Lexical Markers of Common Grounds (SiP 3), Volume 3 (Studies in Pragmatics) (Studies in Pragmatics)

LEXICAL MARKERS OF COMMON GROUNDS

STUDIES IN PRAGMATICS General Editor: Bruce Fraser Associate Editors: Kerstin Fischer, Maj-Britt Mosegaard Hansen

The Studies in Pragmatics series is dedicated to publishing innovative, authoritative monographs, and edited collections from all micropragmatic, macropragmatic and metapragmatic linguistic perspectives. Rooted in the interdisciplinary spirit of the Journal of Pragmatics, it welcomes not only book proposals from linguistics proper, but also pragmatically-oriented proposals from neighboring disciplines such as interactional sociology, language philosophy, communication science, social psychology, cognitive science, and information science. The goal of the series is to provide a widely read and respected international forum for high quality theoretical, analytical, and applied pragmatic studies of all types. By publishing leading edge work on natural-language practice, it seeks to extend our growing knowledge of the forms, functions, and foundations of human interaction.

Other titles in this series: FISCHER

Approaches to Discourse Particles

AIJMER & SIMON-VANDENBERGEN

Pragmatic Markers in Contrast

Forthcoming: CAFFI

Mitigation

Proposals for the series are welcome, please contact the General Editor, Bruce Fraser: [email protected]

LEXICAL MARKERS OF COMMON GROUNDS

EDITED BY

ANITA FETZER University of Lueneburg, Germany

KERSTIN FISCHER University of Hamburg, Germany

Amsterdam • Boston • Heidelberg • London • New York • Oxford Paris • San Diego • San Francisco • Singapore • Sydney • Tokyo

Elsevier The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, UK Radarweg 29, PO Box 211, 1000 AE Amsterdam, The Netherlands First edition 2007 Copyright © 2007 Elsevier Ltd. All rights reserved No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without the prior written permission of the publisher Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford, UK: phone (+44) (0) 1865 843830; fax (+44) (0) 1865 853333; email: [email protected]. Alternatively you can submit your request online by visiting the Elsevier web site at http://elsevier.com/locate/permissions, and selecting Obtaining permission to use Elsevier material Notice No responsibility is assumed by the publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the Library of Congress ISBN-13: 978-0-08-045322-4 ISBN-10: 0-08-045322-8 ISSN: 1750-368X For information on all Elsevier publications visit our website at books.elsevier.com Printed and bound in The Netherlands 07 08 09 10 11

10 9 8 7 6 5 4 3 2 1

Working together to grow libraries in developing countries www.elsevier.com | www.bookaid.org | www.sabre.org

Studies in Pragmatics (SiP) General Editor Bruce Fraser Boston University, USA Associate Editors Kerstin Fischer University of Hamburg, Germany Maj-Britt Mosegaard Hansen University of Copenhagen, Denmark Consulting Editor Jacob L. Mey University of Southern Denmark, Denmark Editorial Board Kent Bach, San Francisco State University, USA Diane Blakemore, University of Salford, UK Shoshana Blum-Kulka, Hebrew University, Israel Claudia Caffi, University of Genoa, Italy Alessandro Duranti, UCLA, USA Marjorie Goodwin, UCLA, USA Hartmut Haberland, University of Roskilde, Denmark William F. Hanks, University of California, USA Sachiko Ide, Tokyo Women’s University, Japan Mikhail Ilyin, Moscow State Institute of International Relations of the Foreign Affairs Ministry of the Russian Federation (MGIMO), Russia Richard W. Janney, University of Munich, Germany Kasia Jaszczolt, University of Cambridge, UK Elizabeth Keating, University of Texas, USA Sotaro Kita, University of Bristol, UK Ron Kuzar, University of Haifa, Israel Alec McHoul, Murdoch University, Australia Brigitte Nerlich, Nottingham University, UK Etsuko Oishi, Fuji Women’s University, Japan Srikant Sarangi, Cardiff University, UK Marina Sbis`a, University of Trieste, Italy Maxim Stamenov, Bulgarian Academy of Sciences, Bulgaria Deborah Tannen, Georgetown University, USA

This page intentionally left blank

TABLE OF CONTENTS

List of Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Anita Fetzer and Kerstin Fischer Part 1

