Structure of centre of attention in a multi-party conversation in Japanese: Based on the data of a review meeting concerning a Science Café held in Hiroshima.
Meta-informative Centering in Utterances Between Semantics and Pragmatics Edited by
André Włodarczyk Université Charles de Gaulle & Université Paris-Sorbonne
Hélène Włodarczyk Université Paris-Sorbonne
John Benjamins Publishing Company Amsterdam / Philadelphia
Table of contents Preface Introduction André Włodarczyk & Hélène Włodarczyk
VII IX
PART 1. Associative semantics and meta-informative centering Roles and anchors of semantic situations André Włodarczyk
3
Frames of semantic situations André Włodarczyk
21
Grounding of the meta-informative status of utterances André Włodarczyk
41
Attention-centered information in language Hélène Włodarczyk
59
PART 2. Neuropsychological evidence for the MIC theory Semantic and episodic memory by reference to the ontological grounding of the old and new meta-informative status Franz J. Stachowiak
103
Tracing the role of memory and attention for the meta-informative validation of utterances Franz J. Stachowiak
121
PART 3. Meta-informative centering in languages It-clefts in the Meta-Informative structure of the utterance in modern and present-day English Ana Elina Martínez-Insua & Javier Pérez-Guerra
145
Discourse coherence and referent identification of subject ellipsis in Japanese Shigeko Nariyama
167
VI
Meta-informative Centering in Utterances
Structure of centre of attention in a multi-party conversation in Japanese: Based on the data of a review meeting concerning a Science Café held in Hiroshima Miki Saijo Verbal aspect in Slavic languages between semantics and pragmatics Hélène Włodarczyk
183 193
The position in the utterance and the melodic realisation of object and reflexive pronouns in classical modern literary Russian Olivier Azam
231
Accented and unaccented pronouns in Ancient Greek: A pragmatic choice by the speaker Jean-Christophe Pitavy
259
Personal subject pronouns and the meta-informative centering of utterances in classical Latin 285 Perrine Vedrenne-Cloquet Glossary of defined terminology
297
Index
303
Introduction André Włodarczyk & Hélène Włodarczyk
Université Charles de Gaulle / Université Paris-Sorbonne, Centre de Linguistique Théorique et Appliquée (CELTA Sorbonne)
1. Information and meta-information At the beginning of the 21st century, the study of communication in human languages still remains under the influence of the theory of information structure (Lambrecht 1994) which elaborates on the concepts of theme, rheme, and communicative perspective. The meta-informative centering (MIC) theory is an alternative framework based on the concepts of centre of attention (CA), meta-information and the meta-informative status of information. Importantly, the MIC theory is built on associative semantics (AS) in which the concept of information is defined as a compound relational structure, which to some extent makes it compatible with the definition of information as often used in the field of computer (information) science. In order to explain how it is possible for the hearer to interpret a linguistic message (i.e. to build a mental representation of the situation spoken about) a mapping must be established between the linguistic form and its content, using some formalised representation of meaning. From the semantic point of view, the logical motivation for the formal representation of situations consists in reducing multi-argument relations solely to unary and binary ones, and combining them in compound configurations. 2. Attention centering The origins of the MIC theory go back to Włodarczyk 1999 when we used the concept of “centre d’intérêt” (Fr. centre of interest) for the first time. This French term has since been replaced by that of “centre of attention” under the influence of the American centering theory (Grosz & Sidner 1986), and it underwent a few revisions and reassessments in the succeeding papers published in English. According to the MIC theory, because of the linear order of speech, no judgment may be uttered without selecting at least one centre of attention; thus centering is considered as a structuring operation not only at the discourse/text level but also at
X
André Włodarczyk & Hélène Włodarczyk
the level of utterance (cf. Chomsky’s binding theory) because of its necessarily linear (sequential) nature. In the American centering theory, centres of attention are defined at the discourse/text level: one constituent of an utterance is treated as a forward or backward looking centre in order to maintain the cohesive flow of information from one utterance to its successor. Forward and backward looking centres make it possible to give an account of the relations which bind utterances together into coherent discourses/texts. Although the centering theory precedes in time the meta-informative centering theory, these theories were created separately from each other. However, the results of the American theory can be straightforwardly integrated into the MIC theory. Since the anaphoric and cataphoric motivations of the meta-informative old and new status of the utterance during the communication act match perfectly well the concepts of backward- and forwardlooking centres, the MIC theory is not an alternative to the American centering theory. Rather, the latter is a complementary approach with respect to the MIC which is more general. Although the main characteristics of utterances