Connexity and Coherence: Analysis of Text and Discourse 9783110854831, 9783110111026


269 15 20MB

English Pages 415 [424] Year 1989

Report DMCA / Copyright

DOWNLOAD PDF FILE

Recommend Papers

Connexity and Coherence: Analysis of Text and Discourse
 9783110854831, 9783110111026

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Connexity and Coherence

Research in Text Theory Untersuchungen zur Texttheorie Editor Janos S. Petöfi, Bielefeld Advisory Board Irena Bellert, Montreal Antonio Garcia-Berrio, Madrid Maria-Elisabeth Conte, Pavia Teun A. van Dijk, Amsterdam Wolfgang U. Dressler, Wien Nils Erik Enkvist, Abo Peter Hartmann f , Konstant Robert E. Longacre, Dallas Roland Posner, Berlin Hannes Rieser, Bielefeld Dieter Vieh weger, Berlin] DDR Volume 12

w DE

G Walter de Gruyter · Berlin · New York 1989

Connexity and Coherence Analysis of Text and Discourse Edited by Wolfgang Heydrich, Fritz Neubauer, Janos S. Petöfi and Emel Sözer

w DE

G Walter de Gruyter · Berlin · New York 1989

Printed on acid free paper (ageing-resistant — pH 7, neutral)

Deutsche Bibliothek Cataloguing in Publication

Data

Connexity and coherence: analysis of text and discourse / ed. by Wolfgang Heydrich ... — Berlin ; New York : de Gruyter, 1989 (Research in text theory ; Vol. 12) ISBN 3-11-011102-0 NE: Heydrich, Wolfgang [Hrsg.]; GT

ISSN 0179-4167 © 1989 by Walter de Gruyter & Co., Berlin 30 Printed in Germany All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced in any form — by photoprint, microfilm or any other means — nor transmitted nor translated into a machine language without written permission from the publisher. Typesetting and Printing: Arthur Collignon GmbH, Berlin Binding: Lüderitz & Bauer, Berlin

Foreword 1. Text-theoretical/semiotic-textological research is, roughly speaking, focused on the connectedness of the verbal stuff of texts, on the connectedness of the states of affairs (in the world fragment) expressed in the texts and on the analysis and description of the relation between that verbal stuff and the world fragment expressed by means of it. In other words: this research concerns both the analysis of the constitution of texts and the semiotic questions of text meaning. In 1982, Wolfgang Heydrich, Fritz Neubauer, and Jänos S. Petöfi, members of the Faculty for Linguistics and Literary Studies of the University of Bielefeld, started to organize and promote research work with international cooperation on the topic of 'text-connectedness' by collecting and systematizing literature, discussing questions on this topic, and publishing the results. (a) Up to the present, four readers have been compiled and published with 50 articles about a number of different aspects of text-connectedness, where the authors represent different research directions and come from different countries (cf. Neubauer (ed.) 1983, Petöfi-Sözer (eds.) 1983, Sözer (ed.) 1985, Petöfi (ed.) 1986). In addition, a reader appeared which provides a systematic overview of the theoretical and methodological questions of the analysis of text-connectedness and text meaning (cf. Petöfi (ed.) 1988). (b) Particular thematic issues have been published in journals: cf. Petöfi (ed.) 1983 with articles about methodological questions of discourse processing, and Petöfi-Vitacolonna (eds.) 1982 with articles about questions of the interpretation of literary texts. (c) In 1983 and 1984, two international and interdisciplinary workshops took place at the Center for Interdisciplinary Research (Zentrum für interdisziplinäre Forschung / = ZiF/) of the University of Bielefeld, which were accompanied by an international conference in Summer 1984 at the Centro Internazionale di Semiotica e di Linguistica of the University of Urbino. In the first workshop az ZiF, besides individual aspects of text connectedness, an attempt at elaborating a thematic and terminological synthesis by Hatakeyama, Petöfi and Sözer was discussed. (The former discussions are included in Heydrich-Petöfi (eds.) 1985, as to the latter cf. Conte (ed.) in publication.) For the second ZiF workshop thematic reports (together with two coreports for each report) were submitted for discussion (cf. this volume).

VI

Foreword

The Urbino conference also had text-connectedness as its general topic. The 36 papers presented there embrace the whole spectrum of this topic (cf. Conte-Petöfi-Sözer (eds.) in publication). (d) In 1986 a workshop took place at ZiF, devoted on one hand to general questions of the text-connectedness of poetic works and their interpretation, and on the other hand, to questions of the analysis of connexity following the method of Jakobson. (As to the former topic cf. Petöfl-Olivi (eds.) 1988, as to the latter cf. Konstantinidou-Ramöller (eds.), in prep.) (e) In addition, two survey volumes are in preparation. The first one provides an overview of the relevant literature on research in text connectedness written and published in different countries. These overviews are compiled by leading researchers from the countries concerned. (Cf. CharollesPetöfi-Sözer (eds.) — the first part of this Survey appeared in 1986.) The second survey volume will contain overviews of questions of interpretation (cf. Bokay-Olivi-Petöfi (eds.) in prep.) 2. The reports and co-reports contained in the present volume attempt to provide a comprehensive picture of the main aspects of text-connectedness, and the problems, results and perspectives of textological research concerning these aspects. Part I contains five reports (those of M. Charolles, K. Heger, L. Lundquist, R. G. van de Velde, and D. Viehweger) together with two co-reports on each of them, and the replies of the reporters to the co-reports. The topics in this first part are text production and text reception, aspects of the connectedness of texts associated with different text sorts, interaction, inference, and various methodological questions of textological research. The three reports of Part II (by P. Bange, Ν. Ε. Enkvist & Μ. Björklund, and Η. Kayser) treat various questions of the constitution of conversation. In these contributions the main stress is also placed on the aspects of connectedness. Finally, in Part III we find a report (by Ν. E. Enkvist) which deals with the problems of terms used in textological research, and a survey (by P. Lohmann) which attempts to supply bibliographical assistance for further research in text-connectedness. The editors wish to thank the Center for Interdisciplinary Research of the University of Bielefeld for the financial support of this conference. They hope that the present volume — together with the volumes referred to in this Forewood — will be a useful contribution to the further development of the methodology of analysis and description of text connectedness. Bielefeld, 1989

Wolfgang Heydrich Fritz Neubauer Jänos S. Petöfi Emel Sözer

Foreword

VII

Bibliography Bokay, Α., Τ. Olivi, J. S. Petöfi (eds.) in Research in Text Meaning. A Survey. Hamburg: Buske. prep. Charolles, M., J. S. Petöfi, Ε. Sözer (eds.) 1986 Research on Text Connexity and Text Coherence. A Survey ( = pt 53.1). Hamburg: Buske. Conte, M.-E. (ed.) in Kontinuität und Diskontinuität in Texten und Sachverhaltskonfigurationen. Diskussion über press Konnexität, Kohäsion und Kohären£ ( = pt 50). Hamburg: Buske. Conte, M.-E., J. S. Petöfi, E. Sözer (eds.) in Text and Discourse Connectedness. Proceedings of the Conference on 'text, connexity, coherpress ence'. Urbino, July 16— 21, 1984. Amsterdam: John Benjamins B.V. Heydrich, W., J. S. Petöfi (eds.) 1986 Aspekte der Konnexität und Kohären% von Texten ( = pt 51). Hamburg: Buske. Konstantinidou, Μ., P. Ramöller (eds.) in Beiträge \ur Diskussion über Jakobsons Methode der Textanalyse ( = pt 63). Hamburg: prep. Buske. Neubauer, F. (ed.) 1983 Coherence in Natural Language Texts {— pt 38). Hamburg: Buske. Petöfi, J. S. (ed.) 1983 Methodological Aspects of Discourse Processing ( = Text, volume 3—1). Berlin, New York, Amsterdam: Mouton Publishers. Petöfi, J. S. (ed.) 1985 Text Connectedness from Psychological Point of View ( = pt 55). Hamburg: Buske. Petöfi, J. S. (ed.) 1988 Text and Discourse Constitution. Empirical Aspects, Theoretical Approaches ( = RTT 4). Berlin, New York: W. de Gruyter. Petöfi, J. S., T. Olivi (eds.) 1988 Von der verbalen Konstitution %ur symbolischen Bedeutung — From Verbal Constitution to Symbolic Meaning ( = pt 62). Hamburg: Buske. Petöfi, J. S., E. Sözer (eds.) 1983 Micro and Macro Connexits of Texts ( = pt 45). Hamburg: Buske. Petöfi, J. S., L. Vitacolonna (eds.) 1983 Analisi e interpretation dei testi letterari ( = Versus, 35/36). Milano: Bompiani. Sözer, Ε. (ed.) 1985 Text Connexity, Text Coherence. Aspects, Methods, Results ( = pt 49). Hamburg: Buske.

Contents

Preface

V

Part I Michel Charolles Coherence as a Principle in the Regulation of Discursive Production. .

3

Hans-Jürgen Eikmeyer On Michel Charolles' "Coherence as a Principle in the Regulation of Discursive Production" Jozsef Andor Strategies, Tactics and Realistic Methods of Text Analysis. Comments on Michel Charolles' "Coherence as a Principle in the Regulation of Discursive Production"

28

Michel Charolles Comments on Eikmeyer's and Andor's Report

37

Klaus Heger Text Coherence in a Dialogical Presuppositional Group: Chapter XXXI of Unamuno's "Niebla"

41

16

Michael Metzeltin Co-report on "Text Coherence in a Dialogical Presuppositional Group: Chapter XXXI of Unamuno's Niebla" by Klaus Heger 100 Wolfgang Raible Some Comments on the Contribution of Klaus Heger: "Text Coherence in a Dialogical Presuppositional Group: Chapter XXXI of Unamuno's Niebla" 107 Klaus Heger Some Comments on the Comments by Michael Metzeltin and Wolfgang Raible 116 Lita Lundquist Coherence in Scientific Texts

122

X

Contents

Klaus Hölker Remarks on Form, Coherence, and Knowledge. Comment on Lita Lundquist's "Coherence in Scientific Texts" 150 Anette Villemoes Remarks on "Coherence in Scientific Texts" by Lita Lundquist

158

Lita Lundquist Some Comments on the Comments by Klaus Hölker and Anette Villemoes 171 Roger G. van de Velde Man, Verbal Text, Inferencing, and Coherence

174

Peter Bosch Coherence and Cohesion: Comments on Roger G. van de Velde's Paper "Man, Verbal Text, Inferencing, and Coherence" 218 Frantisek Danes Report of Roger G. van de Velde's Paper "Man, Verbal Text, Inferencing, and Coherence" 228 Roger G. van de Velde On some Limitations of Interpersonal Understanding (Comments on the comments by P. Bosch and F. Danes) 240 Dieter Viehweger Coherence — Interaction of Modules

256

Maria-Elisabeth Conte Coherence in Interpretation. Comments on Dieter Vieweger's Paper: "Coherence — Interaction of Modules" 275 Roland Harweg Is Coherence a Pragmatic Phenomenon? Comments on Dieter Viehweger's Paper "Coherence — Interaction of Modules" 283 Dieter Viehweger Coherence is Also a Pragmatic Phenomenon. Some Rejoinders to E.-M. Conte and R. Harweg 291 Part II Pierre Bange Constitution of Relationships as a Factor in Interactive Coherence . . . 305 Nils Erika Enkvist andofMartina Björklund Toward Taxonomy Structure Shifts

324

Contents

XI

Hermann Kayser Some Aspects of Language Understanding, Language Production, and Intercomprehension in Verbal Interaction 342

Part III Nils Erik Enkvist From Text to Interpretability: A Contribution to the Discussion of Basic Terms in Text Linguistics 369 Patricia Lohmann Connectedness of Texts: A Bibliographical Survey. (Part II)

383

Index of Names

401

Part I

MICHEL CHAROLLES

Coherence as a Principle in the Regulation of Discursive Production In Charolles (1981 and 1983) I tried to defend the idea that coherence — unlike connectedness — is not an inherent property of discourses but constructed by the interpreters. I argued that the receiver ( I ) a priori does his utmost in order to assign a coherent representation of the meaning to the discursive occurrences he is treating — i. e., an acceptable configuration of relations between the individuals and the states of affairs denoted by the discursive occurrences. To construct this configuration, the interpreter relies, for example, upon the connexity-indicators (connectors) present in the discursive occurrences. The connexity-indicators support conventional instructions guiding the interpretative process. They indicate to the receiver what types of relations the utterer (Lo) establishes between the states of affairs he is talking about. These indicators are marks provided by Lo for /; they signal to / a pure relational scheme. I must fill in this relational framework with the discursive components which can occupy a place in it, in order to have access to the way Lo links individuals and/or states of affairs. Coherence is a principle for the interpretation of discourse in the sense that normally / makes the generous hypothesis that the utterances he receives must be treated (and possibly repeatedly treated) in such a way that, at the end of the interpretative (or re-interpretative) process, he has access to a configuration of relations between individuals and/or states of affairs which seems acceptable to him on the basis of the representation he has in mind of what Lo wanted to mean while he was saying what he said. In the following pages, I will discuss the corresponding principle at the level of discourse production. I will defend the idea that coherence is also a principle governing discursive utterances. In its cooperative generously idealistic version, this hypothesis maintains that Lo gives — in principle and by principle — to I all the relevant relational indications which are necessary for him to catch the coherence of his discourse. This means that Lo is able to manage with efficiency the possibilities of relational indications attached to certain words and constructions belonging to the language used. We shall speak of an intelligent management of these relational means when we are able to establish that Lo distributes interpretative instructions according to a reasoning (or a calculus) that takes into account

4

Μ. Charolles

— his plan (or project) of communication at the precise moment of the discourse-production process, — his representation of Γs interpretative skills. I would like to make two general remarks before examining these points. First: from a "philosophical point of view", the hypothesis that coherence is a principle governing discourses is not particularly extravagant. Considering that the subject who receives some discursive occurrence cannot ignore that it is produced in order to signify something — even if it is nonsense! — similarly, the subject who produces a discursive uterrance cannot ignore that it will be interpreted as expressing a certain meaning. As coherence is the "minimal meaning" that allows access to other interpretations (for instance, indirect ones), it seems perfectly reasonable to imagine that Lo does his utmost so that / succeeds in calculating, at least, the relations he (Lo) establishes between the different parts of what he says (for a discussion see Charolles, 1983). The second remark is somehow different. To establish that the usage of some connexity carriers correspond to Lo's intelligent calculation, it would be necessary to carry out psycho-cognitive investigations. Several devices can be used to this end. In the studies devoted to writing processes, one of the currently practised methods is the protocol, wherein "subjects are asked to say aloud everything they think and everything that occurs to them while performing the task" of writing (Hayes and Flower, 1980: 4). In the following pages I will not resort to such devices. Our purpose consists simply in inducing relevant insights for further psychological observations. The type of approach I will develop pertains to processual analysis (Eikmeyer, 1982) but it aspires only to be a linguistically oriented preparation for more controlled studies (on general problems of "speech production" and "strategies in discourse planning", see Zammuner, 1981).

Intelligent

treatment and strategy

An intelligent management is a management governed by evaluation of what must be done. It presupposes a decision process: the subjects decide to accomplish one or several actions for reasons of achieving a goal. In order to make these notions clearer, let us consider an example of intelligence observed at a moment when this behaviour has noticeable effects on the definition of emission strategies. It has been experimentally established (Just and Carpenter, 1980; Frazier and Rayner, 1982; Rayner, 1983) that readers, when confronted with ambiguous sentences embedded in short texts do not spend more time treating them than in reading unambiguous utterances of a same length. These observations demonstrate that readers do not assign equal importance to the

Coherence as a Principle in the Regulation of Discursive Production

5

two or η theoretically possible interpretations of ambiguous occurrences. They adopt one and only one reading and follow it as long as they can. They do not properly anticipate possible alternatives, they treat the text data which arrive ("on-line") according to the sense of the interpretation they have adopted. This process, called garden-path strategy, implies a decision task which is determined by context and co-text features, but it escapes the subject's consciousness. Moreover, it has also been established that when a reader faces a difficulty — for example, information which does not fit in with the interpretative thread previously adopted — he is able to return to the relevant passages to correct his first interpretation. His re-analysis can be said to be intelligent because in catastrophic situations, the subject does not re-interpret the whole text or does not come back to passages irrelevant to the problem it meets. Concerning the management by Lo of relational indications useful for the access to the coherence of this discourse, we shall speak of an intelligent strategy when Lo takes into account the normal ways of interpretation that I attempts to adopt — i. e., when Lo's decision process includes, as a parameter, for example, Γs natural reading strategy. The following text is an illustration of such a strategy. It is an extract from a French newspaper (Est Republicain, 11. 1. 83). Only the beginning of the text is reproduced, because it is adequate for what we want to show. The English translation is meant to help the understanding of the French version which allows the interesting phenomenon to be observed: (1) "Nouvelle affaire

d'autodefense

ELLE CHERCHAIT SON CHAT: SON VOISIN TIRE ET L'ATTEINT A LA TETE Un magon de 44 ans, Μ. Vicente Duarte, a blesse d'un coup de fusil de chasse sa volsine, qui venait chercher son chat dans son jardin, dimanche, ä Osny (Val-d'Oise), pres de Pontoise. Mme Christiane Boulaire, secretaire de 26 ans, avait penetre dans le jardin de Μ. Duarte." "New self-defense a f f a i r SHE WAS LOOKING FOR HER CAT HER NEIGHBOUR FIRES AND HITS HER IN THE HEAD A mason of 44 years, Mr. Vincent Duarte wounded his neighbour with a gun. She was looking for her cat in Mr. Duarte's garden, Sunday, at Osny (Val d'Oise) near Pontoise. Mrs. Christiane Boulaire, a 26 year old secretary, broke into Mr. Duarte's garden ..."

6

Μ. Charolles

Practically all the subjects to whom this text was submitted declared after reading that Mrs. Boulaire's neighbour caught "elle" (i. e. "Mrs. Boulaire") but not "her cat". Only very few readers pointed out that it would have been possible to understand that Mr. Duarte had shot Mrs. Boulaire's cat, but that this interpretation — perfectly possible in the French version where the pronoun Γ [her in the English version] is ambiguous — does not really agree with the title. If the article had mentioned that Mr. Duarte had hurt Mrs. Boulaire's cat, these readers i) would have realized that they had "naturally" interpreted the title in a particular way which was not the only one possible, ii) would have corrected their previous interpretation and connected /' to son chat (with a possible backward ocular movement towards this region of the text). Such a reading method is typical of garden-path-strategy with intelligent reanalysis in catastrophic situations. The author of the text should have formulated this title in such a way that it avoided any ambiguity. He should, for instance, have chosen: "SHE WAS LOOKING FOR HER CAT HER NEIGHBOUR FIRED A N D WOUNDED HER IN THE HEAD"

"ELLE CHERCHAIT SON CHAT SON VOISIN TIRE ET LA BLESSE A LA TETE"

Thus the writer had at his disposal linguistic devices (in French) to produce a text without any pronominal ambiguity. Nevertheless the ambiguity of the original title is purely theoretical, which is confirmed by the fact that practically none of the readers perceived it. Then the formulation adopted by the author is perfectly efficient because the readers "normally" treat the text in a convenient way — i. e. in a sense which is confirmed by the rest of the text. The expression mood chosen is intelligent — although not very creative — because it serves the writer's communicative goal. This example shows that the notion of efficiency, when applied to interpretation processes, must be defined in relation to the natural conditions regulating the interpretation tasks. Therefore an intelligent management of connexity-carriers — here anaphoric markers — suppose that Lo distributes them in taking into account the "natural conditions" of interpretation (see Eikmeyer, 1983) that determine the most likely strategies (for example, reading) in a precise situation of reception. The situation of reception — more precisely Lo's representation of it — determines the choice of a reading strategy. The utterance: "Elle cherchait son chat, son voisin tire et l'atteint a la tete", if detached like an example in a paper on anaphorics, text (1) would, of course, not be read identically by a linguist. In the original text, the comment "New self-defense affair"

Coherence as a Principle in the Regulation of Discursive Production

7

is also a headline as the resolution of ΐ in eile, because it is commonly acknowledged that such affairs only concern human beings and implies that the subject "metes out justice" to another person. Similarly, the fact that text (1) appeared in a regional newspaper non-specialised in animal defense affairs is relevant for the agreement of the interpretation /' = eile, because the reader infers that the information concerns, as usual, a topic dealing with the lives of human beings. So the way a ho takes or does not take advantage of the situation in which a text is received is significant for the analysis of his emission strategy. Let us now examine two other texts: text (2) is a new illustration of the cooperative hypothesis, whereas text (3) is a counter-example typical of a non-cooperative strategy of writing.

The story of Tancrede and Clorinde The following text is a short abstract of the story of "Tancrede and Clorinde". It was delivered to the audience before a performance of the Monteverdi opera. It was intended to help the audience who does not understand Italian to be able to follow the plot. (2) The story: "Tancrede, the Christian, loves Clorinde, the Muslim. This one, as wild a warrior as beautiful, sets fire, with Argan to one of the Crusader's towers. Tancrede who, of course, did not identifiy her, chases after and challenges her to single combat. The strange struggle begins. The only witness is the night 'which would like to hide in its breast such a gloomy drama'. Three times, the young lady escapes the knight's grasp ('an enemy's grasp, not a lover's one'). But just before the sun rises, she weakens: a terrible cut pierces through her young and tender flesh. Defeated, she asks Tancrede to baptise her. Tancrede goes and gets some water from a nearby brook. He comes closer to his adversary in order to accomplish his pious service. He removes her helmet, recognizes her face and 'remains without any voice and word', while heaven opens to the dead". The short passage interesting for our topic is the following one: "Tancrede who, of course, did not identify her"... We would like to show that Lo's way of writing presupposes an intelligent calculation which is shown by the occurrence of this information (as and where it is), Lo's goal is to spare the readers a labourious re-analysis. First of all, let us point out that the production of this piece of information corresponds to a pure textual necessity. Indeed, if we consider the dramatic performance itself, it is not essential for the understanding of the action. We

8

Μ. Charolles

can for instance, perfectly imagine a producer trying to communicate to the audience the fact that the enemy challenged by Tancrede is not Clorinde in order to lead up to the hero's final tragic discovery. To describe precisely that Tancrede does not know his adversary's identity at the beginning of the combat answers to a "necessity" which is tied to the understanding not of the plot of the opera but of the text (2). How does one conceive of this sort of necessity? If we delete in (2) the clause "who, of course, did not identify her"... we obtain a new text (2') which is still understandable but which implies a reading strategy at an expensive cognitive cost. A reader of (2') has no reason, at first glance, not to interpret that her refers to Clorinde. The same thing occurs for the following definite descriptions "the young lady", "his adversary" and pronouns. The reader who interprets (2') in this way faces a problem of coherence at the end of the text. His problem consists in the question: how can the writer of (2') say that Tancrede suddenly discovers that he has just pierced through Clorinde's young tender flesh, while the preceding passages imply the he knows that he was struggling against her? The inference: "Tancrede struggles against Clorinde (and he knows that)" is invited by the expressive mood adopted in this version of the text. It is not based on proper linguistic matters but naturally invited. So the results of natural conditions of interpretation and the problem of coherence it raises can only be explained by these conditions. Let us take the following example to illustrate this point: (3) Oedipus wants to marry Jocaste. Sentence (3) is ambiguous because of the propositional attitude of the verb to want. So (3) allows two types of theoretical interpretations: an opaque and a transparent one. In the opaque interpretation the proper noun Jocaste is taken over by the experiencer ('S): Oedipus. It cannot be substituted salva veritate by any definite descriptions, particularly by his mother. In the transparent interpretation, the proper noun Jocaste is supported by Lo\ Oedipus wants to marry Jocaste who is also Oedipus' mother in Lo's universe of beliefs. In the opaque interpretation, Jocaste may not be the real name of the person Oedipus wants to marry. Similarly, in the transparent interpretation, Oedipus may not call "Jocaste" the woman he wants to marry. In a transparent content interpretation, Lo and S share the same universe of beliefs: Lo is responsible for the denomination Jocaste, but J would have used the same one if he had to speak of the individual Jocaste. Oedipus is willingly incestuous solely in this last case. (On opacity — transparence, see in addition to the classical literature: Martin, 1983; Kleiber, 1981; Galmiche, 1983). These different interpretations can be described in terms of mental spaces (Fauconnier, 1984):

C o h e r e n c e as a P r i n c i p l e in t h e R e g u l a t i o n o f D i s c u r s i v e P r o d u c t i o n

9

Transparent context interpretation: r = Jocaste

Jocaste

g = Oedipus mother

Oedipus' mother

The figure above must be read as follows: the verb of propositional attitude: to want introduces the mental space of J (IMS) and its parent pertaining to Lo (IRLo). In a transparent context, the proper noun Jocaste represents in IR a role r (the role to be Jocaste) which is equivalent to the role g (to be Oedipus' mother). Each of these two roles has a correspondent r' and g' in IM(S)· r ' = g' in I M(S) like r = g in I R(Lo) . Opaque interpretation: r = Jocaste or χ g = Oedipus' mother

In

IR(LO) r '

Jocaste g ' = Oedipus' mother

corresponds to r or x, but in IM(S) the identity r' = g' is not valid.

Transparent interpretation: r = Jocaste

r' = Jocaste

g = Oedipus' mother

g' = Oedipus mother

r has a correspondent in IM(s)· It m a y be r' or X . In IR(L0) r = g, whereas in IM(S) r' differs from g' Let us now suppose a person who wants to present in a few words the plot of Sophocle's play and says: (4) "Oedipus wants to marry Jocaste. She is his mother". Grammatically speaking the pronoun she refers to Jocaste — we do not take into account here the possible ambiguity between the antecedents Oedipus and Jocaste. Nevertheless, the ambiguity on roles remains and (4) accepts at least the three following paraphrases (the subscripts under Jocaste and she indicate the participants supporting them):

10

Μ. Charolles

(4')

Oedipus wants to marry Jocaste. She is his mosther. J Lo

(4") Oedipus wants to marry Jocaste. She is his mother. Lo Lo (4"') Oedipus wants to marry Jocaste. She is his mother. S + Lo S + Lo In (4") and (4"') (transparent and transparent-context interpretations, respectively), the pronoun refers to a nominal antecedent belonging to the same space(s). On the contrary, in (4') (opaque interpretation) Jocaste belongs to Im(S) whereas she is due to Lo. This remark illustrates the fact that pronouns the antecedent of which belong to a determined space can easily identify a correspondent situated in another space linked to the former (Fauconnier: 55). Moreover, change of mental spaces may not be perfectly grammatically marked as it is the case in (4) when interpreted like (4'). Obviously, it is possible to express (4) in such a way that the change of mental spaces is indicated. For instance: (5) "Oedipus wants to marry Jocaste, but he doesn't know she is his mother", where the verb to know marks that the proper name Jocaste belongs to Im(S), whereas the pronoun she takes place in IM(LO)· Thus, as Lawler (1977: 111) says, "the language disposes of means to resolve the problems it sets". But one can ask how an interpreter behaves when there is no mark of spacechanging in an utterance as, for instance, in (4). Lawler (op. cit) claims that in such circumstances, / complies with the normal hypothesis, i. e. / acts as if the utterance had to be interpreted in a transparent context. Lawler argues that I will decide to interpret a definite description as opaque only if he has some reasons to believe that the context is not transparent. This is the most "inexpensive strategy" (Lawler, op. cit.) and Lo, knowing that this strategy is normal only indicates to I that he is not responsible for a description when he wants to disqualify it. Lawler attributes to expressions like "so-called", "unbeknowst to him", etc., the function of signaling that Lo is not responsible for a definite description or for a denomination. Lawler (111) says also that "when I doesn't perceive any contradiction (...), / interprets the descriptions as if they were the fact of the participant responsible for the last one". This strategy explains what occurs when somebody reads version (2') of Tancrede and Clorinde's story. When the reader arrives at: "Tancrede chases after and challenges her ..." he has no reason to suppose that there is a change of mental spaces at that point of the text. He complies with the normal hypothesis according to which, if the writer does not explicitly mark a change of mental spaces, then he can continue interpreting the text in the same way as in preceding passages. As

Coherence as a Principle in the Regulation of Discursive Production

11

the two first sentences do not contain any mark implying an opaque reading, I treats them in a context-transparent way and does the same thing for the rest of the text. When he comes to: "he (...) recognizes her face ...", I is obliged to reconsider his strategy in order to preserve the coherence of the text. So he goes back to the text (intelligent re-analysis) towards: "Tancrede chases after and challenges her ...". He calculates that there is a change in the manner in which Lo supports the denominations, descriptions and references tied to Clorinde, and concludes that Tancrede does not know that he is struggling against Clorinde at the beginning of the combat. Then he pursues this new interpretation to the end of the text and, seeing that it does not lead to any more problems, he takes for granted that this way of understanding is the one required by (2'). This re-analysis is possible due to the phenomena of linguistic under-determination in changes of spaces (Fauconnier, op. cit.). (2') accepts a reading like the second one envisaged just above because there is no linguistic rule constraining Lo to signal a change of mental spaces in such a context. From a cognitive point of view, (2') requires a relatively extensive reanalysis which may constitute a heavy burden for the reader. Indeed, he must situate a possible point of change of spaces, go back to this point which is rather distant from the catastrophic place, and correct his former interpretation up to this point. If Lo would believe that this re-analysis strategy is too expensive and could see a convenient means well fit to facilitate the reader's task, he would probably use this device, and thus spare the reader such an extensive re-interpretation process. This is probably the reason why the author of (2) mentioned in the original version that Tancrede "of course" does not know his adversary's identity at the beginning of the combat.

Λ thief in the night The next example provides another illustration of the normal hypothesis evoked before. It is an extract from the American magazine New York (Sept. 1, 1980), reprinted from K. Hatakeyama, J. S. Petöfi, Ε. Sözer (in Ε. Sözer ed. (1985)), ( 6 ) A thief in the night. During the recent transit strike, a young man was walking home f r o m work through the park. It was late and he was alone. In the middle of his treck he saw someone approaching him on the path. There was, of course, a spasm of fear: He veered, the stranger veered. But since they both veered in the same direction, they bumped in passing. A few moments later the young man realized that this could hardly have been an accident, and felt for his wallet. It was gone.

12

Μ. Charolles Anger triumphed and he turned, caught up with the pickpocket and demanded his wallet. The man surrendered it. When he got home, the first thing he saw was his wallet lying on the bed. There was no way of avoiding the truth: He had mugged somebody.