1

Grounding and Common Ground . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

Ok, Next One: Discourse Markers of Common Ground . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 ˇ Sherri L. Condon and Claude G. Cech Grounding and Common Ground: Modal Particles and Their Translation Equivalents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 Kerstin Fischer Common Ground in Interaction: The Functions of Medial Doch in German . . . . . . . . . . . . 67 Karin Pittner A Common Ground for Knowledge Through Knowledge of Language: A Computational Research of Consensus-based Meaning in Scientific Papers . . . . . . . . . . . 89 Carlos Rodríguez Penagos Imposing Common Ground by Using Temporal Connectives: The Pragmatics of Before and After . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 Thora Tenbrink Part 2

Common Grounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

The Pragmatics of Common Ground: From Common Knowledge to Shared Attention and Social Referencing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143 François Nemo Reformulation and Common Grounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 Anita Fetzer Whose Common Ground? A Misunderstanding Caused by Incorrect Interpretations of the Lexical Markers of Common Ground . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 Moeko Okada Common Ground, Categorization, and Decision-Making . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 Thanh Nyan Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211

This page intentionally left blank

LIST OF CONTRIBUTORS

Anita Fetzer, University of Lueneburg Kerstin Fischer, University of Hamburg ˇ Claude G. Cech, University of Louisiana Sherri L. Condon, The MITRE Corporation François Nemo, University of Orléans Thanh Nyan, University of Manchester Moeko Okada, Senshu University Karin Pittner, Ruhr-Universität Bochum Carlos Rodríguez Penagos, Universidad Nacional Autónoma de México Thora Tenbrink, University of Bremen

This page intentionally left blank

Lexical Markers of Common Grounds Edited by Anita Fetzer and Kerstin Fischer © 2007 Published by Elsevier Ltd.

INTRODUCTION Anita Fetzer and Kerstin Fischer

1. THE PROBLEM This volume investigates the role of lexical markers with respect to the relationship between grounding and common ground on the one hand and between common ground and different types of common ground on the other. Recent research has shown that the content of utterances cannot be assumed to enter common ground only because of the fact that they are being mentioned. Rather, grounding is a complex, sophisticated mechanism in which minimal signals play key roles (e.g. Gardner, 2001). Besides contributing to the process of jointly construing common ground by ratifying each other’s contributions, lexical markers exercise a similar function as contextualization cues and framing signals (Gumperz, 1992) regarding the activation of particular types of common ground. The relationship between the two processes, grounding on the one hand and activating common ground on the other hand, is elaborated on and examined from the perspective of lexical markers. Lexical markers are conceived of as multifunctional devices whose degree of explicitness may vary. They have an important function in the processes of grounding and of activating common ground. The first part of the volume addresses the relationship between mechanisms of grounding and their reference to common ground. The second part examines different types of common ground. It is shown that the investigation of lexical markers provides a novel perspective for investigating the relationship between grounding, common ground and common grounds.

2. COMMON GROUND IN CONTEXT The multifaceted and heterogeneous category of common ground is central to theories of pragmatics, sociolinguistics, discourse and context. In its common-sense reading, it contains all of the information shared by a speech community, such as facts, attitudes, beliefs, norms, conventions and preferences as well as other types of sociocultural knowledge. In the research paradigm of the ethnography of communication (Saville-Troike, 1989), a speech community’s common ground and their sociocultural knowledge is anchored to both linguistic code and social practice, and has been systematized with respect to the categories of situation,

