaviour are discussed. At first arguments are cumulated speaking in favor of rules. Rules make the unorderly structure of language perspicuous, language seems to be arranged in certain non-random patterns, language has been identified via its expressions, neglecting function, actions, and pragmatics and thus artificially has been cleaned from the inherent "life", energeia (the Humboldtian term again). On this basis the dynamic character was successfully erased. While there are good reasons to talk of implicit rules, Strecker can

Introduction

21

see still better reasons not to do so. First, implicit rules are a purely theoretical construct which are by no means empirically validated or even validable. Secondly, how could language be rule governed before it has ever got started. No rule exists at the beginning. A petitio principii. Strecker thus criticizes the assumption that the very possibility of language depends on there being rules creating (meanings of) expressions which are understood by the hearer. Explanation of linguistic behaviour by means of rules is but the second best thing. Our world does not seem to be completely calculable. For games such as life and language which has no fixed condition there cannot be a fixed set of strategies. Strecker focuses upon the genuine, dynamic, unexpected, i. e. unforeseen and unforeseeable conditions which demand creative use of the traditional rule-governed means, and which may thereby bring about a change of the context and the world. Creativity as context changer. Rules would not merely be of no use but would be a hindrance to success in such non-standard conditions. The third contribution in the group of Language Change and Rules is "Not by Nature nor by Intention (The Normative Power of Language Signs)" by H. Heringer. The author takes a long standing dispute of the arbitrarité du signe as a starting point for investigating the origination of meaning which cannot be of purely natural nor of purely intentional kind. The study of authentic sources proves some of the discussion as a misunderstanding, as so often. Three points have to be kept in mind. First, there are no pre-existent ideas, all is linguistically formed. There cannot, hence, be an extralinguistic foundation or a natural one-one relation of signifiant and signifié. Secondly, arbitrariness does not mean that the speaker has a choice here. Thirdly, de Saussure speaks of the "vérité, que la language se transforme sans que les sujets puissent la transformer". The change of language is a fateful evolution. It turns out to be an essential feature of language, that it evades the will of the single individual as well as the collective will. Heringer comes to conclude that the languages are social institutions of a special nature. They are "ni par nature, ni par intention". The characteristic of a language sign X stands for Y is of normative nature. The objects X and Y are not preestablished, however, they are mutating historical objects. X is a regulating idea, a pattern according to which we classify utterances. In the form X counts as Y in this situation, Heringer considers various reasons of a recipient's tacit assumption: rules, precedence, analogy, coordination. All have normative replicational structure. Signs can only be normative insofar they are arbitrary. But we experience our language as natural. Is there an antinomy? The omniscient and omnipotent creator does not seem to be any longer needed (as Siissmilch has suggested those days). This is contrasted by Menger with his genetic explanation of the origin of values and — since signs are values according to de Saussure — of signs. Menger's explanation is a typical invisible hand explanation, as we have found it with Keller's article. A model of Schelling is discussed which illustrates how spontaneous orders can come about without

22

Th. T Ballmer

planning. This seems to be the point of mystery here. Again a three step process lies at the basis of an invisible hand explanation. The typical ingredients are: (1) acts of individuals, (2) a process having the result of the individuals' acts as a basis and accumulates them, (3) a state which has a meaningful order (not being the aim or direct result of the acts of the individuals). The attempt of the recent philosophy of language is then discussed to explain meaning on the basis of what the speaker intends. This attempt is being refuted. Nor can other theories, one based on the reification of language, the other based on the speaker as active creator, be maintained. Basic questions of a theory of meaning change are asked: How do new meanings originate, how do they spread? Invisible-hand explanations are worked out in some more detail for some specific cases. Heringer takes it to be enlightening to comprehend the change of meaning as a case of cultural evolution. Exactly as in biological evolution, variation (plus mutation!) and selection are essential elements. Semantic evolution, however, is with respect to genetic evolution particularly fast, it is neither blind nor accidental. Moreover, every speaker already takes into account the possible selection in his speaking to the extent that he includes the knowledge of his partner by anticipating common knowledge and by acting accordingly. The fourth group of articles Morpbogenetic Linguistics is characterized by the fact that the dynamics is investigated from a point of view which could best be called substantial. Whereas Procedural Linguistics puts its focus upon the microlevel of language change, namely the particular production and understanding acts, whereas Quantitative Linguistics is interested in numerical (but not structural) enouncements resulting in the formulation of (conditioned) laws and theorems which is considerably different from morphogenetic and morphological propositions, whereas Language Change theories are, like biological evolution theories, specific about the mechanism of change but not about the overall outcome, the form and dynamic structure of the language, it seems that the papers of the forth group aim at substantially comprehending and reconstructing the global hypostatized object of language seen as a dynamic gebilde embedded in the rest of a dynamic world, especially the biosocio-psychological world. The first paper of this group, "Tension and Suspense: on the Biogenesis and the Semiogenesis of the Detective Novel, Soccer, and Art" is written by W. A. Koch. His topic is one of cultural evolution, but it is seen also in a wider perspective, that of biogenesis. The topic concerns the phenomenon of 'Spannung'. In the essay, the theoretical emphasis is on the genetic precedence of process as opposed to state and on the evolutionary transfer of structural elements. Processes are primordial. This conception convenes very much to the dynamic point of view taken in this book. 'Spannung' as a typically process-like structure may be projected onto space or onto time and will accordingly be split into two types: tension and suspense. Koch argues that tension and suspense are intimately related to the faculty of

Introduction

23

organisms of forming (activated) frames, i. e. of forming structured hypotheses. The psycho-physiological counterpart of hypotheses, in the sense of frames for planning and projecting, is thus the sensing of 'Spannung' in one of its modes, spatial or temporal. This highly remarkable hypothesis relates a dynamic and structural notion, namely activated frame ( = hypothesis), with a psycho-physiological phenomenon. It seems thus that the substantial dynamical approaches can be cast into such a state of explicitness that even the body-mind hiatus can be bridged. This bridge allows to draw substantial structural conclusions as to the dynamics of psycho-biotic phenomena. It may seem surprizing that detective novels, soccer, and art are put on one and the same level. Koch demonstrates quite convincingly that there exist common roots of all these domains in the territorial behavior of animals. The threat connected with the defense of the animals' chosen territory against intruding rivals seems to be the origin of emotions like suspense (and tension). It takes a moments reflection and a careful study of Koch's arguments to accept his audacious thesis. But there seems to be a considerable impact of unification in Koch's assumptions. The impact is strengthened by a number of morphologic and morphogenetic considerations having their roots in semiotics being accompanied by a morpho-logical symbolisms put to test at a great many occasions, although not yet mathematized in one of the customary ways. Notions, very useful in the dynamics of physics, seem to play a role also in Koch's conceptions: symmetry, asymmetry, polarity (monad, dyad), phase sequences, equilibria (e. g. between amity and enmity) etc. There are presented, moreover, a great number of philological investigations of texts, reports of biological facts and a great many statements revealing unknown and unexpected interconnections between a number of seemingly different topics. The last contribution to this book, Linguistic Dynamics, in the group Morphogenetic Linguistics, is "The Psychology of Context Change" by Th. Ballmer. The objective of this contribution is fourfold. First, the evolution from logic via linguistics to psycho-logic, a methodological programme with an impact in the recent history of science is displayed. Secondly, there is discussed, as a major result of the displayed historical trend, a thesaurus analysis of natural language leading to what one could call thesaurus logic. Its kernel is a comprehensive analysis of anthropologically relevant processes. Thirdly, there is discussed another product of the displayed historical trend, the context change logic. Whereas the thesaurus logic focuses upon the global structural interrelations of processes, context change logic focuses upon the local dynamic processes. Fourthly, there is pointed out, how, on the basis of thesaurus logic and context change logic, the relation between logico-linguistic matters and psychological matters is to be conceived substantially and procedurally. The bio-genetic fundament of the thesaurus structure and the context change mechanisms provide the safe ground for establishing these results. This paper thus starts out to analyse a recent historical trend

24

Th. T Ballmer

demonstrating that there seems to evolve a convergence between logic and psychology. The analytic investigations of the philosophy of language seem to become, under the influence of linguistic examination, more and more realistic in three ways: first, there is a trend of incorporating more and more of the "inner structure" of linguistic expressions (predicates, quantifiers, adverbial operators, illocutionary elements, etc.). A major point is reached when all lexical items (predicates, operators, particles, ...) are identified and grasped in their mutual analytic relationship. This would be the state of a Thesaurus-logic. Secondly, there is a trend of incorporating more and more of the dynamic processes going along with the generation (synthesis) and the understanding (analysis) of linguistic expressions. The context dynamics caused by linguistic processes is at stake. This is one further aspect in which the psychologic and biologic reality gets gradually more recognized. A major point is reached when a Context Change Logic (a generalization of a Procedural Logic) is realized. Context Change Logics provide the appropriate algebra for dynamic processes including procedural semantics (plus pragmatics), procedural syntax (plus morphology) and procedural phonology (plus graphematics). Context Change Logics can be derived phenomenologically, as it is done in this paper. The consequent pursuing of this programme leads, as is shown, to some major issues of thesaurus logic: the classification of context change processes is reflected in the verb thesaurus of natural language. Thirdly, there is a trend of analysing and incorporating more and more of the structure of the context. This must lead to a Context Structure Logic (including context dimensions, the so called "Marked Model of Nested Worlds", static and dynamic data structures conceived physically or epistemically, context architecture, context dramaturgy, in brief context organization. It should not come as a surprise that a consequent penetration of this topic again leads to the taking into account of major results of thesaurus logic. Context architecture and dramaturgy is organized on the basis of all anthropologically relevant processes (cf. The Principle of Relevance). These three trends sustain forcefully the view that in the whole there is a conspicuous movement towards becoming more and more realistic in the logico-analytic investigations of natural language. This realism pertains, firstly, to psychology and, because of their being a biological fundament, to biology as well. These latter points are verified in the appendix "A Processual Definition of Psychology", but transgress, strictly speaking, the reach of the paper.