The phenomenona we can observe in connection with this text are of the same nature as those described in Tancrede and Clorinde's story. All the pecularity of (6) rests on the fact that the reader falls systematically into the trap and is obliged to re-interpret it. Indeed everybody who glances through the text has no reason, in a first reading, to suspect that the definite descriptione: "the pickpocket" and the expressions "his wallet", "the man surrendered it" have to be understood in an opaque mode. Considering that these expressions are not marked, the reader interprets them according to the normal hypothesis — i. e. in the same mood adopted for the treatment of the preceding passages of the text. Since the preceding passages are easily interpretable in a transparent context mood (as in a novel supported by an omniscient narrator), the reader does not realize that the expressions quoted above need a particular opaque treatment. So, when the reader discovers at the end of the text that the thief is not the one he believed, the only solution for him is to look backwards for the passage where he was misled. This reinterpretation process is practically bound to happen. Besides, it does not only concern the expressions "the pickpocket", "his ^wallet" and "the man surrendered it". The correction also deals with "it was gone" and even with the title. How does I choose the expressions which can or must be corrected? How does he, for instance, determine the scope of the correcting process? The question arises because / should begin to suspect and reconsider other expressions than the relevant one indicated above. But nobody thinks, while re-interpreting (6), after the discovery of the final point, to correct his understanding of "someone" in "he saw someone approaching him on the path". This point is important because it allows us to make more precise assumptions about what happens during the re-analysis phase and why this re-analysis is intelligent. Indeed, one can conjecture that the reader considers the problem which he is confronted with as the basis for the re-analysis process. Thus he picks up and corrects only the information linked to the basis of the problem. Since the problem in (6), for instance, consists in knowing to whom the wallet belongs, the reader is able to correct the only text-data linked to the theme of this question. I should like to add two remarks about (6). First, the author should have spared his readers the re-interpretation process we discuss above. Quotation marks, or expressions referred to by Lawler, would have signaled, for example, the opacity of the "pickpocket". If the writer of (6) had written: "He caught up with 'the pickpocket' and demanded 'his' wallet. The man 'surrendered it'",

Coherence as a Principle in the Regulation of Discursive Production

13

the reader would have been warned that Lo did not want to be responsible for the expressions between quotation marks. Also the headline should have been formulated as follows: "A 'thief' in the night". Thus the reader would have been invited to infer that if the writer did not accept the quoted denominations and descriptions, it was probably because he judged them to be untrue. So the readers would have anticipated the ending of the story. But since (6) is "only" interesting by reason of the oddness of its ending, it certainly explains why the writer chose a particularly under-determined mode of expression. Finally, we can conjecture that he had willingly formulated the sotry in such a way in order to deceive his readers in the same way the young man deludes the socalled "pickpocket" of the story. We can see therefore how the writer cleverly builds upon the reader's intelligence and natural strategies of interpretation. The second point concerns a more linguistic topic. The possibility of reading deception is due here to the phenomenon of under-determination in changes of mental spaces which is rather common in natural-language communication. It is also due to the usage of indirect speech which permits certain ambiguities which are impossible in direct speech. Even if the young man's demand had been expressed as follows: "He caught up with the pickpocket and said to him: 'Give me my wallet'. The man surrendered it ...", the use of quotation marks would have been still possible: "He caught up with the pickpocket and said to him: 'Give me my wallet'. The man 'surrendered it' ...", but the final re-interpretation would not have been easier. Why? The difficulty comes from mentioning that the man surrendered the wallet to the young man. If the man accepts "surrendering" the wallet, the reader is allowed to infer that the man stole it, and the end of the text becomes practically inconceivable. The quotation marks on "surrendered it" would not have changed the problem. In such a context, they would probably have been interpreted like a mark of literal quotation, indicating that the man had actually employed the verb to surrender during the exchange and, consequently, that the had acknowledged the theft of the wallet. In direct speech, the passage examined here becomes practically impossible to recover by an intelligent re-analysis. The use of quotation marks does not facilitate at all the re-interpretation process. This is easily explainable because the use of direct speech does not allow ambiguities of the contents of Lo's universes of beliefs or mental spaces and of the participants involved. If the young man says to his interlocutor "give me my wallet", it proves that he believes the person to whom he is speaking has stolen it and he indicates his belief to

14

Μ. Charolles

him. Then if the interlocutor therefore surrenders the wallet, it proves that he had accepted Lo's belief as relevant for the exchange. So the young man's universe of belief is presented as shared and consequently as legitimated by the participant of the dialogue who is an ontologically decisive witness. Then Lo cannot further mark this universe as untrue. We can thus see how the writer of (6) takes real advantage of the possibilities of underdetermination offered by indirect speech. The fact that he does not use direct speech is another proof of the intelligence and perfidy of his writing strategy which is very far from being cooperative in the classical sense of the word! Concluding

remarks

We have examined two short texts showing how the principle of coherence applies on the level of discourse production. In the first one, Lo, anticipating Γs difficulties of reading prevents an extensive re-analysis. His mood of writing is intelligent because it takes into account the natural strategy I is likely to adopt in order to interpret a text underdeterminated in the distinction of mental spaces. Lo anticipates the normal hypothesis by indicating to I a transition to opacity. Lo's writing step is cooperative in the usual sense of the word. The second text is written in a way well suited to deceiving the reader. The mode of expression adopted by the writer is not cooperative because he does not prevent the natural strategy of interpretation leading to a catastrophe. Considering that Lo possesses the convenient devices to anticipate such a catastrophic deception, one can imagine that the underdetermination is done intentionally. So the intelligence of Lo consists here in taking advantage of the natural conditions of interpretations. The analysis developed here refers to a description of discourse processing. We gave priority to linguistic matters over psychological considerations. We tried to formulate interesting conjectures for further empirical observations which should be carried out in accordance with the methods currently used in psycholinguistic studies. This preparatory stage in processual analysis is necessary when one works on such a critical phenomenon as the opacitytransparence distinction. Bibliographical

references

Carpenter, P. and Just, M. 1977 "Reading Comprehension as Eyes See It", in: Just, M. and Carpenter, P. eds. (1977), 109-139. 1983 "What Your Eyes D o While Your Mind Is Reading", in: Rayner, Κ . ed. (1983), 275-307.

Coherence as a Principle in the Regulation of Discursive Production

15

Charolles, M. 1981 "Coherence as a Principle in the Interpretation of Discourse", Text 3, 1, 71—99. 1983 "Text Connexity, Text Coherence and Text Interpretation Processings", in: Sözer, Ε. ed. (1985), 1 - 1 6 . Eikmeyer, Η.-J. 1983 "Procedural Analysis of Discourse", Text 3, 1, 11—39. Fauconnier, G. 1984 Espaces mentaux. Paris: Minuit. Frazier, L. and Rayner, Κ. 1982 "Making and Correcting Errors During Sentence Comprehension: Eye Movements on the Analysis of Structurally Ambiguous Sentences", Cognitive psychology, 14, 1, 178-210. Galmiche, M. 1983 "Les ambiguites referentielles ou les pieges de la reference", Language Fran^aise 57, 60 - 87. Gregg, L. W. and Steinberg, E. R. (eds.) 1980 Cognitive Process in Writing. Hillsdale, N. J.: LEA. Hatakeyama, K.; Petöfi, J. S.; Sözer, Ε. 1983 "Text, Connexity, Cohesion, Coherence", in: Sözer, Ε. ed. (1985), 36 — 106. Hayes, J. R. and Flower, L. 1980 "Identifying the Organisation of Writing Processes", in: Gregg, L. W. and Steinberg, E. R. eds. (1980), 3 - 3 0 . Just, M. and Carpenter, P. (eds.) 1977 Cognitive Processes in Comprehension. Hillsdale, N . J . : LEA. Kleyber, G. 1981 Problemes de reference: Description definies et noms propres. Paris: Klincksieck. Lawler, J. 1977 "Quelques problemes de reference", Languages 48, 100 — 119. Martin, R. 1983 Pour une logique du sens. Paris: PUF. Rayner, Κ. (ed.) 1983 Eye movement in Reading. New York: Academic Press. Sözer, E. (ed.) 1985 Text Connexity, Text Coherence: Aspects, Methods, Results {= Papers in Textlinguistics 49). Hamburg: Buske. Zammuner, V. 1981 Speech production ( = Papers in Textlinguistics 34). Hamburg: Buske.

HANS-JÜRGEN EIKMEYER

On Michel Charolles' "Coherence as a Principle in the Regulation of Discursive Production" As Charolles I adhere to the opinion that linguistics should not be mainly concerned with the formulation of declarative statements about language, i. e. statements about what a language is, which structures a language has, etc. Instead, linguistics should try to give a dynamic description of language, i. e. observe processes of language production or understanding and it should describe the procedures (programs, instruction lists) users apply during such processes. This does not amount to saying that the declarative or structuralistic view of language is incompatible with the processual/procedural view. In the end also a procedural approach will formulate declarative statements about language. However, the notions and concepts used in these statements will be based upon or derived from procedural notions and models focussing upon dynamic aspects. The procedural framework I have worked out together with Hannes Rieser (cf. Eikmeyer and Rieser, 1980 and 1983) is characterized by a list of so-called naturalness conditions (cf. Eikmeyer, 1983) with the aid of which we try to adjust the theoretical model, in many respects, to the restrictions of natural situations in which natural language is used, produced or understood. A prominent place in this list is taken by the subjectivity condition which denies that linguistic objects have intrinsic, context-free properties. Instead, this condition favours the view that all properties of linguistic objects are properties assigned to the object by a language user during the process of production or understanding. The subjectivity condition may receive special formulations for the respective linguistic levels — phonetic, syntactic, semantic level — and the notions used on these levels. Thus, subjectivity also applies to the notion of coherence and I agree with Charolles on the following points: (i) Recipients start from the generous hypothesis that the language signal they are confronted with is (part of) a coherent whole. (ii) The producer of a discourse has means to control the process of understanding with respect to the ease of constructing a coherent interpretation on the recipient's side. He may exploit analysis strategies he himself invited. These preliminary remarks mean that I cannot be expected to make critical comments concerning the general layout of the ideas Charolles has presented.

On Michel Charolles' "Coherence as a Principle ..."

17

My comments are additions to or cautious elaborations of concepts and notions he has touched upon. I will concentrate upon three major questions. The first two — namely the domain of re-analysis procedures and the depth of re-analysis — both bear relations to a more general problem which may be called depth of understanding. In discussing the third point, I will try to unfold the wide range of possibilities for a behaviour which may be termed uncooperative. The domain of re-analysis

procedures

It is reasonable to assume that processes of language production and understanding are not deterministic in nature but that they come into being by the application of strategies. The application of a strategy always runs the risk of re-doing things already supposed to be settled if a catastrophic situation contradicting the strategy turns up. The question arises as to what objects re-analyses apply, i. e. what the domains of re-analyses procedures are. From Charolles examples and explanations of re-analysis in contexts of reading one could get the impression, that re-analysis always involves another treatment of the language signal in question. In fact, we find re-analyses of this kind; Charolles mentions backward eye movements, well-known from psycholinguistic experiments. Normally these movements span over quite short distances, but we all know from our experience with reading that we even go back several pages in a paper or monograph and re-read whole passages. However, different cases can also be found: In a criminal story entitled Mannen son gick upp i rök (Sjöwall and Wahlöö, 1966), a Swedish journalist seems to have disappeared during a stay in Budapest. A Swedish detective is sent to Hungary in order to solve the case. In the 70 page long description of his stay and his efforts to find the journalist, the authors of the story use the same technique as the author of the thief-story analyzed by Charolles: the reader finds a lot of statements about what the journalist has done in Budapest, whom he has met, where he has been, etc. However, at the end of the story, some 30 pages later, it turns out that the journalist had not been in Budapest at the time he had been supposed to. He had been killed and his murderer had been in Hungary in his place. Since the whole story is written from the perspective of the detective, the authors do not indicate what may be called changes of semantic space a posteriori. This style was chosen in order to retain the tension. Now, when the reader gets to know the solution for the case, he certainly has to do some re-analysis work, however, we cannot assume that he re-reads the 70 page description from the new perspective he had received. Why then do we sometimes re-read passages with quite long distances from the passage we actually interpret and sometimes not? A possible explanation is that the depth of understanding a reader is aiming at may require the

18

Η.-J. E i k m e y e r

re-analysis of the physical stimulus. If it does not, a re-analysis or rearrangement of the semantic model he has built up — his semantic memory — may suffice. When one tries to understand e. g. the proof of a logical theorem, one may be forced to re-read some definitions or other theorems mentioned earlier; in order to understand the solution of the criminal story, in contrast, such extensive re-analyses are not necessary, restructuring the semantic memory will do. In the last case, re-analysis will certainly not entail that every single semantic representation of the statements made, with respect to the journalists' supposed stay in Budapest, are reconsidered. A reader's semantic memory will normally not be complete in this respect and contain more or less condensed macropropositions. Now, let me take a look at spoken interaction with respect to the question whether re-analysis require re-consideration of the physical stimulus. In spoken interaction, the physical stimulus is not preserved by some medium, as soon as it appears, it disappears again. For purposes of re-analysis in such contexts, we can assume some kind of acoustic memory, preserving the acoustic impression a hearer has had, but the impressions are already the physical stimulus ( = language signal)

On Michel Charolles' "Coherence as a Principle . . . "

19

result of some interpretative process and cannot count as authentic. One may object that we have considerable capacities to retain acoustic impressions since we make frequent use of them in meta-communicative arguments. This is correct, but I would assert that it requires considerable effort to retain these impressions. Furthermore, they have to be anchored very soon after the physical stimulus they are representations of has been produced. Hence, re-consideration of physical stimuli in spoken interaction is impossible in principle; to some extent, acoustic memory impressions may be used instead. A third factor determining the domain of re-analysis procedures besides the medium of communication and the entended depth of understanding is the nature of the property which leads to a catastrophe which in turn in the stumbling-block for the re-analysis. With respect to coherence, a semantic property, semantic re-analysis is inevitable. It is either supported by re-analysis procedures operating on other domains or it is worked out independently. The domains of re-analysis procedures may be hierarchically arranged as shown in fig. 1. The concentric circles in this picture indicate a relation of inclusion, i. e. whenever the language signal is to be re-analysed this entails re-analysis on the other levels etc. From this perspective, semantics is the dominant level.

Depth of

re-analysis

From a procedural point of view, understanding entails re-analysis since reanalysis procedures are quite normally applied in the course of a process of understanding. Thus, if one wants to say something about the depth of understanding — which may vary from case to case according to conditions to be specified — then one obviously has to say something about the depth of re-analysis, i. e. the degree of detail a re-analysis process has. Because I argued for the dominance of semantic re-analysis when questions of coherence are concerned, I will try to exemplify what semantic re-analyses of different depths may amount to. For purposes of illustration, I will use the story "A Thief in the Night", discussed by Charolles. I will roughly outline some states that the semantic memory of a recipient reading this story may have. My description will include the detection of the inconsistency and two proposals for semantic reanalyses removing the inconsistency: the first proposal may justly be called a deep re-analysis, the second a shallow one. As representation language for the semantics, I will use lists of predicate logic formulae 1 . I did not decide upon this representational device because I 1

Lists of predicate logic formulae should not be confused with the conjunction of these formulae. In standard predicate logic one cannot reconstruct the catastrophic inconsistency turning up, since the incremental growth of a list of formulae does not have a counterpart.

20

Η.-J. Eikmeyer

regard it as suitable for semantic representations in Principle, but because it simplifies the exposition. In representing the meaning of the story, I boiled it down to the bare facts, again for ease of exposition. For the same reason, I also excluded reference to different points of time since the relevant time relations are clear from the context. The lists of formulae I will present consist of two columns, the left-hand columns contain (oversimplified) semantic representations of the text sentences, the right-hand columns contain inferences (marked by double arrows pointing to the right) and presuppositions (marked by double arrows pointing to the left). Figure 2 shows the state the semantic memory may have at the moment when the recipient notices the contradiction (•£) between the inference HAVE (x, w) in line (8) and the presupposition ι HAVE (x, w) in line (10). In these formulae, the constant χ denotes the protagonist of the story, the constant y the supposed thief and the constant w represents x's wallet. It is not clear whether the inferences in lines (4) and (5) are forward inferences made when the recipient processes the language material corresponding to ι HAVE (x, w) in line (4) or whether they are backward inferences made when the recipient reads the word pickpocket. Consequently, I left the decision open. WALK (x) M E E T (x, y) BUMP (x, y) -. HAVE (x, y)

STEAL (y, x, w) =>

HAVE (y, w)

=>

HAVE (x, w)

PICKPOCKET (y) D E M A N D (x, y, w) G I V E (y, x, w) G O - H O M E (x) F I N D (x, w)

and ρ = < pl5 p Lo > As I understand it, cooperation means to keep the cost as low as possible and to receive as high a pay-off as possible. Technically one can thus regard cooperation as the solution to an optimization problem having the following form: (13) c = min! and ρ = max! Since one generally cannot assume that anything can be gained without the least effort, we have to assume that c has a lower bound. Moreover, natural producers and interpreters have limited resources 2 only and thus c has also an upper bound. The optimization problem (13) would have a trivial solution if c and ρ were independent of one another. However, this is not the case, as e. g. Charolles's discussion of the Tancrede-story has demonstrated: If Lo had produced the version (2') of the story, the efforts of I — i. e. his cost q — would have been much higher in order to receive the same pay-off pi — i. e. in order to arrive at a coherent interpretation. The relationship between cost c and pay-off ρ may be mathematically expressed by the equation: (14) ρ = f(c) We can summarize these remarks as follows: Cooperation means solving an optimization problem (13) with the condition (14). What, then, does non-cooperation mean? Cooperation and non-cooperation should neither be used as a dichotomic pair of notions, nor should they be considered the extremes of a one-dimensional scale on which any kind of behaviour can be judged as more or less (un-)cooperative. In order to demonstrate this, I will play a combinatoric game starting from (14'), a more detailed formulation of (14) which takes into account that according to (12), ρ is an ordered pair. (14') p,

= f,(c) = f 2 (c) From a purely combinatoric point of view, in both f] and f 2 the components q and c Lo of c may occur together or alone. This leads to nine different specializations of (14') which can be consistently (and nicely) arranged in a network (see figure 5). PLo

2

Cf. Eikmeyer (1983) for further discussions of limited resources and its relations to other naturalness conditions.

24

H.-J. Eikmeyer Pi = f, (c,) P1.0 = f2 (ci.„)

P,

=

PL„

= f2

f, (C[.„) (CL)

Figure 5

Now we can start thinking if this theoretical classification bears any relation to reality. In fact, one can quite easily assign characterizations to the nodes of the network, which seem to express relevant traits of communicative behaviour: the center node represents cooperative behaviour, the remaining eight nodes represent different kinds of uncooperative behaviour. The upper corner may be labelled autonomous behaviour, the lower one mutual dependent behaviour. The three nodes to the left stand for I-dominated behaviour, their counterparts on the right hand side for Lo-dominated behaviour. However, all nine elements of this classification can only claim the status of a prototypical case which may be realized to a certain degree. One can easily add a concept of degree to the classification by assigning weights to the parameters in function ^ and f 2 . The current finite classification uses to weights only: 1 if a parameter appears in the function's parameter list, 0 if it does not appear. One can create highly elaborate techniques for the gradation of weights. For reasons of simplicity, I will mention the standard method of fuzzy systems: All real numbers in the closed unit interval [ 0 , 1]

On Michel Charolles' "Coherence as a Principle . . . "

25

will be used as weights and thus the space between the nine original prototypes is filled with graded weight constellations. This new classification contains infinitely many values, certainly enough for descriptive purposes. However, the well-known problem of descriptive adequacy, raised in connection with fuzzy systems using the unit interval, may be started anew 3 . This question should not detain us here because I do not claim that any classification proposed in this contribution has the status of a model for a classification of communicative behaviour. They only serve the purpose as guide-lines for further thinking. It may be useful to add more substance to the notions "cost" and "payoff", freely used until now. It should be obvious that the cost of both Lo and I are complexes of many components. With respect to Ci let me just mention memory space used and processing time needed. It is well-known that these two factors are highly interrelated: If the interpreter makes a lot of memory space available, he can shorten the processing time for data access. On the other hand, if access time is uncritical, memoral space may be decreased. However, storing data in a large memory is time consuming since the suitable place for the data has to be found first. In this respect a garden path strategy favours little memory space. It speeds up processing locally, but globally it runs the risk of an increase in total time. Similar conclusions can be drawn for a producer's cost c Lo : excessive planning slows down production speed; less planning may speed up things locally, but globally it again runs the risk of extended production time due to repairs. Amount of planning, memory space and processing speed are quite manageable concepts to be used for the explication of the term cost. With respect to the notion "pay-off", one is quickly confronted with far more diffuse concepts. I will demonstrate this with the aid of the thief-story: Charolles called the mode of production for this story non-cooperative, but he noticed that the author, nevertheless, took advantage of this mode. The same holds, for the interpreter as well, because the story would have been far less interesting to read if the interpreter had not been trapped. The author gains something too, since he becomes known as a writer of thrilling and challenging stories. The respective pay-off I and Lo receive is to be localized on a higher, aesthetic level and it makes up for the additional efforts counting as cost on the interpreter's side. Thus, if aesthetic gains count — and I maintain that they do — the production mode chosen by the author of the thief-story is clearly cooperative. Another point to be noticed as that the pay-off the producer of a text receives may depend upon the interpreter's pay-off and vice versa. For me, for instance, it is no fun telling a joke to somebody who does not catch the

3

Cf. Eikmeyer and Rieser 1978, p. 2 4 - 3 7 and Przelecki 1976.

26

H . - J . Eikmeyer

point. Taking such dependencies into account, we can further expand the classification. In all nine cases of figure 5 we could have (i) p Lo as a parameter in f,, i.e. p : = f, (..., p Lo ), (ii) pi as a parameter in f 2 , i.e. p Lo = f 2 (---, Pi) or (iii) both as parameters in the functions. One arrives at a classification of 27 cases which can be arranged in the form of a cube shown in figure 6.

Pi

PI PLO

= f| (· ·, PLo)

=

FI (· · , PLO)

= f2 (- • ,

PI„ =

PI)

F2 (· · , P I )

Figure 6

The nine cases with condition (i) above are to be found on the upper side of the cube, those with condition (ii) on the lower side. Condition (iii) cases are arranged on the intermediate plane and the prototype of cooperation is the center node of the cube. Obviously, we could derive a graded classification starting from the cube as we did before. This is still not the end of the combinatoric game, since until now all considerations have been based upon the optimization problem (13) — i.e. trying to minimize the cost and maximize the pay-off. One can imagine other versions. The following example is a very extreme case: Lo tries to — — — —

maximize minimize minimize maximize

Fs cost, Γs pay-off, Lo's cost, and Lo's pay-off.

On Michel Charolles' "Coherence as a Principle ..."

27

Let me stop the game here and add an important remark. I gave a so-tospeak objective description of cooperation as a solution to an optimization problem. In fact, all notions and relations have to be subjected to the subjectivity condition. This means that maxima and minima are extreme values judged by either / or Lo and these judgements may differ. The relation Pi = fi (ci, cLo, p Lo ) e. g. should be expressed more precisely in the following way: P. = J . ( f i ) (J.(ci), Ji(c L o ), M p L o ) ) where J, (x) expresses Fs judgement of x. Concluding

remarks

I argued that re-analysis procedures executed in connexion with the construction of a coherent interpretation are dominantly semantic in nature. Among these, both deep and shallow re-analyses can be imagined; psycholinguistic research should show what type is performed under what conditions. With respect to non-cooperative communicative behaviour, one has to face a wide range of dimensions along which it can differ from cooperative behaviour. References Eikmeyer, H.-J. 1983 "Procedural Analysis of Discourse", Text 3, 1, 11—37. Eikmeyer, H.-J. and Rieser, H. 1978 Vagheitstheorie. USP Mathematisierung der Universität Bielefeld. 1980 Pro^edurale Analyse: Eine Pilot-Untersuchung Kontextwechsel und Sprachdynamik. USP Mathematisierung der Universität Bielefeld. 1983 "Procedural Grammar for a Fragment of Black English Discourse", in: Ballmer, E. ed. (1985), Linguistic Dynamics: Discourse, Procedures, and Evolution. Berlin: de Gruyter. Przelecki, M. 1976 "Fuzziness as Multiplicity", Erkenntnis 10, 371—380. Sjöwall, Μ. and Wahlöö, P. 1966 Mannen som gick upp i röck. Stockholm: Nöhrstedt & Sönners Förlag.

J 0 Z S E F ANDOR

Strategies, Tactics and Realistic Methods of Text Analysis Comments on Michel Charolles' "Coherence as a Principle in the Regulation of Discursive Production" 0. The aim of Charolles' paper is to identify and characterize the processes of strategies and tactics text producers apply to act upon the interpretation processes evoked in the recipients or readers in the course of their understanding and conceiving of the text product itself and rendering coherence to it. He characterizes coherence in a vague sense as a principle for the interpretation of discourse in the sense that normally makes use of the "charitable hypothesis" according to which the text or discourse he faces can really undergo the interpretative or reinterpretation process on the basis of acceptability criteria. This being the case, the following two hypotheses can be made: (i) along such lines coherence is a general principle of text processing and analysis relying on a body of expectations both from the part of the producer and the interpreter; (ii) coherence is partly based on factual knowledge, subjective in nature. In what follows I will dwell on discussing factuality and subjectivity and the relevance of these notions for text processing in greater detail. 1. Charolles states, or, rather, prophesizes in the introductory part of his paper that "coherence is the 'minimal meaning' that allows access to other interpretations". Alas, he fails to specify what he actually means by "minimal meaning". The interpretation he gives right after declaring the principle thus seems to be quite trivial. I do not think there is any reasonable support on empirical or even on theoretical grounds for postulating a minimal-meaning condition for text coherence. I should rather postulate a set of "thresholds" in meaningfulness, but on global grounds. That is, in the sense that understanding a piece of written or spoken discourse highly depends on certain thresholds pertaining to different — co-textual, con-textual and textual — levels of analysis. As regards the structural relation of connexity and cohesion carriers, the threshold sets rather rigid conditions or rules for the process of analysis, which are highly objective in character. This idea is closely related with the nature of the particular components of grammar (primarily with morphology and syntax), which serve as reserves for providing access to

Strategies, Tactics and Realistic Methods of Text Analysis

29

means of rendering such conditions on texts. Thresholds skipped in these domains result in ambiguities or even ungrammaticality in terms of acceptability criteria. It is the domain of a more global nature where the threshold becomes significantly subjective, thus giving a larger scope for multiple-way interpretation, but still based on clear-cut guidelines, which I will discuss below. Strangely enough, Charolles himself tacitly seems to hold this view — in spite of the powerful statement criticized above — when he analyzes connexity (or cohesion) in his passage written in French (Nouvelle affaire d'autodefense), in which he calls attention to the relevance of the process of reanalysis. 2. In analyzing reception activities on the basis of strategies and tactics on the part of the text producer, Charolles identifies a number of methods, emission strategies or decision tasks as he calls them: the "garden-path strategy" closely related to the "intelligent strategy", and the "cooperative strategy" as well as the "non-cooperative strategy". Reanalyses can occur in all of these sequences of actions of text production and reception. It is the domain of such strategies of text construction that subjectivity enters the picture, and connexity relations are interpreted in the light of underlying and supporting conditions of coherence. The so-called "garden-path strategy" assumes a linear sequencing of reading activities on the part of the recipient. Texts constructed in accordance with this approach are said to be cooperative in nature, and, let me also add, such texts are likely to be redundant in the array of their co-textual means of make-up. Thus, for example, with their reliance on the expectations of the reader such texts are likely to have a powerful and redundantly organized lexical make-up. The text producer's hypotheses about the knowledge of the recipient result in too many clear-cut lexical items based on the tacitly acknowledged background general knowledge derived from a tacitly assumed matching of the expectations with respect to the lexical representations of the two partners. If there are any breaks in the connexity or cohesion conditions exemplifying the process, the intelligently managed "garden-path strategy" provides the means for a cooperative processing via a redundant (versus just sufficient or strict) representation. The above is exemplified by Charolles' first sample text: "The Story of Tancrede and Clorinde". It was observed by Halasz (1983: 418 — 424) that in the reception process the interpreter relies on a body of experientially based expectations and schemata stored in his memory, and with their help he oversimplifies the content of the passages, that is, he has a body of standardization devices at his disposal, and these help him to analyze a text linearly by utilizing the strategy that Charolles called "garden-path strategy". On the other hand, to cope with such efforts and tactics of readers, text producers, relying on their expectations and also on their knowledge about the expectation of the receiver, endeavour to enrich the passage in question within the given standards, in

30

J. A n d o r

so far as they intend to cooperate with the reader. This is the point where the idea of subjectivity and objectivity gains ground. Both the text producer and the receiver make use of standardized knowledge structures, scenic knowledge and frame knowledge, to match the former to the latter in their activities of representation and analysis. (Concerning the relation between the two types of knowledge structures see Andor, 1982 and de Beaugrande, 1980). Though these contain mainly stereotypical information in their content, they are individually acquired and applied. Hence, the content of scenes and frames is individually based (though in the course of development the framework of the two knowledge structures is constantly enriched by an ideal, totalistic, but, at the same time, standardized basis). As a result of the cooperative strategy (from the side of the producer) and the intelligent strategy (used by the receiver) there still arises a "struggle", during which failures in the wellbased reception can occur due to differences in the make-up and storage of the scenes and frames in the memory of the partners in spite of the attempts of standardization by both sides, and enrichment and oversimplification, used separately. Thus, the reanalysis strategy that is called "inexpensive" by Charolles assumes a relatively passive recipient, but, on the basis of the above, "inactivity" gains ground at the expense of potential activity from the recipient's side, as well, to oversimplify in the interpretation process. It is on the basis of the above struggle that the producer tries to provide a redundant representation — if necessary — once he intends to apply the cooperative strategy to make his text coherent and accessible for the recipient, as in examples such as introductory and informatory summaries in the beginning of broadcasts of opera or theater performances on the radio, as exemplified by "The Story of Tancrede and Clorinde". 3. Following these remarks, let me discuss, however, Charolles' other sample passage, his example for the non-cooperative strategy at greater length. Let me tell at the start that I am not sure about the source of the extract from the American magazine New York (Sept. 1st, 1980), nor is the reference given precisely. This is a pity, as contrary to the passage about Tancrede and Clorinde the reader does not know the exact origin of the "Thief in the Night" story on the basis of Charolles' paper. One could regard it as ordinary news of a strange kind, although with an occasionally unusual lexical makeup and whith relatively many paragraphs, unless there is something else in the air.1 Right at the start, let me declare that I accept Charolles' point of view according to which the text analyzed exemplifies the so-called non-cooperative strategy, and let me see how the strategy is realized or fulfilled in the writer's 1

Jänos S. Petöfi, one of the sources referred to by Charolles informed me (personal communication) that the story is the result of a competition for anecdotes supposedly about real life in New York City published in the magazine referred to.

Strategies, Tactics and Realistic Methods of Text Analysis

31

tactics purporting to deceive the reader, hence providing the story with a point, making the reader look "backwords for the passage where he was mislead". I will rely on the methodology of frame analysis for my assertions in the following. In identifying possible frames or a dominating frame in the story I will rely on the occurrence of frame triggering lexical items, words or complexes of words in the lexical representation of the events that the story manifests. I have called such items "text organizers" elsewhere (Andor, 1982). It is assumed that text organizers or perspectivizers play a critical role in realizing the "garden-path strategy", as it is these items where the close relation and matching between stereotypical knowledge and expectations are revealed, which lead to a standardized interpretation in a given language community. Of course, the content of some of these organizers is greatly influenced by developmental factors and socio-cultural background, and thus, the set of expectations developed by individuals needs to be handled with care. The sequential ordering of text organizers highly influences the process of interpretation and the identification of frames, and it also relates to the formal representation of a text in terms of paragraphs. It is basic requirement in a standardized interpretation process that the representation in terms of paragraphs be reasonable according to the textually bound thema-rhema matching. I would call this the linearity principle of text representation. Failures of following the principle result in failures in interpretability in a systemic way by following the "garden-path strategy" applied by the writer. On the other hand, once we face the writer's intention to deceive the reader by relying on a non-cooperative strategy, the selection and rendering of frame triggering organizers to given, specific positions in the text product will greatly contribute to the success of the story, which is, what the writer intends to achieve. Let us see now, how these notions work at least in the first part of the sample text, including the title. In order to find empirical support for the ideas presented later, I used the sample text in a test with 80 students as experimental subjects. A questionnaire with the following instructions was given to the subjects (who had never participated in such tests before): 2 "a. Read the passage once. b. Try to identify words or chains of words you think have dominance over the occurrence of others (items that could function as key-words in the passage). c. Try to give an explanation to your selection of words listed in b. d. What is your opinion about the sequences of paragraphs? e. What is your opinion about the title? f. Other remarks."