consist in the fact that the priority in their generation/understanding is given to attention-driven processes, utterances are built choosing the most suitable ready-made verb valence schemata which are used by the speaker in order to communicate the old or new status of expressed content. Therefore, linguistic expressions which correspond to grammatical notions such as subject, object, topic and focus are defined as attentiondriven phrases (ADP); i.e. phrases governed by the centres of attention (1) in base utterances where merely either an old or new status of information can be envisaged for both the subject and predicate in each utterance (but where no contrast between the old and new status of information is expressed), and (2) in extended utterances where the topic and focus are defined as dually opposed concepts; (a) the topic being defined as governed by an old meta-informative status contrasting with the new one of the comment and (b) the focus as governed by a new metainformative status contrasting with the old one of the background. Although the importance of attention for cognitive functions of the brain has recently also been emphasised by neuropsychologists (part 2 in this volume by Franz J. Stachowiak), this faculty of the human brain has not yet been explicitly used to account for linguistic objects in mainstream cognitive linguistics. In the proposed theory, one kind of attention is a component of pragmatics and since the semantic content of linguistic messages is not directly accessible, the interpretation of the truth value of utterances needs to be postponed and priority is given to the resolution of the meta-informative status of communicated chunks of information. In other words, in linguistic communication, truth valuation of the content of an utterance takes place only behind – so to speak – communication, i.e. after the meta-informative old or new status has been assigned using such means
Introduction
of expression as intonation, word order, declension, determination, modality and aspect, etc. It seems important for us to mention the endeavours of the psychologist Russell Tomlin (1997) who tried to check whether attention might be of use in research on languages, but the erroneous interpretation of the results his Japanese collaborators obtained led him to abandon the initial idea, stating that there was no direct link between attention and expression. 3. Between semantics and pragmatics Languages provide speakers with ready-made schemata of expressions having verbs for pivots (known as verb valences), learned by native speakers from the very beginning of their linguistic activity. These schemata enable the speakers to communicate whatever they “mean” about situations of the “world out there” (as these situations are conceptualized in their mind). However, once the schema of an expression has been selected, the speakers’ freedom as regards concentrating their attention on a desired chunk of information as represented in their mind is considerably limited. Nevertheless, they may transform the schema they have “at hand” in such a way that eventually it fits their attentional purposes. For example, they may change the word order or the voice of the verb in case they need the selected chunk of information be expressed by an attention-driven phrase (ADP). Currently, combining notions such as “argument structure” and “verb valence” linguists attempt to capture the mappings between noun phrases in the syntactic plane and the roles enacted by the participants in the semantic plane of discourse. Here, this correspondence is recognised as a more complex problem the solution of which cuts across another dimension of language, namely, the pragmatic one where attention-driven saliences (centres of mental representations) are established independently of semantic roles and situations. It is for this reason that subjects and objects are defined as ADPs, being therefore considered as more than merely ‘formal’ syntactic positions. They are the result of a compromise between the pragmatic motivations of the speaker and the available linguistic resources expressing the effects of globally and locally centred attention in base utterances. Indeed, since the syntactic structure of utterances depends on the pragmatic (meta-informative) constituency, there is a need to insist on the relationship with some important pragmatic issues on the one hand, as well as with morphology and syntax on the other. For this reason, we can observe that the meta-informative structure of utterances and discourse is very closely related to the linguistic form. Most pragmatic approaches are rather ‘distant’ from the linguistic form, they deal with the inferences drawn by the hearer about the speaker’s intentions in order
XI
XII André Włodarczyk & Hélène Włodarczyk