Introduction

25

References Ballmer, Th.: 1978, Logical Grammar. With Special Consideration of Topics in Context Change. North-Holland: Amsterdam. Ballmer, Th.: 1981a, "The Interaction between Ontogeny and Phylogeny: A Theoretical Reconstruction of the Evolution of Mind and Language". In: W. A. Koch (ed.) Semiogenesis. Essays on the Analysis of the Genesis of Language, Art, and Literature. Lang: Frankfurt, p. 81 - 544. Ballmer, Th.: 1981b, "Linguistic Dynamics. A Physio-Phenomenologic Treatise of Linguistic Theory". In: B. Rieger (ed.) Empirical Semantics II. Brockmeyer: Bochum, p. 1—58. Ballmer, Th.: 1982, Biological Foundations of Linguistic Communication. Towards a Cybernetic Theory of Language. Benjamins: Amsterdam. Ballmer, Th.: 1983, "Zur Gewinnung einer Fundamentalklassifikation des menschlichen Wissens I, II". In: International Classification, 10, No. 2/3, p. 69 —77. Brennenstuhl, W.: 1982, Ability and Control. Towards a Cybernetic Linguistics. Benjamins: Amsterdam. Chomsky, N.: 1965, Aspects of the Theory of Syntax. MIT-Press: Cambridge Mass. Chomsky, N.: 1982, Lectures on Government and Binding-. The Pisa Lectures. Foris: Dordrecht. Dresher, E., Hornstein, N.: 1976, "On Some Supposed Contributions of Artificial Intelligence to the Scientific Study of Language". In: Cognition, 4, p. 321—398. Eigen, M., Winkler, R.: 1975, Das Spiel. Naturgesetze steuern den Zufall. Piper: München. Ehlich, K., Rehbein, J.: 1972, "Zur Konstitution pragmatischer Einheiten in einer Institution". In: D. Wunderlich (ed.) Linguistische Pragmatik. Athenaion: Wiesbaden, p. 209—254. Fu, K. S.: 1977, "Introduction to Syntactic Pattern Recognition". In: Fu, K. S. (ed.) Syntactic Pattern Recognition. Springer: Berlin, p. 1—30. Hacking, I.: 1981, "Spekulation, Berechnung und die Erschaffung von Phänomenen". In: H. P. Dürr (ed.) Aufsätze %ur Philosophie Paul Feyerabends (vol. 2). Suhrkamp: Frankfurt, p.126 —158. Haken, H.: 1978, Synergetics, Nonequilibrium Phase Transitions and Self-Organisation in Physics, Chemistry and Biolog). Springer: Berlin. Heege, R.: 1984, "Äquilibration im Lernprozeß". In: K. Kornwachs (ed.) Offenheit, Zeitlichkeit, Komplexität. Campus: Frankfurt. Minski, M.: 1975, "A Framework for Representing Knowledge". In: P. H. Winston (ed.) The Psycholog) of Computer Vision. McGraw Hill: New York, p. 211—277. Paul, H.: 19094, Prinzipien der Sprachgeschichte. Max Niemeyer: Tübingen. Saussure, F. de: 1967, Cours de Linguistique Générale. Publié par Ch. Bally, A. Sechehaye. Payst: Paris, (1916). Schleicher, A.: 18733, Die Darwinsche Theorie und die Sprachwissenschaft. H. Böhlau: Weimar. Schank, R.: 1975, Conceptual Information Processing. North-Holland: Amsterdam. Steinthal, H.: 18822, Einleitung in die Psychologie und Sprachwissenschaft. Ferd. Dümmlers Verlagsbuchhandlung: Berlin. Thom, R.: 1975, Structural Stability and Morphogenesis. Benjamin: London. Vollmer, G.: 1975, Evolutionäre Erkenntnistheorie. Angeborene Erkenntnisstrukturen im Kontext von Biologie, Psychologie, Linguistik, Philosophie und Wissenschaftstheorie. Hirzel: Stuttgart. Winograd, T.: 1972, Understanding Natural Language. Edinburgh Univ. Press: Edinburgh. Winograd, T.: 1977, "On Some Contested Suppositions of Generative Linguistics about the Scientific Study of Language". In: Cognition, 5, p. 151 —179. Zipf, G. K.: 1965, The Psycho-Biology of Language. MIT-Press: Cambridge (1935). Zipf, G. K.: 1972, Human Behavior and the Principle of Least E f f o r t . Hafner: New York.

Part I

Procedural Linguistics

TEUN A. VAN DIJK

Strategic Discourse Comprehension1 1. The Notion of

'Strategy'

Most of the discourse comprehension models now on the market have a structural rather than a strategic character.2 They describe the understanding process completely or predominantly in terms of analysing or constructing representations and on the basis of rules. Such rules operate on structural units or categories, both on those of discourse and on those of knowledge. Although such rules may be motivated from a purely theoretical point of 1

2

This paper was written in december 1980. In the meantime the model sketched here has been developed considerably. For full details we refer to its booklength treatment in van Dijk & Kintsch (1983). The ideas in this paper should therefore be considered as preliminaries to this book. Only on some important new developments we will give some few specifics in the footnotes below. Also, we have here kept the bibliographical references to a minimum. For full bibliographical acknowledgements, we also refer to the book. The present paper was presented as a lecture for the XVth International Congress of the Società di Linguistica Italiana, held in May, 1981, in Santa Margherita Ligure and Genova, Italy. I wish to thank the participants to this congress for relevant remarks about this paper. An Italian version of the paper will appear in the Proceedings of this Conference. Finally, I would like to thank Walter Kintsch of the University of Colorado in Boulder, Colorado, USA, for his long year collaboration in our work in discourse processing. Many of the ideas in this paper would not have been developed without his stimulating suggestions. The present formulation, though, is mine, and I therefore, of course, bear full responsibility for any errors, vagueness or speculativeness. The work on discourse processing which would qualify as 'structural' as I use this term here is too vast to even refer to succinctly in this paper. First, our earlier model with Kintsch (Kintsch & van Dijk, 1978) has this more abstract, structural feature, although also several processual aspects were already built in (such as the notions of cyclical treatment of incoming information, reinstatement searches in episodic memory, the strategic use of world knowledge in the establishment of coherence, and so on). For other work related to this model, see e. g. Meyer (1975), Just & Carpenter (1977), Freedle (1977, 1979), Sanford & Garrod (1981), Graesser (1981), to name only some important recent books. For recent surveys see Ballsteadt et al. (1981) and Groeben (1981), two books which also pay attention to practical (and didactic) aspects of discourse comprehension and reading. For details on current research, the reader should consult the last few volumes of journals such as the Journal of Verbal Learning and Verbal Behavior, Cognitive Psychology, the Journal of Educational Psychology, the Psychological Review, Memory and Cognition. See also the special issues of Poetics (vol. 8, 1980) and Text (vol. 3, 1982), edited by me on story comprehension and new models in discourse processing, respectively.

30

T. v. Dijk

view in a model which accounts for grammars or memory representations of discourse, and although language users to a certain extent may be said to 'know' rules of understanding (semantic interpretation), the actual processes involved rather have strategic nature. In this theoretical note I will briefly review the major reasons why the strategic aspect of discourse comprehension is crucial for a cognitive model. First, however, let us consider the notion of strategy itself. Besides its obvious military implications, the notion has been used above all in the framework of the analysis of games and play, e. g. in mathematical game theory and its applications. Such applications presuppose, first, that there are several players or participants in a game. The general goal is of course to win the game, that is, to have better final results than one's opponents. Games further have rules which determine which actions are allowed and in which order. Such rules may be defined in general terms: they define the particular game which is played. Strategies, however, are variable, personal, uses of the rules, such that a player obtains a game situation that is comparatively better than the game situation of other players. Strategies typically have a certain number of steps. These steps are deliberately planned in order to reach the desired (better) game situation. For instance, such steps may be moves in the game that force the other players to take specific reaction steps, which in turn are conditional for the next steps of the first player. Also we may have strategic steps that dissimulate the 'real' strategic goals of a player: B thinks A is going to do p (reach result R(p)) whereas A wants to reach R(q). When we speak about strategies of language comprehension this metaphor from game theory is only partly useful. First, we normally do not have several players (in understanding monological discourse at least), but one language user, viz. a reader/hearer. Second, there is not a particular final goal being desired, but at most a 'continous' goal, viz. (optimally) understanding a text, or perhaps the speaker or writer of the text. That is, if we would use the strategic metaphor at all, this would apply to uses of the notion that are close to that of problem solving. And in fact, especially Al-orientied models of discourse comprehension try to apply a problem solving metaphor to text understanding processes. 3 The analogy however is still rather weak. First we should have a more or less 'difficult task'. Although some texts may be hard to understand, normal understanding is not usually experienced as a 'difficult task'. Next, there should be a specific goal, viz. the solution of a problem of the performance of (some aspects of the) task. The analogy does hold however in the sense that the language user takes a number of steps in order to perform a complex task.

3

This work is typical for Al-inspired work in discourse processing, see e. g. work by Schank & Abelson, Black, Wilensky, Thorndyke in his later work, and many others. See the two mentioned special issues of Poetics and Text for several papers in this paradigma.

Strategic Discourse Comprehension

31

These steps are not rule governed, but rather depend on the analysis, which may be variable and personal, of the specific data at hand, and a planning procedure which calculates the possible or most probable results of each of the steps. The final result of comprehension being searched for by the language user, is an adequate representation of the 'meaning' of the text, both semantically and pragmatically (which speech act is being performed), and, in a wider context, a representation of the (inter-)action of the speaker, and hence of the 'underlying' intentions, purposes, or motivations of the speaker. On the other hand, the hearer/reader also has more or less speaker-independent goals: (s)he may have the primary goal to use the information in the text within the framework of another task or goal (extension of knowledge, information necessary for own future actions, and so on): Then, language comprehension strategies, just as other strategies, should be optimal. They should yield the best results with a minimum of 'cost'. This 'cost' may be formulated in terms of processing resources (time, attention, memory space). For instance, a reader may decide to read very carefully, but this requires specific attention (and hence does not allow cognitive or contextual 'distraction' or 'noise') and especially time. Hence, a just balance between speed and accuracy will be sought. Clearly, this balance will differ for each text type and context type. For the average citizen, reading the newspaper rapidly hardly has serious possible drawbacks in case some information is not read (understood) at all or only partially. For a politician this may be different. The same holds for contexts in which we should read texts with a didactic goal (e. g. memorize for exams). In other words, strategies crucially differ from text type to text type and from context type to context type. But for each communicative situation the optimal result is striven for: as fast as possible but as accurate as possible. In this respect strategies differ from rules. Rules may yield a perfect analysis of the data, but they may be cumbersome, slow or place too high demands on memory space or retrieval processes. It may be more effective, therefore, to follow a fast strategy which is nearly always correct (yielding good results), and for those cases where wrong results are obtained, to apply another strategy (or rule). From these brief introductory remarks, we may now define a discourse strategy, relative to the goal of semantic, pragmatic and interactional understanding, as a ( p r e p r o g r a m m e d plan for the execution of a sequence of steps, such that the end-result (understanding) is optimal given the (other) goals of the language users and given the cognitive resources available (time, attention, knowledge, interest, etc.). Typically, strategies will try to obtain satisfactory intermediate goals which in turn are (supposed to be) optimal conditions for the subsequent steps. In other words, a complex task such as 'understanding a text' is split up in a number of 'subtasks' (subprograms, subplans, subroutines, etc.) which, together, are expected to yield the desired final result.

32

T. v. Dijk

Note that strategic planning of language comprehension strategies is seldom 'conscious' in normal interpretation situations. Conscious or even explicit planning only takes place when the text is some semantic puzzle, for instance in the case of a riddle, a modern poem, or a scientific paper. Hermeneutics is precisely the study of these 'explicit' strategies in the interpretation of 'difficult' texts. We have assumed that there is not just one set of strategies for discourse comprehension. For each subproblem we may have different sets. Thus, we may have syntactic, semantic or pragmatic strategies. Characteristically, however, unlike rules, linguistic strategies are not just level specific. Thus, syntactic analysis may make use not only of surface structure data (word order, syntactic categories, morphological information), but also and at the same time of (partial, expected or already present) semantic and pragmatic information. In other words, strategies are flexible: they may be adapted to the particular situation at hand. In this note, I will focus attention on semantic strategies of discourse. That is, those strategies that are applied in order to reach an adequate (partial) interpretation of the text. Cognitively this means, traditionally, that — stepwise — a semantic representation of the text is being constructed in episodic memory (EM). The adequacy of that text representation (TR) in EM is measured along several dimensions or criteria. It may mean that the TR must be relatively complete. This completeness may hold both for the local (detail) level and/or for the global (macro-) level of understanding. It may (also) mean that the TR must be optimally relevant: in that case it must contain the information which is needed, i. e. information which is useful for the performance of other tasks. Third, it may mean that the TR is optimally organised, presuming that better organized TR's allow faster retrieval than messy TR's. Fourth it may mean that the TR's is optimally embedded-, if a TR is just a loose part of EM — even well-organized — it may be difficult to find it back if it is not related to other episodical experiences or other cognitive information (knowledge, beliefs, opinions, attitudes or emotions). These are just four adequacy criteria. Maybe there are more. Important to stress, though, is that apparently (semantic) strategies are not only defined in terms of the kind of units or levels or sections they apply to (the categorical aspect of strategies), but also in terms of the kind(s) of adequacy they want to establish for the planned final result(s). Let us now try to review a number of these semantic strategies needed in a model of discourse comprehension. In this note, the strategies will not be formally formulated. Our main concern is first to discern the major kinds of strategies involved. Later we then will have to make these explicit, both within a partial theory and as hypotheses or conclusions of experimental work.