Let me sum up the results of the test initially.

2

The sample text was presented to the subjects of the experiment in a Hungarian translation.

32

J. Andor

Question b: The list below contains items that had significant rate of occurrence in the lists (the first figure indicates the number of occurrences, and the second refers to the occurrence in percentages of all answers). thief (80 — 100) night ( 8 0 - 1 0 0 ) transit strike (12—15) park ( 8 0 - 1 0 0 ) late (63-78.75) alone ( 4 6 - 5 7 . 5 ) path ( 8 0 - 1 0 0 )

stranger (68 — 85) bump (57-71.25) realise ( 4 2 - 5 2 . 5 ) wallet ( 8 0 - 1 0 0 ) pickpocket ( 8 0 - 1 0 0 ) demand ( 4 8 - 5 2 . 5 ) surrender (61 —76.25) mug ( 8 0 - 1 0 0 )

Question c: All the subjects who took part in the survey agreed that the basic principle of their selection was the critical role of the chosen items in their contribution to the following events and bringing the story to a point. 3 Question d: All the subjects felt it unnecessary for paragraph 3 (with the single sentence "It was gone") to form a separate unit in the text and agreed that it had better be attached to the second paragraph (100%). According to 69 subjects (86.25%) the story only needed three paragraphs, the first should end as in the original, the second with the fourth paragraph in the original, and the last separated as in the original. 11 subjects (13.75%) felt it necessary to separate the second paragraph while the majority wanted to place it to the end of the third of the original. Question e: 24 (30%) of the subjects expressed the view that the title was unnecessary, and 47 (58.75%) felt it advisable to eliminate the word thief from the title to get a sharper point of the story ("Strangers in the night" was suggested as a possible title by some of the subjects). The rest 9 (11.25%) did not require any change in the title. In order to gain further empirical support for the analysis, I performed another test, a survey of the associative capacity of some of the "key items" chosen by the participants of the first experiment. The items tested were park, path, pickpocket, surrender and bump. 30 subjects, none of which participated in the first test, took part in the second experiment. The subjects had one minute at their disposal for each item. The results and data gained were the following PARK

3

-

trees bushes green walk sunshine playground

A detailed analysis of the answers given to this test instruction cannot be carried out in the space given here.

Strategies, Tactics and Realistic Methods of Text Analysis

grass chilly fountain flowers benches leaves dry things pebbles thrush dog September spring SURRENDER

-

burglar force key police demand little child robbery sweating forhead tense atmosphere terminal report school

BUMP

-

toss kick foot accidentally excuse angry

PATH

-

narrow winding forest green evasion Indian (!!) dense lose the way dark night trickle of water bush noise danger

34

J. Andor

PICKPOCKET -

wallet money crowd shame beggar tricky skilful putter bus pocket anger shop police argument shouting

On the basis of the data gained, the following remarks can be made concerning the coherence of the text. Beyond any doubt, a thematically arranged lexical make-up can greatly contribute to the coherence of a text or discourse through the stimulating capacity of the elements. The title of the passage in question contains two organizers: " t h i e f ' and "night". Both items readily associate spontaneously with items creating a gloomy, fearful atmosphere. The title of the passage thus clearly reflects the core of the contents of the story: the act of stealing, theft (which is the dominant frame stimulated so early with its scriptal background also represented lexically later on), which could be described in terms of a "garden-path strategy" in a cooperative way. If the item " t h i e f ' was replaced by "park" in the title, due to the high rate of the occurrence of items that get associated with "park" in a pleasant sense, the title would not reflect such a gloomy, negative tone towards the emotional content of the story right at the start, at least not for the ordinary reader. As regards the array of organizers further on, "transit strike" occurs next. As it is irrelevant, though, from the point of view of the rest of the contents of the story, its frame triggering capacity does not effect supporting lexical items, and thus, apart from its — perhaps — relevant causal link with the item "was walking" taken in the strict sense, the item remains vacuous in the passage analyzed. This view is supported by the relatively low rate of rendering "transit strike" the role of organizer in the passage by the subjects of the first experiment. In the same sentence, however, another powerful item, "park" appears. We immediately get a tense tone, but this is highly supported by the two organizers in the title. Were it not for the occurrence of the item " t h i e f ' in the title, the gloomy atmosphere would not develop so early and readily in the story, at least not for the ordinary reader outside the United States. For the American reader, though, "park" does instigate a frame with a gloomy

Strategies, Tactics and Realistic Methods o f Text Analysis

35

content including fear, and this is the case with readers with background (scenic) knowledge o f New York City and its parks (including Central Park) and nights attached to it. For such readers the cooccurrence o f these items has its immediate gloomy and fearful effect, and it is on such grounds that with an additionally introduced supporting organizer, "path" (which has in itself a lot of items with a negative tone in its circle o f associates), the occurrence o f " o f course" (highly debatable otherwise, as remarked by the experimental subjects in their responses given to question f. o f the questionnaire) becomes relevanten the fourth sentence o f the first paragraph, where "spasm o f fear" also appears as a supporting unit to "path". Due to lack o f space I do not intend to carry on checking the role o f organizers and explaining their relevance for the passage further. It should be sufficient to say that it is not by chance that the majority of such items are verbs or nouns: they have the highest relational capacity in terms of argument roles in a cotextual analysis o f the sample. My purpose has been to emphasize the fact that the frame triggering capacity o f particular lexical items that serve as organizers in a passage highly contributes to the conditions o f coherence revealed by the interpreter, but, at the same time, the writer o f the text needs to rely on the frame knowledge o f the recipient matching it with the sociocultural domains in the make-up o f his own frames so that there is a balance of expectations from the point o f view of both partners. I f such an approach lacks failures of understanding will occur with the inability o f a coherent interpretation of some o f the items or the whole. The latter case is demonstrated by the inadequate use o f the items "realized" (instead o f "occurred to", for instance) and "pickpocket" in the sample text, which do not fit the scene represented by this frame of "theft" on empirical gounds (see the list o f associates triggered by "pickpocket" above). The occurrence o f "pickpocket" is ruled out on the basis o f the associative links gained from the test o f spontaneous lexical association, and thus the item fails to match the other organizers of the passage. What is the source o f the non-cooperative strategy, then, which Charolles has diagnosed in the text under discussion? As Charolles states, it is really hidden in the parallel o f the transparent versus opaque interpretation derived from the pronominal element " h i s " in the 11th line of the story. Whether transparency can be established by an intelligent re-analysis as a result o f which the opaquely represented item can be coindexed with the items "someone" in the third line and "somebody" in the last, and whether an intelligent re-analysis would provide methods broader than the scope o f connexity and cohesion, raises wider perspectives, especially analysis based on cognitively and socio-culturally conditioned knowledge structures (frames) which rely on factual knowledge of a similar scope (scenes and scripts). It should be sufficient to say that such knowledge structures play a critical role in setting coherence conditions on the textual representation o f information, and cohesion and connexity are shown to have a realizational role, they map the underlying structures onto the surface representation o f texts.

36

Andor, J. 1982

J. Andor

"A grammatika perspektiväja avagy a perspektiva grammatika ja [The perspectives of grammar or a grammar of perspective]", in: Koväcs, F. ed.: Njelvpedagögiai iräsok, Budapest, 8 4 - 9 8 . 1985 "On the Psychological Relevance of Frames", Quaderni di Semantica 6, 212—221. de Beaugrande, R. 1980 Text, Discourse, and Process. Norwood, N. J.: Ablex. Halasz, L. 1983 Λζ olvasäs: nyomoqas es felfedei^es [Reading: Investigation and Discovery], Budapest: Gondolat Hatekayama, K.; Petöfi, J. S.; Sözer, Ε. 1985 "Text, Connexity, Cohesion, Coherence" (preprint version, Bielefeld, November 1983), in Sözer, Ε. (ed.): Text Connextity, Text Coherence. Aspects, Methods, Results, Hamburg: Buske, 3 6 - 1 0 5 .

MICHEL CHAROLLES

Comments on Eikmeyer's and Andor's Report A. 1) Concerning the first point in Eikmeyer's report, the question of the domain of re-analysis processes, I agree with him when he says that it is necessary to take into account re-analysis strategies involving long distances. Obviously, in such cases subjects do not re-read all the pages separating the breakdown from the relevant passage(s) for its solution. Of course, re-analysis is not synonymous with re-reading. Re-analysis can be described as intelligent in that sense that the subject who accomplishes it is able: — to detect (generally easily) only the preceding passage(s) which is (are) relevant for the problem he is confronted with; — to re-arrange his previous interpretation without re-reading the whole text; — and then to continue his reading on the basis of the re-arranged interpretation. This is not always the case. Sometimes, the subject will effectively re-read the whole text, i. e. from its beginning to the point of breakdown. The example given by Eikmeyer: "When one tries to understand, for example, the proof of a logical theorem one may be forced to re-read some definitions or other theorems mentioned earlier and perhaps to re-treat the entire demonstration" sounds very likely. Of course, it would be interesting to collect observations on that topics. In my opinion the differences we can notice intuitively are certainly explainable from a cognitive point of view. But it would be too long to discuss this matter. 2) Now a remark concerning eye movements. The re-interpretation processes for the modified version of Tancrede and Clorinde's story or for the original version of "A thief in the night" may be accompanied of backward eye movements. But it is not obligatory because since these two texts are not very long the preceding information relevant for the re-interpretation processes may still be present in the reader's short term memory when the breakdowns occur. So the subject does not need to look backwards in the texts in order to search them. Thus eye movements do not always constitute a liable test for detecting re-analysis processes: they are only indicators.

38

Μ. Charolles

3) Two remarks concerning spoken

interactions·.

a) In spoken interactions the stimulus is not preserved by some medium, so re-interpretation processes can be more difficult than in written texts. Nevertheless one must pay attention to two facts: — the assumption above supposes that the short-term visual memory has the same capacity as the short-term acoustic memory (if we can speak likewise) but this is far from being sure! — the comparison supposes also that during reading there is no oralisation or sub-oralisation, which is also debatable, particularly in problematic situations of interpretation like that one envisaged here. b) In spoken interactions there is co-presence of Lo and / and it deminishes in principle the risks of difficulties resulting from the absence of any physical record. First / can ask questions to Lo, this is not always possible but it happens very often. Anyway the utterer receives "signs of understanding" from the receiver(s), so he can adapt (accomodate) his discourse to his audience. This is, for example, very interesting to observe in classrooms. There are important differences between written and spoken communication but in oral situations the receivers are not so disavantaged as we can imagine in theory. 4) Concerning the question of the depth of re-analysis Eikmeyer's remarks are particularly suggestive. To my mind the distinction between depth and shallow methods or strategies of re-analysis is certainly very useful if we want to describe what really occurs in many contexts. Eikmeyer's definition "of shallow re-analysis" ("a shallow re-analysis is conterfactual in that sense that it is not, itself, really worked out, but only the information one would need to work it out is supplied" — p. 21) seems to me particularly interesting. Nevertheless in his presentation of a deep semantic re-analysis there is a point which sounds curious to me. This point is the following: In figure 1 Eikmeyer transcribes the nominal expression "the pickpocket" with a predicate "PICKPOCKET" and the argument Y. My question is: why does Eikmeyer not similarly analyse, for example, the first nominal mention of the individual X, i. e. with a predicate ("YOUNG MAN") and an argument ("X")? Why is "the path" not transcribed in the same manner ("PATH(Z)")? And so on for (at least) all the definite nominal expressions of the text. This point is very crucial. The line 6 in figure 1 anticipates the solution proposed at the line 6' in figure 2. The notation adopted in the first figure (line 6) makes the solution possible. But the decision to analyse "the pickpocket" with a predicate and an argument cannot be motivated by any theoretical argument. It is because we know that we will have to introduce the doxastic modality to believe in the second figure that we accept to decompose the definite nominal expression "the pickpocket" in a Russellian manner.

Comments on Eikmeyer's and Andor's Report

39

This decision is not innocent but we must keep in mind that "in reality" the reader has to determine precisely which nominal expressions he has to opacifiate. 5) Concerning the last part of Eikmeyer's paper (the chapter entitled Cooperative and un-cooperative behaviour) the idea to use a system of formal description with the two categories pay and cost and to envisage the questions of cooperativity like an optimisation problem is very exciting and promising. Although there is no confusion in Eikmeyer's paper, it seems to me that there is a risk of misunderstanding in the terminology adopted: the reason is that in its ordinary meaning to pay is not the opposite of to cost. But this is a detail unimportant as regards the proper content of Eikmeyer's suggestions. B. Concerning Andor's report I would like to comment on only one point, namely the role of scenes or frames in connection with strategies of reading and reanalysis. Frames play a crucial role in the activation of "routine" interpretation. For instance, if a text is connected with a frame in the reader's mind he will be able to interpret it more easily. This assumption sounds intuitively likely and it has been experimentally confirmed. Thus frames have certainly something to do with gardenpath strategies: we can assume that they reinforce such strategies. Now, concerning re-analysis processings, in the case of deep semantic retreatment (see Eikmeyer's report), we can engage in the hypothesis that frames facilitate the detection of relevant data in the part of the text treated previously. If these data are wellknown and conventionally ordered, the reader has no difficulty in finding them again. But we can also argue the opposite hypothesis. Considering that frames encourage superficial reading strategies, it may be the case that it makes access more difficult to previous information in the preceding passages of the text connected with the frame. Let us suppose for instance a story with a "restaurant episode": John entered in a restaurant. He sat down and waited for the waitress. There were many persons in the dining room. A few minutes later a young girl came to him and gave him the menu ... Let us imagine a version of the text in which the expression "a young girl" is written in quotation marks. We can conjecture that certain readers will not perceive these marks: observing the standard framelike structure of the text, they will treat it in a shallow way and not pay attention to this "detail" of expression. Then, it might be the case that this information achieves a crucial importance for the understanding of the following part of the text and we

40

Μ. Charolles

can therefore conceive that the mood of treatment adopted by these readers makes retrospective access to "a young girl" more difficult. On the contrary, we can argue that the frame "restaurant" is helpful for the receivers to detect this passage since, during the re-analysis phase, they have no difficulty to distinguish the precise sub-episode wherein the waitress occurs. We could develop a comparable argument about the role of the so called "super-structural" knowledge because this knowledge is of a framelike type — the only difference being that it concerns prototypical organization of texts and not prototypical organization of sequences of facts. As regards the analysis of text organisers presented by Andor it seems to me very interesting, but, in my opinion, it suffers from three major defaults: a) The decomposition in key-words is purely intuitive, how could it be justified? b) The empirical investigations of the associative nets connected with the key words provide stimulating results but the information delivered has been obtained independently of the context. So the question arises if the associations would have been the same if the words had been tested in connection with the text "A thief in the night"? c) Finally, the remarks on text organizers and associative nets are not properly connected with the problem I tried to deal with, namely the underdetermination of opacity in the nominal expression "the pickpocket". My question would be: how can the observations presented have relevant consequences for the examination of the problem I wanted to treat? It has to be taken into account that this problem concerns a very specific linguistic point the resolution of which engages local interpretative skills.

KLAUS HEGER

Text Coherence in a Dialogical Presuppositional Group: Chapter X X X I of Unamuno's "Niebla" 0. Introduction The purpose of this paper is twofold: by applying a given theoretical apparatus to a given text in order to find out the elements that guarantee the coherence of this text, some particular features of this text and its coherence shall be analyzed, and at the same time the apparatus applied in this analysis shall be tested as for its usefulness. With its analytical aim, this paper thus belongs to the wider field of what commonly is referred to as text linguistics. With its more theoretical aspirations, however, the theoretical apparatus applied and tested in this analysis being the hierarchy of ranks of meaningful language units such as has been proposed in Heger (1976) (quoted in the following as MWST) and re-shaped in Heger and Mudersbach (1984) (quoted in the following as AAA), it is rather the complementarity of language universale and language typology in Hansjakob Seller's sense that is envisaged (cf. Heger (1985)). The hierarchy of ranks which is conceived of and constructed as a noemic tertium comparationis that is meant to serve in comparative semantics (cf. Heger (1983) and Heger (1985)), will be briefly summarized in the first section. For all details that would pass beyond the frame of the present paper, as well as for the development that has led to the conception of this hierarchy, the reader is referred to MWST and AAA. One point, however, should be stressed right from the beginning: this conception has its roots in what semantics has been in European linguistics at least since Michel Breal vulgarized the term in the nineties of last century (cf. Heger (1986), § 5); it is not interested in existential quantifiers, truth values, and the like, that hardly seem appropriate to classify an utterance like "no existes" as anything but an inacceptable contradiction. The second and main section of this paper will deal with the analysis of chapter XXXI of Miguel de Unamuno's novel "Niebla". This text — which is quoted according to the 1951 edition in the Coleccion Crisol — has been chosen because of a number of characteristics that seem to render it particularly favourable as object of an analysis of text coherence in dialogues: — nearly the entire chapter consists of one dialogue, — there is no problem of delimitation since throughout the whole dialogue there are only two partners, and

42

Κ. Heger

— the dialogue invites the reader to interpret it on different levels simultaneously since it takes place between the protagonist of the novel and its author. The first part of section 2 therefore will have to give an explicit analysis of the position of chapter XXXI within the novel as a whole — not only for the sake of making exhaustive use of the possibilities offered by the theoretical apparatus that is to be tested by its application, but also in order to introduce, in a sufficiently well-defined way, the distinction of the different levels of interpretation. Only after this preparatory step, the text itself will be presented and analyzed in the second part of section 2. This analysis will be exhaustive down to the ranks of what might be considered as sentences. Since intra-sentential coherence does not belong any more to the object of this paper, and since in the particular case of the text chosen it can be taken for granted, analyses that descend to still lower ranks will only be added where they contribute to detect further aspects of inter-sentential coherence. Taken together, the two sections thus imply that a theoretical instrument which itself has been constructed in an ascending way starting with the smallest meaningful units, is made use of in a descending analysis of a given text as its starting point. As for this complementarity as well as for many other aspects, this analysis benefits a lot from an earlier attempt of mine that also aimed at analyzing a given text (cf. Heger (1974)), and in particular from the critical comments it has received from Wolfgang Raible (cf. Raible (1974); and Gülich and Raible (1977): 136-151). 1. Hierarchy of ranks of signemes 1.1. General In of to be

principles

order to enable a hierarchy of ranks of signemes to meet the requirements its task as a tertium comparationis in comparative semantics, the way how construct such a hierarchy is bound to a number of conditions which can formulated as the following principles.

1.1.1.

Exhaustivity

Any hierarchy of ranks of whatsoever entities should be constructed in a way that does not introduce arbitrary limitations of the domain it is supposed to cover. This means that its construction has either to start with the minimal rank and proceed in an ascending way, or inversely to start with the maximal rank and proceed in a descending way. 1.1.2.

Ascendance

Since, in the case of meaningful language units, it seems less adventurous to consider as well-defined, or at least as well-definable, minimal units rather than maximal ones, only one of the two possibilities pointed out in § 1.1.1

Text Coherence in a Dialogical Presuppositional Group

43

really does exist. It necessarily leads to a decision in favour of the construction of hierarchies that starts with the minimal rank, proceeds in an ascending way, and compels the hierarchy to remain open-ended as for the question of what might be considered as a maximal meaningful language unit (cf. below § 1.2.5.5). It should be stressed, however, that this preference for ascending constructions is a secondary one and does not imply any disqualification of descending constructions as such. 1.1.3 Specificity In order to prevent hierarchies from remaining mere devices of recursive addition, it is recommendable to remember Karl Bühler's classification of the relation that holds between word and sentence as an unrepeated and unrepeatable one (cf. Bühler (1934): 258). Translated into the principle that is of interest here, this means that any hierarchy of different ranks of meaningful units has to be constructed in a way that guarantees that each transition from one rank to the immediately following one is given a definition that is qualitatively different from the definition of any other transition within this hierarchy (cf. Heger (1986), § 2.2); in other words, that each such transition is defined, and distinguished from all other transitions, by its particular differentia specifica (cf. Gülich and Raible (1977): 138). Introducing such differentiae specificae into the construction of a hierarchy has two important consequences. As soon as — and this is the normal case — a differentia specifica allows for further subspecifications, it does not only define one particular transition between two ranks that immediately follow each other, but also provides a common denominator for a whole group of what in an ascending construction are the next higher ranks that can be defined by means of those subspecifications. Thanks to this possibility, the eleven ranks of the hierarchy to be presented here, can be subdivided into five homogeneous groups. The second important consequence is in a way complementary to the first one. Since any new differentia specifica by definition creates heterogeneity as compared with what has been specified by the preceding one, and since on the other hand, there is not the least guarantee that all differentiae specificae that might be made use of, are of equal importance in all languages — there is rather a guarantee for the contrary —, the corresponding possibilities of rank-shift must be provided for within the hierarchy to be constructed. In the case of the hierarchy to be presented here, this is done by — re-interpreting the well known terms of 'free' and 'bound' in a way that classifies as 'free on rank Rn' any unit of rank Rn that can appear without any addition as a unit of rank Rn + 1 ( = ascending rank-shift), and as 'bound on rank Rn' any unit of rank Rn that can appear on rank Rn + 1 only when accompanied by one or more units of rank Ri with 1 < i < n; and

44

Κ. Heger

— admitting what I have come to call 'rank-reduced designations of sememes of higher ranks' ( = descending rank-shift, cf. MWST, § 3.0.2.3; cf. also below § 2.2.4). 1.1.4

Semanticity

Since the hierarchy to be presented here is one of meaningful language units, it has by definition to be semantically based. In the sense of Ferdinand de Saussure's signifiant: signifie dichotomy, it refers to the signifie side of language units and, more precisely, in the sense of Andre Martinet's two articulations, it refers to units of his premiere articulation. These references imply a number of important consequences: — With reference to another famous Saussurean dichotomy, the hierarchy is referred to the higher level of abstraction on which it might be appropriate to speak of langue systems (for more details cf. Lieb (1970); Heger (1982), § 3.2; and Heger (1985), § 1.3); in order to insist hereupon, I shall use from here on the term 'signeme' instead of the neutral 'meaningful language unit'. As in earlier publications of mine (cf. MWST: 39 — 41 and 339), the signeme is defined as meaningful unit on the abstraction level of langue systems. — Contrary to what in former times was familiar in European linguistics, neither the term 'semantics' nor the term 'signeme' imply any limitations whatsoever that might restrict their applicability to the lower — lexical — ranks of the hierarchy to be presented: the langue equivalents of sentences and texts are just as well signemes as are those of words and parts of words (cf. Gülich and Raible (1977): 137-138). — Contrary to what is familiar in logical semantics, the term 'semantics' and its derivations are referred not to extensional interpretations of the relation between signs and things denoted by them, but exclusively to the signifie side of signemes and to what might be classified as their intensional analysis (cf. above Introduction; and Heger (1981)). 1.1.5.

Extra-monolinguality

Since the hierarchy of ranks of signemes to be presented here is meant to serve as a tertium comparationis in comparative semantics, it has to meet the conditions that hold for any tertium comparationis of this kind: it has to be extra-monolingual (cf. Heger (1985), § 1.4.1) in the sense of being sufficiently independent of particular languages in order to be a real tertium and, at the same time, of being sufficiently related to language(s) in order to be capable of serving as a tertium comparationis. This condition is met by the fact that all definitions of ranks higher than rank R1 refer directly, not to the signeme itself, but to its sememe(s) as the only point within the trapezium model of the language sign (cf. MWST: 3 8 - 6 0 ; and Baldinger (1980): 260) where the

Text Coherence in a Dialogical Presuppositional Group

45

monolingual signifie side and the extra-monolingual side of the noeme can be directly mapped upon each other. 7.2. Ranks of signemes 1.2.1. Lexical ranks (cf. also Baldinger (1980): 2 6 3 - 2 6 6 ) The whole hierarchy starts off with the definition of the minimal signeme as the minimal unit of Andre Martinet's premiere articulation from where I have also taken over the term 'moneme'. Monemes can be subdivided — into monemes with exclusively reflexive-metalinguistic sememes (such as French pomme2 with the only information that it forms part of pomme de terre "potato", a signeme of a higher rank; or as the Latin conjugation class vowel -a- in laud-a-mus) and monemes with non-exclusively reflexivemetalinguistic sememes (such as French pomme\ "apple" or Latin laud- and -mus)·, and — into lexemes (such as French pomme with either sememe, or Latin laud-) and grammemes (such as Latin -a- and -mus) (cf. MWST: 79 — 80 and 218-220). From these subdivisions, subspecifications can be obtained that allow for the definition of two more ranks: Rank R l : Moneme = minimal signeme. Rank R2: Minimal Autosemic Unit = one lexeme + monemes with exclusively reflexive-metalinguistic sememes, bound to this lexeme. Rank R3: Flectional Form = one minimal autosemic unit + grammemes bound to this minimal autosemic unit, including monemes with exclusively reflexive-metalinguistic sememes that are bound to these grammemes. These three ranks are the only ones where finite inventories can be expected to exist in a given language; they are therefore best known as the domains of lexicology and lexicography on one hand, and of morphology on the other. 1.2.2. Prepositional

ranks

The transition from rank R3 to the next group requires a device that allows for an explicit description how two or more lexemes contained in the corresponding number of flectional forms, can be combined with each other. The most generic form to which all such combinations can be reduced in the last instance, is that of a quality being attributed to an individual or to a set

46

Κ. Heger

of individuals. As a means of representation of such attributions, the actant model has been developped (cf. MWST, chapter 4; and AAA, chapter 2; for some examples cf. below § 1.3) out of its three basic components which consequently are the actant that stands for an individual, a set of individuals or the like; the relator that stands for a quality; and the predicator as the component that establishes the attribution of a relator to an actant. It should be stressed that the actant model, in spite of its similarity with and its undoubted indebtedness to Lucien Tesniere's dependency grammar, represents actants and relators as noemic categories; it does not, however, represent directly monolingual signemes that can only in a second step, and after exhaustive semasiological analysis, be mapped onomasiologically onto these noemic categories (cf. below § 2.2.4). Since the qualities represented by the relator, can be either one-place or two-place attributes, there is at least one possibility for introducing a further subspecification together with the new differentia specifica represented by the predicator. Another subspecification can be introduced at this point thanks to a specific characteristic of this hierarchy (although at least implicitly present in many others as well): since the great majority of systemic specifications that in many languages of the world are expressed by grammemes, specify entities that can be located only on ranks R4 and R5, and since these specifications so far have been taken account of only in the case they are expressed by grammemes, it is the appropriate moment to re-introduce them as such after rank R5. Although from the purely noemic point of view they, again, constitute a new differentia specifica, from the point of view of their virtual integration already in the definition of rank R3 it seems preferable to consider them as forming one group together with ranks R4 and R5 (cf. MWST, chapter 5; and AAA, chapter 3). This second group of ranks can thus be defined as follows: Rank R4: Compositional Form = one flectional form + signemes of rank Ri with 1 < i < 3 that are bound to this flectional form by the attribution of one-place relators. Rank R5: Propositional Form = one compositional form + signemes of rank Ri with 1 < i < 4 that are bound to this compositional form by the attribution of two-place relators. Rank R6: Specified Propositional Form = one propositional form + signemes of rank Ri with 1 < i < 5 the sememes of which are systemic (quantitative, deictic, classifying, etc.) specifications of components of the actant model representing this propositional form.

Text Coherence in a Dialogical Presuppositional Group

47

1.2.3. Sentence ranks The differentia specifica that allows to pass beyond rank R6 is assertion — interpreted as the speaker's taking over the communicative responsability for the fact that there is full agreement between his proposition and the state of affairs corresponding to it in the extensional domain of denotata to which he is referring. Assertion serves as a new differentia specifica when it is — either given by the speaker, which is the case of normal declarative sentence, — or explicitly marked as missing and in need of being supplemented. This latter case covers most of what has become commonly known as speech acts, and a first step towards systematization of speech acts can be taken by starting off from the two distinctions between — whom the speaker O E expects to supplement the missing assertion: himself (OE), the hearer (OE), or somebody else (OE), and — how the missing guarantee for agreement between proposition and extensional state of affairs is expected to be given: either by adding an assertion to the respective proposition ( = lingual) or by creating the respective state of affairs (— non-lingual). The combination of these two distinctions yields the following six-place system of basic speech acts (cf. chart 1); it goes without saying that this system is open for further subspecifications that might seem useful. Chart 1: assertion to be supplemented by: form:

lingual:

non-lingual:

speaker

hearer

OE

OE

OE

rhetorical question to be answered by the speaker

question

rhetorical question not to be answered

INRH

INT

INOP

promise, menace, etc.

request, command, prohibition, etc.

wish

PROM

IMP

OPT

The third group of ranks can thus be defined as follows: Rank R7: Speech Act Specified Propositional Form (Σ 7 ) = one specified propositional form

48

Κ. Heger +

signemes of rank Ri with 1 < i < 6 bound to it, the sememes of which are speech act specifications of this specified propositional form.

Speech act specifications will be marked by the abbreviations introduced in chart 1 (e. g. I N T for questions), speech act reactions by the same abbreviations with a superscribed arrow added (e. g. I N T for answers). Rank R8: Asserted Propositional Form (Σ 8 ) either = one specified propositional form + signemes of rank Ri with 1 < i < 6 bound to it, the sememes of which are assertions of this specified propositional form, or = a pair consisting of one speech act specified propositional form and the corresponding speech act reaction. Assertion will be marked by E. For more details of the notational representation of signemes Σ 7 and Σ 8 , cf. AAA, §§4.1.1 and 4.1.2. 1.2.4. Presuppositional Groups Asserted propositional forms of rank R8 and, at least for certain criteria, speech act specified propositional forms of rank R7 can share certain characteristics and in this case be assembled in groups that presuppose this common characteristic. In turn, this common characteristic therefore can serve to define a corresponding presuppositional group of the respective signemes Σ 8 or Σ 7 and thus allows to pass beyond rank R8. Since, however, different presuppositional groups cannot only be combined with each other but also be subordinated to each other without any restriction, all of them have to be located on the same rank R9 — and this in a way that allows for all kinds of recursive mutual combinatory relations (for an earlier attempt of mine to attribute them to different ranks, cf. the well-founded criticism of Raible (1974): 1 4 0 - 1 4 4 ; and Gülich and Raible (1977): 1 4 3 - 1 4 4 ) . The following presuppositional groups are provided for in the present state of the hierarchy: 1.2.4.1. Assertory Presuppositional Group (Σ9Β) defined as a chain of signemes Σ 8 or of signemes Σ 7 and Σ 8 each non-first member of which refers in some way — normally by consent or dissent — to the assertion of a preceding member (for the notational representation cf. AAA, § 4.2.1).