2 Lexical Markers of Common Grounds that is physical setting and psychological scene, participants, viz. speaker, hearer and audience and their statuses in the participation framework, ends, namely the goal and the purpose of the speech event from a sociocultural viewpoint, act sequence, that is how something is said as regards message form and what is said as regards message content, key, that is mock or serious, instrumentalities, viz. channels (spoken, written or email) and forms (vernacular, dialect or standard) of speech, norms of interaction and interpretation, and genre (Hymes, 1974). The notion of common ground also plays a prominent role in philosophical and cognitive conceptions of knowledge in the research domains of linguistics, pragmatics, philosophy of language and cognitive science. Here, common ground serves as background for reasoning and for retrieving speaker-intended meaning and other types of implicit meaning, such as indexical expressions, implicatures and indirect speech acts. Common ground is further understood as presupposition, which is categorized into two distinct domains, viz. semantic and pragmatic presupposition. The former is defined in a truth-conditional frame of reference and applies to any context, and the latter is conceived of as context-dependent. Presuppositions are organized and administered in the framework of context sets (Stalnaker, 1999), which serve as common ground in communication. In the field of computer science, common ground and world knowledge have frequently been conceptionalized as a database. While common ground and background are generally conceived of as context-independent notions which obtain in all possible scenarios and in all possible worlds, pragmatic presupposition, unlike semantic presupposition, is context-dependent. It is generally used as a cover term referring to the necessary and sufficient extra-linguistic conditions of communication, such as the social-context categories of participant, time and location. Pragmatic presuppositions are accommodated in speech act theory’s felicity conditions (Austin, 1980; Searle, 1969), which are considered as linguistic and social context categories and their satisfaction is assigned the status of a default configuration (Sbis´a, 2002). Pragmatic presuppositions are negotiated in and through the process of communication, and because of their defeasibility (Levinson, 1983), they can be cancelled, should a particular contextual configuration arise. Common ground in the sense of background furthermore plays a fundamental role not only in natural language communication, but also in natural language processing, for instance in dialogue system modelling. According to Vanderveken (2002), participants negotiate the compatibility of background with utterances and their felicity and satisfaction conditions in and through the process of communication. It has to be pointed out, however, that background and context are not identical because possible contexts of utterance can have different backgrounds. As a consequence of that, background contains not only mutual knowledge of facts about the conversational background, but also knowledge about the world and of the world, such as ethical norms and sociocultural values, thus transcending the common-sense notion of context. Searle (1995, 1999) considers background to be a necessary condition for both literal and non-literal meaning thus assigning it the status of a basic premise for felicitous communication. He defines background as an open-ended set of skills, pre-intentional assumptions and practices which are not representational but rather enable intentional acts and states to manifest themselves. For this reason, the conditions of satisfaction for attempted illocutionary acts depend on background and its mutual knowledge, and because of the background, what is said is undetermined by linguistic meaning.

Introduction 3 In relevance theory (Sperber and Wilson, 1996), common ground is conceived of as a common set of premises for inference rules. The relevance-theoretic conception of common ground differs from the traditional notion of mutual knowledge by its attempt to avoid the logical consequence of infinite regress, which follows from the code model of communication. To avoid infinite regress, Sperber and Wilson base their theoretical framework on an approximation of mutual knowledge, namely cognitive environments and mutual manifestness. The context-dependent refinements of mutual knowledge and common ground are as follows. A cognitive environment of an individual is conceived of as “a set of facts that are manifest to him” (Sperber and Wilson, 1996: 40). It “is merely a set of assumptions which the individual is capable of mentally representing and accepting as true. (…) We will argue that when you communicate, your intention is to alter the cognitive environment of your addressees; but of course you expect their actual thought processes to be affected as a result” (Sperber and Wilson, 1996: 46). They specify the consequence of communication with respect to mutual manifestness and its social relevance: “Mere informing alters the cognitive environment of the audience. Communication alters the mutual cognitive environment of the audience and communicators. Mutual manifestness may be of little cognitive importance but it is of crucial social importance. A change of the mutual cognitive environment is a change in their possibilities of interaction” (Sperber and Wilson, 1996: 61). But what exactly is manifestness, and what consequences does it have for a context-dependent conception of mutual knowledge and common ground? To use Sperber and Wilson’s (1996: 39) own words: To be manifest, then, is to be perceptible and inferable. An individual’s total cognitive environment is the set of all the facts that he can perceive or infer: all the facts that are manifest to him. An individual’s total cognitive environment is a function of his physical environment and his cognitive abilities. It consists not only of all the facts that he is aware of, but also all the facts that he is capable of becoming aware of, in his physical environment. The individual’s actual awareness of the facts, i.e. the knowledge that he has acquired, of course contributes to his ability to become aware of further facts. Memorised information is a component of cognitive abilities. Relevance theory extends the domain of reference of manifestness from facts to all assumptions. As a consequence of the extended frame, manifestness requires a scalar interpretation, invoking different degrees of manifestness. Regarding its semantics, manifest is weaker than know, and manifest is weaker than assume. In order to communicate in a felicitous manner and to avoid coordination problems, communicators presuppose and construct mutual cognitive environments. Thus, relevance theory assumes a cognitive context which is assigned the status of a necessary condition for fleshing out the explicatures of a communicative contribution which are required for the calculation of the relevant implicatures. Only then can the contribution be made optimally relevant. Other theories of context which have been formulated in the fields of cognitive science, linguistics, pragmatics, sociopragmatics and sociolinguistics (Akman et al., 2001; Bouquet et al., 1999; Duranti and Goodwin, 1992; Fetzer, 2002, 2004) consider context as an unbounded entity which requires delimitation. Against this background, it categorizes into cognitive context, linguistic context, sociocultural context and social context. Analogously to the difficulty of accommodating infinite regress in a definition of common ground, context suffers similar delimitation problems. That is to say, context cannot really be delimited and therefore cannot