Strategic Discourse Comprehension

2. Sentential Semantic

33

Strategies

Since we focus attention on textual comprehension we will be brief about sentence based strategies, i. e. the understanding of words, phrases, clauses or whole sentences. These have been extensively discussed in the psycholinguistic literature. 4 The upshot of theory and experiments at this level is that (i) semantic interpretation makes use of various kinds of surface structure data (syntactic structures, lexical items, stress and intonation, etc.), semantic data (previous interpretations), world knowledge (expectations, probable facts), and pragmatic information (what speech act is now probably being performed?), (ii) these data are not reviewed level by level, but in an optimally combined way, and (iii) they are monitored by the final result (a semantic interpretation) aimed at, or rather by the probable schema of such a result (conditions on 'meaningful' interpretations). This means, for example, that if the final representation should feature the semantic category of an Agent (e. g. because the text and context suggest an action discourse, see below), surface structure will be scanned for word order (initial noun phrases will often express Agents), categories (Agents are usually expressed by noun phrases), lexical information (Agents are often human individuals), and so on. It will be assumed here that semantic representations have schematic FACT structure. 5 That is, we do not simply represent the meaning or content of sentences in terms of lists of (atomic) propositions, but rather in terms of a functional diagram, featuring a major predicate (representing a state, process, event or action), the sequence of Participants, such as Agent, Patient, Object, Instrument, etc., together with their Modifiers, and finally categories of a 'circumstantial' nature, representing the possible world, time, place or conditions of the major event, state or action. Such a FACT is a cognitive representation of a fact denoted by the sentence (or clause). This may mean that the linguistic information as such may not be sufficient to fill the slots of the schema: world knowledge or contextual (episodic) knowledge may be required to fill empty slots if necessary.

4

5

The notion of (grammatical) strategy, especially for syntactic decoding has been used especially by Bever (see Fodor, Bever and Garrett (1974). For a survey, also of semantic strategies of sentence comprehension, see Clark & Clark (1977). In order to avoid confusion with the ontological notion of a 'fact', we have in the meantime abandoned the use of the cognitive notion of FACT, and have simply replaced it by '(complex) proposition'. This means that such a complex, schema-like knowledge structure may have (atomic) propositions as its terminal categories. The notion of FACT was originally used in order to account for the intuitive insight that people have some cognitive representation of 'facts' in the world, and such a cognitive representation is usually more complex than the propositions we know from philosophy or logic. Both for theoretical and practical reasons, we will assume that a FACT, or what we now call a complex proposition, is verbally expressed as a non-complex sentence or clause in natural language.

34

T. v. Dijk

Now, important for our discussion is the assumption that sentential strategies are co-operating with textual strategies. Sentences usually do not come alone but are part of monological or dialogical discourse. Hence, after the first sentence, their interpretation, and hence the strategies involved, will depend on (i) information expressed by or implied by previous sentences (ii) expectations about 'what will come next' derived from previous sentences (FACTS), (iii) strategies which have proven to be successful in previous interpretations of the same text, (iv) over-all goals (in terms of final interpretations and their adequacy) of the text as a whole (v) over-all goals of the speech act or interaction as a whole. These strategies go vice versa: in order to interpret specific sentences we need textual information (or expectations), and in order to understand larger textual structures, e. g. sequences, we need information from the individual sentences. In both cases (directions) the processes will be both top down and bottom up (data driven), because world knowledge and structural knowledge will monitor understanding both at the sentence- and at the text-level, whereas in both cases also the actual information of the text is needed to fill the schematic hypotheses or to match the predictions or expectations.

3. Local

Coherence

As soon as we go beyond the sentence (or clause) boundary, the language user will have to solve a first strategic problem, viz. to establish coherence between the clauses or sentences interpreted along the (strategic) lines suggested briefly above. Of course, the major monitoring strategy is the very assumption that the subsequent sentences of a text are coherent. This assumption is the so-called 'normalcy' condition on semantic interpretation, which is so strong that even in the face of lunatic, literary, children or different culture's discourses, which need not be coherent-for-us, we at least try to establish an interpretation, i. e. to link sentences meaningfully. The general processing model for these local coherence strategies assumes that each 'new' clause or sentence, after interpretation, is being linked to a or the previous sentence representation. If such a link is possible and adequate (satisfies a number of coherence conditions) the conclusion is that that piece of the text is coherent, after which new information may be decoded and interpreted and linked to the previous information, and so on. This process is cyclical because of the memory limitations of the STM buffer, the capacity of which does not allow directly available storage of more than a couple of sentences (say two or three — complex — FACTS). Once a coherence link has been established, the oldest information may now be transferred to episodic memory, from where it may be retrieved immediately if needed for further coherence links of a more indirect nature.

Strategic Discourse Comprehension

35

This is more or less the theoretical process involved. It however is still very rough and not very strategical. In order to allow fast coherence interpretations we should assume that the process is more complex and more flexible at the same time. Before we specify some details of these local semantic strategies, it should however be noted what kind of textual representation we are after. We have seen earlier that this final 'goal' of understanding is a major monitoring device for the application of strategies: if the language user does not know what to construct at all, how can (s)he possibly do it, let alone do it optimally? So, I will here assume that understanding discourse is also and at the same time understanding the world. This aspect of semantics, somewhat neglected in cognitive psychology, is important for several reasons. Also, we need comprehension of the referential aspects of a text, i. e. what it is about. That is, it should be borne in mind that the speaker not (only) wants that (s)he is understood, or that the text is understood, but also that the hearer/reader changes his/her knowledge about the world according to the information of the text (and speech act performed by the utterance of the text), and possibly that the hearer changes his/her opinions or attitudes with this knowledge and/or starts planning some action. In other words, it is essential that the reader/hearer not only constructs a good representation of the text 'itself — so that later reference can be made to what was actually said — but also an adequate model of the situation (MS), that is of the complex sequence of states (a scene) or sequence of events and/or actions (an episode), as denoted by the text or otherwise, e. g. paratextually, referred to by the speaker. 6 Compared to the situation model thus being constructed by a reader/hearer, the textual representation may be relatively incomplete or fragmentary. That is, we have processes of various kinds that retrieve or construct MS's from TR's together with contextual and more general world knowledge. After all, it will happen often that a text is about the same or a similar model, and in that case we only need to retrieve that model, or actualize an instantiation of it from LTM in case the model is stereotypical (restaurant, parties, etc.). Thus, we assume that in episodic memory a language user will principally try to 6

The notion of a 'model of the situation' is developed in detail in van Dijk & Kintsch (1983). It is a cognitive correlate of earlier ideas, in model theoretic semantics of discourse, e. g. in Ballmer (1972), van Dijk (1977) and Petofi (1979) among others. In psychology and AI, the idea has been also briefly put forward by Stenning (1978) and Johnson-Laird (see e. g. Johnson-Laird & Gamham, 1978; Johnson-Laird, 1983). There are many important cognitive arguments why such a separate model of the situation (besides a textual representation) is imperative, e. g. an account of reference and co-reference, the use of previous experiences, the integration of knowledge, the frequent implicitness of discourse, and so on. Also, models which have often been used and therefore 'verified' provide the basis for learning, viz. by generalization and decontextualization. Conversely, general semantic knowledge about episodes or situations may in turn, by instantiation provide the 'general' backbone of a situation model, and therefore may serve as expedient retrieval strategic information to find the relevant model in episodic memory.

36

T. v. Dijk

construct a semantic model by means of the TR also being constructed. Of course, instead of a full 'double representation' system, we may assume that the semantic TR is some form of 'graph' within the situation model, possibly signalled with various surface structures (and schematic structures, see below), which tell us what is 'new' in the model. For our local coherence strategies this means first of all that we will not simply try to connect two FACTS (or propositions) according to the semantic rules of discourse, but also, and above all, that we check in the model we already had before (experience, previous texts) or which we are now constructing, whether the two or three FACTS involved can be connected at all in the situation model. Sometimes this connection is so obvious, or has been pre-programmed (episodically or in LTM, e. g. as more general script fragment), that further analysis of TR or of the model (or some variant of the model), is not necessary. This is of course a very fast strategy, but may lead to obvious errors if no minimal side-checking is performed (same world?, same situation? same participant?, etc.). Against this background of episodic memory representations of both text and world, we may now try to distinguish a number of strategy types aimed at 'drawing a good picture', i. e. a picture that is clear, has good likeness with the 'real world', is more or less complete, detailed, and useful (relevant). 1. A first local strategy is obvious: if a sentence in a textual sequence should be linked coherently with a previous sentence, then it is not plausible that it will first be interpreted completely. On the contrary, as soon as some first semantic category has been established, mostly a Participant (e. g. an Agent) or a Circumstantial (time, location or manner), this category may be matched with the representation of the previous sentence. Thus, a pronoun or definite noun phrase may be linked to the same individual represented in the model — e. g. some person — as an individual already identified by a noun phrase in the previous sentence. The pronoun and especially a full (definite) noun phrase will in general be sufficiently specific to (re-)identify this discourse referent in MS. If not, further information must be gathered to identify the correct referent, e. g. by analysing the most probable predicate which would fit the individual in MS, i. e. the FACT as a whole in which the individual plays a role. Note that if the next sentence is introduced with a 'new' individual expression (signalled with an indefinite noun phrase), this may mean that the individual should be constructed in MS for the first time: no need to reidentify an individual already in MS. Similarly, time or place expressions may signal that we no longer are within the same episode, but have to change to another situation (and — see below — should construct another macroproposition): the next day, in London, in the other room, thus, are expressions which may change the scene for the facts being talked about. Hence, we have partial interpretation of a next sentence, identification of the proper semantic category, and then as soon as possible the establishment