Text Coherence in a Dialogical Presuppositional Group

49

1.2.4.2. Actantial Presuppositional Group (Σ9Α) defined as a chain of signemes Σ 7 and/orX 8 (and applicable in a wider sense to a chain of such chains as well) each member of which contains at least one actant that is referentially identical with an actant contained in a preceding (anaphorical identification) or following (cataphorical identification) member (for the notational representation cf. AAA, § 4.2.2). 1.2.4.3. Monological Presuppositional Group (C— Σ9Κ,) defined as a chain of signemes Σ ^ 8 with at least one member of rank R7 or R8, all members of which are addressed by the same speaker (sender^ Ο Ε ^ π ) to the same hearer (receiver) ΟΕ* μ π ) or to the same group Ο Ε ^ π ) of hearers (receivers) (for the notational representation cf. AAA, § 4.2.3.1). 1.2.4.4. Dialogical Presuppositional Group (Σ9 Κμ) defined as a chain of ρ monological presuppositional groups Ο - Σ 9 1 ^ · " to C —Σ9κ and < 8 4 > and has to be read as 'the dialogical presuppositional group Σ 1 0 Κ ι consists of three subunits which, with reference to communicative level K 2 , are three monological presuppositional groups C — Σ1014*1·1', C — Σ10Κ(1·2), and C - Σ10κ), and 4. the actantial presuppositional group Σ',^2 that consists of three signemes Σ 8 '. 2.2.2. Text In this section, the entire text of chapter XXXI ( = signeme Σ'^ 2 ) is reproduced and marked for all signemes down to those of rank R7. In order to avoid unnecessary repetitions, the following simplifications of marking these signemes will be made use of: 81 9 Al — Components Σ31 of signemes Σ31 with j = 1, 2, or 4, will be marked as: Κ 2 : j : i. — Signemes Σ^ 3 . with 1 < i < 192 will be marked as: Κ 2 : 3 : i. — Signemes C — Σ®,1^2·» with 1 < i < 93 will be marked as : j :, followed by the number marking their components; in the text itself, verba dicendi and their specifications that serve to designate these monological presuppositional groups as frames — in other words, that correspond to any component other than N2 of model 4 which has been given as an example in § 1.3 —, will be marked by double underlining. — Signemes Σ$ 3 « with 5 < i < 191, 1 < j < 93, and 7 < k < 8, will be marked as: k : i (automatically preceded by the corresponding j :); in case of k = 7, speech act specification will be marked as follows: 7a = INRH 7e = INT 7i = INOP 7o = PROM 7u = IMP 7y = OPT corresponding speech act reactions are marked with " (ä, e, ϊ, ö, ü, and y respectively). Because of their particular interest for the question of text coherence, transitions between subsequent signemes Σ^ί2·') will be marked — with +, if they connect two or more signemes Σ 7 that belong to the same monological presuppositional group and have identical speech act specifications,

68

Κ. Heger

— with *, if they connect two signemes Σ 7 the second one of which conveys the speech act reaction of the first one and which thus form one common signeme Σ 8 , and — with if they connect two or more signemes Σ 7 and/or Σ 8 that form one common signeme Σ 9 Ε by consenting to, dissenting from or questioning a preceding assertion. In cases of ambiguous interpretation, decision on where to delimit signemes Σ 7 and Σ 8 has been taken according to the criterium of sentence initial capitals. For an English translation of the text cf. Appendix. K2: 1 : K2 : 1 : K 2 :: 1 : K 2 :: 1 :

K2;: 1 : K2:: 2 : K 2 :: 2 : K2;: 3 : K2:: 3 :

K2:: 3 : K 2 :: 3 : 1:8: 1:8: 1 :7e:

1 Aquella tempestad del alma de Augusto termino, come en terrible calma, en decision de suicidarse. 2 Queria acabar consigo mismo, que era la fuente de sus desdichas propias. 3 Mas antes de llevar a cabo su propösito, como el näufrago que se agarra a una debil tabla, ocurriösele consultarlo conmigo, con el autor de todo este relato. 4 Por entonces habia leido Augusto un ensayo mio en que, aunque de pasada, hablaba del suicidio, y tal impresiön pareciö hacerle, asi como otras cosas que de mi habia leido, que no quiso dejar este mundo sin haberme conocido y platicado un rato conmigo. 5 Emprendio, pues, un viaje acä, a Salamanca, donde hace mäs de veinte ano vivo, para visitarme. 1 Cuando me anunciaron su visita sonrei enigmäticamente y le mande pasar a mi despacho-libreria. 2 Entrö en el como un fantasma, miro a un retrato mio al öleo que alii preside a los libros de mi libreria, y a una sena mia se sentö, frente a mi. 1 Empezö habländome de mis trabajos literarios y mäs ο menos filosoficos, demostrando conocerlos bastante bien, lo que no dejö, jclaro estä!, de halagarme, y en seguida empezö a contarme su vida y sus desdichas. 2 Le ataje diciendole que se ahorrase aquel trabajo, pues de las vicisitudes de su vida sabia yo tanto como el, y se lo demostre citandole los mäs intimos pormenores y los que el creia mäs secretos. 3 Me miro con ojos de verdadero terror y como quien mira a un ser increible; crei notar que se le alteraba el color y traza del semblante y que hasta temblaba. 4 Le tenia yo fascinado. 5 — iParece mentira! -reoetia-. iparece mentira! V A no verlo no lo creia ... 6 7 No se si estoy despierto ο sonando ...

2 : 7e : χ·8 — Ni despierto ni sonando -le conteste. X 3 : 7e : 9 — No me lo exolico ... no me lo explico -anadio-: mas Duesto aue usted oarece saber sobre mi tanto como se yo mismo, acaso adivine mi propösito ... * 10 4 :7e: — Si -le diie-, tu -v recalque este tu con un tono autoritario-, tü. abrumado Dor tus desgracias, has concebido la diabölica idea de suicidarte, y antes de hacerlo, movido por algo que has leido en uno de mis Ultimos ensayos, vienes a consultärmelo. K 2 : 3 : 11 El pobre hombre temblaba como un azogado, mirändome como un poseido miraria. K 2 : 3 : 12 Intentö levantarse, acaso para huir de mi; no podia.

Text Coherence in a Dialogical Presuppositional Group

69

Κ 2 : 3 : 13 No disponia de sus fuerzas. 5 : 7u : 14 — jNo, no te muevas! -le ordene. *

6 : 7 e : 15

— Es que ..., es que ... -balbucio.

*

7 : 7e : 16 — Es que tu no puedes suicidarte, aunque lo quieras. Si 8 : 7e : 17 — ,;C0mo? -exclamo al verse de tal modo negado ν contradicho. *

9 : 7e : 18 9 : 7e : 19 * 10 : 7e : 20 X 11:8: 21 X 12 : 8 : 22 13 : 8 : 23 14 : 7e : 24

+

14 : 7e : 25

-

Si. Para que uno se pueda matar a si mismo, ^que es menester? -le pregunte.

— Que tenga valor para hacerlo -me contestö. — No -le dije-, jque este vivo! - jDesde luego! — jY tu no estäs vivo! — jComo que no estoy vivo? ,;Es que he muerto? -v empezö. sin darse clara cuenta de lo que hacia. a palparse a si mismo.

*

15 : 7e : 26 15:8: 27 X 16 : 7e : 28

+

16 : 7e : 29

— jNo, hombre, no! -le replique-. Te dije antes que no estabas ni despierto ni dormido, y ahora te digo que no estäs ni muerto ni vivo. — jAcabe usted de explicarse de una vez, por Dios! jAcabe de explicarse! -me suplico consternado-. porque son tales las cosas que estoy viendo y oyendo esta tarde, que temo volverme loco.

*

17 : 7e : 30

X 18 : 7e : 31

— Pues bien; la verdad es, querido Augusto -le dije con la mäs dulce de mis voces-, que no puedes matarte porque no estäs vivo, y que no estäs vivo, ni tampoco muerto, porque no existes ... — ,;C0mo que no existo? -exclamo.

*

19 : 7e : 32

— No, no existes mäs que como ente de fiction; no eres, pobre Augusto, mäs que un producto de mi fantasia y de las de aquellos de mis lectores que lean el relato que de tus fingidas Venturas y malandanzas he escrito yo; tu no eres mäs que un personaje de novela, ο de nivola, ο como quieras llamarle. 19:8: 33 Ya sabes, pues, tu secreto. X 20 : Al oir esto quedose el pobre hombre mirändome un rato con una de esas miradas perforadoras que parecen atravesar la mira e ir mäs allä. miro luego un momento a mi retrato al oleo que preside a mis libros. le volvio el color ν aliento. fue recobrändose. se hizo dueno de si. apovö los codos en mi Camilla, a que estaba arrimado frente a mi. v. la cara en las palmas de las manos ν mirändome con una sonrisa en los ojos. me diio lentamente: 7e : 34 — Mire usted bien, don Miguel ..., no sea que este usted equivocado y que ocurra precisamente todo lo contrario de lo que usted se cree y me dice.

70

21 : 71:

Κ. Heger

35

— Υ V

Έ' /

crearse 189 sonarse 48

Text Coherence in a Dialogical Presuppositional Group

Model 28:

Designations:

creador 178, 184

Designations:

criatura 88, 132 producto 32, 65

2.2.4.4. Thematic element W, (= domains of denotata j W fl (fictional 'world', corresponding to level 3) Designations: fantasia 32, 62, 65 ficcion 32, 36, 65, 66, 68, 114, 177, 181 ficticio 83, 114, 183 nivola / nivolesco 32, 65, 71, 117, 183, 184 novela / novelesco / novelista 32, 41, 65, 68, 69, 81

88

Κ. Heger

Wit ('world' of literature, corresponding to level 2) Designations: pretexto (para que la historia llegue al mundo) 37 real 43 W 0 ('real world', corresponding to level 1) Designations: realidad 36, 127 (de) veras 65 verdad 30 W ri ('world' of religion, interpretable as level 0) Designations: celestial / cielo 126 Dios 28, 126, 137, 138, 142, 156, 160, 171, 172, 179 Nuestro Senor 126 religion 126 Cf. also Unamuno 51, p. 257. 2.3. Levels of interpretation Again and again, analyses in §§ 2.1 and 2.2 have been faced with the phenomenon of simultaneous reference to different domains of denotata·, in § 2.1.1, classification of the overall signeme Σ ] ο υ as a nivola was the first information given by the author OE 1 about a peculiarity that was going to be confirmed by the facts — that in § 2.1.2, one of the authors OE 2 occurring on the communicative level K 2 , at the same time belongs to level 3 of the domain W fl of the particular fictional 'world' of the novel proper; — that in § 2.1.3, the inventory given in chart 2 comprises not only — what has to be expected — actants that belong to level 3 of the domain W fl , but also the very author Unamuno (N t j) belonging to level 1 but acting as well on level 3; — that, illustrating what has been observed in § 2.1.3, in §§ 2.2.1 to 2.2.3 a conversation is analyzed in which two interlocutors are participating who not only belong to different levels of domains W; (N (> , and N 3 1 ) , but who in addition freely move up and down within the hierarchy of three levels, to which a fourth one is added by the invocations of God (cf. e. g. signemes Σ ^ > 2 and Σ%> 9 ); and — that in § 2.2.4, designations f o r ' d ' , Έ ' , and W, overlap in a characteristic way (cf. e. g. sonar f o r ' d ' and Έ ' ; fingir / ficcion / ficticio for Έ ' and W fi ; real(idad) for WQ and W lt ; or Nuestro Senor Don Quijote in Wri or WitP). What has to be added to this enumeration, are the apparent contradictions that turn out to fit perfectly and coherently into this frame of simultaneous reference to different levels. Beside the "no existes" (cf. signeme Σ quoted already above in the Introduction, there are many more examples of such apparent contradictions as e. g. the fact that Unamuno threatens to kill

Text Coherence in a Dialogical Presuppositional Group

89

Augusto Perez (cf. signeme Σ^ 2 ^) — and in the end does make him die — after having asserted that he is not alive at all (cf. signeme Σ^2·1^). With all these means of refined structuring, Unamuno charges the reader N, 2 in his different functions as OE 1 and OE 2 with the task of reconstructing the perfect coherence of a text of which he has to prove to be an adequate "foster-father" (cf. Raible (1983)).

3. References Baldinger, Κ. 1980 Semantic Theory — Towards a Modern Semantics, Oxford. Bühler, Κ. 1934 Sprachtheorie, 1) Jena 1934; 2) Stuttgart 1965. Gülich, E; Heger, K. Raible, W. 1974 Linguistische Textanalyse — Überlegungen %ur Gliederung von Texten, Hamburg 1) 1974; 2) 1979. Gülich, E.; Raible, W. 1977 Linguistische Textmodelle, München. Heger, K. 1974 "Signemränge und Textanalyse", 1) in: Gülich, E.; Heger, K.; Raible, W., 1974, 1—71; 2) in: Grammars and Descriptions, eds. van Dijk, Τ. Α.; Petöfi, J. S., Berlin, New York 1977, 2 6 0 - 3 1 3 . Heger, K. 1976 Monem, Wort, Sat% und Text, Tübingen, (quoted as: MWST.) Heger, K. 1981 "Außersprachlichkeit — Außereinzelsprachlichkeit — Übereinzelsprachlichkeit", in: Logos Semantikos — Studta linguistica in honorem Eugenio Coseriu 1921 —1981, eds. Geckeier, H.; Schlieben-Lange, B.; Trabant, J.; Weydt, H., Berlin, New York 1981, vol. II, 6 7 - 7 6 . Heger, K. 1982 "Verhältnis von Theorie und Empirie in der Dialektologie", in: Dialektologie — Ein Handbuch %ur deutschen und allgemeinen Dialektforschung, eds. Besch, W.; Knoop, U.; Putschke, W.; Wiegand, Η. E., vol. I, Berlin, New York 1982, 4 2 4 - 4 4 0 . Heger, K. 1983 "Zum Verhältnis von Semantik und Noemantik", in: Zur Semantik des Französischen — Beiträge %um Regensburger Romanistentag, eds. Stimm, H.; Raible, W.; Wiesbaden (Zeitschrift für französische Sprache und Literatur, Beiheft 9) 1983, 40—44. Heger, K. 1985 '"Concepts' and 'Noemes'", in: Language Invariants and Mental Operations — International Interdisciplinary Conference held at GummersbachjCologne, Germany, September 18- 23, 1983, eds. Seiler, H., Brettschneider E., Tübingen 1985, 9 7 - 1 0 1 . Heger, K. 1986 "Fifty Years of Linguistics — Four Examples", in: Folia Linguistica 20, 1986, 251-264. Heger, Κ.; Mudersbach, Κ. 1984 Aktantenmodelle — Aufgabenstellung und Auflauregeln, Abhandlungen der Heidelberger Akademie der Wissenschaften, Philosophisch-historische Klasse, Jahrgang 1984, 4. Abhandlung (quoted as: AAA). Lieb, Η. Η. 1970 Sprachstadium und Sprachsystem, Stuttgart.

90

Κ. Heger

Raible, W. 1974

"Vergleich mit der von Klaus Heger durchgeführten Textanalyse", 1) in: Gülich, E.; Heger, K.; Raible, W., 1974, 127 — 149; 2) in: Grammar and Descriptions, eds. van Dijk, Τ. Α.; Petöfi, J. S., Berlin, New York 1977, 1 6 0 - 1 7 3 .

Raible, W. 1983

"Vom Text und seinen vielen Vätern — oder: Hermeneutik als Korrelat der Schriftkultur", in: Schrift und Gedächtnis — Archäologie der literarischen Kommunikation /, eds. Assmann A. and J.; Hardmeier, Chr., München 1983, 20—23. de Unamuno, M. 1951 Niehla; Abel Sänche%, Madrid (Aguilar: Collecciön Crisol)

4. Appendix:

English translation

of chapter xxxi of Unamuno's

"Niebla"

The following English version of the text analyzed in the preceding sections is taken from the translation by Warner Fite (published in 1928 by A. A. Knopf, New York/London, and re-edited in 1973). Marking of signemes and of transitions between subsequent signemes is the same as in § 2.2.2. Κ2:1: Κ2 :1: Κ2 :1:

Κ2:1:

Κ2:1: Κ2:2: Κ2:2 : Κ2:3:

Κ2 :3 :

Κ2:3: K 2 :: 3 : 1 : 8: 1 : 8: 1 :7e:

1 The storm in the soul of Augusto ended in a terrible calm: he had resolved to kill himself. 2 He wanted to put an end to that self which had been the cause of all his misery. 3 But before carrying out his plan it occurred to him, like a drowning sailor who grasps at a straw, to come and talk it over with me, the author of this whole story. 4 Augusto had read an essay of mine in which I made a passing reference to suicide and this, together with some other things of mine that he had read, had evidently made such an impression upon him that he did not wish to leave this world without having met me and talked with me for a while. 5 Accordingly, he came here to Salamanca, where I have been living for twenty years past, to pay me a visit. 1 When his call was announced I smiled enigmatically, and I had him come into my study. 2 He entered like a ghost. He looked at a portrait of me in oil which presides over the books of my library, and then at a sign from me he took a seat opposite me. 1 He began by speaking of my literary works, in particular of those that were more or less philosophical, showing that he knew them very well; which, of course, did not fail to please me. Then he began to tell me of his life and of his misfortunes. 2 I interrupted him by telling him to spare himself the trouble; I was as familiar with the vicissitudes of his life as he himself; and this I demonstrated to him by citing some of the most intimate details, and in particular some things that he thought to be utterly hidden. 3 He looked at me with genuine terror in his eyes, as one looks at some incredible being. I seemed to see a change in the colour and in the lines of his face, and I saw that he even trembled. 4 I had him fascinated. 5 "It hardlv seems true." he kept reoeatine. "it hardlv 6 7

I shouldn't believe it if I had not seen it. I don't know whether I am awake or dreaming —"

Text Coherence in a Dialogical Presuppositional Group

2 : 7e : 3

: It

4

•. It

8 X 9

:

91

"Neither awake nor dreaming," I replied. "I can't explain it — I can't explain it," he went on. "But since you seem to know as much about me as I know myself, perhaps you guess my purpose in coming."

*

:

K 2 :: 3 : K 2 :: 3 : K 2 :: 3 : 5 :7u:

10

"Yes," I said. "You" — and I gave to this "you" the emphasis of authority — "you, oppressed by the weight of your misfortunes, have conceived the diabolical idea of killing yourself; and before doing it, impelled by something you have read in one of my last essays, you have come to consult me about it." 11 The poor man shook like a drop of mercury and looked at me with the stare of one possessed. 12 13 14 "No, don't move," I commanded him. *

7 Ü ·• 15 6 •• 'e

" D o you mean — do you mean —" he stammered.

*

7 : 7e : 16 X 8 : It •. 17

"I mean that you cannot comit suicide even if you wish to do so." "What!" he cried, finding himself so flatly opposed and contradicted.

*

9 : 7e : 18 9 : 7e : 19

"Yes, if a man is going to kill himself what is the first thing that is necessary?" I asked him.

*

10 : 7e : 20 SC 11 :; 8 : 21 X 12 ;: 8 : 22 13:: 8 : 23 X 14;: 7e : 24

"That he should have the courage to do it," he replied. "No," I said, "that he should be alive." "Of course!" "And you are not alive." "Not alive! What do you mean?

+ 14:: 7 e : 25

D o you mean that I have died?" And without clearly knowing what he was doing he began to pass his hands over his body.

*

15 : 7e : 26 15 : 8 : 27

"No, my man, no!" I replied. I told you before that you were neither waking nor sleeping; now I tell you that you are neither alive nor dead."

16 : It

"Tell me all of it at once; God,

:

28

+ 16 ;: 7 e :

29

17:: It

•.

30

18::

It:

SC 31

tell me all," he begged me in terror. "With what I am seeing and hearing this afternoon I am afraid of going mad."

*

"Very well, then. The truth is, my dear Augusto," I spoke to him the softest of tones, "you can't kill yourself because you are not alive; and you are not alive — or dead either — because you do not exist." "I don't exist! What do you mean by that?"

92

Κ . Heger *

19 : 7e :

32

19:8:

33 X

20 :

7e :

34

21 : 7§ :

35

" N o , y o u d o n o t exist except as a fictitious entity, a character of fiction. My p o o r A u g u s t o , you are only a p r o d u c t of my imagination and of the imagination of those of m y readers w h o read this story which I have written of your fictitious adventures and misfortunes. You are n o t h i n g m o r e than a personage in a novel, or a nivola, or whatever you choose t o call it. N o w , then, you k n o w your secret." U p o n hearing this the p o o r man continued to look at me for a while w i t h one of t h o s w e perforating looks that seem t o pierce your o w n gaze and go beyond: presently he glanced for a m o m e n t at the portrait in oil which presides over my books, then his colour returned and his breathing became easier, and gradually recovering, he was again master of himself. He rested his elbows o n the arm of the sofa opposite me. against w h i c h he was leaning; and then w i t h his face in the palms of his hands he looked at me with a smile and he said slowly: "Listen t o me, D o n Miguel — it can't be that you are mistaken, and that what is h a p p e n i n g is precisely the contrary of w h a t you think and of w h a t you have told me?"

*

" A n d what d o you mean by the contrary?" I asked, rather alarmed t o see him regaining his self-possession.

*

22 •.!%•. 36

"May it n o t be, my dear D o n Miguel," he continued, "that it is you and not I w h o are the fictitious entity, the one that does n o t really exist, w h o is neither living n o r dead?

+ 22 \Ί%·.

37

23 : 7e :

38

May it not be that you are n o t h i n g m o r e than a pretext for b r i n g i n g my history into the w o r l d ? "

*

"Really this is t o o m u c h ! " I cried, n o w becoming irritated.

X 24 : 7u : 39

"Please d o n ' t get so excited, Senor de U n a m u n o , " he replied. " K e e p calm.

24 : 8 :

40

You have expressed d o u b t s a b o u t my existence —"

25 : 8 :

X 41

X 26 : 7u : 42 26 : 7e :

43

27 : 7e :

44

" D o u b t s ? N o ! " I interrupted. " A b s o l u t e certainty that you do not exist outside of the novel that I have created." "Very well; then please d o n ' t be disturbed if I in turn, d o u b t your existence rather than my o w n . Let us come to the point. Are you not the person w h o has said, n o t once but several times, that D o n Q u i x o t e and Sancho are not only real persons but m o r e real than Cervantes himself?"

*

X 28 : 7u : 45 28 : 7e : 46

"I can't deny it, but the sense in which I said it was —" "Very well, never m i n d in what sense. Let us come to another point. W h e n a man w h o is lying asleep in his bed dreams of something, which is it that m o r e truly exists, he as the consciousness that dreams or the dreams themselves?"

*

29 : 7 | :

47

" A n d w h a t if the dreamer dreams that he himself exists?" — I turned the question on him.

Text Coherence in a Dialogical Presuppositional G r o u p

30 : 7 | :

93

48

"In that case, friend D o n Miguel, my own question would be, in what fashion does he exist? As the dreamer who dreams o f himself or as something dreamed by himself?

30 : 7u : 49

And note this, moreover: in entering this discussion with me you are already recognizing me as an existence independent of yourself."

*

31 : 7ü : 50

" N o t at all. N o ,

31 : 7ü : 51

not at all," I said quickly.

31 : 8 :

52

"In entering this discussion I am merely satisfying a private need of my own. Apart from discussion and contradiction I am never alive; and when there is no one outside of me to question and contradict me I invent some one to do it within me.

31 : 8 :

53

My monologues are dialogues."

32 : 8 :

54

" A n d perhaps the dialogues that you fabricate are nothing more than monologues."

33 : 8 : 33 : 8 :

X 55 56

34 : 8 :

X 57

34 : 8 :

X 58

+

MC 35 : 7u : 59 *

"It may be. But in any case I tell you, and I wish to repeat it, that you do not exist outside of me —" " A n d I will again suggest this to you, namely, that you do not exist outside of me and of the other characters that you think you have invented. I am certain that D o n Avito Carrascal would be of my opinion and the great D o n Fulgencio —" " Y o u needn't mention them — "

36 : 7{J:

60

36 : 7e :

61

"Very well, then, I won't; but you shouldn't make fun of them. And now let us see; what do you really think about my suicide?"

37 : 7e :

62

37 : 7e :

63

And that settles it."

38 : 8 :

64

38 : 8 :

65

" Y o u r saying that 'it does not really suit me' is very Spanish, Sefior de Unamuno, but it is far from edifying. Moreover, even granting your strange theory that I do not really exist and that you do, that I am nothing but a character of fiction, the product of your imagination as a novelist — or as a 'nivolist' — even so there is no reason why I should submit to 'what really suits you,' that is, to your caprice.

38 : 8 :

X 66

39 : 8 :

67

*

+

" I think, then, that since you do not exist except in my imagination — as I tell you again — and since you neither ought nor are you able to do anything but just what I please, and since it does not really suit me that you should kill yourself — well, you are not going to kill yourself.

For even those whom we call characters o f fiction have their own inwrought logic - " "Yes, I know that s o n g . "

94 40 : 8 :

Κ. Heger Χ 68

"But, really, it is a fact that neither a novelist nor a playwright is at all able to do anything that happens to occur to him, to a character that he creates. That a fictitious character in a novel should do that which no reader would ever expect him to do, is forbidden by all of the established principles of art

>: 41 : 8 :

69 X 42 : 7e : 70

"Doubtless a character in a novel —" "Well, then?"

*

43 : 7e : 71

"But a character in a nivola — a 'nivolistic' character —"

44 : 7u : 72

"Let us drop these buffooneries. They are offensive, and they wound me where I am most sensitive. Whether of myself, as I think, or because you have given it to me, as you suppose, I have my own character, my own manner of being, my own inwrought logic, and this logic demands that I kill myself —"

>:

44: 8 :

73

45 : 8 :

74 3C 46 : 7e : 75

+

"You may think so, but you are wrong." "Let us see. Why am I wrong?

46 : 7e : 76 +

And where am I wrong?

46 : 7e : 77 +

Show me where my mistake lies.

46 : 7e : 78

Since the most difficult knowledge that there is, is knowing yourself, I may easily be mistaken and it may be that suicide is not the logical solution of my problem. But prove it to me. For though this self-knowledge be difficult, Don Miguel, there is another kind of knowledge that seems no less difficult —"

46 : 8 :

79

47 : 7e : 80

"And that is? —" I asked.

*

48 :

He looked at me with a smile that was shrewd and enigmatic and then he said slowly: 7e : 81 "Well, that which is more difficult than self-knowledge is this: that a novelist or playwright should know the characters that he creates or thinks he creates." K 2 : 3 : 82 These sallies of Augusto were beginning to make me uneasy and I was losing my patience. 49 : 8 : 83 "And I insist," he added, "that even granting you have given me my being — a fictitious being, if you please — even so, and because it is so, you cannot prevent me from killing myself just because, as you say, it does not really suit you." 50 : 7u : 84

+

"Very well, that will do -

50 : 7u : 85 +

enough!" I cried, bringing my fist down on the sofa.

50 : 7u : 86 +

"Hold your tongue!

50 : 7u : 87

I don't wish to hear any more impertinence!

Text Coherence in a Dialogical Presuppositional Group

95

+ 50 : 7u : 88

And from a creature of mine, too!

50 : 7o : 89

And since I have had enough of you and I don't know, moreover, what more to do with you, I have now decided, not that you may not now kill yourself, but that I shall kill you.

+ 50 : 7o : 90 +

You are to die, then, but soon!

50 : 7o:

Very soon!"

91

51 : 7e : 92

+ 51 : 7e : 93

Aueusto was horror-struck. "What!" he cried. " D o you mean that you are going to let me die, make me die — you are going to kill me?"

*

52 : 7e : 94 X 53 : 7u : 95

"Yes, I am going to cause you to die."

53 : 7u : 96 +

never!

53 : 7u : 97

never!" he shrieked.

+

"Oh, never!

*

54 : 7ü : 98

"Ah!" I said, looking at him with mingled pity and rage.

54 : 7e : 99 +

"And so you were ready to kill yourself, but you don't want me to kill you?

54 : 7e : 100

And you were about to take your own life, but you object to my taking it?"

*

55 : 7e : 101 X 56 : 8 : 102 X 56 : 8 : 103

X 57 : 7a : 104

"Yes, it is not the same thing —" "To be sure, it is not. I have heard of cases of that kind. I heard of a man who went out one night armed with a revolver, intending to take his own life, when some thieves undertook to rob him. They attacked him, he defended himself, killed one of them and the others fled. And then, seeing that he had bought his life at the cost of another's, he renounced his intention." "One can understand that," observed Augusto. "It was a matter of taking the life of somebody, of killing a man; and after he had killed another, what was the use of killing himself?

*

57 : 7ä : 105 X 58 : 8 : 106 58 : 7e : 107

Most suicides are frustrated homicides; men kill themselves because they have not the courage to kill others —" "Ah! now I understand you, Augusto, I understand. You mean that if you had had the courage to kill Eugenia [ = N 3 4 of chart 2] or Mauricio [ = N 3 ω of chart 2], or both, you would not be thinking of killing yourself, isn't that so?"

*

59 : 7e : 108 3S 60 : 7e : 109

"Let me tell you, it is not precisely of them that I am thinking — no!" "Of whom, then?"

96

Κ. Heger

61 : 7e : 110

"Of you" — and he looked me straight in the eve.

62 : 7e : 111

"What!" I cried, rising to my feet. "What!

+

62 : 7e : 112

Have you conceived the idea of killing me? You? And of killing me?"

*

63 : 7u : 113

"Sit down and keep cool.

*

Do you think, Don Miguel, that it would be the first case in which a fictitious entity, as you call me, had killed him whom he believed to have given him his being — his fictitious being, of course?" *

64 : 7e : 115

+

"This is really too much," I said, walking up and down my study. "This passes all limits.

64 : 7e : 116

This couldn't happen except —"

65 : 8 : 117 Si 66 : 7u : 118

"Except in nivolas," Augusto completed with a drawl.

+

"Very well, enough!

66 : 7u : 119 +

enough!

66 : 7u : 120 +

enough!

66 : 7u : 121 +

This is more than I can stand.

66 : 7u : 122 X 66 : 8 : 123

You came here to consult me — me, you understand — and you?

X 67 : 7u : 124 68 : 7ü : 125 X 68 : 8 : 126

69 : 7e : 127

You begin by desputing my own existence, forgetting that I have the right to do with you anything that suits me — yes, just what I say, anything that may happen to occur to —" "Don't be so Spanish, Don Miguel —" "And now this too, you idiot! Well, yes, I am indeed a Spaniard, Spanish by birth, by education, Spanish in mind, body, and language and almost by professions and occupation; Spanish above everything and before everything. Spanishism is my religion, the heaven in which I wish to believe is a celestial and eternal Spain, and my God is a Spanish God, the God of our Lord Don Quixote, a God who thinks in Spanish and who said in Spanish, Let there be light! Sea — his word was a Spanish word —" "Well, and what of it?" he interrupted, recalling me to reality.

*

70 : 7e : 128

"And now you have conceived the idea of kiling me.

70 : 7a : 129 +

Of killing me -

70 : 7a : 130 +

me?

70 : 7a : 131

And you?

Text Coherence in a Dialogical Presuppositional Group

+

70 : 7 a : 132

97

A m I to die at the hands of one of my creatures?

*

70 : 7 ä : 133

I'll stand no more of it.

70 : 7o : 134

And so to punish you for your insolence, and to put an end to these disintegrating, extravagant, and anarchistic ideas with which you have come to me, I hereby render judgment and pass the sentence that you are to die.

+

7 0 : : 7 o : 135

+

7 0 : : 7 o : 136

Ό. 71 : 7u :: 137 * 72 •• '7οÜ ·· 138

+

72 •• '7οÜ •:• 139 X 73 : 7 u : : 140

As soon as you reach home you shall die. You shall die, I tell you, you shall die." "But — for God's sake!" cried Augusto. now in a tone of supplication, pale and trembling with fear. "There is no God that can help you. You shall die." "Yes, but I want to live, Don Miguel, I want to live — I want to live —"

*

74 •• 7' eÜ ·· 141

"Weren't you just now thinking of killing yourself?