4 Lexical Markers of Common Grounds be described completely, as has already been explicated by Wittgenstein (Wittgenstein, 1978). Depending on the frame of investigation, context is delimited to the global surroundings of the phenomenon to be investigated, for instance to an institution or to a sociocultural setting, in which the discourse is taking place, or it is delimited to a communicative genre, of which the phenomenon to be investigated is a constitutive part. Here, context is not restricted to linguistic material, which is referred to as linguistic context or co-text. It also includes social and cultural phenomena, and is referred to as social context and as sociocultural context. Prototypical social-context constituents are participants, their psychological dispositions, their social and discursive roles, the physical location of the discourse, time and preceding discourse. From a sociocultural-context perspective, these universal constituents are given a culture-specific interpretation and are reframed, such as a more individualistically or a more collectively oriented conception of a participant, and a more polychronically or a more monochronically oriented conception of time. In its narrow definition, context is delimited to the local (or immediately adjacent) surroundings of the phenomenon to be investigated and can refer to the immediately adjacent surroundings of a phoneme, morpheme, phrase or lexical item, sentence or utterance, for instance. It can also refer to the participants and their immediately adjacent surroundings, and to the setting and its immediate surroundings. If the surroundings are cognitive material, for instance a proposition, a mental representation or an assumption, they are called cognitive context. If the surroundings are of an extra-linguistic nature, that is non-cognitive and non-linguistic material, they are called social context, and if the surroundings are language material, they are called linguistic context. Both context and common ground are dynamic concepts as there is always new information and thus new propositions to be added to the common ground or to the context. To sum up, common ground can be differentiated with regard to a number of perspectives. First, it is seen as comprising a set of true propositions which serve as a resource for the understanding of utterances. Regarding its function, participants presuppose its validity and fall back on it when they retrieve implicatures. In that frame of reference, common ground is looked upon as objective and for this reason, it represents a context-independent category, which is true in all possible contexts. In the truth-conditional approach, common ground is implicit. It can be explicated via true propositions, thus spelling out its underlying presuppositions. In order to describe the underlying mechanisms regarding the joint construction of a shared basis of true propositions, the notion of grounding plays a key role. In contrast, in an integrated frame of reference, grammatical constructions and lexical expressions encode not only propositional or factual information, but also discursive information anchored to the textual and sequential organization of discourse and interpersonal information anchored to the interpersonal and illocutionary domains. The extension of frame requires a multifunctional approach to grammatical constructions and a multifunctional approach to lexical expression, where propositional, discursive and interpersonal meanings are realized and expressed in a simultaneous manner. Against this background, communicative contributions contain anaphoric, cataphoric and generic grounding devices (Givón, 1993). Communicative contributions therefore not only create common ground (Lambrecht, 1994), but at the same time indicate to which domain of common ground they are to be attributed. Thus, the act of grounding is intrinsically connected with information management, with the participants’ common ground and with the administration of common ground, but what is being grounded, and more precisely, where is it being grounded?