to interpret utterances. Obviously, such approaches, along with praxemics (where language is used in a perlocutory act), also belong to the study of language. Although these aspects of pragmatics are not characteristic of the MIC theory, the centrality of pragmatics in this approach leads to an important consequence concerning the compositionality principle which has been largely admitted in contemporary linguistic theories since the work of Montague R. Compositionality is often spelt out as follows: “The meaning of a complex expression is determined by the meaning of its structure and the meanings of its constituents.” (May 2013, http://plato.stanford.edu) It is however obvious that in an important number of uses of language such as that of stereotypes, this principle may be questioned. In the MIC-theoretical framework this problem can be treated in an innovative way by establishing a direct relationship between syntax and pragmatics (less directly with semantics as is often intuitively taken for granted). We consider that the constituency structure of utterances is a direct mirror of the relationship between the subject (the globally centred attention-driven phrase) and the predicate (any information communicated about the subject). We wish to emphasise that in our view, both subject and predicate are concepts belonging essentially to pragmatics. However, what makes communication extremely fast and efficient is that there exists a default relationship between the subject (as attention-driven phrase) and one of the semantic roles which varies depending on language family. In nominative languages, the subject of an utterance with a verb in the active voice (which is the “unmarked” voice in this case) is by default related to active semantic roles. Thus, when a speaker selects a verb in order to refer to a situation (a state of affairs in the world out there) s/he can retrieve from his/her memory a syntactic structure containing the information about the semantic role enacted by its subject; e.g. in English, the verb eat in the active voice requires as its subject the choice of a noun referring to an animate being which plays an active role (eater). This default relationship is different in ergative languages where speakers are provided with default information concerning the semantic passive role expressed by the noun phrase in the ergative case. Nevertheless subject remains a pragmatic concept, which makes it possible for languages to use alternative verb voice forms or verbal lexemes in which the subject does NOT play the default semantic role relevant to a given type of languages (for example, the active role in nominative languages). Among verb categories, aspect and tense are essential for metainformative strategies since they enable speakers to choose a point of view or a point of reference concerning the situation they talk/predicate about. In other words, these categories indicate which chunk of information (stage or moment, in the case of aspect and past, present or future, in the case of tense) is being/has been chosen to fit the centre of attention.
Introduction XIII
In conclusion, restrictions on the compositionality principle ultimately boil down to pragmatic restrictions, or putting it more explicitly, semantic interpretations are dependent on the context of the utterance, be it linguistic or extralinguistic. Thus semantic theories cannot deal with the problem without including pragmatic considerations. The MIC theory – which incorporates pragmatic motivations into the constituency structure of utterances – is an attempt to provide a systematic account of the limitations of compositionality that can be observed in human languages. For the same reason, we claim that the pragmatic validation of utterances as conveying information with either new or old communicative status precedes in natural languages the truth logical valuation of sentences (true or false), the latter being the main preoccupation in formal languages. According to the pragmatic relevance theory (Sperber & Wilson 1986), linguistic communication depends more on inferences about speakers’ intentions and representations than on the decoding of linguistic expressions. We claim however that the pragmatic dimension of utterances cannot be modelled without the syntactic and semantic dimensions. Nevertheless, although semantic content is the main motivation of linguistic communication, utterances do not express all the information humans would probably be willing to transmit, all the more so since speakers cannot communicate content other than from their point of view and to the best of their knowledge. The attention-driven meta-informative, hence subjective, dimension of natural languages is represented by the rich choice of paraphrases of various kinds; e.g. passive or active voices, personal or impersonal (anonymous) constructions, topicalisation and focalisation, etc. Therefore, human languages should be treated using more sophisticated theoretical frameworks than the ones which were elaborated, in an effort to build a bijective (non ambiguous) mapping between form and meaning, for artificial and formal languages, preserving expression from subjectivity and ambiguity. The not fully compositional conception of natural language semantics lying at the heart of the AS and MIC theories brings them, in a sense, closer to the relevance theory than might at first glance seem to be the case. The syntax and morphology of natural languages make it possible to encode semantic content which is inextricably interwoven with the pragmatic attitudes of the speaker. However, in order to analyse expressions made up of heterogeneous elements (belonging to various dimensions of language), the proposed theoretical framework is an attempt to explain how morphological forms and syntactic structures of a particular language are mapped into semantic and pragmatic representations. Although some very important concepts (such as subject and predicate) sound astonishingly familiar and hackneyed, they are integrated in an innovative coherent theory and their scope is fixed by strict terminological conventions.