*

75 : 7e : 142

75 : 7 u : : 143 *

76 •• 7' eÜ ·• 144 *

77 : 7e : 145 X 77 : 7u : 146

+

7 7 : 7u : 147 * 7 8 : 7ü : 148 X 7 9 : 7u : 149

"Oh! if that is why, Don Miguel, then I swear to you that I will not kill myself, I will not take away this life which God, or yourself, has given me; I swear it to you — now that you wish to kill me I myself want to live — to live — to live —" "What a life!" I exclaimed. "Yes, whatever it may be. I want to live even though I am again to be mocked at, even though another Eugenia and another Mauricio tear my heart out. I wish to live — live — live —" "Now it cannot be — it cannot be —" "I want to live — live — I want to be myself, myself, myself."

*

8 0 : 7ü : 150 X 81 : 7u : 151

+

81 :: 7 u : 152

"But what if that self is only what I wish you to be — ?" "I wish to be myself — to be myself! I wish to live!" and his voice was choked with sobs.

*

8 8 : : 7ü : 153

"It cannot be — cannot be —"

8 3 : 7u : 154

"Listen, Don Miguel, for the sake of your children, of your wife, of whatever is dearest to you!

+

83 :: 7u : 155

Remember that you will then cease to be yourself — that you will die —"

98

Κ. Heger +

83 : 7u : 156

He fell on his knees at my feet, begging and imploring me: "Don Miguel, for God's sake! I want to live. I want to be myself."

*

84 7ü

157

"It cannot be, my poor Augusto," I said, taking him by the hand and lifting him up. "It cannot be.

+ 84 7Ü 158 84 8 : 84 8 : 84 8 :

159 160 161

5C 85 8 : 162 X 86 8 : 163 86 8 : 164

87 7e

X 165

I have now decreed it — it is written — and irrevocably; you can live no longer. I no longer know what to do with you. God, when he does not know what to do with us, kills us. And I do not forget that there passed through your mind the idea of killing me —" "But, Don Miguel, if I —" "It makes no difference. I know where I stand. And really I am afraid that if I do not kill you soon, you will end by killing me." "But are we not agreed that —?"

*

88 7e 88 8 :

166 167 X 88 8 : 168 X 88 8 : 169 X 88 8 : 170 X 89 7u 171

"It cannot be, Augusto, it cannot be. Your hour has come. It is now written, and I cannot now recall it.

And for all that your life can now be worth to you —" "But — good God!"

*

90 7ü 90 7u

172 173

Ü 91 7'a

174

"There is neither 'but' nor 'God' that can avail you. Go!"

*

91 91



'a



'a

"And so you won't?" he said.

+ 175

"You refuse?

+ 176

You are unwilling to let me be myself, come out of the mist and live, live, live; to see myself, hear myself, touch myself, feel myself, feel my own pain, be myself — you are unwilling, then?

+ 91 : 7a : 177 *

And so I am to die as a fictitious character?

91 : 7ä : 178

Very well, then, my lord creator Don Miguel, you too are to die, you too! And you will return to that nothing from which you came!

+ 91 : 7ä : 179 +

God will cease to dream you!

91 : 1'i : 180

You are to die, yes, you are to die, even though you do not wish to; and die shall all of those who read my story, all of them, all, all, without a single exception.

Text Coherence in a Dialogical Presuppositional Group

91 : 7ä : 181 X 91 : 8 : 182 X 91 : 8 : 183 X 91 : 8 : 184

3S 92 : 7e : 185

99

Fictitious entities like myself — just like myself! All of them are to die; all, all. And it is I who tell you this — I, Augusto Perez, a fictitious entity, a 'nivolistic' entity, just like yourself. For as for you, my creator Don Miguel, you too are only a 'nivolistic' entity, and your readers are 'nivolistic' entities, just like me, just the same as Augusto Perez, your victim —" "Victim!" I cried.

*

93 : 7e : 186

+

93 : 7e : 187 93 : 7o : 188 X 93 : 8 : 189 Χ 93 : 7o : 190 X 93 : 7u : 191 Κ 2 : 3 : 192 Κ2 :4 : Κ2:4: Κ2:4:

"Victim, yes! To create me only to let me die! Well, you too are to die! He who creates creates himself, and he who creates himself dies. You will die, Don Miguel, you will die, and all those who think me, they are to die too!

To death, then!" This supreme effort of the passion for life — of the thirst for immortality — left poor Augusto utterly weak. 1 I pushed him towards the door. He walked out with his eyes fixed upon the ground, 2 passing his hands wonderingly over himself as if he were uncertain of his own existence. 3 I wiped away a furtive tear.

MICHAEL METZELTIN

Co-report on "Text Coherence in a Dialogical Presuppositional Group: Chapter XXXI of Unamuno's Niebla" by Klaus Heger 0. Preliminary

Note

As co-author of a Text Theory (see: M. Metzeltin and H. Jaksche (1983), and: M. Metzeltin (1981)) I will first of all refer to this theory as a comparative basis for the discussion of Heger's model and analysis.

1. Purpose, aim and object of the examination The purpose of Heger's examination is the analysis of Text Coherence in a text rich in dialogues. The aim of his examination is to find out whether a theoretical apparatus proves to be suitable. The object, to which the analysis and the apparatus will be applied, is chapter XXXI of Unamunos' novel "Niebla". This chapter is especially rich in acts of dialogues that can be easily delimited from each other. Besides, it is a very interesting fact, that the 'fictional world' of the characters in the novel joins the 'real world' of the author in such a way, that the reader will be urged "to interpret it (= the dialogue) on different levels simultaneously" (page 42). The choice of purpose, aim and object is clearly formulated and efficient. None the less some questions remain unanswered to the co-referee: — The term 'text coherence' gets neither explained nor discussed (as it was done in Metzeltin and Jaksche (1983), chapter 4, and Metzeltin (1981), chapter 11). What does the referee mean by 'text coherence', how does text coherence practically come about in his opinion, and what value does it have for the analysis and interpretation of texts? — A theoretical apparatus only is interesting, when its practical achievements get tested. But in the course of that I would first apply the apparatus to some normal cases and later on, if the results are positive, to special cases.

Co-Report on "Text Coherence ..."

101

"Niebla" tends to be a kind of Anti-Novel, in which even the realisation of coherence is different from the one in the traditional realistic Spanish novels (by authors such as Perez Galdos, Pardo Bazin etc.). To me the actual test-procedure has apparently been skipped over. — In my opinion text coherence means the coherence of a whole text. The analysis of the coherence in a textual fragment (e. g. a chapter) can be a good exercise, but may hardly lead to the expression of anything essential on the topic of text coherence. A linguist can, of course, analyse the continuous coherence between the complex sentences of textual fragments, in order to find out the rules for the connection of complex sentences. In that case the connection of complex sentences is the central problem, — and not that of text coherence. The referee is especially concerned with text coherence in dialogues and the mixing of the 'fictional world' with the 'real world' (see pages 1—2). Especially by means of these two devices, Unamuno removes from the rules of the traditional novel, as can be deduced from the following sample passage: — Si, que empezaras creyendo que lo llevas tu, de tu mano, y es facil que acabes convenciendote de que son ellos los que te llevan. Es muy frecuente que un autor acabe por ser juguete de sus ficciones ... — Tal vez, pero el caso es que en esa novela pienso meter todo lo que se me occura, sea como fuere. — Pues acabarä no siendo novela. — No, sera ..., sera ... nivola. — (Υ que es eso, que es nivola? — Pues le he oido contar a Manuel Machado, el poeta, el hermano de Antonio, que una vez le llevo a don Eduardo Benot, para leerselo, un soneto que estaba en alejandrinos ο en no se que otra forma heterodoxa. Se lo leyo, y don Eduardo le dijo: "Pero, jeso no es soneto! ..." "No, senor — le contestö Machado —, no es soneto, es ... sonite." Pues asi con mi novela, no va a ser novela, sino ... jcomo dije?, Navilo ..., nebulo, no no, nivola, eso es, jnivola! Asi nadie tendrä derecho a decir que deroga las leyes de su genero ... Invento el genero, e inventar un genero no es mäs que darle un nombre nuevo, y le doy las layes que me place. j Y mucho diälogo! — j Y cuändo un personaje se queda solo? — Entonces ... un monologo. Y para que parezca algo asi como un diälogo invento un perro a quien el personaje se dirige. (chapter XVII).

It would be useful to discuss how significant dialogues and possible 'worlds' are to text coherence and what consequences Unamunos' techniques have for the coherence of the novel.

2. The theoretical

apparatus

On principle the theoretical apparatus comprises more or less complicated actant models and types of signemes.

102

Μ. Metzeltin

A signeme consists of one or several monemes. Heger distinguishes five main types of signemes which mainly correspond a monemic surface description with the following text subdivisions: (ranks of signemes)

(text subdivisions)

lexical ranks propositional ranks sentence ranks presuppositional groups text ranks

lexeme and syntagma sentence complex sentence paragraph text

The signem rank Rn is achieved by the combination of a signeme of rank η + 1 with specific monemes or groups of monemes. Thus the signeme rank 7 for instance, is formed by: "one specified propositional form" ( = R6) + "signemes of rank Ri with 1 < i < 6 bound to it, the sememes of which are speech act specifications of this specified propositional form" (page 48). The signemes can be analysed with the help of the actant models. These consist basically of a two-place functor, -specified either as a positive or a negative one, — which establishes the attribution of a relator R (a quality, for instance) to an actant (that might stand for an individual term) — (page 46). In the paper the functors Ρ (quality functor), Τ (temporal functor) and C (causal functor) are employed. With the help of this apparatus Heger succeeds in covering not only sentences, but also speech acts, — not only complex sentences, but also rather large text units, — not only the text itself, but also the title, the author and even the possible 'worlds', to which the intensional noemes refer. This theoretical apparatus raises a number of questions to the co-referee. In part these might be pure questions of comprehension, which are caused by the conciseness of presentation, the concise exemplification and the constant pressure to decode the basically distinct, but not always simple notations. It is out of question, that the surface structure of a rather long text consists of various hierarchically connected parts (such as Communicate (

\

Title Complex (

Author

Text ϊ Title

(

Chapter 1 ... Chapter η (

λ

Paragraph 1 ... Paragraph η

)

and that these parts can formally be described by ranks of signemes.

Co-Report on "Text Coherence . . . "

103

But how relevant are the various types of signemes to the constitution of the text? Metzeltin and Jaksche have shown that the pairs "Proposition ( = nucleus of proposition + complex of transmission) / Realisation of Proposition" and "Textoide ( = a certain configuration of basic propositions) / Realisation of the Textoide as a Pattern" play an important part for the existence of a text. The classification of the expressed propositions into complex sentences, paragraphs and chapters, the mention or not-mention of the sender (transmitter), etc. are problems of pragmatic text arrangement, in other words, of the arrangement of the textual surface structure, which has to ensure, that the reader understands the intended reference and the intended function of the 'communicate' the way, the sender actually expects him to do. The exact semantic and syntactic nature of the signemes and signeme parts in several cases does not appear sufficiently concrete to me. — The propositional rank, on principle, consists of an actant and the attribution of a relator. An actant "stands for an individual, a set of individuals or like" (page 46); a relator "stands for a quality" (page 46) (1st definition). A "Compositional Form" (Rank 4) consists of "one flectional form + signemes of Rank i with 1 < i < 3 that are bound to this flectional form by the attribution of one-place relators" (page 46). (2nd Definition). According to this 2nd definition a statement like Der Flug ist schön ( = 'Die Personen P,_ n teilen den Personen Qi_ n mit, daß P,_ n schön fliegen') very probably is a signeme of rank 4. But that cannot semantically be possible, for, although Flug is a 'flectional form', 'Flug' is not an 'individual', but a 'quality'. — The signemes of rank 8 are defined as, e. g.: "a pair consisting of one speech act specified propositional form and the corresponding speech act reaction" (page 48); the signemes of the subsequent rank are defined as: "a chain of signemes Σ 8 or of signemes Σ 7 and Σ 8 each non-first member of which refers in some way — normally by consent or dissent — to the assertion of a preceding member." (page 48) What exactly is the difference between a speech act reaction and a consent / dissent? — The possible world (that of a certain religion, e. g.) to which the "intensional specifications" (page 50) of a presuppositional group relate, can be, in my opinion, mentioned by certain lexemes (which may be related to that certain religion, e. g.). How would the semantic function of such a repetition by seized signemically? Even more generally formulated: How can connotations and evocations of a certain 'ambiente' be seized by signemes?

104

Μ. Metzeltin

Heger holds the opinion that the hierarchy of signemes is not limited at the top, in other words, that one can always find more complicated types of signemes. It is correct, that one can relate a certain text to other texts of the same author or the same genre. Heger's model shows, in a very concrete way, that the unit 'sentence' is an unsuitable object to serve as a basis for the delimitation of linguistics versus literary criticism. Does the analysis of texts with often clearly marked beginning and end, none the less, require a certain methodology, — as the arrangement of textual typologies and 'text-stories' demand certain other methodologies? The unlimited signeme hierarchy seems to negate a methodological division between text linguistics and history of literature. As far as the constitution of texts is concerned, the relation between the propositions, — according to Metzeltin and Jaksche (28 — 30): tautologies, accumulations, disjunctions, successions and causalities are meant, — play an important part. Heger's report does not pay much attention to these semantic elements. Are these connections always functors? Which is the monemic and signemic status of a functor? Which rank/ ranks are they related to?

3. The Analysis of the Novel In paragraph 2.1. the referee translates the facts, that — — — — —

the the the the the

author presents a text to the reader, author has given a certain title to the text, author makes it clear, that the text belongs to the genre of the nivolas, text consists of three prologues and the actual novel, actual novel consist of 33 chapters and 1 epilogue,

into signeme notations. The translation appears to be exact. But the meagre notations in words have a much more positive effect on the reader than the notations in symbols. I cannot make out what advantages this signemic description offers for the comprehension of text coherence. That a post-prologue represents an answer to a prologue and that a third prologue serves as an answer to both, — and that the terms 'question' and 'answer' correspond in some way, — ought to be obvious to anyone who is concerned with the disposition of texts. None the less it is unusual to write three prologues for one novel; the 'prologue inflation' would rather endanger the coherence of the text.

Co-Report on "Text Coherence . . . "

105

But the signeme analysis does not show what the specific coherence among the prologues and between the prologues and the novel-text consists in. The analysis remains very vague. As Heger wants to concentrate on chapter XXXI, important donors of coherence as the dog Orfeo, the part of the a^ar and the theme of marriage are left unconsidered.

4. The Analysis of Chapter

XXXI

In paragraph 2.2 Heger concentrates on the examination of chapter XXXI of the novel. He distinguishes four parts: — one first actantial presuppositional group (5 signemes Σ 8 ), — a second actantial presuppositional group (2 signemes Σ 8 ), — a third dialogical presuppositional group (192 signemes Σ 7 ' 8 ) and a remaining actantial presuppositional group (3 signemes Σ 8 ). The referee obviously uses a pure graphic device in order to delimitate these groups: the paragraph's border. But it is neither syntactically nor semantically imperative, that, for instance, the complex sentence "Este supremo esfuerzo de pasion de vida, de ansia de inmortalidad, le de jo extenuado al pobre Augusto." belongs rather to the third than to the fourth group. From the analysis of the dialogical group one can gather (pages 75 — 76), that coherence is based on the following factors: — one common sender (transmitter) for different statements — action and reaction in the area of performance — consent and dissent. These factors certainly play a part as donors of coherence within the paragraphs, but hardly between the chapters. Besides they are not only typical for chapter XXXI, at all. The excessive use of the performative area rather donates coherence as an intentionally applied 'styleme' than a hierarchically binding element. It is not discussed, further on, how complex sentences of the same group cohere. A new term and coherence-donating factor is introduced only at the end of paragraph 2.2: the "thematic links". But these are not defined concretely, — it is only said, that they "can be interpreted as constituting an actantial presuppositional group Σ 9Α " (page 77) and that it can be concluded from the given examples ("alive", "asleep", "existing", 'fictional world', 'world of literature', 'real world', 'world of religion') that they do correspond to the isotopoi of Greimas and the isosemies of Metzeltin and Jaksche in some way. The latter have shown (Metzeltin and Jaksche, 1983: 66 — 69), that the isosemies can be closely related to the textoides and that they are one of the

106

Μ. Metzeltin

most important devices for the construction of text coherence. This whole problematic nature does not get treated thematically by Heger, — maybe, because he merely concentrates on one chapter. Therefore it is hard to judge, if the interpretation of the "designations" (pages 82 — 87) is always adequate to the text. Inmortalidad, e. g., could rather belong to model 14 than to model 15, — suicida does not need another interpretation than that of suicidarse, — estar dormido and estar sonando may have to be distinguished, and so on. In paragraph 2.3. Heger almost casually calls attention to the use of apparent contradictions, which "fit perfectly and coherently into this frame of simultaneous reference to different levels (the various possible worlds of reference are meant)" (page 88). Here coherence apparently means 'consistence'. It is not mentioned, that the contradiction in the novel probably is a styleme and only as such a donor of coherence (see for example the numerous inversions such as ",|No es acaso todo creacion de cada cosa y cada cosa creacion de todo?", chapter VII). 5.

Conclusion

The preceding ideas are only meant as constructive notes and questions. Heger's model is, for several reasons, significant for linguistics in general and text linguistics in particular. It makes it possible to seize a text (author and title included) as a whole; it calls attention to the performative area, without neglecting the propositional nucleus; it treats the problematic nature of the possible 'worlds' thematically. But it would gain a lot, if text semantics was more intensely taken into consideration. (The English translation of this contribution is due to Robert Kluge.) 6.

References

Metzeltin, Μ.; Jaksche, Η. 1983 Textsemantik. Ein Modell %ur Analyse von Texten, Tübingen 1983. Metzeltin, M. 1981 Introdu^äo ä leitura do Romance da Raposa. Ciencia do texto e sua aplicafäo, 1981.

Coimbra

WOLFGANG RAIBLE

Some Comments on the Contribution of Klaus Heger: "Text Coherence in a Dialogical Presuppositional Group: Chapter X X X I of Unamuno's Niebla" Klaus Heger analyzed a novel of the Spanish author Miguel de Unamuno, or, to put it more precisely, he analyzed important parts of this novel. In order to show the background and the general framework of this analysis, I shall try to answer two questions: (1) Why did Heger choose to analyze exactly this novel and not, for instance, a novel by William Faulkner or Joseph Conrad? (2) What are the analytical instruments enabling a linguist to cope with the difficulties of an extremely intricate literary text? I Miguel de Unamuno (1864—1936) was one of those gifted and versatile men of letters who have contributed to the fame of Spanish thought. He held a chair of Greek, later on a chair of history of Spanish. For long years, he was chancellor to the University of Salamanca, but also had to spend six years of his life in exile on the Canaries and in France. He was a great writer of essays, plays, and novels. One of these novels plays a most important part in Spanish literature, and this novel is exactly the novel analyzed by Heger. This novel is not only important for Spanish culture, is is even a landmark in the history of novel. There are at least three reasons for the importance of this text. First, persons existing in novels use to have a fictitious existence, they are fictional beings. For Unamuno, things are different. Already in the first of three prologues to this novel, the reader becomes acquainted with a person named Victor Goti. Goti is not only the author of the prologue — he is also a friend of the novel's protagonist. This is tantamount to saying that Goti, the author of the prologue, is a fictitious person, too. — Second, in the novel itself we frequently have the pleasure to come upon a person named Miguel de Unamuno, the author himself. The third point perhaps the most important: In chapter 31 — the chapter analyzed by Heger — there is a very long and intriguing dialogue between the author and his protagonist Augusto Perez. It is an agön logon in the true sense of the word, i. e. a life-and-death struggle with words, the duellists being the author and one of his fictitious beings.

108

W. Raible

Augusto Perez, the protagonist, intends to commit suicide. He comes to the house of the author because Unamuno the essayist had written a text on suicide, an essay which Perez, the protagonist, happened to read. So he intends to discuss the problem with a competent person. The three points just mentioned make obvious what confers particular interest to this novel. It is the mixing-up of things or levels which used not to be mixed up before. We are confronted with a transgression between different 'worlds'. II It is perhaps not by pure accident that this novel was written by a Spanish author because there is — nevertheless — a certain tradition in the transgression of worlds since the work of a seventeenth-century author — Miguel de Cervantes. (One of the most famous writings of Unamuno is, by the way, the essay "The Life of Don Quijote and Sancho", written in 1905.) In order to make visible the congeniality between the two authors, I would like to recall two points: (1) Don Quijote is the novel of a hero who read too many romances of chivalry and who, having done this, went out of his mind. His craziness results exactly in the fact that he mixes up literature with the real world because he thinks the real world in which he lives is a world of chivalry. (2) The first volume of Don Quijote was extremely successful (and this against the expectations of the author). As life goes in literature, another author tried to take profit from the success of a colleague. So he wrote a continuation, a sequel, of Don Quijote. Now, when, some years after the first volume, Cervantes wrote the (true) second volume of Don Quijote, he confronted his hero with the bewildering fact that there existed already a spurious novel containing the adventures of a false Don Quijote. That is to say, Cervantes confronts his fictitious hero with the real fact that there exists a false fictitious Don Quijote. These two points might be a hint at the existence of a certain Spanish tradition in this line of thinking. Perhaps they make understandable that the most ingenious living author, Jorge Luis Borges, a man who has exploited all or nearly all possibilities of literature, is an author of Spanish tongue, too. This summary consideration of the literary and cultural background shows that Heger has very well chosen the subject of his analysis. This statement holds also for the importance of the novel Niebla outside the world of Spanish literature. Here the discovery of Unamuno's was very successful, too, because what he created or invented, especially in chapter 31 of his novel, is most intriguing for the reader. So Unamuno had many successors out of which I would like to mention only two: one is Luigi Pirandello with his play Sei personaggi in cerca d'autore (1921). The persons of this drama are offsprings of the imagination of Pirandello's calling for life, that is for publication. The

Some Comments on the Contribution of Klaus Heger

109

second author, Andre Gide, four years later wrote his novel Les fauxmonnayeurs. In this text there is not only a person, Edouard, writing a novel with the title Les faux-monnayeurs. There are also several persons commenting on the novel. The recursive procedure of Andre Gide has become known in literature as a procedure called "mise en abime". This literary and cultural background makes out of Heger's contribution the answer of a linguist to the challenge of an intriguing literary text.

Ill My first question was: Why did Heger analyze exactly this novel? Having answered this question I can tackle the next one: What are the instruments which enable the linguist to describe what is intriguing 'normal' readers of Niebla? In order to do this, I do not intend to describe once more the noematic system of Heger's. I did this elsewhere and Heger himself gave an outline of his system at the beginning of his contribution. I would just like to mention the three components which are most important for the problem under discussion. First, Heger's system has hierarchy. That is to say, Heger considers texts as wholes consisting of functional parts having other parts themselves. His system is able to handle the phenomena of level-skipping, rank-shift, and mutual combination and embedding of parts of the text. Second, another important component of Heger's system is the notion of 'communicative responsibility' which is linked to the notion of assertion. If a speaker asserts a sentence or a larger part of the text, he takes over the communicative responsibility with respect to the hearer. By doing this, he creates the possibility to be contradicted. If a speaker does not assert what he says, the assertion can, may, or has to be delivered subsequently by the speaker himself, or by the hearer. In this way, Heger is able to cope with sequences like question/answer, order/fulfillment, promise/keeping of the promise, and so on. The possibility to gainsay is an important characteristic of the verbal part of our communication. We cannot, for instance, contradict the non-verbal communication (laughing, crying, and the like) of our partners. A third component facilitating Heger's task is the distinction between different levels of communication and between different worlds linked to these levels. As regards these levels and the corresponding worlds, we have the following situation: — On the first level of communication, the author addresses himself to the reader, author and reader being entities of the real world. What is communicated is the novel Niebla. This novel is a 'nivola', that is to say a novela (novel) which is nebulous. — On the second level of communication, there is a person of the novel, Victor Goti, who addresses himself to the reader, too. What is commu-

110

W. Raible

nicated is the first prologue and this first prologue is part of the world of literature. Goti, belonging to the world of fiction, behaves as if he were a person belonging to the real world. — On the third level of communication, the author addresses himself to the reader. He communicates the text of the novel Niebla, this novel belonging to the world of fiction. — A fourth level of communication is formed by the persons of the novel itself. On this level, we have to place the long dialogue in chapter 31 in which the author, belonging to the world of fiction this time, addresses himself to his hero, and vice versa. The following scheme is thought to visualize the embedding of the various levels of communication: Level,

author

reader "Niebla"

real world

'nivola' reader

Prologue by G o t i author

reader Niebla

Dialogues in Niebla

IV In order to outline what Unamuno does with the different levels of communication, I shall introduce some supplementary considerations. Let us first look at the relation which holds between the author and his work (a relation which, to a certain extent, is similar to the relation holding between God and the world created by God). The author knows all about the fictitious world he has created; the persons created by the author do not know anything about their creator. Like God, the author can bring persons to life and to death, he is master of his fictitious world. This relation between the author and his work is visualized in the following scheme. L, means 'first level', L 2 'second level'. The double arrow shows who knows about whom. Somewhat different is the relation holding between the dreaming person and the dreamed world. In contradistinction to the author, the dreamer is not conscious of the fact that the dreamed world is nothing but a dream. This is shown in the following scheme. The broken double arrow indicates that the dreamer is, to a certain extent, the 'author' of the dream, but is not conscious of the fact that he is dreaming.

Some Comments on the Contribution of Klaus Heger

111

Things become different when the dreamer is awake. At this moment he realizes that he was dreaming. This is to say that the former dreamer, who by now is awake, knows that, when he was asleep, he was dreaming. In this way the person who was dreaming has to split himself (herself) up into two persons: the dreamer, who is awake and knows that, when he was asleep, he had a dream. This in shown in the following scheme:

Scheme 2 b

Now the model of the dreamer applies also, in a certain sense, to the author. In scheme la, we were confronted with an author creating a work of literature.

112

W. Raible

When this author reads what he wrote, he is on level 0 (L0). On this level, the author can comment on what was written by himself on L,. This is the distinction between the author (level 0) and the narrator (level 1), introduced into literature by Miguel de Cervantes in his Don Quijote. These relations are shown in the following scheme 2a:

Authors of literary works (and not only authors) are always exposed to a certain danger. It may happen to them that they undergo the fate of the magician's apprentice. That is to say that the persons they created become virulent and obsessive; they leave the realm of level 2 and intrude into level 1, the level of the author. This is shown by the simple broken arrow in scheme 2a. Pirandello has described exactly this escaping of persons from level 2 into level 1. In this situation, it is obligatory that the author construct a level 0. If he did not so, we would have to suppose that he went out of his mind, like Don Quijote who mixes up the real world with the world of chivalry. But there is still another danger: As long as I keep my dreams in my mind, as long as I enclose the persons I create in my own imagination, dreams and persons may merely be obsessive. But as soon as we publish what we have dreamed of or what we created, and as soon as others come to know it, things become different. Works of fiction may be so successful that their protagonists become part of the general knowledge of a society or culture. In a certain sense, they become alive, they may even outlive their authors, like Ulysses, Don Quijote, or Sherlock Holmes. Readers may even compel the author to publish more about the life of fictitious persons.

Some Comments on the Contribution of Klaus Heger

113

So the author of a literary work is exposed to a twofold danger: (1) Persons may evade f r o m his imagination and intrude into the world of the author (L,); (2) Persons may escape f r o m the world of fiction and become part of the world of the readers (also Lj). So we come to the following scheme:

world of the conscious reading/commenting author = author of a nivola

world of the author and of the persons who have read his work

Scheme 3 a

T h e simple broken arrow means: Persons escape f r o m the imagination of the author and make themselves at home in their author's world. The simple arrow means: Persons of literature become part of the world of readers. Unamuno's novel describes this twofold danger: When persons f r o m level L 2 escape into level Lj, i. e.into the world of the author and, later on, perhaps also of his readers, this makes a 'nivola' out of level Lj. T h e author on level zero, describing and observing levels L t and L 2 , is the author of a nivola.

V The schemes outlined in section IV are not useless and superfluous inventions of the commentator, they are rather presupposed by the dialogue in chapter 31 of Unamuno's Niebla. In fact, Augusto Perez is one of those persons w h o escaped f r o m level L 2 into level L,, i. e.into the world of the author, and this splits the author into t w o persons, one on level L 0 , w h o is aware of the evasion, and one on level L, speaking and acting with his hero. But not only

114

W. Raible

can the entire situation be mapped on scheme 3a. We can see as well that the antagonists use some of the other schemes in their argumentation. (I shall use the segmentation of Heger in order to localize the parts of the text I refer to.) At the beginning, the author, on level L 0 , tells us ( K 2 : 1 :4) that Augusto Perez escaped from level L 2 : he read an essay in which Unamuno treated the subject of suicide. He comes to his author and wants to discuss his plan before carrying it out. When Perez begins to speak about the works of Unamuno he read, the author (on level L,) points out to him scheme la: As the godlike author he knows all about Augusto Perez — who is frightened to death (K2 : 3 : 2 and 3). He thinks he is in the state outlined in scheme l b (1 : 7e: 7). But the author tells him that he is neither awake nor dreaming (because he belongs to level L 2 of scheme 2a (2: 7e: 8)). Then he terrifies Perez once more with his author's knowledge: he even knows that Perez came to him because he intends to commit suicide. What follows is a long dialogue in which the author tells Augusto that he cannot commit suicide because he is only a fictional being, entities of fiction being neither alive nor dead. Augusto Perez is nothing but a person in a novel — or in a "nivola" (19 : 7e : 32 — "tu no eres mis que un personaje de no vela, ο de nivola, ο como quieras llamarle"). That is to say that the author is based on scheme 2a. But now, Augusto Perez starts a counter attack exploiting the possibilities of scheme 3a. Persons of fiction can be much more alive than real persons. Perez refers to Don Quijote and Sancho (26: 7e: 43). Even Unamuno knows this: "?No ha sido usted el que no una sino varias veces ha dicho que Don Quijote y Sancho son no ya tan reales, sino mas reales que Cervantes?" So he, Perez, has an existence of his own (30: 7u: 49). He refers to the inner logic of literary beings and states that according to this logic he has to kill himself. He argues so well with the Unamuno of level L, that Godfather Unamuno, the author on level L 0 , becomes angry and decides to kill Augusto himself. As an author in the sense of scheme la he is in a position to do this because the novel is still unpublished. This means that Augusto Perez, at the moment of the dialogue, is only part of L, of the author, but not yet part of L, of the readers. Only an existence as part of L t (or even L 0 ) of the readers would have given him the chance of immediate immortality. So Unamuno's threat means a decisive change of the situation. Augusto Perez does not like at all to be killed by his author, he wants to live — but the author, by virtue of scheme la, is stronger.

VI To conclude my comments on the contribution of Klaus Heger, I would like to call attention to the following points.