Introduction 5

3. GROUNDING AND DIFFERENT TYPES OF COMMON GROUND Common ground has been examined on the one hand as a set of true propositions, which serve as the foundation for reasoning and retrieving implicated meaning. To this set, not only new information is added in communication but the information contained is also updated in the communication process. Here, researchers have taken a product perspective. In the research paradigms of interactional sociolinguistics and ethnomethodology on the other hand, common ground is looked upon from a process viewpoint and thus is constructed and reconstructed in and through the process of communication. Here, relevant variables – for instance solidarity, power, ethnicity and gender – are seen as jointly constructed in interaction by a particular employment of the linguistic code and by a particular employment of a social practice (Garfinkel, 1994; Gumperz, 1977, 1992). To bridge the gap between a conception of common ground as either process or product, common ground has to be assigned the status of both product and process, such that it is not only presupposed in communication but also interactionally organized. So where do the different conceptions of common ground meet, and where do they depart? All of the research paradigms examined above share a conception of common ground as a presupposed common knowledge base which is required for felicitous communication. To that base, participants anchor their communicative contributions and they fall back on it, should they require further information which may not be explicitly encoded. So, common ground is implicit and presupposed. Furthermore, the majority of conceptions of common ground share some kind of update operator which administers common ground. But there is no general agreement with respect to what type of information common ground contains. While the philosophy-of-language and linguistic conceptions of common ground tend to favour a factual knowledge-based type of common ground, cognitive science employs factand assumption-based formats. However, in order to be felicitously integrated into a theory of natural language communication, the prevailing concept of common ground needs to be delimited and contextualized with respect to the basic pragmatic premises of intentionality (Austin, 1980; Searle, 1969) and current purpose (Grice, 1975). Thomason’s notion of conversational record (Thomason, 1992) is closely connected with the philosophy-of-language concept of common ground: it contains factual information since it is public, and its construction takes place in the intentional exchange of communicative acts. In Thomason’s framework, speaker meaning is equivalent to a sort of coordination-oriented intention. It is intrinsically connected with Lewis’s concept of accommodation (Lewis, 1979), which regulates the removing of obstacles to achieve the desires and goals participants attribute to others when they communicate. This is due to the fact that speaker meaning is neither produced nor interpreted in isolation. Rather, it is anchored to a conversational record which contains (1) information that is public, (2) presumptions and (3) an update operator. Regarding the nature of the connectedness between common ground and common grounds, Thomason’s concept of conversational record can be considered as an important context-dependent subset of common ground. It allows for both the accommodation of new information and a commonly shared basis. The former is due to its update operator, while the latter is due to its status as public information. But do conversational records and common ground really contain public information only?

6 Lexical Markers of Common Grounds Clark’s approach to communication is both social and cognitive. Correspondingly, his conception of common ground is also both social and cognitive, or to use his own words: “two people’s common ground is, in effect, the sum of their mutual, common, or joint knowledge, beliefs and suppositions” (Clark, 1996: 93). An interpersonal outlook on language and language use requires an extended frame of investigation with common ground as one of its constitutive parts. In such a framework, common ground can no longer be conceived of as consisting of one domain only. Instead it must be differentiated into (minimally) two interacting dynamic domains, namely a participant’s personal common ground where her/his subjective and privileged experience is stored, and a participant’s communal common ground where social or joint experience is stored. Only then is it possible to cooperate and coordinate actions and perform joint actions. The two types of common ground are interconnected, constantly updated and, if necessary, revised. The concepts of conversational record (Thomason, 1992) and of personal and communal common ground (Clark, 1996: 110ff.) are further refined with regard to their participant orientation and their embeddedness in context. Fetzer (2002, 2004) examines particular subsets of common ground, namely context-dependent dialogue common grounds. Dialogue common ground is a dynamic sociocognitive construct, which differentiates into an individual dialogue common ground, that is a participant’s representation of her/his dialogue common ground, and a collective dialogue common ground, that is a participant’s representation of the set of participants’ representation of dialogue common ground. The former contains and administers the information of one individual’s dialogue regarding force, content, formulation, and interpersonal and interactional presuppositions. It thus captures an individual’s processing and interpretation of information. The latter contains and administers the ratified information of the collective. In communication, the participants’ individual dialogue common grounds may differ, sometimes to a large extent. In addition to a participant’s construction of an individual dialogue common ground, he or she also constructs a collective dialogue common ground which is anchored to the set of participants and to the set of their ratified contributions. The collective dialogue common ground contains and administers set-specific information, such as collective we-intentions, collective communicative goals, collective inferencing strategies and collective coherence (Fetzer, 2004). The set-specific values function as a filter, which canalizes what-has-been-said accordingly by guiding the interpretation and production of intersubjective meaning in a prespecified manner. Collective dialogue common ground intersects with individual dialogue common ground and with other individual and collective dialogue common grounds. In natural language communication, the individual and collective dialogue common grounds are permanently updated, and all of the postulated contributions and their presuppositions are allocated to the individual dialogue common grounds. Correspondingly, all of the ratified contributions and their ratified presuppositions are allocated to the collective dialogue common grounds, which are expected to be almost identical for the participants because of the necessary condition of ratification. Naturally, the collective dialogue common grounds are permanently updated. Furthermore, the collective dialogue common grounds serve as a foundation for the participants’ inferencing processes. Because of their set-specific status, ratified presuppositions are assigned a co-suppositional status. Analogously to the differentiation between individual dialogue common ground and collective dialogue common ground, presuppositions and co-suppositions subcategorize along similar lines. With regard to the validity of dialogue common ground, it is only valid in the restricted domain of a particular