XIV André Włodarczyk & Hélène Włodarczyk
4. Overview of the contributions to this volume The first part of this volume contains the presentation of essential issues of the associative semantics (AS) and meta-informative centering (MIC) theories, the second part is an attempt at a neuropsychological foundation of this theoretical framework, and the third part deals with its application to several linguistic phenomena in different languages. Due to its origins dating back to investigations on typologically different languages and because it is attuned to contemporary research in the field of computer science, the ASMIC theoretical framework exhibits both theoretical and experimental characteristics. 4.1 Associative semantics and meta-informative centering The AS and MIC theories are characterised by an interdisciplinary approach; therefore linguistic problems are tackled in connection with neuropsychological approaches and computer science (artificial intelligence: multi-agent communication) approaches. As regards language theory, our theoretical approach calls into question the autonomy of syntax not only in relation to semantics but also to pragmatics. The first chapter concerns the problem of semantic roles which constitute the core of the semantic layer of natural languages. The second chapter is devoted to a formalised semantic description of types of situations in diverse languages, applying a subset of a universal ontology: this issue constitutes the basis for aspect studies (Chapter 10). The grounding of the meta-informative old and new status in discourse and in ontology is tackled in the third chapter, whereas the fourth chapter concerns attention centered information in different types of languages. 4.2 Neuropsychological evidence for the MIC theory The second part is an attempt by a neurolinguist Franz Stachowiak, to bring evidence from neuropsychological research on memory and attention to bear on the meta-informative centering in language. The chapter entitled Semantic and episodic memory with respect to the ontological grounding of the old and new metainformative status aims at integrating a MIC component into a neurolinguistic model of language production and understanding. The second chapter of this part is devoted to attention centering in communicative processes and to its relation to word order and syntax. Stachowiak claims that “verbal processing of information is strongly influenced by or even rests on the capacity and mode of operation of working memory and other types of memory and is intricately related to attentional processes, which play a role in directing the interest of a communication partner in spoken or written language.”
Introduction XV
4.3 Meta-informative centering in languages In the third part of the volume are presented the fruits of the application of the MIC theory to various languages regarding a number of problems which until now have found no satisfactory solution in approaches where syntax is not connected with meta-information. 4.3.1 Discourse coherence in English and Japanese The MIC theory provides a set of definitions of meta-informative operations independently of their expression in different languages, be it by lexical, grammatical, syntactic and/or prosodic markers. Thus, careful distinctions are made between meta-informative operations and their expression(s). Indeed, probably in any natural human language, topicalisation is a sort of meta-informative operation, it is characterised by the old status of the topic contrasting with the new status of the comment part of the utterance, whatever the linguistic means of expression. For example, left dislocation is one of them but, as a syntactic device, it may also be used to express a focus. Unfortunately, as a result of the difficulty of separating the ‘form’ from the ‘content’ in linguistic analyses, syntactic constructions have often been used to refer to the meta-informative operations they represent; because of the lack of a formalised axiomatic theory it is impossible to identify common structures among different languages. In the chapter by Ana E. Martínez-Insua & Xavier Pérez Guerra, the English itcleft is “presented as a device of focalisation used for establishing a meta-informative contrast with the second part of the utterance, making it possible, in a Strict Word Order (SWO) language as English, to put this focalised constituent at the front of the utterance (thus contradicting the neutral order: given-before-new)”. This corpus based study brings evidence for the quite recent consolidation (in the history of English) of the mechanism of clefting as a meta-informative strategy. In her paper about subject ellipsis in Japanese Shigeko Nariyama discusses two ellipsis-related issues using the MIC theory: (1) ellipsis as a discourse coherence marker, which answers the question: why use ellipsis? (i.e. the pragmatic issue); and (2) referent identification of ellipsis, which answers the question: how do we know whom/what a zero encoding of ellipsis refers to? (i.e. a semantic issue). It shows that centre of attention plays an important role in providing an adequate and consistent explanation of the mechanism of ellipsis. It also attests to the prominence of implicit lexical knowledge in Japanese that contributes to the referent identification of ellipsis. The chapter by Miki Saijo is an attempt to use the concept of centre of attention for the analysis of a multi-party conversation in Japanese. It is shown that the CA structure in the multi-party conversation in this case is identical to the
XVI André Włodarczyk & Hélène Włodarczyk