Some Comments on the Contribution of Klaus Heger

115

1. When we use the distinction, established by Petöfi, between connexity, cohesion, and coherence, the description given by Heger does not concern coherence, but rather the cohesion and connexity of the text under discussion. Especially in sections 2.2.3 and 2.2.4, Heger gives excellent descriptions of the most important cohesive devices playing a part in the text he has analyzed. 2. The essential point in my comments is the splitting of a person, here the splitting of the author, into two or more persons (here: level L] and L0). L 0 is the level of communication corresponding to the initiated, knowing reader. So I think that when Victor Goti, in his prologue, addresses himself to the readers, the intended readers are those of level 0 and not those of level 1. 3. Heger describes the dialogue by marking the changes in the turns of the speakers. Now, when we read the dialogue, we become aware of the fact that there are larger stretches of text forming a certain unit, such a unit being, for instance, the sequence 5 : 7u: 14 to 19 : 8 : 33. How can we know, in the analysis, that this sequence forms a unit? How could we detect a functional sub-unit of this larger unit? How could we analyze this dialogue as an argumentation, for instance in the sense outlined in the Topika of Aristotle? Nevertheless, the description given by Heger is an excellent framework for further interpretation, among others a framework in which we can locate the difficulties and the points crucial for the interpretation. I would like to add such an interpretation, i. e. an attempt to establish coherence in the sense of Petöfi: (1) in chapter 31 we learn that persons committing suicide are people who would have liked to kill someone else, but did not dare to do this (57: 7a: 105 "Los mäs de los suicidas son homicidas frustrados; se matan a si mismos por falta de valor para matar a otros ..."). (2) Unamuno himself thought about suicide. He wrote an essay on this topic. So perhaps he himself intended to commit suicide. (3) Knowing (1), Unamuno makes up his mind to kill someone else instead of killing himself. So he intends to kill ... Augusto Perez. (4) Perez is a part of Unamuno himself. So, while killing Perez, Unamuno commits homicide and suicide at the same time. Now, since this homicide and this suicide take place in the world of literature, nothing at all has happened in the real world (apart from the fact that a masterpiece of literature was created). Unamuno found a literary solution to a personal problem, putting at the same time his renown for posterity on a firm footing. So all's well that ends bad.

KLAUS HEGER

Some Comments on the Comments by Michael Metzeltin and Wolfgang Raible 0. The first comment I have to make on the comments that my paper Text Coherence in a Dialogical Presuppositonal Group: Chapter XXXI of Unamuno's "Niebla" has been judged worth while, is the expression of my gratitude towards the two commentators for having taken the trouble of thoroughly analyzing the rather condensed text of which I am guilty. In particular, I have to thank Wolfgang Raible for having once again accompanied my attempts with the critical support and decisive suggestions I have now been owing him since more than twenty years; and I have to thank Michael Metzeltin for the effort of entering into discussions which, precisely because of their going on for more than two decades, suffer from the bad habit of taking for granted what is far from being so for everybody. It is only on the background of these introductory remarks that I'll try to answer questions and to comment on comments in the following sections. In quotations, the following abbreviations will be used and accompanied by § numers and/or other indications: Η Μ R R : PT

= = = =

my paper under discussion, Metzeltin's comments on H, Raible's comments on H, and Raible's paper "Phänomenologische Textwissenschaft — Zum Beitrag von Katsuhiko Hatakeyama, Jinos S. Petöfi und Emel Sözer (Text, Konnexität, Kohäsion, Kohärenz)", in: Conte, Maria-Elisabeth (Hg.): Kontinuität und Diskontinuität in Texten und Sachverhaltskonfigurationen. Diskussion über Konnexität, Kohäsion und Kohären Hamburg: Buske.

1. A first type of questions refers to the choice I have made by using Unamuno's "Niebla" as object of my analysis. Raible's comments start by asking "why did Heger choose to analyze exactly this novel" and by giving to this question an answer (R §§ I and II) for which I am all the more grateful since, by insisting on Unamuno's importance in the history of literature, it covers an aspect which had been completely neglected in my paper. At the same time, this highly appreciated supplement to my description allows me to refute the impression of which Metzeltin seems to be afraid

Some Comments on the Comments by M. Metzeltin and W. Raible

117

when he suspects that "the unlimited signeme [rank, Κ. H] hierarchy seems to negate a methodological division between text linguistics and history of literature." (M § 2, p. 104): first of all, the fact that I have completely neglected the perspective of history of literature, does not mean that I consider it as replaced or replaceable by that of text linguistics; on the contrary, I consider it so different from the latter — in the sense of different disciplines differing from one another not in their objects, but in the kind of questions they ask — that, within the limited space of my analysis, I preferred to strictly remain within one of these prespectives rather than to mix them up in a not sufficiently controlled manner. Besides, as for the respective domains of objects and their mutual delimitations, the principle of specificity introduced in Η § 1.1.3, guarantees the possibility of defining exactly the point where, within my hierarchy of ranks of signemes, one enters the domain which, seen from a different point of view, might be identical with that of history of literature: as soon as there is a text-assertion (cf. Η § 1.2.5.2), occurrences of respective signemes Σ 1 0 Τ may be — but do not have necessarily to be — considered objects of history of literature. Another question of Metzeltin's that has to be answered in this context, suggests that it might have been better to choose a more 'normal' text than Unamuno's "Niebla" (Μ § 1, question 2, p. 100). Apart from the fact that the analysis given in my paper has not been the first attempt of testing my hierarchy of ranks of signemes for its usefulness in the analysis of signemes of text ranks (cf. the references given at the end of Η § 0), I think that examples chosen for such tests should not only be simple enough not to introduce too many complications at once, but also complex enough to make the analysis worthwhile. 2. Perhaps the most important question asked by both Metzeltin (Μ § 1, question 1, p. 100) and Raible (R § VI 1), refers to the missing definition of the sense in which I use the term coherence. If coherence has to be defined as a "text-eternal property of text" (R : PT § III), Raible is certainly right in stating that my "description ... does not concern coherence, but rather the cohesion and connexity of the text under discussion" (R § VI 1), since I have not made use of any "text-eternal" criteria at all. If, however, coherence is defined as a property of the domain of denotata to which a text refers and, in case the text under discussion refers to different domains of denotata successively or simultaneously, as a property of the relations that hold between these domains of denotata (cf. R : PT § VI), things seem to look different. Domains of denotata Wi are introduced, in Η § 1.2.5.1, as the constitutive criteria of the definition of the Specified Presuppositional Group (Σ 10δ ) of rank R 10; and four such domains of denotata (or 'possible worlds', in order to answer at the same time the questions raised in Μ § 1, question 4, p. 101, and in Μ § 2, question 3, p. 103) have been distinguished among what Unamuno's "Niebla" is referring to more or less simultaneously: the "real

118

K.Heger

'world' W 0 of the author and his readers" (H § 2.1.1, p. 56), the " 'world' of literary activity" W lt (H § 2.1.2, p. 58), the "fictional 'world' of the novel proper" W fi (H § 2.1.2, p. 58), and — though in an admittedly not very explicit way - "the domain Wri of religion" (H § 2.2.4, p. 88). The first three of these domains of denotata correspond to the four levels distinguished in R § III, p. 110, in the following way: Raible's first level is identical with W„, with the author OE1 and his readers OE1, according to Η § 2.1.1, model 5; Raible's second and third levels both belong to W k and are distinguished by the authors OE^ 1} ( = Victor Goti, according to Η § 2.1.2, model 8) and OEl ( = Unamuno, according to Η § 2.1.3, model 11) who are addressing themselves to the respective readers O E ( U ) and OEl; and Raible's fourth level is identical with W fi with its different senders OEf and receivers OEj. Consequently, what Raible insists upon as "the splitting of a person" (R § VI 2), corresponds to the fact that the reader Ν, 2 appears as OE1 in model 5, as OE^t) in model 8, and as OE^ in model 11 in my paper; and that the author N^i ( = Unamuno) appears not only as OE1 and OE2, respectively, but also — though, in my paper, only implicitly — as OEj throughout the whole of chapter XXXI. Now, considering coherence as a property of these domains of denotata as well as of the relations that hold between them, seems to me quite compatible with a definition that assigns, according to R : PT § VI, connexity to the signifiant, cohesion to the signifie, and coherence to the domain(s) of denotata. In this interpretation, however, my description does concern coherence as much and as well as cohesion and connexity; but it does so in a way that differs in an important detail from more current descriptions of coherence, by giving preference to a paradigmatic approach, on a higher rank Rn + i, to things that habitually are the object of a syntagmatic approach on a lower rank Rn. This paradigmatic approach enables me to set up, besides possible specific scales of coherence, cohesion, and connexity respectively, an additional overall scale of what I may call for a moment coherence in a wider sense and what can be measured according to the rank of the signeme(s) Σ1 that is (are) promoted to Σ 1 0 8 by the addition of a specification according to domains of denotata and to Σ 1 0 Τ by the addition of a text-assertion. The highest point on this scale corresponds to the specified and text-asserted Polyorational Presuppositional Group C — Σ 9 Κ as it is represented in a particularly interesting form by Unamuno's "Niebla", and one of its lowest points corresponds to such specified (according to domains of denotata) and text-asserted ordered sets of asserted pairs Minimal Autosemic Units Σ 2 as a telephone directory. This overall scale of coherence in a wider sense also allows to answer Metzeltin's question "how relevant are the various types of signemes [which I understand as different ranks of signemes, Κ. H] to the constitution of the text" (M § 2, p. 103), or rather to show that an answer to this question necessarily depends on the type of text in question. There certainly are types of texts for which coherence, cohesion, and connexity respectively are a prerequisite on all ranks

Some Comments on the Comments by M. Metzeltin and W. Raible

119

from R1 up to R10S; on the other hand, it certainly would not be very promising to expect any kind of cohesion or connexity on ranks R 3 to R 9 in a telephone directory. 3. Another type of questions refers to sub-units of the text that are not — or not sufficiently — taken account of in my analysis. I certainly do not hesitate to admit a number of simplifications which have turned out to be inevitable, were I not to overcharge my paper. Two of these simplifications have been explicitly mentioned as such in Η § 2.1.3, p. 62; they both concern the object of Metzeltin's remark that "the analysis of the coherence in a textual fragment (e. g. a chapter) ... may hardly lead to the expression of anything essential on the topic of text coherence" (Μ § 1, questions 3, p. 101). This remark is only too true, and the warning it expresses, has therefore been exactly the reason why the title of my paper reads "Text Coherence in [and not: of] a Dialogical Presuppositional Group"; and it has also been the reason why the analysis of chapter XXXI of Unamuno's "Niebla", announced in the rest of the title, is preceded by a detailed — though admittedly still "very vague" (M § 3, last paragraph) — analysis of this chapter's position within the novel as a whole (H § 2.1, p. 54). The same reproach of over-simplification holds even more true in the case of Raible's question (cf. R § VI 3) how to detect, by means of the analysis I have proposed, the sub-units of the Dialogical Presuppositional Group Σ'^ 2 each of which comprises more than one Monological Presuppositional Group C - Σ ^ Ρ » with 1 < j < 93 (cf. Η § 2.2.1, model 13). This way of reformulating R § VI 3, reveals two different questions: first, what are the criteria that allow the analyst to delimit one such sub-unit from the following one? And second, what is, within the framework of my hierarchy of ranks of signemes, the adequate means to account for such sub-units? As for the first of these two questions, model 13 (H § 2.2.1) shows only a tripartition of Σ®^2 by the insertion of signemes Σ^ 3 . between 4 : 7e : 10 and 5 : 7u : 14, and between 48 : 7e : 81 and 49 : 8 : 83. More such insertions could be obtained by re-interpreting as signemes Σ®,.,, the verba dicendi and their more or less extensive specifications in 20 : 7e : 34 and 69 : 7e : 127, thus leading to the following five distinguishable sub-units of the dialogical parts ο£Σ$: (1) from 1 : 8 : 5 to 4 : 7e : 10; (2) from 5 : 7u : 14 to 19 : 8 : 33 (corresponding to Raible's example cf. R § VI 3); (3) from 20 : 7e : 34 to 48 : 7e : 81; (4) from 49 : 8 : 83 to 69 : 7e : 127; and (5) from 70 : 7e : 128 to 93 : 7u : 191.

120

K.Heger

Although this result seems to give a satisfactory answer to Raible's question, it is evident that — it does not solve the problem of other delimitations which one might be interested in drawing, for example between 36 : 7{j : 60 and 36 : 7e : 61, or between 90 : 7u : 173 and 91 : 7" : 174; and — it certainly does not answer the second question with what kind of means the sub-units thus obtained are to be accounted for, since insertions of signemes Σ^ 3 , between two signemes Σ^2·'* have not the slightest influence on Σ'^ 2 being considered as one Dialogical Presuppositional Group, and not as three, or five, or any other number of such Σ', 1 ^. What, however, these insertions and the delimitations suggested by them do hint at, are sub-units that are held together by certain thematic links (cf. Η §§ 2.2.3 and 2.2.4) as seen in the perspectives of the speakers OEf. Whereas the thematic elements that constitute these links have been summarized in Η § 2.2.4, chart 4, and whereas the interplay of 'v', 'd', and Έ ' , on one hand, and the different domains of denotata WQ, W lt , W fi , and W rl , on the other, has at least been vaguely mentioned, I have not given any answer at all to the important question as to what these thematic elements turn out to be in the perspectives of the speakers Ο ΕJ Miguel de Unamuno and Augusto Perez. Only a well founded and explicit answer to this question would have allowed for delimiting sub-units of Σ®^2 according to the moments when a speaker OE? switches from one perspective to another, and for describing these subunits of Σ93ξ,2 as that what they have to be considered to be: as Actantial Presuppositional Groups Σ t h a t have to be placed hierarchically between Σ ^ 2 and the respective Monological Presuppositional Groups C — Σ^' 2 · 0 . 4. Among the points raised by Metzelin and not yet taken into consideration, there are some questions which are fairly easy to answer: (1) I fully agree with Metzeltin's remark at the end of Μ § 1 that "it would be useful to discuss how significant dialogues and possible 'world' are to text coherence": as for the 'possible worlds', I think I have sufficiently insisted on the importance of the domains of denotata as which they are introduced in the definition of the Specified Presuppositional Group Σ 1 0 δ (cf. above section 2); and dialogues have been precisely the object of a contribution that pretends to analyze coherence in a Dialogical Presuppositional Group. (2) Question 2 in Μ § 2, p. 103, asks "What exactly is the difference between a speech act reaction and a consent/dissent?"; it is answered, on one hand, by the definitions given for — speech acts as different forms of explicitly insisting on the fact that an assertion is missing and needs being supplemented, and

Some Comments on the Comments by M. Metzeltin and W. Raible

121

— speech act reactions as the respective forms of supplementing (or refusing to supplement) the missing assertion, and, on the other hand, by the interpretation of consent (dissent) as forms of agreeing (disagreeing) with an assertion that has been given previously. (3) Question 3 in Μ § 2, p. 103, asks how the repetition of lexical elements can be adequately represented; in the given context I suppose the appropriate answer is the possibility of making them figure in charts like chart 2 (H § 2.1.3) or chart 4 (H § 2.2.4) which serve to describe Actantial Presuppositional Groups Σ 9 Α (cf. Η § 1.2.4.2). (4) Although delimitations of paragraphs are considered as 'purely graphical' criteria by Metzeltin (Μ § 4, p. 105), I don't see any reason not to take them as seriously as any other information given by a text (or the edition of a text): in a written text, any information is due to 'purely graphical' criteria. Other points are apparently results of misunderstandings. This seems to me to be the case with formulations such as "monemic surface description" (M § 2, p. 102), "seized signemically" (M § 2, p. 103), "the monemic and signemic status of a functor" (M § 2, p. 104), and "signeme notations" (M § 3, p. 104), which I find difficult to make any sense of on the basis of the definitions according to which I am used to understanding the terms moneme, signeme, and functor, it must be the case with the re-ordering of the actant models I have used in Η §§ 2.2.4.1 to 2.2.4.3, which is proposed in Μ § 4, p. 105, but which would be incompatible with any definition I have ever given of actant models; and it is certainly the case with the classification of the example "Der Flug ist schön." (Μ § 2, question 1, p. 103) which, in this isolated form as well as in any other context, can only be understood as being asserted and therefore corresponds to a signeme of rank R 8 and not of rank R4.

LITA LUNDQUIST

Coherence in Scientific Texts Abstract Texts being real, actualized manifestations of natural language in social contexts, this article gathers some concepts from the domain of realistic semantics, in order to point out how divergences in the coherence structures in two texts, dealing with the same subject domain, result from the general principle of the division of linguistic labour. 1.

Introduction

Text theories about text coherence have lately been characterized by much abstract model building, and little concrete text analysis. This is quite understandable when one considers the latest theories about text coherence which all point to the creation of coherence as the result of a subjective, knowledgebased interpretation. It is this subjective element that prevents a strict, rigid objective analysis of how coherence in natural language texts comes into being in an actualized reading process, and this is probably the reason why text researchers prefer abstract theories about coherence. In this article, I will venture upon the hazardous task of undertaking a concrete analysis of coherence, as it can be interpreted in the reading of two texts, both treating the scientific domain of the natural kind term "blood". I shall not propose an objective, intersubjectively verifiable model of analysis, but content myself with a rather intuitive analysis, as I think there is a need in text theories for results gained by analyses of real texts actualized in natural language and produced in social contexts. I shall first make general remarks about text types dealing with topics from natural sciences and the semantic particularities of such scientific texts, then continue with an analysis of the coherence evident in the text forms of the two text-samples, and conclude with a comparison of the coherence structures of the two texts. The ultimate aim of the analysis and comparison of the two texts, which although they deal with the same object domain, are aimed at two different receiver-groups — the one a group of experts, the other a group of non-

Coherence in Scientific Texts

123

experts — is to show how the context-dependent factor coherence manifests itself in the form of the two texts, as well in their lexicon as in the syntactic ordering of their textual linearisation. 1.1.

Coherence

Before describing and analyzing coherence in scientific texts, some recent theories about coherence in texts in general should be mentioned. First, some of the latest works on text theories present coherence not as a typically linguistic problem, nor as a typically textual problem, but as a general principle in the interpretation of all human activities, and therefore automatically applying to the interpretation of texts: Whenever a discourse presents itself materially as forming an entity, the receiver automatically supposes it to be coherent (Charolles, 1983: 77).

Secondly, text coherence is regarded as a property not inherent in texts, but assigned to them by the interpreter. Petöfi, who considers coherence analogously to textuality, states: I do not consider textuality (coherence) as an inherent property of a verbal object, I believe rather that it is a property assigned (or not assigned) to a verbal object, in whatever form, in a special context by an interpreter (Petöfi, 1983: 266).

And thirdly, coherence is seen as a function of the background knowledge of the interpreter: the interpreter must use all his knowledge/beliefs/assumptions, which he thinks can be related to the text to be interpreted, in order to be able to interpret it (Hatakeyama et al., 1984: 19).

This background knowledge constitutes the cognitive universe of the interpreter, his "mental models", which are essential in the interpretation process, as they mediate between the text-world fragments on the one side and the real-world fragments on the other. By his mental models the reader relates the textworld fragments to the extralinguistic correlates (see Hatakeyama et al., 1984) as shown in Figure 1, (where A J

: epistemological knowledge Γ ° °

In order to compare and shed new light on the different coherence establishing mechanisms of the two texts, I shall, in the conclusion, introduce a new perspective, viz. the procedural perspective of text understanding. 4.1 Coherence of scientific texts in a procedural

perspective

The procedural analysis of discourse (see Eikmeyer 1983) is a natural extension of the realistic, diachronic and dynamic semantics of Kripke and Putnam, as it is, itself, dynamic and realistic; it is dynamic, because it strives to model, by theoretical constructs, the decisions taken in the process of interpretation, and it is realistic, because it represents the processes of production and understanding of natural language by real speakers by adding the "naturalness conditions". The first of these natural assumptions is the subjectivity assumption which, implying "that discourses do not have meanings, but that speakers assign meanings to discourses" (Eikmeyer, 1983: 32) agrees with recent text theories (see our pp. 123 — 124).

Coherence in Scientific Texts

145

The theoretical constructs used to describe the procedures that are followed in the decisions of the interpretation process are borrowed from programming language. The process of understanding, which is equivalent to the progressive establishment of coherence, is seen in its unidirectional temporality. Here instances of interpretation are frozen in order to delimit the linguistic data that function as input for the incremental process of interpretation to take place: 1

>

instance 1

instance 1 + 1

inputs

inputb Figure 10

The procedure is described by means of a conditional: if a then b else c Figure 11

"a stands for conditions which have to be fulfilled in order to take decision b. If a is not fulfilled, the alternative c is pursued ... the a-part of a conditional may be called the structural knowledge involved in a decision, the b- and cparts are procedural knowledge". (Eikmeyer, 1983: 26) The structural knowledge involved in the procedures permits us to segment, delimit, identify and categorize the elements of the linguistic input, whereas the procedural knowledge makes it possible to predict what comes next. What is important to underline here, as a textual perspective on coherence creation, is, on the one hand, the process of identification (which includes segmentation, delimitation and categorization), and on the other, the process of prediction. On the morphological-syntactical level of the sentence, structural and procedural knowledge is fairly easy to seize and describe (see Beaugrande 1980, ch. 2 on "Sequential Connectivity"). Passing from sentence level to text level, the structural knowledge of text structures, such as knowledge of thematic structures stated p. 142 may function as rules which govern and control the processes of textual identifications and predictions. In Text 1 (Le Monde), a structural knowledge of the thematic progression in a text helps first to delimit and categorize an element as a rheme, and next to predict, with a certain probability, and to identify the (sub)theme in the following example:

146

L. Lundquist if

a text begins with "Le sang ... forme de deux

then

sous-ensembles",

the subthemes to reappear will be the two

"sous-ensembles".

Figure 12

A structural knowledge of logical-temporal signalling of text-ordering helps first to delimit and categorize "D'abord" as a connector, and then to predict and identify following items: if

"D'abord

then

will be mentioned later the other "sous-ensemble"

identification:

un sous-ensemble"

"Le second sous-ensemble est liquide". Figure 13

Because of this thematic structuring and signalling of logical coherence between text elements, and because of the syntactic linearization and proximity of identifiable elements: Example:

("Les hematies ou t globules rouges Τ dont le role ... est de Τ se charger.)

predictions are easily made and immediately fulfilled by clear identifications. In the second text, the expert text from the biological review, no clear thematic structuring, nor any explicitly marked logical coherence, helps the non-expert reader to classify, to form predictions and to locate identifications. The identification of items in this text relies, from the very start, on the construction of an appropriate semantic model in the mind of the reader, in which can be placed progressively the objects mentioned in the text. Only by an appropriate frame activation will the reader be able to predict the semantic-thematic structure of the text, to identify subsequent elements, and to perform the default-assignment (see Minsky, 1980: 11), necessary to create a coherent interpretation. The procedures to be executed in the interpretation of this text thus involve data of a more global conceptual nature, so that a decision process may have the form:

Coherence in Scientific Texts if

then

147

"l'aprotinine est un inhibiteur ... utilise en clinique dans les syndromes fibrinolytiques" we will probably hear something about "les effects inhibiteurs de l'aprotinine" in other circumstances. (This then should of course be specified in the appropriate scientific terms.) Figure 14

Furthermore, for the interpretation, and the establishing of coherence to be successful, global rules about language use, contexts and text types, i. e. rules of a more pragmatic and in this case, epistemological nature, are needed in order to classify, create the right predictions and permit the correct identifications. The input and the epistemic procedures could be formulated as follows: if

in a scientific text, one of the initial statements goes: "Experimentalement, (l'aprotinine) agit favorablement sur r e v o l u t i o n du choc h e m o r ragique chez le rat."

then

p r o o f s and/or c o u n t e r p r o o f s will be put f o r w a r d . Figure 15

if

the text is a report f r o m natural sciences

then

p r o o f s and/or c o u n t e r p r o o f s will be based o n the reliability of experiments, observations, etc. Figure 16

Only if the global predictions are created can the links of the following items to the preceding ones be identified. The conclusions to be drawn from this procedural perspective on the coherence establishing mechanisms in scientific texts are, first, that the textual form of Text 1 satisfies naturalness conditions (see Eikmeyer, 1983: 31—36) to a larger extent than Text 2; the natural assumptions about the limited resources of the memory capacity of a natural interpreter, of his understanding segmentwise, immediately and upon the information available are met by the syntactic linearization and logical ordering of text elements. The second conclusion to be drawn is that in Text 1, the creation of coherence is based on the application of local rules, involving knowledge of syntactic, thematic and semantic structures, whereas coherence in Text 2 relies on global rules, such as knowledge of conceptual and epistemological structures.

148

L. Lundquist

This distinction of local interpretation techniques, which "are oriented towards the past" (ib.: 35), in the creation of coherence in Text 1, and global techniques, which "push the interpretation ahead" (ib.: 36) allows for the introduction, on a text level, of the principle of the division of linguistic labor which Putnam proposed on the word-level; (see also Eikmeyer, 1983: 36). It might be claimed that texts reflect this division of linguistic labor in their coherence structures and procedures, coherence in non-expert texts being processed by local rules, coherence in expert-texts by global rules. These rules are thus dependent on different contexts; a successful, coherence-creating textualization of precise expert-terms of criteria-defined concepts into a non-expert context relies on syntactic and thematic rules as well as on semantic rules, whereas, in the expert context, conceptual and epistemological rules are needed to form the relevant predictions necessary for the appropriate identifications. References Bastide, F. 1982

Exploration du "croire" dans le domaine scientifique, Documents de travail et prepublications, Universitä di Urbino. de Beaugrande, R. 1980 Text, Discourse and Process. Toward a Multidisciplinary Science of Texts. London. Charolles, M. 1983 "Coherence as a principle in the interpretation of discourse", Text 3 (1). Danes, F. 1970 "Zur linguistischen Analyse der Textstruktur", Folia Linguistica 4. Eikmeyer, H.-J. 1983 "Procedural Analysis of Discourse", Text 3 (1). Eikmeyer, H. J. and Rieser, H. 1981 "Meanings, Intensions and Stereotypes. A new approach to linguistic semantics", in: H.-J. Eikmeyer and H. Rieser (eds.), Words, Worlds and Contexts, Berlin—Ν. Y.: W. de Gruyter. Gardin, J. C. 1981 "Vers une epistemologie pratique", in: La logique du plausible, essais d'une epistemologie pratique. Greimas, A. J. 1983 a "Le contrat de Veridiction", in: Du Sens II, Paris. Greimas, A. J. 1983 b "Le savoir et le croire: un seul univers cognitif", in: Du Sens II, Paris. Grize, J. B. 1977 "Logique et discours", in: T. A. van Dijk and J. S. Petöfi (eds.), Grammars and Descriptions, Berlin—Ν. Y.: W. de Gruyter. Hatakeyama, K. et alii. 1984 "Text, Connexity, Cohesion, Coherence", in: E. Sözer (ed.). French version in: Documents de travail et prepublications, Universitä di Urbino. Kripke, S. A. 1972 "Naming and Necessity", in: Davison and Harman (eds.), Semantics of Natural Language.

Coherence in Scientific Texts

149

Lundquist, L. 1980 La Coherence Textuelle, Kebenhavn: Nyt Nordisk Forlag. Minsky, M. 1980 "A Framework for Representing Knowledge", in: Metzing (ed.), Frame Conception and Text Understanding, Berlin—N.Y.: W. de Gruyter. Petöfi, J. S. 1983 "Text, Signification, Models and Correlates", in: G. Rickheit and M. Bock (eds.), Psycholinguistic Studies in Language Processing, Berlin —Ν. Y.: W. de Gruyter. Putnam, H. 1975 "Explanation and reference", in: Mind, Language, and Reality, vol. 2, Cambridge. Putnam, H. 1975 "The meaning of meaning", in: Mind, Language, and Reality, vol. 2, Cambridge. Searle, J. R. 1969 Speech Acts, London. Sözer (ed.) 1985 Text Connexity, Text Coherence, Aspects, Methods, Results, Hamburg: Buske.

KLAUS HÖLKER

Remarks on Form, Coherence, and Knowledge Comment on Lita Lundquist's "Coherence in Scientific Texts" "Blut ist ein ganz besonderer Saft." Goethe, Faust I, Studierzimmer"

0. In her paper "Coherence in Scientific Texts", Lita Lundquist, henceforth L. L., analyses the beginnings of two medical articles in French on blood (T (1983); Τ (1984)). Τ (1983) was written for experts, Τ (1984) for non-experts. Her aim is to show "how the context-dependent factor coherence manifests itself in the form of the two texts, as well in their lexicon as in the syntactic ordering of their textual linearisation" (p. 123). Context-dependence here means dependence on the reader's knowledge, whether it is an expert knowledge or a non-expert knowledge. This is an interesting enterprise, because very little serious empirical research work on this subject has been done. To my mind, research having the aim defined above, demands the following: [1] some clarification of the central terms involved, in particular of those terms whose uses are still unsettled: coherence, knowledge, expert, scientific text etc. [2] hypotheses concerning the relations between coherence and knowledge [3] hypotheses concerning the relations between coherence and formal, lexical, syntactic etc. structures of texts [4] empirical tests of what follows from the hypotheses in [2] and [3]. Needless to say that nobody can deal with all of the aspects mentioned in a single paper. L. L.'s contribution to such an enterprise comprises the following: concerning [1] she deals with the concepts of coherence, knowledge, an expert and a scientific text (chs. 1 & 2); concerning [2] she presents the basic ideas on word semantics developed in Kripke (1972) and Putnam (1975) (cf. ch. 2). If I understand L. L. correctly, she considers these reflexions as essential for the clarification of the relations between coherence and knowledge; concerning [3] one finds reflexions on the coherence of scientific texts from a procedural perspective in the concluding chapter (ch. 4), and applications of her theoretical considerations to the analysis of two texts can finally be

Remarks on Form, Coherence, and Knowledge

151

found in chapter 3, where she investigates the thematic and syntactic structures, the roles of the expert, semi-expert and non-expert terms, and the pragmatic characteristics of the two texts Τ (1983) and Τ (1984) already mentioned. 1. As can be seen from these preliminary remarks, I perfectly agree with L. L.'s general way of dealing with the relations between form, coherence, and knowledge, but I disagree on a number of details, some of which I shall mention in what follows. But before doing so, I should like to warn the reader: it may well be that I have missed the gist of L. L.'s paper. In that case I beg L. L.'s and the reader's pardon. 1.1.

First of all the concept of coherence, whose clarification is necessary for achieving the goals that L. L. has set herself: it is not at all clear to me how she uses the term coherence. Let me present a list of the contexts in which she uses this term: [1] Coherence may manifest itself in the form of texts (cf. p. 123) [2] "... coherence ... as a general principle in the interpretation of all human activities, ..." (p. 123) [3] "... coherence ... as a property not inherent in texts, but assigned to them by the interpreter." (p. 123) [4] "... coherence ... as a function of the background knowledge of the interpreter ..." (p. 123) [5] "The establishing of coherence, which is a condition for a successful interpretation, may be regarded as an adaptation of new knowledge (i. e. new relations) to established knowledge, as an act of relating new ideas to old." (p. 124) (Where is the difference then between establishing coherence and learning?) [6] "... instrumental experiments and ... placing of the newly discovered phenomenon in a theoretical framework ... are decisive parameters in the kind of knowledge which is the indispensable background for the establishment of coherence in scientific texts." (p. 126) [7] Coherence preexists as a cognitive structure (cf. p. 133) [8] There are two types of coherence: coherence for an expert and coherence for a non-expert (cf. p. 134) [9] The three fundamental speech acts (reference act, act of predication, illocutionary act) "generate" coherence structures, (cf. p. 135) [10] There are three types of coherence structures: thematic structures, semantic structures, and pragmatic structures (cf. p. 135).