Introduction 7 dialogue, but may, if strengthened through ratification in other dialogues, become valid in an extended frame of reference. Unlike presuppositions, co-suppositions administer information about something which is the case and about something which is not the case. Against this background, co-suppositions are only valid in the restricted domain of a particular dialogue, but may, if strengthened through ratification in other dialogues, become valid in a more extended frame of reference. Thus, a necessary condition for the allocation of a contribution and its presuppositions to a collective dialogue common ground is their ratification. Because of the dynamic conception of common ground and its subcategorization into individual and collective dialogue common grounds, participants are constantly involved in displaying their understanding of the current status of the relevant common grounds. Functional and cognitive approaches to language and language use understand language and language use as interconnected and as firmly anchored to context (Clark, 1996; Givón, 1993). In communication, participants do not simply exchange information but rather realize their communicative contributions by encoding information in particular linguistic surfaces which are composed of a finite number of lexical expressions and of a finite number of grammatical constructions. Since communication is a context-dependent endeavour par excellence, the information encoded can never be completely new but it is always connected with prior information which is a constitutive part of the participants’ common ground, and with new information which is not yet part of their common ground. Grounding is thus intrinsically connected with common ground and with the participants’ administration of common ground. Regarding its cognitive status, the act of grounding is anchored to a logic of upward completion and downward evidence: “Levels of action form what I have called action ladders, which have the properties of upward causality, upward completion, and downward evidence” (Clark, 1996: 389), which is also found in abductive processing of information, in abductive processes of hypothesis formation and in abductive reasoning (Givón, 1989; Levinson, 1995). The contextualization and delimitation of common ground requires the fundamental premise of communication to be conceptualized in a dynamic network-based frame of reference which accommodates the act of grounding. However, the act of grounding is more than a simple update operation as it administers not only the process of adding new information to common ground, but rather the addition of new information to the context-independent category of common ground and to the context-dependent category of different types of common grounds, which may be anchored to individual participants and their assumption of what they consider to be part of their common ground, or to individual discourses or dialogues. A dynamic conception of common ground thus requires the accommodation of a permanent negotiation of common situations in which participants display to each other – and to possible audiences – what they consider to be the common ground of their interaction. In this approach, participants therefore not only display their understanding of each other’s contributions, but are also involved in construing and displaying the assumed shared basis. One such mechanism consists of the use of particular lexical markers that serve speakers as presentations of what they assume to be common ground.