Strategic Discourse Comprehension

37

of a coherence link with a previous sentence representation. This may mean, for instance, that the next FACT, now under construction, takes place in the same situation (time, location, world) and features the same participant(s). Of course, minimal (neither sufficient nor necessary) coherence is important as a first step (!) in the linking process of the subsequent FACTS as a whole. Note that it is precisely this strategy which presupposes the linguistic topiccomment articulation of sentences. The topic contains typically the semantic information that serves as the first-coherence-link with previous discourse: usually expressed by pronouns or definite articles, it mostly will (re-)identify the individual(s) that will also appear in the FACT being constructed, whereas the comment then will denote that part of the new (next) FACT that has not yet been identified before, and for which, therefore, another kind of coherence strategy is required. This kind of link between sentential surface structure (word order, syntactic categories, but also intonation and special morphemes), semantic representation of sentences and coherence strategies not only holds for (expressions of) individuals, but also for propositions or (embedded) FACTS as wholes, for which subordinate initial clauses, nominalizations, etc. may be used to signal that this piece of information denotes part of the MS already denoted before. In general then, topic-comment articulation, presupposition and similar phenomena, serve as discourse coherence strategic data, which 'point at' the location of MS the text (or TR) is now 'about'. This first local strategy has all the typical characteristics of a strategy: (i) it is monitored by structural characteristics of its 'goal', viz. TR/MS (i. e. identification of some Participant) (ii) it uses various kinds of data, also surface structure data of the (next) sentence, now being processed, (iii) it is hypothetical, because the assumption may be revised after further information of the sentence, (iv) it is effective, because it serves as a first (provisonal) coherence link between two FACTS even if the second FACT is not yet entirely known. 2. What next? What is the second strategic step in the establishment of semantic coherence between subsequent sentences of a text? Assuming that a provisional link has been established by the (re-)identification of one of the participants in the next FACT, the other elements of this FACT should now come into the picture. In this stage it is important that this next FACT gets its usual hierarchical structure. This means that the major predicate must be identified, i. e. the state, event or action in which the (given) participant is involved. For English this implies that the (main) verb of the Verb Phrase is being interpreted, establishing at the same time the over-all schematic structure of the FACT, that is the various roles (cases) which can be expected. After a brief check with MS whether the first-linked participant can indeed be in a role — e. g. Agent — of this schema, a link must be established between the main predicate and the previous FACT. A first information we have comes from morphology: the tense ending of the verb will indicate whether we still are within the same time dimension of the situation (scene

38

T. v. Dijk

or episode) established by the previous FACTS, whereas person indication (singular, plural, 1st, or 2nd/3rd person) will indicate whether the participant is subject of the verb, and hence maybe indeed Agent of the action, or Experiencer of the event denoted by the verb. These are the usual sentence comprehension strategies, but it is interesting to note that they are textsensitive: they do not merely indicate time or role, but also 'same time' or 'same participant' in relation to previous FACTS denoted by the text. Proper predicate coherence establishment is a strategy that needs world knowledge. In order to decide whether some event or action can be linked to a previous one, we must know what possible relations in the world can obtain for such events or actions. Although there is a variety of such relations, we may summarize them as conditional and functional relations, respectively. Conditional relations are those that enable, make probable or necessitate facts, given other (previous) facts. Functional relations are rather intensional; they specify what kind of information linkage there is (e. g. general vs. specific, statement vs. example, contrast, etc.). For the conditional relations, the strategy then must establish the link both ways: is the previous predicate (or FACT) a possible condition, a probable condition or a necessary condition for the actual predicate (or rather FACT-schema we are now constructing), and, conversely, is the actual predicate, or rather the actual event or action or state, a possible, probable or necessary consequence of the previous fact(s)? This answer may be supplied in a variety of ways, some strategically based, others rule-based. First, the relation may match with an expectation already generated by inference from previous facts/FACTS: if somebody is said to have a gun with him (in a crime story), the expectation is that it may actually be used, an expectation which may simply be 'verified' by the interpretation of the predicate of the actual FACT. Secondly, episodic information in which the FACT is embedded may provide this kind of match: we simply remember from previous texts that there is a world picture fragment in which Americans may use their power when their oil supplies are menaced, so if we read about army movements of the USA after some blockade of a crucial tanker line, this link will be verified as 'coherent'. Third, more general world knowledge — derived from episodic knowledge — may provide more systematic and stereotypical links between the predicate and the previous facts, e. g. in the case of frames and scripts: usual properties of previously identified individuals are now matched, or subsequent stereotypical events or actions in which they are engaged (in the usual restaurant, party, supermarket or breakfast scripts). This is where the strategic data originally come from, but part of them may already be part of MS (actualization of episodic memory or explicit expectations derived from previous parts of the text). This means that in many cases the coherence check for the predicate can be read o f f directly, so to speak, from MS. In other words, this fragment of TR can be accepted as coherent if the predicates are already linked (conditionally or functionally) in MS or if such a link can be established in MS given further data from EM or LTM (scripts, etc.).

Strategic Discourse Comprehension

39

Note that predicate coherence, as we suggested, goes both ways: the strategy must check whether the actual event or action is a possible (or probable or necessary) consequence of the previous one, and at the same will check whether the previous one was a possible, probable, or necessary condition for the actual action or event: (1) (2)

John fell from a chair. He was dead. John fell from the Eiffel Tower. He was dead.

In the first case, coherence presupposes that the second fact (state) is a possible consequence of the first. This is the case, but a rather unlikely one. In the second example, John's being dead is a consequence of a rather unusual condition. In other words, in (2) the consequence may easily be predicted given the first sentence. In fact, most texts will not even bother to mention it. In (1) this prediction is highly unlikely, and we will after the second sentence expect information about what happened exactly: how he fell, why he died, etc. In (2) we will want more information not about the consequences of the first sentence but about its conditions: how was it possible that John could fall from the Eiffel tower? But once given the first sentence of (2) the coherence match for the second one is easy: it will simply check a probable expectation inserted into MS. We may on the basis of this interpretation strategy, and for a given person or group or culture, define the news value of each textual continuation in terms of the probability that a link between the facts already has been or can be inferred from the previous FACTS, together with other episodic and general knowledge information. The higher this probability the lower the news value. This strategy presupposes other local coherence strategies of a more general nature. We already assumed that tense and person endings (morphemes) signalled same time and same participant. We must also assume, however, that next sentences denote next events or actions (or resulting states). If not, this must explicitly be signalled by appropriate tenses, adverbials or connectives (earlier, before, the previous day ...). Later mentioned previous conditions are functionally interpreted (if world knowledge does not provide a possible consequence link) as in: (3)

I had a car accident yesterday. The brakes did not work.

where the situation of the brakes is not interpreted as a consequence but as condition of the accident (by scriptal information), which means that the coherence link is functional: the second sentence is used as an explanation of the (fact denoted by the) first sentence. In less clear cases appropriate connectives must signal the particular relation between the predicate (new FACT) and the previous FACT: for each conditional relation (both ways) we have several connectives (depending also on subordinate/main clause structure). They typically will introduce the sentence or the clause, even before the topic (Subject, Agent, Noun Phrase)

40

T. v. Dijk

information re-identifying the relevant old or new participants. This means that first a link is established for the FACTS as wholes, in other words that the kind of relation is now relatively important (or not predictable or inferrable given the two FACTS). Strategically, connectives are very important. Neither identical participants nor possible consequences as such are sufficient for the establishment of coherence: what we need is a relation between the FACTS as wholes; and the connectives, all other things being equal, are the most effective signals for the nature of this relationship between the facts (the FACTS in TR or MS). Note, furthermore, that the link between a predicate and previous FACTS need not be with an immediately preceding FACT, i.e. with the FACT denoted by the last clause or the last main clause, although this may well be a condition for the strategy, not for the rule of linear coherence. So, if no link can be established with the last (main) clause (FACT), there should be a strategy that effectively finds the appropriate FACT, either in the STM buffer, or by reinstatement from EM. With the double information at hand (looking for the most appropriate condition for the actual consequence, and vice versa) this search (in MS/TR) may be effective because there will be few possible candidates at such a local range for this kind of 'double bind'. Of course at longer text ranges this number of possibilities will be much greater. But in that case, there is a textual production strategy which 'recalls' previous facts, by presupposition or in general subordinate clauses: (4)

Because the brakes of my car didn't work, I was arrested by the cops.

As for already introduced individuals such initial, subordinate clauses will simply be used as a strategic cue for the reinstatement (or search in MS) of the relevant condition of the new FACT now being constructed. 3. The next strategy, or set of strategies, will complete the coherence analysis of the next FACT. That is, once established a link for the main predicate, we also have the provisional schema of the FACT as a whole. Essentially, this means that the other participants in the FACT are reviewed for possible coherence links. Again, definite noun phrases or pronouns will be cues in this case. I don't know whether there is some ordering in the strategic construction of the (next) FACT and hence in the establishment of (further) coherence with the previous FACT. Maybe normal surface ordering of English (direct object, indirect object, complements, representing semantic Object or Patient, Goal and other roles respectively) is a cue for strategic ordering. Once analyzed the sentence as a whole, and constructed the full next FACT, the foregoing strategies may briefly be checked for adequacy. That is, was the correct coherence established as assumed or did the full FACT yield information that should revise the original hypothesis? After all the first noun phrase (subject) might turn out not to be the Agent (or not

Strategic Discourse Comprehension

41

the sentence topic) and hence not the primary participant being the 'bridge' for the coherence relation at all. 4. Local coherence is a function of global coherence. Subsequent propositions or FACTS may well be connected, according to the strategies mentioned above, but this is no guarantee for over-all coherence: there may be no line or orientation in the coherence or, in other terms, there is no topic or theme 'under' which the respective sentences are connected. This means that the next local strategy is a check with the 'ongoing' macroproposition: does the next FACT belong to the same topic, or is another topic being initiated? In fact, if no immediate coherence can be established by the strategies mentioned above, this may mean that a new topic is introduced, and that the next sentence/FACT is the first of a next episode. Part of this strategic control on global coherence has already been performed: if the next FACT is expected on the basis of world knowledge this may mean that this FACT is indeed also globally coherent if dominated by the same macroproposition as the (more general) one dominating the script in world knowledge. Concretely, we know that 'John bought a ticket' is not only possible after the sentence 'John went to the station', but also that it is a next component of the series of actions together defining the global action of 'John made a train trip'. This global coherence check may again be strategic in the sense that not the full sentence must have been understood before part of the strategy already applies. Given an Agent and the main Predicate, this may perhaps already give a cue for provisional macro-match, e. g. if the Agent of the macro-action is the same person, and if the predicate appears in the same script. 4. Global Coherence:

Macro-Strategies1

In the last strategy of the establishment of local coherence already appears the necessity of establishing also global coherence. According to the theory of discourse understanding, this means that language users will try to derive a global theme or topic from fragments of the text. In other words, they will construct semantic macrostructures. Abstractly, macrostructures are derived from the text base by the application of macrorules to sequences of propositions, both those expressed by the text 7

The notion of global coherence, theoretically reconstructed in terms of semantic macrostructures, has been dealt with in detail in much of my previous work, e. g. in van Dijk (1977) and especially in van Dijk (1980a). We here merely suggest, rather succinctly, that our actual view on macro-structure formation, or macro-inference, is much more strategic: we already derive macrostructures (or themes, or discourse topics, or gist) as soon as we have understood a first clause, or even as soon as we have understood the actual communicative context.

42

T. v. Dijk

and those inferred from general knowledge. However, strategically this would be very impractical: a reader will seldom wait until the end of a passage or whole text before inferring the theme or topic. Not only (s)he will want to know as soon as possible what the text is about, but also a provisional macrohypothesis is necessary to do the control, mentioned above, of local coherence. Finally, there are memory capacity constraints that force the reader to 'reduce' the textual structure as soon as possible, which means that the text base will as soon as possible be further organized by means of higher level macropropositions. The first question is: when and how does this 'provisional topic assignment' take place? Strategically of course this would require 'as soon as possible'. If a first sentence of an episode or text already allows a probable global topic, this topic will be taken as the provisional topic for the sequence as a whole. In general this may mean, again, the application of world knowledge because if the first sentence may be matched with a script (or episodic information), the corresponding macroproposition may be derived immediately: (5)

John went to the station. (...)

This would perhaps generate already the macroproposition 'John is making a trip by train', a hypothesis confirmed by a next sentence like 'He bought a ticket' and disconfirmed by the next sentence 'He bought a newspaper at the newsstand and went home again'. This will in general hold only for those sentences that express propositions which are components of stereotypical scripts. If the first sentence does not yet allow a full macro-inference, the strategy may also be partial: (6)

John got up at seven o'clock.