152

Κ. Hölker

[11] Coherence structures may be (partly) signalled by the surface structures of texts (cf. pp. 135, 140). [12] Coherence may be established along the main thematic chains (based on referential, syntactic, and semantic features, e. g. semantic isotopies) (cf. pp. 135, 137). [13] The contrast in the semantic load of two expressions may convey coherence (cf. p. 140). [14] "... knowledge of the different accrediting systems such as techniques, tools, instruments, observations, measurements etc. is required in order to grasp the coherence in the argumentation structure of the text [T 1983]..." (p. 143) [15] "... what conveys coherence seems to be background knowledge and not linguistic form." (p.143) [16] "Lack of adequate conceptual framing and absence of epistemological knowledge made the text incoherent and precluded interpretation." (p. 143) [17] "... linguistic form is an essential factor in the creating of coherence." (p. 144) [18] Background knowledge and "textual competence" are dominant factors in the creation of coherence (cf. p. 144). [19] "The process of understanding ... is equivalent to the progressive establishment of coherence ..." (p. 145) [20] "... for the interpretation and the establishing of coherence to be successful, global "rules" about language use, contexts, and text types ... are needed ..." (p. 147) [21] The creation of coherence may be based on the application of local rules (rules involving knowledge of syntactic, thematic, and semantic structures) or global rules (rules involving knowledge of conceptual and argumentative or persuasive structures) (cf. pp. 147, 145). [22] Coherence structures reflect the division of linguistic labour in an extended Putnamian (1975) sense (cf. p. 148). As I have already said, these and other contexts tell me very little about L. L.'s use of the term coherence. To say it in terms of a slightly revised KripkePutnam-semantics (cf. Hölker (1977)): L. L. does not clearly state which criteria are most central for the identification of coherent texts. I suppose that coherence is used here in a very broad sense, comparable to what Hatakeyama et al. (1985) call continuity. To my mind, it would be much more useful to draw nice distinctions within such a broadly conceived concept as coherence is, something L. L. in fact does in nuce when she distinguishes between coherence based on local rules and coherence based on global rules (cf. pp. 147-148).

Remarks on Form, Coherence, and Knowledge

153

1.2.

This lack of (pertinent) distinctions within the concept of coherence contrasts with a plethora of concepts concerning knowledge. L. L. distinguishes between: expert knowledge, non-expert knowledge (cf. p. 126), knowledge a priori, knowledge a posteriori (cf. p. 127), general knowledge, specific knowledge (cf. p. 134), knowledge of the accrediting systems specific to natural sciences (cf. p. 142), generalized knowledge of textual structures (cf. p. 144), structural knowledge, procedural knowledge (cf. p. 145) etc. These terms seem to me to be sufficiently clear for L. L.'s purposes, whereas the uses she makes of terms referring to knowledge in general and the organization of knowledge are less perspicuous: "background knowledge", "cognitive universe", "mental models", "cognitive frame work", "ideas" (p. 123), "structured universe of the interpreter" (p. 125), "historical or causal chain which links the sufficient a priori properties of a term to the necessary essential a posteriori properties of the same term" (p. 128), "instantiation of background knowledge" (p. 144). 2.

Because of the terminological difficulties mentioned, it is not quite easy for me to understand L. L.'s statements on the relation between coherence and knowledge. Nevertheless, it is clear that L. L. thinks that it is necessary to dispose of a realistic (diachronic, social, dynamic) semantics in order to clarify the relations between form, coherence, and knowledge, for she writes: In order to describe how the surface form of a scientific text yields coherence, we need a semantic model that is realistic since it is to describe how newly discovered scientific knowledge (expert knowledge) is encoded in the lexicon of a natural language as compared to nonscientific (non-expert) knowledge, (p. 126)

This is a very bold statement. L. L. should have shown that other theories of meaning cannot account for these relations. Moreover, is it in fact true that there is a relation between the surface form coherence of scientific texts and the (way of) encoding of newly discovered scientific knowledge in the lexicon of a natural language? In her analyses of the two texts on blood, L. L. tries to show that a realistic Kripke-Putnam semantics is needed to analyse the semantic/conceptual relations existing between the expert terms in Τ (1983) (cf. pp. 140 — 141). As these relations reveal coherence, the relation between coherence and knowledge and the mediation between them by means of a stereotype semantics become at once apparent. It is even trivially direct, because lexical knowledge reappears as part of a text competence seen as an instantiation of background knowledge (cf. pp. 143 — 144). The same is true of the relation between coherence and the knowledge she calls "epistemological" knowledge (cf. pp. 143-144).

154

Κ. Hölker

L. L. says of these two types of coherence (based on global rules) that they can be grasped by an expert only. The non-expert can grasp the coherence of a scientific text only at a more superficial level. She therefore proposes to apply the principle of the division of linguistic labour also to texts, not only to words as Putnam (1975) did (cf. pp. 147-148). I think that this proposal is rather unmotivated. What is the meaning of "division of linguistic labour" as applied to texts? As far as I understand the principle of the division of linguistic labour, it says that there are at least two groups in a linguistic community that use the same expressions, though some of them in different ways, and that one of the groups (e. g. non-experts) relies on the other group's (e. g. experts) use of the expressions used in a different way, because the first group does not know the exact extensions of the expressions. This is for instance true of terms like "gold" or "copper" and their extensions but not of scientific expert texts in so far as associated epistemological knowledge and terms restricted to the expert's use are concerned, because a scientific expert text is not the sort of thing both experts and non-experts use, though in different ways, such that the non-experts rely on the experts' use of the text (or of the terms restricted to experts' use that occur in it), because the non-experts do not have the proper epistemological knowledge (or conceptual knowledge associated with the terms restricted to experts' use). The non-expert can dispense with scientific expert texts, but he cannot dispense with words like "gold" or "copper". To my mind, suggestions like these are due to one of the dogmas textlinguists cannot get rid of, in this case the dogma that texts are hierarchically structured: words — sentences — text and that one can pass from one level of the hierarchy to another without hitting upon any catastrophic changes.

3. As to the relation between form and coherence, L. L. writes: "... coherence manifests itself in the form of the two texts, ... in the syntactic ordering of their textual linearisation." (p. 123), "... the surface form of a scientific text yields coherence, ..." (p. 126) or "... [in Τ (1983)] coherence is established by a surface identification process only, ..." (p. 140). It is not clear to me what L. L. means by "coherence manifests itself', "form" or "surface form". It seems to me that she refers in the quotations to some type of coherence, something she calls later on "coherence based on local rules" (p. 147). Let us call it surface coherence. Surface coherence is described by L. L. in terms of thematic chains/semantic isotopies, i. e. in terms of coreferential relations and paradigmatic and syntagmatic relations between meanings of lexical items, provided that these relations are signalled in the surface structure by morphological (e. g. "l'aprotinine ... II"), lexical (e.g. "oxygene ... oxyder") or microsyntactic means

Remarks on Form, Coherence, and Knowledge

155

(e.g. "tissu complexe") (cf. p. 135). I suppose that it is "for the sake of simplicity" (p. 135) that L. L. does not take into account the relations that exist between lexical items on behalf of the non-expert stereotypes associated with them at that stage of her investigation. The result of the analyses of Τ (1983) and Τ (1984) in terms of surface coherence is that in Τ (1984) there is a structure of blocks, each block symbolizing a short distance linearization of semantically closely related items (cf. p. 137). Each block is integrated into a more complex thematic structure by coordination or subordination (cf. p. 140). Τ (1983), on the other hand, has a surface coherence structure in which the semantic isotopies are merely "accumulated", as L. L. puts it. I. e. there are no blocks; semantically closely related items are rather interspersed in the text. L. L. explains these structural differences by saying that in Τ (1983) "coherence is established by a surface identification process only, and not by integration of the context into a structured cognitive framework." (p. 140). One is probably allowed to infer from this that if one takes into account scientific conceptual knowledge and epistemological knowledge in order to reconstruct the coherence (based on global rules) of Τ (1983), one will get at a deep coherence structure which is about as dense as the surface coherence structure of Τ (1984), because in this case the number of semantic relations between items is greater and there are then short distance linearizations of semantically closely related items, too. I think that though this block structure has less formal correspondence in the surface structure of the text than the block structure in Τ (1984), if one takes into account the morphological, lexical, and microsyntactic means only, the block structure in Τ (1983) is nevertheless matched by the division into paragraphs and their order. It should be noticed that surface coherence, in the sense discussed here, involves two types of coherence: formal surface coherence and thematic/ semantic surface coherence. Using this terminology, L. L.'s results can be stated like this: there is a high degree of thematic/semantic surface coherence in the scientific non-expert text Τ (1984) and a low degree of coherence based on global rules or deep coherence; there is a high degree of deep coherence in the scientific expert text Τ (1983) and a low degree of thematic/semantic surface coherence. This result is not surprising, because deep coherence is defined as coherence based on (conceptual and epistemological) expert knowledge. Secondly, in the scientific non-expert text Τ (1984), there is a high degree of correspondence between the formal surface coherence structure based on morphological, lexical, and microsyntactic means and the thematic/ semantic surface coherence structure; in the scientific expert text Τ (1983), there is only a low degree of correspondence between its formal surface coherence structure based on morphological, lexical, and microsyntactic means and its deep coherence structure. But one might add that in Τ (1983) there is a close correspondence between its formal surface coherence structure based on OTöiro-syntactic means and its deep coherence structure.

156

Κ. Hölker

These differences concerning the relation between form and coherence are largely due to the fact that there is much explaining of terms in Τ (1984) (cf. p. 137). I do not know whether this is typical of scientific non-expert texts. Consequently I do not know whether the results can be generalized. 4. In what follows, I shall try to sketch my own views concerning the relations between form, coherence, and knowledge. The basic idea is to distinguish between several types of continuity (i. e. coherence in a broad sense) and to restrict the use of the term coherence, something that has also been done in Hatakeyama et al. (1985). Let us look at the following passage of a doctor-patient dialogue: D: P: D: P: D: P: D: P: D: P:

Vous voulez vous des verres qui, euh, ... D'accord Tenez, qui se foncent, oui. le voila. Vous reglez 9a Je peux vous faire un cheque? (...?) a la caisse, qui est ä Γ entree, s'il vous plait.

Without any further information about the context, nobody would probably be willing to call this extract coherent, though it contains certain features of continuity. Even if the interpreter knows that the text was produced at an oculist's, the impression of discontinuity would not vanish, because it would still remain unclear why the patient puts a question concerning the mode of paying. Only if the interpreter is an expert of French oculists' consultingrooms, he/she knows that minor sums are paid at once, and then the question is meaningful to him/her, and the impression of discontinuity disappears. I hold the above extract to be a typical case of coherence/incoherence (in a restricted sense), because facts mentioned in the text are at stake here. The linguistic presentation of the facts is unimportant. In order to resolve the discontinuity a non-expert interpreter is confronted with, he/she must have knowledge of certain facts, his/her linguistic knowledge does not help him/ her to resolve the discontinuity. Looking at Τ (1983) and Τ (1984) from this perspective once again, one finds the following: in Τ (1984), the beginning of a non-expert text, the author gives facts about blood in order make the rest of the text coherent for a reader who is not an expert in the field of medicine:

Remarks on Form, Coherence, and Knowledge

157

Le sang est un tissu complexe schematiquement forme de deux sous-ensembles. D'abord un sous-ensemble fait de cellules. ... Le second sous-ensemble est liquide: c'est le plasma, ...

The effect on the form of the whole text is that it gets a binary structure: in the first part, background information about blood is given to the non-expert reader, in the second part, the main information is given. In Τ (1983), one finds the same binary structure: the first part presents background information about experiments with aprotinine already done, and the second part deals with new experiments, the main information. But, of course, no text must be structured like this. The background information might be interspersed in the text. The two texts are different with respect to the knowledge they presuppose. That is why Τ (1983) is incoherent for a non-expert, but it seems that neither coherence nor knowledge have an effect on surface form. One might object that the distinction between knowledge of facts and linguistic knowledge is untenable, and one might even mention Kripke (1972) and Putnam (1975) in order to give support to this objection, for one of their main theses is that the knowledge that assures reference is knowledge of facts. But though this may be true, it is certainly not true that any odd knowledge of facts is used to establish reference. For speakers of, let us say, German, a stereotypical swan is probably white, but it is a fact for at least some of them that there are black swans (in Australia). In order to resolve incoherences in texts the interpreter needs knowledge of facts that is not also reference establishing knowledge. Further explications of concepts concerning continuity can be found in Hatakeyama et al. (1985) and more recently in Petöfi/Sözer (1985). References Hatakeyama, K., Petöfi, J. S., Sözer, Ε. 1985 "Text, Connexity, Cohesion, Coherence", in: E. Sözer (ed.), Text Connexity, Text Coherence, Hamburg: Buske, 36 — 105. Hölker, Κ. 1977 "Über einen Typ von Lexikoneinträgen für gemeinsprachliche Lexika", in: J. S. Petöfi and J. Bredemeier (eds.), Das Lexikon in der Grammatik, die Grammatik im Lexiken, Hamburg: Buske, 91 —105. Kripke, S. A. 1971 "Naming and Necessity", in: D. Davidson and G. Harman (eds.), Semantics of Natural Language, Dordrecht: Reidel, 253 — 355. Petöfi, J. S. and Sözer, E. 1985 "Static and Dynamic Aspects of Text Constitution", in: J. S. Petöfi (ed.), Text and Discourse Constitution, Berlin, New York: de Gruyter, 1988, 440 — 477. Putnam, Η. 1975 "The Meaning of 'Meaning'", in: H. Putnam (ed.), Mind, Language, and Reality, vol. 2, Cambridge: C. U. P., 2 1 5 - 2 7 1 . Τ (1983) "Agregation in vitro de plaquettes de rat par doses cumulees d'ADP", Societe de Biologie, Montpellier. Τ (1984) "De la transfusion totale ä la biotechnologie", Le Monde, 15 —16th April.

ANETTE VILLEMOES

Remarks on "Coherence in Scientific Texts" by Lita Lundquist In her above article Lita Lundquist (LL) ventures an analysis of two concrete texts. The texts chosen for analysis both deal with a subject from natural science, but are intended for two different receiver groups, Text 1 being intended for a group of non-experts (readers of Le Monde) and Text 2 for a group of experts (readers of a medical magazine). Through this change in context, i.e. in the receiver's background knowledge LL intends to show "how the context-dependent factor coherence manifests itself differently in the form of the two texts, as well in their lexicon as in the syntactic ordering of their textual linearisation" (LL p. 123). I shall therefore first comment on A: lexicon in the two texts and B: syntactic ordering of lexical elements, i. e. textual coherence structure of the two texts. These comments will lead to C: a discussion of the concept of scientific texts. I shall finally comment on D: the conclusions drawn by LL. A: Lexicon As stated by LL (p. 137) the lexical items of Text 1 are primarily of a nonexpert nature and to a smaller degree of a semi-expert and expert nature. Text 2 on the contrary is dominated by expert terms and only a few semiexpert and non-expert terms are present (cf. LL p. 140). This difference as to predominance can be related to LL's Figure 3 (p. 129) representing Kripke's ideas of the diachronic evolution of words. If these ideas are "frozen" to give a synchronic picture (cf. Putnam's social semantics LL p. 130 ff) of the linguistic reality of a natural language at the moment of production of the two texts, in this case French at the beginning of 1984, we allow for the ideas of the dynamic semantics (cf. LL p. 132 ff) to be incorporated into the model. A new model thus emerges, containing a static characterization of the use of lexical items combined with a description of the dynamic processes (see below) in which the lexical items may take part.

Remarks on "Coherence in Scientific Texts" by Lita Lundquist

159

Β Non-expert use:

Semi-expert use:

Expert use:

based on general and sufficient knowledge characterized by indexicality, fuzzy sets, stereotypes and vagueness; known to all members of the community

based on specific and essential knowledge gained by scientific discoveries and characterized by specificity; known only to a subset of the community.

continuous change and adaptation of lexical items to social context

scientific knowledge undergoing changes due to discoveries of new distinctive features

CI

C2 new scientific discoveries

Text 1:

Text 2:

> χ 3ο 3σ· sυ u ^ u O' cJ υ„ uυ üja ·3> ü S ϊ" 5«r, 'CS SP ö Jdε vic) cS So .2 ö - ο « £ Όc -Ι)S3 '3u Μ Ο ·- 'S G C £ u « ü Ο u Ο υ S δ s | - vi w C g3 s ε S s ubp g2 (ü "Ο C >· ε« ου 3S -c«υ CL, 3

(Α JJ

cο ο. a Ο ο

οο Ο-, εο υ

3 -σ3 •8 C

no

3 Ο Έο 30

^

oui le train le . la ligne

16 17

s.

\

\

\

—>

F:

oui oui je sais . (vite) non mais c'etait pour avoir une

19

Ε:

idee euh + . ^ .

13

ν

mieux moi

14

F:

15

E:

16

F:

17

E:

enfin on a couche che2 eux lä ce week-end de oui

Noel.parce qu'il y avait du monde a la maison

euh elle oui

prend un bain seule elle s'habille elle fait son lit tout seule elle

18 19

F:

20

E:

oui elle peut se debrouiller \ . oui

elle se debrouille quand meme hein

21

F:

22 23

(cl (claquement de doigt du pere de Ε) + \ \ \ bon il faut que je te quitte Papa (rit) m'appelle E:

24

F:

(rit legerement) alors j't'embrasse

E: A

λ

hein + et puis bonne annee . embrasse tes parents

27 28

d'accord eh bien tu embrasses tout le monde de notre part hein

25 26

+

F:

. .

.

v...

o u i (toi aussi?)

\ .

au revoir

Constitution of Relationships as a Factor in Interactive Coherence

1

F:

Emma

ä bientot

—•

2

E:

et puis tu

tu nous ecris d'accord

3

F:

4

E:

t'ecris courant janvier hein s . hein . allez au revoir

5

F:

hm

)

e

(vite) au revoir ä bientot +

\ . oui

323

NILS ERIK ENKVIST AND MARTINA BJÖRKLUND

Toward a Taxonomy of Structure Shifts* 1.1. Our paper deals with one of the most conspicuous features of impromptu speech, namely the devices used by a speaker who wants to shift from one structure to another. We use the term structure shift to cover all those types of correction or improvement in which a speaker leaves a structure uncompleted and either abandons it altogether or shifts to another structure. 1.2. To identify incomplete structures is not as simple a task as one might think at first blush. For instance in a piece of dialogue such as Where did Susie go? To Paris., to Paris is syntactically a fragment but in its context accepted as complete. In dialogue situations of this kind, a common strategy is to begin with crucial new information and to omit, or possibly give later, such old information as one's conversation partner can readily recover from the context and situation. The crucial-information-first strategy is thus different from the old-information-before-new strategy common for instance in expository prose. In this paper, syntactic fragments whose use apparently owes to the strategy of putting crucial information first, and which to us seemed 'normal' and acceptable in their dialogue context, have not been regarded as incomplete in the sense cited in our definition of structure shift. 1.3. Identifying structure shifts thus presupposes that one can decide which structures should be regarded as incomplete. Another of our assumptions is that when a speaker abandons a structure, consciously or unconsciously, he does so for a purpose. Our stance might be labelled as an instance of 'teleological optimism': when people speak impromptu and are compelled to process discourse in real time, it is important for them to have the means of shifting from one structure to another when need be. A similar telos applies to the various mechanisms of hesitation that often occur as part of patterns * This study was begun, and the basic conceptual frame set up, by Ν. Ε. E., and tested on texts from the London-Lund and Talsyntax corpora, whose use we owe to the kindness of Professors Jan Svartvik, Lund, and Bengt Loman, A b o . The analysis of texts and classification of data were carried out by Μ. B., w h o further improved the taxonomy of structure shifts. The report was written by Ν. Ε. E. and Μ. B. jointly. We owe grateful acknowledgement to the Academy of Finland for a grant enabling Ν. Ε. E. to employ Μ. B. as a half-time Research Assistant in January —June, 1984.

Toward a Taxonomy of Structure Shifts

325

of structure shift. We are therefore looking at hesitation and structure shift, not as instances of regrettable human imperfection but rather as important devices helping people to manage in spite of the strains inherent in real-time processing of discourse. They are worth studying as integral parts of spoken communication and discourse which help speakers to maintain cohesion and coherence and to adapt their text to the requirements of receptors and situations. 1.4. Thanks to a number of studies and collections (e.g., Fromkin 1973 and 1980, Cherubim 1980, Dechert and Raupach 1980, Cutler 1982, and, on pauses, O'Connell and Kowal 1983), linguists are now well aware of the nature and importance of hesitation phenomena and slips of the tongue, pen and hand. One of our principles which differs from those of some earlier studies is our insistence on the importance of reckoning with hesitation as part of the total structure-shift mechanism. On the other hand our purpose here will be limited to surface descriptions of patterns of hesitation and structure shift. We shall not probe deeper into psycholinguistics, for instance by trying to find out under what conditions and circumstances hesitations and structure shifts occur. 2.1. Structure shifts presumably result either from an inability to continue a structure already begun, or from monitoring and rejection of a structure entered into. There is little difficulty in drawing simplistic block diagrams showing the genesis of discourse, for instance in terms of (i) a Situation block presenting participants, their roles and relations, their environment, their distance, available media, and other potentially relevant factors of situational context; (ii) a Strategy and Choice block listing repertoires of communicative purposes; facts, opinions and attitudes for textualization; available text types, styles, politeness levels and the like; illocutionary patterns permitting choice between various, direct or indirect, speech acts; and the like; (iii) a block for Text Patterning, including mechanisms for deciding on the form of textual macropatterns, coherence strategies including those distinguishing between old and new information, cohesion markers, and syntactic text formation; (iv) an Execution block explaining the physical actions involved in the act of speaking or writing, including para- and extralinguistic features; and (v) a Monitoring block. But the drawing of actual flow charts of these complex phenomena inevitably commits the linguist to arrange the various subsystems of choices into a definite, fixed, hierarchic, sequential, or parallel order. Yet many of these processes presumably interact in very complex ways which may even vary from one type of communication and discourse to another. Such flow charts

326

Ν. Ε. Enkvist, Μ. Björklund

are valuable as heuristic aids, compelling the linguist to face a number of specific problems in the ordering and linking of various subsystems of strategies and choices. But they remain little more than pedagogical aids until they can rely on firm psycholinguistic data. 2.2. With Laver (1980), the monitoring component can be modelled as a device comparing two inputs: what the Execution block was supposed to do, and what it actually produced. In monitoring, a speaker matches his output with his intentions. If the two fail to match according to a set of specific criteria of similarity or identity, a speaker can either go on, pretending or assuming that the communication will succeed all the same, or he can interrupt an ongoing structure and produce a structure shift. 2.3. Such a mismatch between intentions and output can arise at many points in the decision systems involved in discourse production. The output can fail to match decisions made in the Strategy and Choice block, Text Patterning block, and Execution block. Any such mismatch can then provoke a revision of decisions made within that particular block, and the revision is exposed on the textual surface through structure shift. The classic controversy between Meringer and Freud about the origin of slips of the tongue was essentially an argument about the level at which discrepancies between plans and output arise. Freud thought that such slips provide clues to subconscious thought, whereas Meringer attributed them to malfunctions of linguistic processes, notably in execution. It may well be that there are slips of both kinds, Freudian and Meringerian. Undoubtedly there are slips or structure shifts that look small on the surface (such as exchanges of single phonemes, of single pronouns of address such as the formal ni and informal du of Swedish, and so forth) but nevertheless reflect major shifts in communicative strategy or in thought. Minor corrections on the textual surface may thus reflect major changes of strategy, and, conversely, extensive corrections may owe to small adjustments of certain text-strategic principles. 2.4. The present paper is devoted to structure shifts rather than to slips. Thus we shall not deal with exchanges of segments and with spoonerisms {the nipper is yarrow for the Ripper is narrow. Butterworth 1981), blends ([grasstli] for grisly or ghastly·. Crompton 1981: 693), it doesn't make any matter (from it doesn't make any difference and it doesn't matter·. Fay 1981: 727), and omissions, malapropisms and haplologies where they were not corrected by the speaker. 3.1. The teleological optimism implying that hesitation and repair phenomena are essential features of impromptu speech and necessary to help speakers overcome the strains of real-time, on-line language processing has been shared by a number of recent investigators.

Toward a Taxonomy of Structure Shifts

327

3.2. Thus Schegloff was interested in the relevance of self-initiated repair for 'syntax-for-conversation' (1979: 262). He noted that repair can substantially affect the form of a sentence, by expanding a noun phrase, changing what was a main clause into a subordinate one, turning a question into an assertion, changing what was to be a wh-question into a yes-no-question, and so forth (ibid. 264, 270). But the components and mechanisms of repair were, as he put it, "orderly in their operation" (ibid. 272). Thus repair within a word was initiated by a cut-off (typically a glottal or other stop) which signalled repair of an element already produced and was therefore post-positioned. Uh and pause commonly occur as initiators of repair outside a word or sound boundary; if they are used after repair has been initiated they can be regarded as prepositioned, forward-looking. 3.3. A taxonomy for hesitation phenomena has been presented, among others, by Hieke (1981). Hieke emphasizes the distinction between prospective and retrospective tactics, that is, the forestalling of problems and the correction of errors actually committed. He thus distinguishes between stalls, consisting of silent pauses, filled pauses, prospective repeats (where a speaker repeats to stall and get time to organize what is to come), and prolongations; and repairs, consisting of false starts and retrospective repeats (which provide cohesion between elements of interrupted constituents, retracing to the beginning of their own constituent boundary and thus helping the hearer to localize the repair) (1981: 154). Repairs can occur in phonology, syntax, and in rhetoric (cohesion, bridging), and syntactic errors are classifiable into substitutions, additions, and restructurings (p. 155). 3.4. Levelt's impressively thorough article (1983) builds on a corpus of spontaneous self-repairs obtained in a laboratory situation where subjects were asked to describe visual patterns consisting of coloured dots connected by arcs. According to Levelt, a common repair type such as Go from left again to, uh ..., from pink again to blue (44) consists of, first, the original utterance containing the trouble spot or 'reparandum' (here, left)·, secondly, the moment of interruption with 'editing', the distance between point of interruption and reparandum being called 'delay' (here: again to)·, the editing phase (uh + hesitation); and the repair. A repair can start directly at the reparandum, or it can retrace to an earlier point: here the span of retracing is from. Levelt illustrates this with a diagram. Levelt classifies repairs into five types depending on what question has provoked the repair: D-repairs replacing a message with a different one (because the question 'do I want to say this now?' was answered in the negative); Α-repairs replacing an expression with a more appropriate one ('Do I want to say it in this way?'), Ε-repairs or error repairs ('am I making an error?'), C-repairs or covert repairs which are exposed through hesitation

328

Ν. Ε. Enkvist, Μ. Björklund

original utterance (OU)

editing phase

repair (R)

#

# G o from left again to

reparandum

delay d = 3

uh ...

editing term (ET)

f r o m pink again to blue

span of retracing s = 1

alteration

(reprinted from Levelt 1983: 45) Figure 1

rather than explicit repair; and R-repairs or the rest category of repairs too confusing for simple classification. Levelt's paper is valuable not only for its clear classification but also for its use of repair data as evidence for processing phenomena. 3.5. Related structure shifts are sometimes collectively labelled as anacolutha, as in Rath 1975. Rath writes about repairs resulting in anacolutha which involve interruptions and shift of structure {Abbruche), improvements (Verbesserungen) where individual words are changed, and repetitions (Wiederholungen) which can exist both in correct constructions and in 'anacolutha'. Among the unrepaired anacolutha are contaminations; so-called Drehsät^e (anacolutha proper or blends of the form a-b-a-c, as in ob da überhaupt die Ehe als das Ursprünglichste und Natürlichste der Welt die Ehe geführt wird, p. 6); and other types of 'anacolutha'. 3.6. Another student of repair who uses the term 'anacoluthon' in a wide sense is Boardman (1977), who studied materials from the London-Lund Corpus to find structures either abandoned completely or abandoned to be reformed. Among other things Boardman studied at what rank and level structures were abandoned (word, phrase, verbal group, nominal group). Boardman's classification was into abandoned structures followed by a parenthesis, which either continued the abandoned structure after the parenthesis, or resumed with a different construction; and abandoned structures not followed by a parenthesis, which fell into three subcategories: those followed by a new construction with a meaning similar to that of the abandoned structure; those followed by the same syntactic structure but a new meaning; and those followed by a construction different in both meaning and structure.

Toward a Taxonomy of Structure Shifts

329

3.7. A stricter definition of anacolutha was adopted by Per Linell (1981). To him an anacoluthon is a structure where a speaker starts out with one construction and changes into another which overlaps with the original. Such structures, also called 'blends' by some English linguists, could be exemplified with a sen gick ju (den här Nynäsexpressen) gar ju alldeles breve här (p. 177) Here the central constituent is shared by both the initial peripheral string and the final peripheral string: we have an overlap between two constructions. Anacolutha can then be subclassified into two categories depending on whether the peripheral strings both contain elements from the same syntactic category or even the same lexemes; or whether the two peripheral parts are syntactically and/or lexically incompatible. 3.8. Such a sampling of some of the relevant studies reveals, among other things, considerable variations in terminology. For instance the term 'anacoluthon' has been used in a very wide sense to cover most if not all kinds of repair, correction and hesitation phenomena (Rath 1975, Boardman 1977), or on the contrary in a very specific sense to refer only to blends with a shared overlap (Linell 1981). The emerging caveat suggests that one should define one's terms and exemplify one's categories with sufficient, concrete detail. In a forthcoming paper, Enkvist suggests the term 'true anacoluthon' for anacolutha in the stricter sense, especially in contexts where confusion might otherwise result. 4.1. Our data come from texts S. 2.11, S. 6.3, S. 6.4, S. 6.6, S. 10.6, S. 10.8, S. 11.3 and S. 12.5 of the London-Lund Corpus of Spoken English. The analysis started out from a classification of all the structure shifts in text S. 6.6. This classification was then modified and refined in the light of data from the other texts. The analysis was performed on the paper output of computer tapes of versions 3 and 4 of the texts, and the numbering of tone units therefore sometimes differs from that in the versions printed in Svartvik & Quirk 1980. 4.2.1. Hesitation Signal is a cover term for all those mechanisms whose effect is to delay the further verbalization of an utterance and thus give the speaker more time for planning, either how to continue a structure already begun, or how to correct or repair a structure regarded as a failure. These mechanisms comprise pause, vocalised hesitation (schwa), repetition, stretching (syllable prolongation), and the use of verbal fillers and hedges (such as you know, I mean, you see, sort o f ) . In the present context only two types of hesitation signals, namely vocalised hesitation and repetition, will be dealt with. Pauses and stretching were awkward to deal with from written transcripts only. And verbal fillers and hedges are multifunctional and should be studied in a more

330

Ν. Ε. Enkvist, Μ. Björklund

complete semantic and pragmatic context (cf. Ostman 1981; Svartvik and Stenström forthcoming). 4.2.2. Vocalised hesitation consisting of schwa with phonetic variations (which in turn may reflect the 'basis of articulation' or a kind of 'centre of gravity' of articulatory movements in the language or dialect in question) seems to be universal across languages as a hesitation signal (cf. Levelt 1983: 74). Schematically, vocalised hesitation could be depicted as follows:

» S. 6.6 39

|---uh and they soon

[a:]

—-» helped me tremendously/**

Figure 2

As we can see, a structure is started and then interrupted at a certain point, whereupon the "uh" follows. After this hesitation signal the structure is continued where it broke off. The speaker may, as it seems, hesitate practically at any point in the generation of a linguistic structure, even at the point of starting a new structure. Further examples: S. 6.6 384—387

when we reached the front in Gradisco/ [3:] I was amazed to find everything so quiet/ S. 6.6 1258 — 1259 there were conflicting [a] opinions amongst them/ S. 6.4 9—11 [3:] grampy/ you were forty-two/ when I was born/ S. 6.4 633 — 635 [a] and [a]/ those who didn't like it/ [a?] didn't drink it/ The above conception of vocalised hesitation corresponds to Schegloff s forward repair (repair with prepositioned initiator, cf. 3.2), Levelt's most minimal form of covert repair (cf. 3.4) and Hieke's stalls in the form of filled pauses (cf. 3.3). 4.2.3. Repetition can be illustrated in the following way:

S. 6.4 1 1 5 - 1 1 6

and I th

I think this also shows something/ Figure 3 ** Diacritics have been removed f r o m our quotations f r o m the texts. Tone unit boundaries are marked by /.