extended utterance structure of the MIC theory, as a result of analysing discourse examples collected from the actual meeting conducted in Japanese in a framework of topicalisation as well as central and peripheral kinds of attention. 4.3.2 Verbal aspect in Slavic languages between semantics and pragmatics The main function of the linguistic category of aspect is perfectly reflected by the traditional term “aspect” or “view” which means that the speaker chooses a view of the situation s/he is speaking about. This view of a situation, or “viewpoint”, is first of all reflected by an analysis of the internal parts of the situation: moments and stages. This necessary choice can be compared to that of attention centering with the purpose of building an utterance (cf. the definitions of subject and object in Chapter 4 in this volume). As such, aspect is an essential component of the metainformative structure of utterances. The internal view of the situation is further completed by external view parameters concerning its repetition, the modification of its flow or intensity, and the composition of several situations into one complex situation. The ASMIC theory is very helpful in dealing with the blurred boundaries between semantics and pragmatics as far as the category of aspect use is concerned, making it possible to propose a tentative but comprehensive way out of endless debates in the field of Slavic aspectology: the problem of aspect pairs, the difference between aspect and Aktionsart, the amazing differences in the use of imperfective verbs in Slavic languages and the use of the imperfect tense in French or progressive forms in English, etc. 4.3.3 Tonic and atonic personal pronouns in modern Russian, classical Greek and Latin The last three chapters bring an innovative explanation to personal pronoun usage in modern literary Russian, classical Greek and Latin. Personal pronouns are markers of discourse coherence both on the semantic and pragmatic levels. On the one hand, they are used for the identification of entities spoken about (taking part in world situations) in relation to the participants of the discourse situation. On the other hand, most Indo-European languages have at their disposal two series of personal pronouns: one of them is unstressed, used in base utterances, and the other one is stressed, used in extended utterances. Historically, this distinction corresponds to that of atonic and tonic pronouns in classical grammar, but with time the use of tonic pronouns as stressed and atonic as unstressed was blurred either by the disappearance of atonic forms (in modern Russian) or the incorrect use (with respect to the previous norm) of atonic pronouns as stressed (e.g. in contemporary Polish, cf. Włodarczyk H. 2012). Therefore, the theoretical basis provided by the MIC theory makes it possible to distinguish methodically
Introduction XVII
between morphology (tonic vs. atonic), prosody (stressed vs. unstressed) and the pragmatic status (old vs. new) of different types of pronoun uses in discourse. As pointed out by Jean-Christophe Pitavy concerning classical Greek: “the use of the accented personal pronoun in conjunction with the verbal form allows the speaker to construct an extended utterance.” Similarly, Perrine Vedrenne-Cloquet shows that in Latin: “the pronouns ego and tu featured in discourse have a highly important role in the construction of meta-informative coherence, since they can be used as topicalisations, but also as focalisations. Moreover, they can be used as additional markers: their second position gives to the first position element a pragmatic emphasis which it would not otherwise have.” The three chapters about pronouns emphasise that reference to the old and new meta-informative status makes it possible to distinguish between topicalised and focalised uses of tonic pronouns. Moreover, the description of word order in utterances containing topicalised and focalised pronouns provides evidence for the fact that topic is the global CA of extended utterances. Olivier Azam concluded from the observation of his Russian corpus that “topicalisation does indeed appear to be incompatible with postpositioning”. Let us add (this is mentioned in passing in Chapter 4) that the ellipsis of pronouns and zero-form pronouns (often referred to indistinctly as pro-drop) should be thoroughly distinguished: the latter can be attested to only in languages in which the person is relevantly marked by verb endings (most Indo-European languages) whereas ellipsis is possible both in Indo-European languages and in languages with no person category in verb flexion (e.g. Japanese). The volume is closed by a glossary of AS and MIC terms.
References Grosz, Barbara J. & Sidner, Candace L. 1986. Attention, intentions, and the structure of discourse. Computational Linguistics 12(3): 175–204. Lambrecht, Knud. 1994. Information Structure and Sentence Form. Topic, Focus, and the Mental Representations of the Discourse Referents. Cambridge: CUP. Sperber, Dan & Wilson, Deirdre. 1986. Relevance: Communication and Cognition. Oxford: Blackwell. Tomlin, Russell S. 1997. Mapping conceptual representations into linguistic representations: The role of attention in grammar. In Language and Conceptualization, Jan Nuyts & Eric Pederson (eds), 162–189. Cambridge: CUP. Włodarczyk, André. 1999. La validation informative des énoncés et la quantification linguistique. Études cognitives/Studia kognitywne 3: 121–133. Warszawa: SOW, PAN. Włodarczyk, Hélène. 2012. L’ emploi des pronoms personnels en polonais (par contraste avec le russe et le français). Études à la mémoire de J. Breuillard sous la dir. de S. Viellard, Revue des Études Slaves, t.LXXXIII, fasc.2–3, 617–648. Paris: Institut d’Etudes Slaves.