In this case we may of course first construct a macroproposition about the getting up or breakfast ritual, but it is unlikely that a story will be about that. So, these may at most be sub-topics introducing a major topic. As a provisional hypothesis however it may be assumed that John is a participant, even perhaps Agent, in the main macroproposition (MACROFACT). Of course the story may after all be especially about Mary, but still it would be strange if such a story would be begin with (6). So, we assume that a macroproposition is inferred as soon as possible from the first sentence or sentence pair, or that even parts of this macroproposition may already be constructed if the first sentence(s) does not yet allow complete macro-inference. We have observed also that world knowledge is crucial in this strategy, because it must provide expectations about what may come locally and/or what 'is the case' globally, against the background of scriptal or otherwise known information about well-known components of a larger theme.

Strategic Discourse Comprehension

43

Macro-strategies are very complex. Let us merely give a summary of the most plausible ones. As we see they not only are based on semantic information, but also on other kinds of information, such as surface structure or contextual information: a. Expectations of the hearer/reader about the probable topic of the discourse inferred from knowledge of the hearer, the specific situation (e. g. a class, train, or doctor's), the interaction context (the approaching stranger in the street apparently having lost his/her way). b. Information from previous texts in memory in similar situations, of the same person. c. Title, headings, summary, lead, etc. at the beginning of the text or the episode, all expressing directly a macroproposition or part of it. d. Thematical sentences, key words, italics, etc. expressing or signalling what the passage is about. e. Rhetorical structures of the passage, which may partly predict aspects of (next) semantic content at a global level. For instance a Contrast may organize opposite themes, a Parallelism may organize similar themes, and so on. f. The schematic structure of the text, if stereotypical and hence preprogrammed in the reader/hearer, will normally impose constraints on its macro-content. This means that we may have expectations about at least the nature of the macropredicate or participants (e. g. an action, human agents, etc.). This holds not only for narrative structure, but also for argumentative structure, newspaper stories or psychological research reports. We see that the basis for the inference about what to expect globally is rather comprehensive: seldom a hearer/reader will only have to judge after the first sentence alone; the context type, the text type and several signals (announcements, titles, etc.) already give partial or full suggestions. Once established a partial macrostructure, the inference of further macropropositions is of course still easier, because then the usual 'local' coherence conditions come into play: each macroproposition must of course, for each macrostructural level, be normally connected with previous ones. Inference from world knowledge and these previous macropropositions will, together with the other cues of the strategy, yield a still better prediction of what will come 'globally'. Clearly, this is all a combination of top down and bottom up inferencing. The 'real' assignment of the (next) topic or theme will of course be possible only after having read part or all of the sentences of the episode that 'expresses' the topic or theme. In other words, at each step the hypothesis is more (or less) confirmed, and finally accepted (or rejected).

44

T. v. Dijk

5. Schematic

Strategies8

In this model of discourse processing it will further be assumed that at least certain classes of texts have so-called 'schematic' (super-)structures. This means that macropropositions of the text may have a global, conventionalized 'function'. Conversations may have introductory categories (e. g. Greetings) or final categories (Leavetaking), stories may have Setting, Complication, Resolution, Evaluation and Coda, and psychological papers have conventionalized global categories such as Theory-Hypothesis, Experiment (Subjects, Materials, etc.), and Discussion. Arguments which have been brought forward against the development of formal systems (e. g. 'grammars') for this kind of schemata are not valid. Conventional categories of this kind cannot be explained away by reducing them to local or global constraints on, for instance, action structures, such as motivations or goals. Local constraints on action descriptions are not the same as the global categories of a more conventional nature. If so, an action description such as (7)

John was hungry, so he bought himself a hamburger

would be a well-formed story, because we have a motivation and the realization of a goal. That stories have a Complication category is not motivated by the action structure, but a conventional fact of narrative discourse in our culture. Thus, we assume that language users will try to fit the macrostructure of a text — which has little further structure apart from linear coherence connections — into such a conventional schema. This means that as soon as possible they will try to decide to which schematic category a part of the text belongs. Since the schematic structure is conventional, a first hypothesis is that the text has canonical, that is normal global structure. If the first sentences of an episode can be interpreted as a macroproposition which fits a first category of the schema, e. g. the Setting of a story, then it will be assumed that the present part of the text is indeed the Setting. Next, schematic categories will need some signalling, because the language user must know where the first schematic category ends and the next is operating, e. g. a Complication after a Setting in a story. This means that if a sentence or a sequence of sentences (an episode or part of it) does no longer fit the macroproposition which has the schematic function of the first category, a next category may be actualized (since the schema is conventional in a given 8

For the notion of (schematic) superstructure, such as the formal structures defining the conventional and categorical over-all 'schema' of stories, see e. g. van Dijk (1980a, b). For psychological discussions, see e.g. the papers collected in the Poi/iVj-issue (van Dijk, ed. 1980). There is a fierce debate between proponents of the cognitive relevance of such schemata ('story grammarians') and those who prefer to account for story structures uniquely in terms of action and event structures and goals, notably Al-inspired researchers (see e. g. Schank & Abelson, 1977; Wilensky, 1978, and the Poetics issue).

Strategic Discourse Comprehension

45

culture this processing has top down aspects). The inferred macroproposition will be matched with the constraints of the schematic category. Thus, a Complication must be filled with a macroproposition that denotes a problem, a difficult task, an event or action preventing the normal development or goals of a person or group. This semantic aspect of the strategy may be completed, as is usual for strategies, with surface signals. The episode may be marked, first, with paragraph indentation, pauses, or chapter segmentation, possibly with appropriate titles, headings, or announcements. Secondly, a next episode may be introduced with global connectives such as Unexpectedly, Suddenly, But, etc. 6. Strategies for Knowledge Use9 It has become a truism that discourse understanding is impossible without the actualization of large amounts of knowledge. Texts typically are incomplete, or, in other words, much of the information is 'presupposed' or implicit in them. To understand sentences, to establish local coherence relations, to derive macrostructures or superstructures, we need information from our world knowledge as it is stored in LTM. It also hardly needs to be repeated that to address these vast amounts of information effectively, this knowledge must be organized. Concepts, thus, are related to other concepts that belong to the same 'domain'; clusters of concepts be organized in frames, scripts, episodes, scenarios and similar higher order structures. These structures, such as all those of LTM (semantic memory) have a general, abstract and stereotypical nature. They must be applicable to many situations, so they may not be too particular, too rigid or too precise. They must be flexible and general, allow fast search or flexible actualization. What we know little about though is how precisely this organized knowledge is used in discourse comprehension. We do not know how much is actualized (or activated), what strategies are involved in search and application, what information is not only 'used' but also 'represented' in a text or world representation in EM, and so on. So, on the one hand we have the lower bound of comprehension requiring a minimal activation of knowledge so that sentences can be understood and 9

The structures and functions of knowledge representations, and in particular their fundamental role in discourse comprehension, has undoubtedly been the major contribution of Al-research on discourse. The classical reference for this kind of work has become Schank & Abelson (1977), that is for the notion of 'script', although other notions have been used before, such as 'schema' already by Bartlett (1932) and, later, by Norman & Rumelhart (1975). See also Bobrow & Collins (1975). In the meantime, the notion of 'script' has undergone several modifications. In fact, it has become more flexible, and Schank and others now also use the notion of MOP, for Memory Organization Packages, which have a more episodic nature, and which can be used to actually 'construct' stereotypical, script-like episodes.

46

T. v. Dijk

minimal coherence established. On the other hand we have the upper bound, rather flexible and depending on subject and situation, consisting of the maximum amount of knowledge being activated for 'normal' understanding (interpretation) of a given text. A text contains expressions for many concepts but we know that it is extremely unlikely (or even impossible) that a language user will actualize everything (s)he knows about the concepts or the schemalike structures in which the concept is involved. To wit, in a story about a birthday party, we may have it that juice is lacking and that John briefly goes to the supermarket to get some. In order to understand the text we only need to know in this case that a supermarket is a public shopping place where one can get food. All we know about supermarkets or shopping is not necessary in the understanding of the story. Apparently, then, a first knowledge use strategy is that knowledge is addressed in so much detail as may be used for the understanding of sentences belonging to the topic (macrostructure of the text). In our example this means that we need to actualize much more about the birthday party-script. As soon as, unexpectedly, an apparently unimportant episode appears to become more important (e. g. John forgot his money when he went to the supermarket, or there was a hold-up when John was there) the script may be actualized after all. I will call this strategy the level correspondence strategy: if in a text only high level descriptions (of events or actions) are given, only high level script information is needed, and if details are given in the text, knowledge will be addressed until the corresponding level of detail in the script. The correlate of this hypothesis runs conversely: if a concept or proposition is a higher macroproposition of the text (a more important theme), then the knowledge scripts or frames associated with this concept will be addressed 'deeper' (in more detail). Yet, this is not sufficient. We may indeed distinguish between 'activating' (addressing) knowledge on the one hand — so that it is ready for use — and the actual use or actualization, or application of knowledge items. Many aspects of stereotypical knowledge will thus be activated without being used. Another question pertains to the constructive aspects of used knowledge: how much may be (partially) constructed as part of a representation of a text or of a situation? Again, this requires a strategic solution. Too powerful and too costly, apparently, is a strategy actualizing all the addressed knowledge and constructing all of this in some representation. On the other hand, if we would only have ex post factum actualization of script-like information, i. e. only if really needed for a match or the establishment of minimal coherence, script information would be of little help in understanding, and no predictions/ expectations or top down processing would be possible or plausible. So, the expedient strategies must be operating between these two extremes. Before we can answer these questions, two remarks are necessary about issues a bit overlooked in script theory. First, a script is necessarily a general,

Strategic Discourse Comprehension

47

abstract, and stereotypical kind of knowledge structure. So, direct application of scriptal information (e. g. propositions) is impossible. What we first need is instantiation or particulari%ation\ the variables of the scripts must be replaced by the constants of the text (according to global constraints on possible values for each category): that a person may get drunk at a party should be instantiated into the possibility that John gets drunk. Secondly, the information of world knowledge, by its stereotypical nature, will in general (in instantiated form) be only 'background information' in a story or other text type: no need to state the obvious or well-known. The few propositions which are expressed in the text will in general just frame the other ones. If Peter goes to the station, buys a ticket, or chooses a seat in the train this is just to (i) signal that he is making a train trip, (ii) to ensure that a more or less plausible description of the 'normal world' is given, but above all (iii) to be able to identify the really interesting facts of the story, e. g. that he gets an accident, falls in love with a girl he meets, or something else narratable. In other words, the realization of a script will in general precisely not define a real text. They are not the text but rather the background texture that makes the other facts understandable, coherent, etc. Against this background we may try to answer the tricky questions about the amount and manner of actualization and integration of knowledge, i. e. its use in STM and storage in EM, respectively. Corresponding to the level strategy, which states that only that level of knowledge representation will be addressed that corresponds to the level of description, we will next assume that the 'amount' of knowledge actualized at each level corresponds to the degree of completeness of the given passage. 10 That is, if the passage apparently gives much detail (instead of just selecting some details) — at each level! — then it is a plausible strategy that a corresponding 'degree' of pre-programmed information is actualized and used for creating possible expectations. Thus, a detailed description of a restaurant scene will simply make use of more known details of the restaurant script. Of course this completeness strategy is linked to the descriptive level strategy: the lower (more precise) the level of description the more details we will usually get. Yet, the link between the two strategies is not direct: there may be stylistic variations in the degree of completeness for each level. Then, we have the integration problem. The question is: is knowledge once actualized and used in STM for any task of discourse comprehension, for instance local coherence establishment, also integrated into the episodic text representation? In the Kintsch & van Dijk (1978) model it was assumed that all and only those propositions are integrated that are necessary to establish minimal coherence, i.e. for the identification of referents or for necessary conditions (presuppositions) for interpretation of propositions expressed by 10

The notions 'level of description' and 'degree of completeness' have been introduced in van Dijk (1977) in order to characterize properties of semantic representations of discourse.