Toward a Taxonomy of Structure Shifts

331

Here a structure is started and then interrupted to be continued only after repetition of the last element(s) pronounced before the interruption. Elements at any point in the generation of a construction and at any level, from the phonological to the phrasal, seem to be 'repeatable'. The same item can be repeated several times. Examples: phonological level S. 6.6 28 — 31 English people in those days/ [nn] knew nothing/ or very little/ about Russia/ S. 6.4 224 — 225 one's school/ [wa] was of prime importance to one/ morphological level S. 2.11 1430 sort of no no nobody knew quite what to do word level S. 6.6 670 — 673 and they went/ without being told/ not even waiting/ to to get a a cup of tea/ S. 2.11 95 — 96 there were two two two mares in the field/ word + phonological level S. 6.4 226 — 228 this is something/ which I think the [m] the modern school/ is not/ phrase level S. 6.4 269 I think I I think I ought to say something S. 2.11 1150 it's very it's very successful/ Repetition as hesitation signal in the above sense corresponds to Levelt's covert repair (cf. 3.4) and Hieke's stalls in the form of prospective repeats (cf. 3.3). Schegloff talks about "marking time" in this connection and seems to mean that this kind of repetition is a form of backward repair; "When a try at a bit of talk is the same as the prior try (...), we may speak of MARKING TIME. This does happen, but regularly the second of these two tries adds an uh, marking a more overt entry into a search and converting the repair type from redoing what preceded (my underlining, Μ. B.) to a forward repair" (Schegloff 1979: 279). 5.1. Structure shift is here the cover term for five types of mechanisms allowing the speaker to shift from one linguistic structure to another in the production of an utterance. The five types of structure shift to be described in the following sections are: correction (5.2.1), improvement (5.2.2), insertion (5.2.3), abandon (5.2.4), and anacoluthon (5.2.5). Section 5.2.6 is a rag-bag category for confused cases. The different categories of structure shift overlap to some extent, which the description will also show. 5.2.1. Correction, which makes it possible for the speaker to correct errors in his speech, can be pictured in the following way:

332

Ν. Ε. Enkvist, Μ. Björklund

S. 6.6 8 8 3 - 8 8 4

1 2 3 4 5

and they would

lie

and they would

place

= = = = =

pre-corrigendum corrigendum post-corrigendum point of interruption editing signal

them/uh/

them 6 7 8 9

= = = =

in side by side/ pre-correction correction post-correction continuation

Figure 4

In this example all possible components of a correction are present. An error (lie = corrigendum) is detected, but the structure is interrupted only after the next word (them — post-corrigendum). Then by a process of retracing the pre-corrigendum (and they will) is repeated, whereafter the correction {place for lie) is made and the post-corrigendum repeated. Now the structure can be continued and possibly completed. All these components are not always present. The process of retracing may amount only to the correction itself (i. e. there is no post-corrigendum and the pre-corrigendum is not repeated, if there is one — the first element pronounced may also be erroneous), and it may consist of repetition of pre-corrigendum + correction (in cases with no post-corrigendum), or of correction -+- repetition of post-corrigendum (i.e. the pre-corrigendum is not repeated, if there is one). Other possible variants of the example above would then be: and they would lie place them in side by side and they would lie they would place them in side by side and they would lie them place them in side by side Corrections occur at different levels: phonological, morphological, lexical and syntactic. Examples: phonological level S. 6.6 595 — 596 those who were retreating/ from the [fant] front line/ S. 6.6 968 — 969 they couldn't have [prandau] they couldn't have pronounced a wors((t)) verdict/ morphological level S. 6.6 117 — 118 I had [nju:] known that beautiful country home/ lexical level S. 2.11 798 oh I've [si:] I've read that S. 10.6 1067 — 1069 and Prince Grace/ of Monaco/ Princess Grace of Monaco/

Toward a Taxonomy of Structure Shifts

syntactic level S. 6.6 213 S. 2.11 114—115

333

he had was already on her staff/ so we'd been so we had a very swift supper/

In syntactic corrections the speaker starts a syntactic construction which brings him to a dead end, and which he then replaces with another syntactic construction. Correction in the above sense largely corresponds to Schegloffs backward repairs (repairs with postpositioned initiation, cf. 3.2), Levelt's Ε-repairs (cf. 3.4), Hieke's repairs (cf. 3.3), Rath's Verbesserungen (cf. 3.5) and possibly some of Boardman's abandoned structures not followed by a parenthesis. 5.2.2. Improvement involves exactly the same processes of retracing as correction. The difference is, however, that the improvement or melioration, in relation to the meliorandum, is not correction of an error but addition of an equally correct, but otherwise more appropriate, unequivocal or emphatic version or expression. There are, of course, borderline cases where it is difficult to decide between correction and improvement. Improvements, at least those without repetition of the post-meliorandum, are accepted and used also in writing, which might be an argument for establishing improvement as a separate category. Figure 5 below illustrates the process in improvement.

S. 6.6 2 0 6 - 2 0 9

the

the

first

first

lazaret/

lazaret or hospital

opened

opened

to the wounded/

1 = meliorandum 2 = melioration Figure 5

Further examples: S. 2.11 471 —472 S. 6.4 504—506

oh yes I'm getting definitely getting very long in the tooth a n d / w e we didn't we certainly didn't fling our arms round his neck/ S. 6.6 185 — 190 when we knew/ without doubt/ that the the situation/ the war situation/ was very very complicated/ we left the countryside/

The above category of improvement corresponds to Levelt's Α-repairs (cf. 3.4), and it is only in his article that this kind of shift is treated as a separate

334

Ν. Ε. Enkvist, Μ. Björklund

category. In Hieke cases of improvement are included in "addition", which is a subcategory of "repair in syntax" (cf. 3.3). Rath has a sub-category called "Verbesserungen" within the category "Korrekturen mit Anakoluthen zur Folge" (cf. 3.5). This subcategory seems to correspond to both correction and improvement. 5.2.3. Insertion here means a structure shift where a construction is interrupted, another construction inserted, and the first construction then returned to, continued, and completed. Insertions occur both in spoken and in written language. In writing we can for instance use parenthetic remarks, which could be called insertions with congruous continuation. In impromptu speech we find not only such correct insertions but also two other variants, here called insertion with repetition and insertion with incongruous continuation. All these three types of insertion in impromptu speech will be illustrated below. a) insertion with congruous continuation

S. 6.6 56 — 58

Moscow was/

[a]

as you know/

a very very ancient city/

Figure 6

Another example: S. 6.4 1208—1213

I remember/ Jeanie Oxley/ you'll go and see her/ saying oh Milly go on/

b) insertion with repetition

S. 6.6 46

I

I

they tell me

[s] speak six languages Figure 7

As the example shows, in insertion with repetition the interrupted structure is repeated after the conclusion of the inserted structure, whereafter it is continued and completed. In many cases the repetition consists of only a part of the interrupted structure or a modified version of it.

Toward a Taxonomy of Structure Shifts

335

Examples: S. 6.6 197 — 201

we were [asja] and Nadia were so young/ they knew nothing about it I I too knew nothing about it/ but we were all/ anxious most anxious to learn/ S. 6.6 398 — 407 on [dhi] Wednesday the eleventh of March/ nineteen fifteen/ all of the dates I give you/ you must understand are Russian dates/ they are thirteen days/ behind/ your calendar/ the English calendar/ (break) so on the eleventh of March [wisa] we started/ S. 6.4 406 — 410 I mean I even remember can I just tell a story here/ because I remember when I was about fifteen or sixteen/ As a variant of insertion with repetition one could treat cases where a structure started by one speaker is interrupted by a comment from another speaker, whereupon the first speaker repeats his interrupted structure before continuing and completing it. Example: S. 2.11 1264-1267

B: and then they A: oh God almighty/ B: then they categorise the risk for you/

c) insertion with incongruous continuation

mS. 6.6 3 0 0 - 3 0 8

it seems that a large dish/ of beautifully made red crosses/ made out of material/ but beautifully made/

I still have mine/ it is a great treasure/ it always will be/

were placed/ on the altar in the evening/

Figure 8

As the example shows the speaker continues the interrupted structure in a way which is grammatically incongruous with the way it was started. The articles summarized above in section 3 only deal with insertion with repetition. Thus, to Schegloff this kind of construction is a repair whose effect on the syntactic form of the sentence is: reorder the elements of projected talk, inserting into a current sentence what might have been planned for a later one (Schegloff 1979: 265). In Levelt's terms insertion with repetition is called Drepairs (cf. 3.4 and Levelt 1983: 51). Hieke counts at least some insertions with repetition as repair in syntax — addition (cf. 3.3). In Boardman's paper abandoned structures followed by a parenthesis correspond to insertion with repetition.

336

Ν. Ε. Enkvist, Μ. Björklund

5.2.4. Abandon makes it possible for the speaker to interrupt a structure he has started and shift to a totally different structure, leaving the first one unfinished. Abandon could be pictured as in figure 9 below.

S. 2.11 1517 —1522

but actually couldn't/

doing

it/

I

and exhausting/for one thing/ any sort of teaching/ don't you find/

Figure 9

As the example shows, the first structure is left unfinished, and another, different one is started. In this example the second construction too contains breaks classifiable as instances of structure shift. Sometimes the structure after the abandon links up with the text before. In other instances the whole idea may be discarded along with the structure (cf. example S. 2.11 1230—1232 below). There are, however, cases where the abandoned structure is obviously necessary for the receiver's understanding of what follows (cf. Polany 1977, 1978). For instance in example S. 2.11 1218—1226 (given below), the abandon after the noun phrase might be interpreted as an instance of thematization by left dislocation. Those who need discrete classification might define the difference between correction and abandon by the principle that a correction ought to show what should be replaced by what, whereas abandon just leaves an unfinished construction behind as the flow of speech goes on. A few examples of abandon: S. 2.11 1218 — 1226

and it was basically given by a management consulting company but unfortunately the two guys who gave ((them)) the talk one of them was still working for the management consultant and the other one who was much the better of the two in terms of presentation and interest and general all round dynamism had been a member of their company but was now working for a a real live organisation S. 2.11 1230 — 1232 if you're any good you don't [ji] it you know you move on from management consultancy right no I don't know that any moral can be drawn really S. 2.11 1231 — 1237 but [3:m] there are about thirty basic courses/ [3:m] half of them I you would if you went into their shop in in Regent Street/ today/ half of them would be the old ones/ The last example (S. 2.11 1231 — 1237) is confusing and open to different analyses. It could be an insertion with repetition if half of them would is

Toward a Taxonomy of Structure Shifts

337

regarded as repetition of half of them\ you would although the object of the interrupted structure then has become subject in the repetition + continuation. It could also be a syntactic correction if the if-clause + half of them would is regarded as correction of half of them/ you would. One solution would be to say that half of them/ you would is abandoned. Note, however, its importance to the information structure of the text. Abandon corresponds to Boardman's "abandoned structures not followed by a parenthesis" (cf. 3.6) and Rath's "Abbruche" (cf. 3.5). Levelt probably counts abandons as syntactic Ε-repairs (cf. Levelt 1983: 54 "Sometimes, syntax becomes fully scrambled for some reason, and the speaker starts all over again—"). 5.2.5. The term anacoluthon is here reserved for cases of structure shift where pre-shift and post-shift structures contain a shared overlap area. This is illustrated in figure 10:

S. 6.6 5 9 - 6 0

and

its

many many

of its

ancient fortresses

Figure 10

If we assume that the repetition of many was purposeful and not hesitant, the example shows that the speaker starts out planning for "and its many many ancient fortresses", but after many many he has either changed to the plan for and many many of its ancient fortresses, or thought he actually started according to this latter plan, and produces of its ancient fortresses. The result is a structure shift where the post-shift structure needs a part of the pre-shift structure to be complete. Further examples: S. 2.11 1526—1527 I have been for the last year I've been doing that thing S. 6.6 363 — 368 to thee/ [flora:ns]/ child of God/ servant of the most high I this token of faith hope and charity/ is given to you/ S. 11.3 122—123 he looked no longer looked like Diaghilev/ The above notion of anacoluthon corresponds to Rath's Drehsätze (cf. 3.5) and Linell's first type of anacolutha, dubbleringar. 5.2.6. Confused shifts. Practically all instances of structure shift in the present materials could be classified as correction, improvement, insertion, abandon, or anacoluthon, with a certain amount of overlapping. There were, however, a few odd cases where the shift is of a very unclear nature. Sometimes it may even be unclear what the speaker is trying to say. Just one example:

338

Ν. Ε. Enkvist, Μ. Björklund

S. 6.6 123 — 133

I can't describe those silver birch forests/ they were lovely by [le] day/ but they were glorious/ by night/ especially in the moonlight/ when [ev] the moonlight would play hide and seek/ amongst those dappled white and [s] the browny/ [3:] trunks/

One approach to such instances is to try to explain them in terms of several, perhaps overlapping or interwoven, structure shifts. 6.1. A taxonomy such as that attempted here is of course merely a first step towards a functional study of the hesitation-and-structure-shift complex. Hesitation and structure shift have attracted attention especially because they might provide us with a window to message processing. Among the questions future functional studies should attempt are: (a) Do hesitations and structure shifts tend to occur at syntactically definable places (junctures between sentences, clauses, phrases, constituents) rather than in places that fail to coincide with syntactic junctures? (b) Do hesitations and structure shifts tend to occur in places where a speaker faces a particularly large number of choices, for instance choices of one out of an open rather than closed class of items (a verb or substantive, for instance, rather than a pronoun or conjunction)? (c) To what extent do hesitations reflect non-linguistic, that is, pragmatic and personal stresses rather than linguistically definable ones? 6.2. At the present state of the art, one of the plausible hypotheses seems to be that hesitations and structure shifts arise through excessive processing loads. Different individuals show considerable differences in critical thresholds: some produce many hesitations and structure shifts in situations where others produce fluent, non-hesitant, and non-shifted speech. Once the critical threshold has been overstepped, however, different people may resort to different hesitation and structure-shift mechanisms, though they still conform to a repertoire of patterns such as that presented here. This repertoire may even be expected to be to some extent universal, though one may also suspect influence from communication situations, text type, style, and even language type. It would thus be interesting to find out whether speakers of richly inflected or agglutinative languages, Finnish or Turkish for instance, show patterns of hesitation and correction different from those common to speakers of, say, English. Perhaps such differences might owe to the different lengths of planning spans that may be determined or at least influenced by the typological structure of the language. 6.3. It is also obvious that a model of impromptu speech — a 'performance model' as some linguists would call it — must provide a mechanism for simulating hesitation and structure shift. One possible approach towards building such a model would be to construct a special hesitation-and-correc-

Toward a Taxonomy of Structure Shifts

339

tion module which could be inserted into the chain of operations involved in speech production whenever the monitor shows a mismatch between plans and execution ('plans' here being a cover term for discourse-producing operations at many levels). For descriptive economy such a model might be constructed to involve a maximal number of stages: point of interruption, hesitation, editing signal, pre-correction (or pre-melioration), correction (or melioration), and post-correction (or post-melioration). The different types of correction or melioration could then be explained in terms of omissions of certain operations from this full repertoire. Abandon thus results if the speaker leaves out the entire pre-correction — correction — post-correction complex. And different types of correction result from different treatments of pre-correction, hesitation and editing signals. Metaphorically one could visualize such a hesitation-and-correction module as a piece of generative apparatus that can be inserted at need into the sequence of generative operations, and which can be used maximally to produce all the stages in full correction process, or less than maximally to omit some of the stages in this complete process. To what extent such a model is worth implementing in detail is a question well worth discussing. Language teachers often point out that such a model would be useful for practical teaching purposes. One of the problems for language learners, including prospective public speakers using their own language, is how to hesitate and to correct oneself in ways that seem acceptable and 'canonical'. And here we come back to teleological optimism, to the view of hesitation and correction, not as something undesirable or evil, but as one of the devices that make impromptu speech and dialogue possible. The impression of fluency does not arise only from an absence of hesitations and structure shifts, but more often of a skilful use of these mechanisms.

Bibliography Boardman, G. M. 1977

A Study of Certain Kinds of Anacolutha in a Corpus of Spoken English. In Bald, W.-D. and Ilson, R., eds. Studies in English Usage ( = Forum Linguisticum 6) Frankfurt a.M.: Lang, P.; Bern: Lang, H., pp. 1 8 3 - 2 2 8 . Butterworth, B. 1981 Speech Errors: Old Data in Search of New Theories. Linguistics 19.7/8: 627 — 662. Also in Cutler 1982. Cherubim, D., ed. 1980 Fehlerlinguistik. Tübingen: Niemeyer. Crompton, Α. 1981 Syllables and Segments in Speech Production. Linguistics 19.7/8: 663 — 716. Also in Cutler 1982. Cutler, Α., ed. 1982

Slips of the Tongue and Language Production. Berlin etc.: Mouton. The Reliability of Speech Error Data. In Cutler 1982: 5 6 1 - 5 8 2 .

340

Ν. Ε. Enkvist, Μ. Björklund

Dechert, Η. W. and Raupach, Μ., eds. 1980 Temporal Variables in Speech ( = Janua Linguarum, Series Maior 86). The Hague etc.: Mouton. Enkvist, Ν. E., ed. 1982 Impromptu Speech: Α Symposium ( = Publications of the Research Institute of the Äbo Akademi Foundation, 78). Abo Akademi. Forthcoming. A Note on the Definition and Description of True Anacolutha. In Stockwell Festschrift. Fay, D. 1981

Substitutions and Splices: A Study of Sentence Blends. Linguistics 19.7/8: 717 — 749. Also in Cutler 1982. Fromkin, V. Α., ed. 1973 Speech Errors as Linguistic Evidence. The Hague and Paris: Mouton. Fromkin, V. Α., ed. 1980 Errors in Linguistic Performance. Slips of the Tongue, Ear, Pen, and Hand. New York etc.: Academic Press. Garnham, Α.; Shillcock, R. C.; Brown, G. Α.; Mill, A. I. D. and Cutler, A. 1981 Slips of the Tongue in the London-Lund Corpus of Spontaneous Conversation. Linguistics, 19: 7/8, 8 0 5 - 8 1 7 . Gasparov, Β. M. 1978 Ustnaja rec' kak semioticeskij ob"ekt. In Semantiha nominacii i semiotika ustnoj reli ( = Acta et Commentationes Universitatis Tartuensis 442). Tartu. Hieke, A. E. 1981 A Content-Processing View of Hesitation Phenomena. Language and Speech, 24: 2, 147-160. Hotopf, W. Η. N. 1983 Lexical Slips of the Pen and Tongue. In Brian Butterworth, ed., Language Production, vol. 2, London etc.: Academic Press, pp. 147—199. Laver, J. 1980 Monitoring Systems in the Neurolinguistic Control of Speech Production. In Victoria A. Fromkin, ed., Errors in Linguistic Performance. Slips of the Tongue, Ear, Pen and Hand, New York etc.: Academic Press, pp. 287 — 305. Levelt, W. J. M. 1983 Monitoring and Self-Repair in Speech. Cognition 14: 41 — 104. Linell, P. 1980 Svenska anakoluter. In Sigurd Fries and Claes-Christian Elert, Svenskans heskrivning 12 ( = Acta Universitatis Umensis 37), pp. 173—183. Linell, P. 1982 Speech Errors and the Grammatical Planning of Utterances. In Koch, W., Platzack, C. and Gunnel, T., eds., Textstrategier i tal och skrift ( = Acta Universitatis Lundensis, Humaniora 38), pp. 134—151. Linell, P. 1983 How Misperceptions Arise. In From Sounds to Words. Essays in Honour of ClaesChristian Elert ( = Acta Universitatis Umensis, Humanities 60), pp. 179 — 191. Nosek, J. 1969 Pause and Repetition in Modern Colloquial English. Acta Universitatis Carolinae, Philologica 3 ( = Prague Studies in English 13), pp. 35 — 58. O'Connell, D. A. and Kowal, S. 1983 Pausology. In Sedelow, W. A. and Sedelow, S. Y., eds., Computers in Language Research 2 (Trends in Linguistics. Studies and Monographs 19), Berlin etc.: Mouton, pp. 2 2 1 - 3 0 1 . Östman, J.-Ο. 1981 You know: A Discourse-Functional Approach ( = Pragmatics and Beyond, II: 7). Amsterdam: John Benjamins Β. V.

Toward a Taxonomy of Structure Shifts Polany, L. 1977 Polany, L. 1978

341

Not So False Starts. Sociolinguistic Working Paper Number 41, Austin, Texas: Southwest Educational Development Laboratory. False Starts Can Be True. Meeting of the Berkeley Linguistic Society, 1978, pp. 628-639.

Rath, R. 1975

Korrektur und Anakoluth im Gesprochenen Deutsch. Linguistische Berichte 37 — 75: 1-12. Russkaja, R. R. Akademija Nauk SSSR: Institut Russkogo Jazyka. Moskva: Izdatel'stvo Nauka, 1973. Schegloff, Ε. A. 1978 The Relevance of Repair to Syntax-for-Conversation. In Givon, T., ed., Discourse and Syntax. Syntax and Semantics, Volume 12, New York etc.: Academic Press, pp. 261-286.

Shattuck-Hufnagel, S. 1982 Three Kinds of Speech Error Evidence for the Role of Grammatical Elements in Processing. In Obler, L. K. and Menn, L., eds., Exceptional Language and Linguistics, New York etc.: Academic Press, pp. 133 — 142. Sternberger, J. P. 1983 Inflectional Malapropisms: Form-Based Errors in English Morphology. Linguistics 2 1 - 4 : 573-602. Svartvik, J. and Quirk, R., eds. 1980 A Corpus of English Conversation ( = Lund Studies in English 56). Lund: C W Κ Gleerup. Svartvik, J. and Stenström, Α.-Β. Forthcoming. Words, Words, Words: the Rest is Silence. Paper read at the AILA Congress at Brussels, August 1984. Zemskaja, E. A. 1979 Russkaja ra^govornaja re?: lingvistiieskij anali^ iproblemy obulenija. Moskva: Izdatel'stvo Russkij Jazyk. Zemskaja, Ε. Α.; Kitaigorodskaja, Μ. V. and Sirjaev, Ε. N. 1981 Russkaja Ras>govornaja r e f . ObStSie voprosy, Slovoobra\ovanie, Sintaksis. Moskva: Izdatel'stvo Nauka. Zwicky, A. M. 1982 Classical Malapropisms and the Creation of a Mental Lexicon. In Obler, L. K. and Menn, L., eds., Exceptional Language and Linguistics, New York etc.: Academic Press, pp. 1 1 5 - 1 3 2 .

H E R M A N N KAYSER

Some Aspects of Language Understanding, Language Production, and Intercomprehension in Verbal Interaction* 0. Introductory remarks Looking at natural language communication throughout the notions of language understanding and production does not mean to consider questions of interpretation and language use only from the standpoint of structural linguistic description, but to point to the mental processes and operations going on inside a speaker's and a hearer's mind during verbal interaction. One methodological approach providing representation models for the description of linguistic structures as well as of mental processes and states underlying the natural language occurrences, is the 'information processing' or 'computational' paradigm. One of the prime assumptions of the Artificial Intelligence and Cognitive Science approach is that human language use can be simultated by a computer, which means that comprehending and producing natural language utterances can be regarded as an information processing where meaning is conveyed throughout linguistic patterns that are produced by cognitive processes operating on a knowledge base stored in memory. In the last years, research in linguistics (computer- and psycholinguistics) dealing with the interpretation of natural language texts and discourse understanding has found useful to adopt the 'compile- and -execute' strategy of the computational paradigm in order to develop dynamic procedural models for the representation of text and discourse processing. These models have also found application in the analysis of dialogues (cf. B. Grosz 1977, Eikmeyer/Rieser 1982, Metzing 1980). A theoretical approach of the analysis of dialogues and conversation which is quite opposite to the computational paradigm is Conversational Analysis, which has its origins in sociology and ethnomethodology. Methodologically, Conversational Analysis uses a purely inductive approach in order to describe the elementary organization of verbal exchange in certain types of conversation. Though, at the first glance, cognitive processes relevant to spoken language occurences are not a topic of Conversational Analysis and Ethnomethodology, comprehension and pro-

* Sections 1 and 2 of this paper are a modified version of a contribution to the conference "Interactions Conversationnelles" at Urbino (Italy) in July 1985.

Language Understanding, Language Production, and Intercomprehension

343

duction processes of speakers are implicitely taken into account throughout some of its basic concepts. The aim of the present paper is, first, to show that the 'computational' and the 'conversational' approach, though quite different in their theoretical apparatus, have some relevant theoretical concepts which could be linked together under the question: What enables interlocuters to understand each other in verbal interaction1? After having given a brief survey on Cognitive Science- and Computational Linguistic's modells of understanding and production processes in (cooperative) communication (section 1), I shall bring forward, in terms of understanding and production, some fundamental concepts of Conversational Analysis in order to show in which way structural conversational properties of verbal interaction could be related to mental processes taking place in 'interactive' understanding (section 2). In the aim to examine further this particular phenomenon of verbal interaction, I shall introduce the notion of 'intercomprehension' (section 3). This notion shall designate the (abstract) state of well-balanced mutual understanding between interlocutors as well as the cognitive operations interlocutors accomplish and work out in conversational activities in order to generate that (cognitive) state, the assurance of which seems to be necessary to take successful functioning of verbal interaction for granted.

1. Language understanding and production as cognitive processing in cooperative communication According to recent work in the fields of Cognitive Science and Linguistics, the study of meaning is "the study of the relationship between linguistic objects and the mental states and processes involved in their production and comprehension" (Winograd 1976: 236), and since we can assume that the utterance of meaningful discourse is the primary purpose of language use, the cognitive mechanisms underlying the production and understanding of utterances are the principal objects of the study of language use. 1.1. In a general way, comprehension and production may be analyzed as mental activities which are based on specific goals, on beliefs and on knowledge, and which take place in a communicative situation of language use where producer and comprehender are processing information. More specifically seen, comprehension and production are a cognitive processing going on inside producers and comprehenders, where linguistic objects (sound 1

The notion 'verbal interaction' labels a type of verbal cooperative communication which can be analyzed relative to its conversational organization (organization of the local level of linguistic and conversational structures) and its interactional organization (organization on the global level of cognitive processes guiding the conversational organization). For further explanation cf. Bange/Kayser 1985.

344

Η. Kayser

patterns) are somewhat linked by complex operations to mental representations which can be viewed as parts of a mental model speakers and hearers develop relativ to their knowledge about the external world. Of course, linguistic and cognitive aspects of understanding and production can not be seen separately: a mental model of the world is essentially based on a speaker's knowledge and belief, just as the linguistic form of an utterance or a sentence depends on decision processes determined by a speaker's knowledge of the language and his communicative goals. However, for reasons of analysis, a model that outlines the nature of cognitive processing may be sketched out, tracing separately processes of understanding and of production. Such a process model is generally supposed to consist of four phases (though there are also two- or three-step models; cf. Winograd 1983, Johnson-Laird 1983) which, in fact, may overlap. 2 Their sequential disposition corresponds only to a processing dominance of one phase with respect to the others. I shall first develop this model for the producer: in a very general way. 1. The phase of 'planning' In this initiating step, the producer intends globally to persue some goal by means of a (spoken or written) text. The kind of text which shall be produced, is selected according to the criteria of efficency relative to a given context. 2. The phase of 'ideation' On this stage of production, which succeeds goalsetting and textselection, appears a mental thematic conceptualisation, functioning as a control denter for the development of a thematic idea. 3. The phase of 'development' This stage contains the further elaboration of the thematic idea, which means the seeking and retrieval of particular knowledge spaces (conceptual organized knowledge) stored in memory. These knowledge retrieval processes generate a 'prepositional representation' which is close to the surface form of a sentence (an utterance) (cf. Johnson-Laird 1983: 408 ff), mapping a mental model or imagery which is some kind of mental picture (cf. Paivio 1971: 450 ff). 4. The phase of 'parsing' (Mental) propositional representations, passing through a stage of 'mental parsing', become linguistic manifestations in the surface form of a text, i. e. in a phonemic or graphemic codation.

2

The presentation of this model follows the one brought forward in Dressler/de Beaugrande (1981) which is a synthesis of different approaches, mainly of de Beaugrande (1981), Flower/ Hayes (1979) and Meyer (1979). For a detailed and complex study on cognitive aspects of (discourse) understanding and production cf. Foss/Hakes (1978), Freedle (ed.) (1977), Johnson-Laird (1983) and Paivio (1979).

Language Understanding, Language Production, and Intercomprehension

345

The process, taking place in reception, can be roughly viewed as operating in analogue phases but in the opposite direction: it will start with a phonemic or graphemic representation which encodes the sounds of an utterance and which is perceived by the language user. Then there will be a propositional representation which is close to the surface form of the utterance, and finally a mental model will be constructed according to what the hearer beliefs or knows about the external world. This phase in the comprehension process which corresponds to the 'development' phase in production, and in which the mental model construction takes place, is called 'concept recovery phase': propositional representation units provoke the activation of knowledge concepts stored in memory which means that the comprehender seeks to mentally reconstruct and identify the concepts developed and coded by the producer. Though such a model had been mainly developed to sketch out the cognitive processes involved in the production and reception of written or spoken monological texts, it is also supposed to be applicable to the production and comprehension in dialogues. Yet, an application to the field of verbal interaction demands modifications and is more complicated. Consequently, one cannot claim that there is an identity correspondance between process phases of production and process phases of comprehension. Neither will there be strong succession: understanding processes of a comprehender do not start when the producer has finished his utterance, but reception begins already before the producer's operations end. Rather, comprehension and production have to be regarded as simultaneoulsy processed by hearer and speaker, respectively. Moreover, with regard to the hearer and his processes of comprehension, we might argue that they partly overlap with the initiating phases of speech production. This is mainly valid for the 'concept recovery' and the 'planning/ideation' phases, since, while (re-)constructing the meaning of the utterance-in-production, the comprehender may already be orienting his mind (more or less consciously) towards his own next utterance (though this may also not be the case, e. g. in verbal interactions with more than two interlocuters). Thus we may claim that the communication process involves simultaneously related activities by the producer and the comprehender throughout the internal cognitive understanding and production processing. 1.2. The cognitive processing model sketched cut, can be integrated in a general model of cooperative communication (cf. Winograd 1983: 13 ff) which shows from the standpoint of information processing between a producer and a comprehender the main components involved in communication and their relation to the cognitive processing. Figure 1 illustrates the overall structure of such a model. It indicates the gobal communication process, i. e. the way communicative goals and different types of knowledge are operating throughout the phases of cognitive processing of production to convey meaning which is received

346

Η . Kayser

!

J3 ω Q

ω

i l hJ < tΡ 2 Ο Ζ

Ο Ζ

c -- ΙΛ"U S υ Ο »α Γ b υ u CL n

Γ

I

,

ζ




L

I

§

Ο ω >

§ υ ζD 2 ο υ

ώ > Η Ζ

Ο Ο ο