48

T. v. Dijk

the text in general. Although this seems theoretically correct, the strategies involved may be more flexible. Instead of (re-)identifying a referent or a fact from the TR in episodic memory, the language user may simply check quickly whether 'this kind of referent or fact' may appear in 'such a situation', that is directly check with scriptal knowledge, without however integrating this knowledge into TR. Only for very specific, non-stereotypical information the reader may be forced to check with the preceding part of the TR. This strategy is faster if we assume that search in stereotypical knowledge in indeed easier than a check in a unique, perhaps very complex TR. To solve this problem I have assumed earlier, that a double system of representation seems necessary in EM: both a TR and a representation of the situation-event-action denoted by the text. In that case the TR may be relatively economical, and the world representation (MS) may contain (i) a scriptal framework, actualized and instantiated from LTM and used to search for memory traces of the same kind in EM, (Jii) detailed inferences induced by MS-information, (iii) personal associations and evaluations or previous experiences with previous situations or texts about such situations of the same kind. This MS, unlike the TR, need seldom be constructed from scratch. On the contrary, it consists of actualized or 'remembered' fragments, scriptlike or more accidental memories which need only be enriched with the actual information of the current text — as represented in TR. Reference, missing links, relations between FACTS — and hence conditions of coherence, and macropropositions, may be supplied by SM: as long as the picture itself remains consistent with LTM knowledge and previous experiences, the text will be acceptable and understandable. 11 The TR itself in that case may be much more idiosyncratic, feature surface structure (stylistic) phenomena that are interesting, specific schematic or semantic organization that is not pictured in MS (e. g. literary or rhetorical transformations), etc. If these more specific textual structures do not have links with MS or with other relevant (opinion, attitude, emotion, evaluative, interactional, etc.) functions, memory for it will be short-lived, and after longer delays only the MS-based semantic information will be retrievable as has been shown in all discourse experiments since Bransford & Franks (1972), or even since Bartlett (1932). Indeed, those elements of MS not corresponding to elements in TR will also be 'falsely' remembered. Instead of constructing, each time we read a (new) text, a new TR which also features a lot of knowledge and a lot of personal memories,

11

This hypothesis has been tested, in contrast with the predictions of the Kintsch & van Dijk (1978) model und those of Schank & Abelson (1977), by my colleagues in the Amsterdam project on Text Processing. See den Uyl & van Oostendorp (1980). It was found that, depending on the kind of text, the kind of task and the kind of reader, the strategies used may vary, viz. 'complete' integration (in TR) of the activated knowledge (as according to Schank & Abelson), only relevant activation (as in Kintsch & van Dijk), or fast, 'sloppy' checks on available stereotypical knowledge structures.

Strategic Discourse Comprehension

49

it therefore seems plausible to account for these phenomena with a combination of textual and world representations of which the latter is partly 'old'. The precise format of this double representation does not matter here. Instead of assuming two independent networks it might be more realistic to assume that the two structures are combined in some intelligent way. This means that some nodes and/or paths of a representation tree for the situation may be assigned an extra signal 'text', meaning that this aspect has been expressed by the text (in a specific way). In addition those nodes or paths not yet known or remembered are now established for the first time by textual information. The picture we get then may be as follows:

Schema 1

In this schema the solid lines represent links between concepts already known from general knowledge (scripts, frames) or known from previous experiences, whereas the dotted lines are those added by the TR. Similarly, circles denote concepts already known and triangles concepts now being inserted for the first time into the combined TR/MS. One argument against this kind of combined representation is the possible transformations in the text. Of course, these may be 'reduced' to normal form in TR/MS, but experiments have shown that both surface structure style, some syntax and lexical items as well as schematic ordering may be recalled or recognized as such, for at least some time. 12 In that case we would need a teal double system with appropriate lines of correspondence. As soon as some concept or proposition is missing in TR it is checked in MS whether it occurs there. The major strategy we now seem to have is apparently not one by which world information is integrated into text representations, but rather text information that is 12

The specific case of 'surface structure' memory for text has been explored in many studies in the last few years (see van Dijk & Kintsch, 1983, for references). See Keenan (1976). See Mandler & Johnson (1977) for details about memory for (narrative) structure.

50

T. v. Dijk

integrated into world representation. Although the precise details of this strategy are still obscure, the over-all approach seems rather plausible: we hardly ever read texts 'for their own sake' (except in literary communication or ritual situations), but rather in order to know something about the actual context or the world in general. The prediction in that case for the d i f f i c u l t y or readability of a text is that the larger the correspondence between TR and MS the easier it is to understand the text. In other words, texts about situations we have little general knowledge about but also situations we have not read about or experienced at least partially before will be more difficult to understand, because in that case (i) there is no 'background knowledge' for the numerous matches each text comprehension process needs, (ii) the construction of MS must be nearly from scratch. So, a story about an everyday accident or incident will be easy to understand because it features many actions, events, objects, locations, and participants we already are familiar with. But a description of a more technical nature or a description of an expedition into strange countries or worlds (as in science fiction) will be on the whole more difficult, even if the concepts as such are not unknown. Not only or not so much the concepts themselves define the difficulty of a text, but rather the newness of their combination in certain episodes or scenes. In fact, the same holds for grammatical structure, semantic structure, schemata, style, and so on. Although this criterion of 'familiarity' in wellknown, it has not yet been integrated explicitly and from a strategic point of view in actual discourse comprehension models. It seems to be a consequent extension of the frame- or scriptbased models of knowledge use, and at the same time it yields a sound basis for the very acquisition of frames and scripts: once combined several textual or other 'experiences', the common features will slowly be generalized and abstracted and integrated as 'general' knowledge into LTM. The model of knowledge actualization we here sketch also provides a plausible answer to the fuzzy problem of the actualization and integration of so many personal 'associations'. It would make a TR very heavy indeed if all these associations were integrated into it. But in an MS we simply may combine TR-information with all we care to activate not only from LTMscripts (if any), but also from previous experiences (or previously read texts). Thus, reading about the recent earthquake in southern Italy, during several days in the newspapers, the representation I have of the actual respective texts will be very fragmentary or irretrievable. I simply have constructed an integrated picture of the whole event, consisting of e. g. MS: (i) general knowledge about earthquakes (and Italy) — movement of the crust of the earth — occur in certain parts of the world, where crust plates meet

Strategic Discourse Comprehension

— — — — —

51

have damaging effects on buildings and roads have different strenghts, measured on a Richter scale heavy earthquakes range around 7 on this scale and up can sometimes be sensed before by animals Italy is a country in Southern Europe, etc.

(ii) episodic events — recent earthquake in Algeria — possible other earthquakes (iii) episodic associations — memory of the villages and landscape near Naples visited last year during a vacation — memory of other scandals of Italian bureaucratic inefficiency (iv) opinions and attitudes — sad about the many victims — furious about lack of effective help (v) emotions — afraid about possible fate of friends living or visiting there TR: (i) Macrostructure:

Heavy earthquake in southern Italy Many poor people died Help very late and ineffective Barren circumstances etc.

(ii) Microstructures:

Camping cars transported from North to South Tent and blanket supplies from all over Europe Goods could not be distributed Goods were stolen A church fell down on some 80 children ect.

(iii) Other aspects:

Newspaper article, first page Pictures of villages destroyed etc.

We now assume that at least some aspects of TR may be integrated into MS so as to make the picture more complete. This will enrich our information about earthquakes, and about desasters in Italy, — and about what to expect when we visit this region a next time. Note that much of the MS may plausibly be activated and even actualized (thought about) during reading, without necessary links with TR. Yet, if we know more about earthquakes

52

T. v. Dijk

and about Italy, that region and Italian bureaucracy, it will be easier to read the text. Also since the MS is richer, it will be easier to remember it later. As we will see below, the occurrence of opinions, attitudes and emotions in MS further 'engraves' events in EM even if the actual TR maybe nearly entirely forgotten. 7. Subjective

Strategies

What we have reviewed so far are strategies — or rather sketches or types of strategies — in rather well known areas of a cognitive model of discourse comprehension. Although such a model is still very fragmentary, at least the kind of concepts and the kind of problems are by now familiar. Yet, strategies of discourse comprehension are not limited to the well known areas of the map of cognition. In the last few pages we already mentioned several elements, such as personal memories or associations in EM, as well as opinions, attitudes and emotions, that also play a role in the comprehension process.13 That is, they may be actualized and integrated into our picture of a situation described by a text. We thereby enter the vast but only vaguely explored continents of what has been called 'hot cognition' or what might also be called 'soft cognition': opinions, attitudes, values, norms, feelings or emotions, interest, etc. Whereas the actualization of general knowledge and the various processes involved in 'general' text understanding may be said to hold for most readers of a culture — with variations according to knowledge, and experience — these soft cognitions will rather define the more personal and subjective ways a reader responds to a text. The problem therefore is: what are the strategies involved in this kind of subjective interpretation of the text and the construction of 'personal pictures' of the situation described by it? In order to even begin to formulate the possible forms of such strategies we need of course some idea about the geography of these lands of soft cognition. To stay in the same metaphor: it is of course not likely that this geography is completely different from the lands we already know a bit about, e. g. that of knowledge, goals, problem solving, etc. This means that our opinions and attitudes will probably also be organized hierarchically, that there are frame- or scriptlike stereotypical organizations of this kind of soft cognition, that the activation and actualization strategies will be similar, and so on. These representational aspects will be neglected here. I will merely assume that the make-up is again layered: Basic is a system of physiologically 13

For somewhat more detail about the role of opinions and attitudes in discourse comprehension, see van Dijk (1983), also for further references to the cognitive und social psychological literature. A more specific example, viz. the role of ethnic prejudice in discourse, has been explored in van Dijk (1984).

Strategic Discourse Comprehension

53

founded emotions/feelings. These have their systematic and strategic links first with more general systems, such as those of attitudes and general opinions, but also directly with individual memories (e. g. the fate of the victims in Italy). Then we assume that attitudes are organizations of general knowledge and opinions about socially relevant issues, such as abortion, nuclear energy or revolutions in South America. These attitudes are in turn organized in more general and basic ideologies. Thus, in discourse understanding particular opinions are formed on the basis of more general opinions, which are fragments of more complex attitudes, which involve more general values and norms and a basic ideology. The specific personal profile of all this is part of 'personality'. Note that just like knowledge, these general opinions, attitudes and especially ideologies, are socially 'based', that is, formed and transformed in situations of social interaction. Discourse understanding is of course a crucial aspect of this social aspect of both hard and soft cognition. Social learning of our various cognitive 'sets' has been rather underestimated by cognitive psychology, but will not further detain us here. But now back to discourse understanding and its strategies. A first question to be asked, then, is whether subjective beliefs, opinions, etc. determine actual understanding of a text. Do we have a 'better' TR or MS, are we able to supply missing links, interpret certain concepts, establish macro- and micro-coherence or detect schemata with this additional subjective apparatus? One aspect clearly emerges: our picture (MS) of the situation described by the text will be 'richer', and hence structurally more complete, and hence better retrievable, according to basic principles of information processing. And also, given emotionally based attitudes, opinions and wants or desires, we have of course firmer general motivation for the very action of reading and understanding the text: we simply have more interest, and hence more attention, and hence a more exclusive focus on the information of this particular text; we are involved and want or need the information, and so on. These two general principles already are sufficient to explain many discourse understanding phenomena, e. g. better and longer recall: But still, we may ask whether subjective understanding is 'better' or more 'effective' understanding, all other things (richer MS, more motivation) being equal. I think so, but it is not easy to prove, and of course it depends on our notion of 'understanding'. First, a distinction should be made perhaps between understanding proper and evaluation, together defining the notion of interpretation. An evaluation involves the various soft cognitions mentioned above: it is the assignment of a evaluative concept (good, bad, funny, beautiful, stupid) to textual information. Yet, this is too vague. We should distinguish between an evaluation of the text proper and of the information or rather the situation denoted by the text. Textual evaluation may be about the style, rhetorics, semantic coherence or completeness, schematic ordering, etc. as such, typically

54

T. v. Dijk

so in literary communication but also elsewhere. When we read newspaper texts this will be less common. We will rather evaluate the events denoted themselves. Hence, we have TR-evaluations and MS-evaluations, to be represented, indeed, separately. Let us provisionally limit ourselves to the latter kind of evaluation and see whether it may determine understanding proper. As a crucial example let us take local coherence assignment. Now, assume also that such coherence links always presuppose general knowledge about relations between facts. These facts may have subjective properties, e. g. opinions of people, terrible events, and so on. Now further assume that we not only know the opinions of other people, but already have opinions, attitudes, ideologies or emotions regarding certain facts or fact types. Then, if in a text two facts are related that involve precisely those emotions or attitudes, we must have the same kind of facilitation as when we have 'more' knowledge about a subject. For instance, in the Italian earthquake story, we can predict or expect better how the people will feel, and hence information about this in the text, if we share their opinions and attitudes towards the corrupt politicians. If we say that we do not 'understand' those people who demonstrate against nuclear energy or for free abortion or against American aid to fascist countries in South and Middle America, this means among other things that we do not share their opinions, attitudes or ideology. But, we have assumed earlier that understanding the world and understanding discourse are very similar things. Hence, any discourse that is about precisely these actions or events will be also better understood in the proper sense, because we are able to supply more coherence relations. Thus, we may interpret a but or because better if the contrast holds for us and the cause is a cause in our opinion: (8)

There are fascist countries in Latin America, because the USA protects its military and economic interests in that part of the world.

Understanding that discourse fragment seems easier if we already have the basic condition underlying the use of because. Thus, in general, we may assume, just as for knowledge, that beliefs, opinions, attitudes and ideologies will influence understanding. The next question then is: what specific strategies may be involved in this kind of subjective understanding? Is it possible for instance that language users make handy shortcuts via these subjective areas, bypassing for instance cumbersome knowledge activation and actualization? This indeed seems to be the case. Take again discourse fragment (8). A 'full' understanding of that fragment would involve actualization of rather complex data structures about international policy, American foreign policy, military and economic aspects of capitalism, and so on. Yet, if a reader just would have an attitude about the USA featuring rather high in the tree the

Strategie Discourse Comprehension

55

proposition 'American governments are usually pro-fascist', or 'American governments are capitalist', and more generally 'Capitalists are never systematically anti-fascist', we have a couple of general opinions that directly may be used to establish the required coherence. The mere association of the concepts 'USA' and 'pro-fascist' may be sufficient for understanding the coherence of this text fragment. In fact, imagine that we would have although instead of because in (8): for most of us the text fragment would in that case become incoherent, because we cannot even imagine how military and economic dominance of the USA would normally favor anti-fascist regimes, at least not in this kind of easy shortcut way, unless we go by general opinions such as 'America is liberal', so therefore also their military and economic interests must be liberal (helping people), and hence would conflict with fascist regimes. A similar story will of course hold for global coherence and the establishment of themes. Given our opinions and attitudes about women's lib, we will also globally read and understand a sexy story by Boccaccio differently than without these opinions or interests. The main topic may in that case be biased towards a topic in which the position of women in the middle ages is present. The strategy, monitored by the general opinion and attitude components, will in that case affect the macrorules of Selection, Generalization, and Construction. Scenes in which women are present and subject to harassment by men will be more selected, generalizations will be made more easily about their strong or weak position and a global picture will be constructed for scenes (e. g. of seduction) which may have different global outcomes. From these few examples it follows that not only soft cognition influences discourse comprehension in a general way — supplying more information to construct TR's and MS's — but that also the strategies themselves may be 'subjective' in the sense that other processes may be short-cut or biased, so that specific 'wanted' representations are constructed, or that otherwise incomprehensible coherence connections may be established rapidly or not at all. It goes without saying that this is all still extremely vague, but in the light of the previous sections it is not hard to imagine the role of this kind of strategies. 8. Speech Act Understanding,

Social Interaction,

and Cognitive

Strategies

Finally we will make some remarks about the proper 'place' of discourse and discourse understanding, viz. as part of the understanding of speech acts and social interaction. 14 Discourses do not occur in isolation, so understanding 14

Cognitive models of speect act understanding and social interaction are still rare. The pragmatic component, both of a theory of discourse and of a cognitive model, has been discussed in

56

T. v. Dijk

them is also understanding their social functions and effects, and vice versa. Many aspects of textual structures, e. g. style and rhetoric or schematic structures, but also semantic content and coherence, depend crucially on various categories and conventions regulating the social context of communication. It is plausible, therefore, that discourse understanding strategies will ultimately be based on interaction strategies or simply be interaction strategies. This interdependence is so ill explored that we only can give some illustrative examples for the various subdomains involved. First speech act understanding. Given the utterance of a sentence or text in a given social context, the problem is that the reader/hearer must assign one or more speech acts to this utterance, that is, interpret it pragmatically. Such an interpretation is not straightforward and not precisely rule-governed: the surface form or the semantic content of the utterance are of course important clues but are neither sufficient nor necessary in the assignment procedure. Hence we need strategies, again, taking many types of textual and contextual information, to assign this interpretation. The strategy involved is necessarily very complex. I assume that in the interpretation of an utterance as a (social) action, a language user will first make an effective analysis of the relevant context. This means analysis for precisely those features that can make speech acts (in-)appropriate. These context features involve some cognitive ones (does speaker, S, know p, want p, etc.?) and social ones (what is the social position of S, what is my social role, etc.?). This analysis will already narrow down the possible speech acts even before something was said (we do not expect orders from friends, assertions from someone about something we know he doesn't know anything about, advices from young children, etc.). That is, the speech act, taken as a social act, must be coherent with the social context and with the ongoing social interaction. This means also that the final situation of the previous action of the speaker or the hearer counts as input for the actual speech act. After this pre-analysis of the social context we have the analysis of the text itself. Again, this will not take place for all possible details for the text as a whole: only those features that are possible signals for pragmatic interpretation the papers collected in van Dijk (1981). Micro-sociologists of various schools do operate with several 'cognitive' notions, such as 'interpretation', 'understanding', 'categories', or 'representation', but they use these notions more or less intuitively, and do not provide an explicit cognitive model of social interaction and its strategies. This will undoubtedly be one of the major tasks of psychology in the coming years, so that a really explicit bridge can be built between language use, discourse, and cognitive processing on the one hand, and the account of speech acts, interaction, situation and their cognitive basis, on the other hand. Thus, whereas micro-sociology has not in fact been consequently cognitive, cognitive psychology has not sufficiently been 'social': discourse understanding strategies as well as the acquisition, the use, the change and even the structures of knowledge, cannot fully be accounted for without an explicit cognitive model of social interaction and social situations.

Strategic Discourse Comprehension

57

will be reviewed. Although as a rule we might assume that first semantic and then pragmatic interpretation takes place, this is not a necessary strategy. Maybe surface structure information together with especially paratextual information (gestures, facework, movements, distance, etc.) are sufficient for a quick guess about the most probable speech act. An angry face or shouting in some specific social situation may mean, directly, a speech act of criticism, accusation, etc. For the definite assignment of course the semantics should be passed, at least partially, because we cannot not identify a speech act without its propositional base. At the same time as the surface structure and paratextual analysis takes place we therefore will probably have it that (as we saw earlier) the semantic interpretation already starts, which enables the pragmatic interpretation to start also, even before a sentence (let alone a text) is finished. Surface structures will contain signals for speech act class (interrogative sentence forms, for instance, in order to perform requests or questions), or more specific stylistic features may indicate evaluation (e. g. congratulation). Finally, semantic content will be analysed for main participants, especially the agent and patient of the speech act and the acts to which they pertain (orders, questions, accusations, promises, etc.), as well as for time and place (promises should be about future actions of the speaker, accusations about past actions of the hearer). With this complex set of multi-layered schematic analysis of both context and text we have the respective clues for the strategy of speech act comprehension. It hardly needs to be said that world knowledge, episodic knowledge and soft cognition, as described above, again play an important role in the interpretation of action. 'Correct' interpretations in principle try to represent the intention(s), plans, purposes/goals and underlying motivations of the speaker — and a representation of how the speaker will probably represent the actual 'state' of the hearer. This kind of analysis of mutual representations would be impossible without mutual knowledge and opinions. Important for our discussion is the assumption that all this hardly obeys precise rules, but rather flexible strategies, which can make use of various types of information from text and/or context. The rules only say which general conditions of the text (e. g. certain semantic constraints, mentioned above) and the context must be satisfied. The strategy will then try to find clues as soon as possible without first going through the full analysis of each level. If the context or the paratextual performance and some very partial surface information or semantic interpretation already yields a good pragmatic guess, this may already be used 'top down' to see whether the rest of the data match this hypothesis. Unfortunately, linguistics has not yet even systematically described all the possible textual features that may be pragmatically relevant. What we also ignore is how speech act concepts occur in semantic knowledge: are they just a list of (some 1000) speech acts, or is this list organized, so that activation and actualization can be effective? Or is this list of concepts part

58

T. v. Dijk

of the lexicon of a language? Or is it part of the knowledge structures we have about social actions in general? Maybe a combination of these aspects is most likely: we, in general, have expressions and hence conventional 'meanings' for the speech acts of our culture, but at the same time the analysis of utterances as speech acts requires an analysis of action and context structures, which also involves 'social' or 'interaction' knowledge. Then, we have macrospeech acts. Just as we may map sequences of propositions of a text, by the macrorules, on sequences of macropropositions, the themes, of a text, we assume that sequences of speech acts are mapped onto sequences of macrospeech acts. That is, we may take a sequence of statements, as a whole, as a threat or accusation. Again, strategies are necessary to infer such global speech acts. That is, during the comprehension of a text or conversation, the hearer/reader will, on the basis of the pragmatic interpretation of some speech acts, try to make a guess about the pragmatic point of the sequence as a whole: what is (s)he driving at? Does he/she want to invite me, threaten me, or accuse me? In that case the respective local speech acts will be analysed also with respect to their possible end-results as conditions for the global speech act ("if I answer that I am free, after a question about whether I am free tonight, this may be a sufficient condition for him/her to want to invite me ...", etc.). Language users are extremely perceptive in this kind of global interpretation strategies of speech acts. If for instance they do not want to have the responsibility of 'refusing' something the speaker is globally aiming at, the component speech acts may already, as a counter-strategy, be such that the necessary conditions for the global speech act cannot be established ("No, I am not free tonight, I have to work"). Note that the 'core' of the global speech act, e. g. the threat or request, need not be expressed in the text or conversation at all. This may either happen indirectly, or else simply by construction of the other speech acts, just as in macro-interpretation at the semantic level. Therefore, this kind of macroanalysis is very powerful: it yields important interactional information although this is not directly expressed by the respective speech acts at all. The macro-speech act by definition is the most important speech act, the pragmatic and hence interactional 'upshot' of what is said: local speech acts may be easily forgotten, but the global one is the major input for further (reactions of the hearer. This is why strategies, not only for interpretation, but also for interaction, are so important at this level. They also involve, therefore, strategies for adequate production of speech acts, viz. of those which are optimally consistent with the own wishes and goals of the (next) speaker (including the goal to cooperate or not with the last speaker). We here, finally, touch the proper social psychological and social implications of discourse strategies. Clearly, understanding a discourse correctly but also fast and easy is one thing: it makes communication efficient. Yet, at the same time the strategies, both of understanding and of production, must be socially effective. They must contribute to an optimal realization of the

Strategic Discourse Comprehension

59

'personality', the wishes, wants, goals, or interests of the speaker or hearer. This will mean, among other things that we want to influence the hearer's knowledge, opinions, attitudes, and possibly, indirectly, also his/her actions based on these, as much as possible according to one's own (speaker's) wishes and goals. This is the basic principle underlying the various production strategies. But also in understanding these social implications are important. We have seen that biased, subjective understanding is a possible or even necessary consequence of the role of soft cognition. In a social psychological perspective, this implies, among other things, that a hearer can interpret a text or part of a conversation according to his/her own opinions, attitudes and ideologies (or not, depending on the goal or reading). Conflicting consequences, e. g. a necessary change in that system, can in that case be avoided (maintenance of cognitive balance, avoidance of cognitive dissonance, etc.). Similarly, not only biased understanding but also purposeful misunderstanding or not-understanding can be strategically useful in interaction. We all know the scene of the high school kid who does not 'understand' the criticism of the teacher, or our lack of 'understanding' of the policeman who wants to give us a ticket. Not understanding means that we do not understand the speech act, and hence the goal of the speech act, and hence try to»avoid the conditions for our actions as they are wanted by others, and which we cannot avoid by other means. This means that the principles of cooperative social interaction (Grice) may only hold in conflict-free social situations. If our wishes and goals are different, we may not participate in the purposes and hence the intentions and hence the speech acts and hence the utterances of the speaker. Strategies of comprehension can regulate this social 'participation' in a refined way. 9. Provisional

Conclusions

From this review of some of the types of strategy used in discourse comprehension it may first of all be concluded that at all levels of comprehension, strategies play a primary role. They solve the various problems posed by the 'formal' rules of interpretation: they provide meaning where apparently there is no meaning, they establish coherence before enough data have been obtained, they derive global themes that can be used to process further information top-down — and establish global coherence. In general then, the strategies provide easy, fast, and flexible means to handle vast and complex amounts of textual and communicative data. They are not tied to specific levels or units, but are able to process various kinds of data at the same time, yielding provisional hypotheses of interpretation. Then, they are necessary to retrieve, address, activate and actualize precisely that kind and that amount of general or episodic knowledge that is needed

60

T. v. Dijk

to understand the text. They provide a subjective background for interpretation such that the results of the interpretation can be controlled by the own wishes, goals, opinions, attitudes or ideologies. Finally they specify what acts can or should be understood to be performed by the speaker and how this (not- or mis-)understanding can be used as input for further interaction, for the (non-)change of the own cognitive systems, for the avoidance of conflict, difficulties or other socially undesired reactions that may harm our cognitive balance on the one hand or our social (selfesteem on the other hand — due to negative evaluation by others. In brief, strategies of comprehension are not only cognitively efficient but also socially, because they allow us to quickly, correctly, easily and above all flexibly act and react to the communicative context. Our discussion of these strategies is tentative and incomplete. Moreover, it is informal, sometimes speculative and therefore sometimes also vague. The function of the review is merely to discuss possible types of strategy and to link various strategies for the respective levels or dimensions of text and communicative interaction. Explicit theories and especially systematic experimental testing will be necessary to formulate and to empirically assess the nature and the functions of these strategies.

References Ballmer, Thomas: 1972, "Einführung und Kontrolle von Diskurswelten". In: D. Wunderlich (ed.), Linguistische Pragmatik. Athenäum: Frankfurt, pp. 183—206. Ballstaedt, Steffen-Peter, etal.: 1981, Texte verstehen, Texte gestalten. Urban & Schwarzenberg: Munich. Bartlett, F. C.: 1932, Remembering. Cambridge U. P.: London. Bobrow, Daniel & Collins, Allan (eds.): 1975, Representation and Understanding. Academic Press: New York. Bransford, John D. & Franks, Jeffrey J.: 1972, "The Abstraction of Linguistic Ideas. A Review." Cognition 1, pp. 2 1 1 - 2 4 9 . Clark, Herbert H. & Clark, Eve: 1977, Psychology and Language. Harcourt, Brace, Jovanovich: New York. van Dijk, Teun A.: 1977, Text and Context. Longman: London, van Dijk, Teun A.: 1980a, Macrostructures. Erlbaum: Hillsdale, N.J. van Dijk, Teun A.: 1980b, Textwissenschaft. Niemeyer: Tübingen, van Dijk, Teun A.: 1981, Studies in the Pragmatics of Discourse. Mouton: The Hague, van Dijk, Teun A.: 1983, "Opinions and Attitudes in Discourse Comprehension". In: J.F. LeNy & W. Kintsch (eds.), Language and Comprehension. North Holland. Amsterdam, pp. 35-51. van Dijk, Teun A.: 1984, Prejudice in Discourse. Amsterdam: Benjamins, van Dijk, Teun A. (ed.): 1980, Story Comprehension. Poetics 9, nrs. 1—3. van Dijk, Teun A. (ed.): 1982, New Models of Discourse Processing. Text 2, nrs. 1 — 3. van Dijk, Teun A. & Kintsch, Walter: 1983, Strategies of Discourse Comprehension. Academic Press: New York. Fodor, J. A., Bever, T. G. & Garrett, M. F.: 1974, The Psychology of Language. McGraw-Hill: New York.

Strategie Discourse Comprehension

61

Freedle, Roy O. (ed.): 1977, Discourse Production and Comprehension. Ablex: Norwood, N . J . Freedle, Roy O. (ed.): 1979, New Directions in Discourse Processing. Ablex: Norwood, N. J. Graesser, Arthur C.: 1981, Prose Comprehension Beyond the Word. Springer: New York. Groeben, Norbert: 1982, Leserpsychologie. Textverständnis-Textverständlichkeit. Aschendorff: Münster. Johnson-Laird, Philip: 1983. Mental Models. London: Cambridge U. P. Johnson-Laird, Philip & Garnham, A.: 1978, Descriptions and Discourse Models. University of Sussex: Brighton. Just, Marcel & Carpenter, Patricia (eds.): 1977, Cognitive Processes in Comprehension. Erlbaum: Hillsdale, N. J. Keenan, Janice: 1976, The Role of Episodic Information in the Assessment of Semantic Memory Representations for Sentences. Ph. D. Diss. University of Colorado, Boulder, Colorado. Kintsch, Walter & van Dijk, Teun A.: 1978, "Toward a Model of Text Comprehension and Production." Psychological Review 85, pp. 363—394. Mandler, Jean M. & Johnson, Judith S.: 1977, "Remembrance of Things Parsed". Cognitive Psychology 9, pp. 111 — 151. Meyer, Bonnie J.: 1975, The Organisation of Prose and its Effects on Memory. North Holland: Amsterdam. Norman, D. A. & Rumelhart, D. A. (eds.): 1975, Explorations in Cognition. Freemann: San Francisco. Petöfi, Jänos S.: 1979, "Structure and Function of the Grammatical Component of the TextStructure World-Structure Theory". In: F. Guenthner & S. J. Schmidt (eds.), Formal Semantics and Pragmatics for Natural Languages. Reidel: Dordrecht, pp. 303 — 338. Sanford, Anthony J. & Garrod, Simon C.: 1981, Understanding Written Language. Wiley: Chichester. Schank, Roger C. & Abelson, Robert P.: 1977, Scripts, Plans, Goals and Understanding. Erlbaum: Hillsdale, N. J. Stenning, Keith: 1978, "Anaphora as an Approach to Pragmatics". In: M. Halle, J. Bresnan & G. A. Miller (eds.), Linguistic Theory and Psychological Reality. MIT Press: Cambridge, Mass., pp. 1 6 2 - 2 0 0 . den Uyl, Martijn & van Oostendorp, Herre: 1980, "The Use of Scripts in Text Comprehension". In: van Dijk, T. A. (ed.), Story Comprehension, Poeties 9, nrs. 1 — 3 (special issue), pp. 275-294. Wilensky, Robert: 1978, Understanding Goal-based Stories. Yale University, Dept. of Computer Science: New Haven.

CHRISTOPHER U. HABEL

Referential Nets as Knowledge Structures Some Structural and Dynamic Properties* 1. Knowledge-Based

Understanding

What does it mean to understand language 1 ? On the one hand, up to now complete and satisfactory answers to this question do not exist and there are a lot of controversies and open questions for linguists, philosophers, psychologists and scientists working in the Artificial Intelligence or Cognitive Science paradigm about the basic principles and processes in language understanding. On the other hand there exist some widely accepted assertions 2 about human natural language processing, i. e. comprehending and generating natural language utterances. The most fundamental and important of these assertions are: (l.a) processing natural language is possible only on the basis of a large amount of knowledge and belief, linguistic knowledge as well as knowledge and belief about the world, (l.b) natural language processing is context and situation dependent, (l.c) theories of natural language processing have to consider the procedural character of comprehension and production and the dynamics of language. In the present paper the knowledge dependence of natural language processing, i.e. assertion (l.a) is used as a starting point 3 . After some remarks on representation of knowledge in general — this is a / the central area of AI

* The author's research on Referential Nets has been supported by the German Research Foundation (DFG). I wish to thank my friends and colleagues Siegfried Guenther, Helmar Gust, Siegfried Kanngiesser, Jim Kilbury, Carola Reddig and Claus Rollinger for the comments and suggestions they gave through the years on various points of Referential Nets. I am also grateful to Thomas Ballmer, Peter Bosch, Herb Clark and Bonnie Webber for helpful comments. 1 Compare on this fundamental and general question: Winograd (1980). 2 In chapter 4 I will describe some ideas (and papers) which are related to the present paper. Some of them strongly influenced my own view of the subject. 3 This view is widely accepted by researchers in Artificial Intelligence; cp. Winograd (1980).

Referential Nets as Knowledge Structures

63

—, and on different types of knowledge a speaker-listener uses and needs for language understanding and producing in particular, the problems of knowledge change, i. e. the dynamic aspects of language will be considered with respect to the problem area of referential expressions. In representing the knowledge of a speaker-listener I distinguish three types of knowledge 4 : — knowledge of facts or assertions, e. g. of states of the world, events, etc. — knowledge of rules, e. g. of rule-like relations between objects or states of the world (or classes of such entities, see below), — knowledge of objects, e. g. of persons in the world. These three types of knowledge are abbreviated by 'F', 'R' and 'O', respectively. Thus, the knowledge of the participants in a communicationevent can be described with the triple (i refers to the specific participants) (2)

K.i =