4. LEXICAL MARKERS In communication, participants exchange information by encoding it in lexical expressions and grammatical constructions in a methodical way (Garfinkel, 1994), and all linguistic choices

8 Lexical Markers of Common Grounds can be understood as related to context in one way or other. With respect to paralinguistic cues, Gumperz has introduced the notion of contextualization cue, which is anchored to the field of interactional sociolinguistics: They serve to highlight, foreground or make salient certain phonological or lexical strings vis-á-vis other similar units, that is they function relationally and cannot be assigned context-independent, stable, core lexical meanings. Foregrounding processes, moreover, do not rest on any one single cue. Rather, assessments depend on cooccurrence judgments (  ) that simultaneously evaluate a variety of different cues. When interpreted with reference to lexical and grammatical knowledge, structural position within a clause or sequential location within a stretch of discourse, foregrounding becomes an input to implicatures, yielding situated interpretations. Contextualization cues enter into the inferential process at several degrees of generality. Minimally, it is necessary to recognize three distinct levels. Firstly, there is the perceptual plane (  ). They serve to provide information on such matters as possible turn-construction units (  ). The second level is that of local assessments of what conversational analysts call “sequencing” and what from a pragmatist’s perspective one might refer to as “speech act level implicatures” (  ) what I have called “communicative intent”. (  ) Third, there is the more global level of framing. (Gumperz, 1992: 232, 233) Analogously to the discursive function of paralinguistic cues which organize discourse both locally and globally by indicating the nature of the connectedness between a communicative contribution and context, a lexical marker contributes to organizing common ground by indicating the nature of the connectedness between particular information communicated by a particular piece of discourse with the discourse common ground and with other types of common ground. Lexical markers connect not only the domains of declarative knowledge and procedural knowledge, but also the domains of explicit knowledge or Thomason’s notion of public information, which can easily be shared, and implicit knowledge or individual dialogue common ground knowledge, which is highly personal. As feedback signals, they may even contribute directly to the grounding of the conversational record. Recurrent background is of immediate relevance for the lexicalization and grammaticalization of new meaning (Lehmann, 1995; Traugott, 1995; Wischer and Diewald, 2002), and recurrent background is at the origin of new meaning. It constitutes a necessary condition for the creating of new lexical expressions or new grammatical constructions, as is also explicated by Vanderveken: “For recurrent non literal meanings in recurrent forms of life of background tend to be lexicalized or realized syntactically after a while” (Vanderveken, 2002: 59). For a lexical expression to be assigned the status of a lexical marker, recurrent background and repeated acts of grounding are required. Because of its high frequency in a particular context, the lexical expression which is to be assigned the status of a lexical marker acquires the additional meaning of an instruction to administer a particular piece of discourse information in a particular domain of a particular common ground. Thus, a lexical marker may be multifunctional: it may carry its original semantic meaning while at the same time providing an administrative instruction. In the socio-cognitive framework, language is used deliberately in an interpersonal setting for social purposes. This is reflected in Clark’s relational conception of meaning which

Introduction 9 is communicated by the participants’ intentional and goal-directed employments of signals: “Signals are built on signs that speakers deliberately create for their addressees – words, gestures, noises, and more” (Clark, 1996: 156). Moreover, Clark postulates the existence of a communal lexicon which is assigned a prime function in the construction of communal common ground. But is communal common ground constructed through language or through language use; is it constructed by utterances or by sentences? Clark provides us with the following answer: “Utterances, however, are not sentences. Recall that signs are types, and they signify types of things, not individual things. Whereas sentences are entirely symbolic, utterances of sentences can never be, because they are particular occurrences and are used to refer to particular objects, states and events” (Clark, 1996: 161, 162). Utterances therefore combine different methods: “In conversation, most utterances are composites of the three methods – describing-as, indicating, and demonstrating – not just one or two” (Clark, 1996: 183), which he schematizes with regard to the basic communicative functions of linguistic and non-linguistic signals, which he calls methods, with regard to the linguistic and non-linguistic signs created, with regard to the cognitive memory resources and with regard to their basic cognitive processes as follows: Method

Sign created

Memory resource

Basic process

describing-as

Symbols

activating rules

indicating

Indices

demonstrating

Icons

mental lexicon grammatical rules representation of spatial, temporal surroundings memory for appearances

locating entities imagining appearances

Source: Clark (1996: 184).

The descriptive function of language relies on the employment of linguistic symbols which are retrieved from the mental lexicon. They refer to objects, states, and events (Clark, 1996: 162). Indices, in contrast, express relational meaning. They serve to indicate particular spatial and temporal entities and therefore are of immediate relevance to the construction of personal and communal common grounds. Icons serve to demonstrate appearances. They map onto collocational and idiomatic expressions and evoke particular cultural meanings; without these contextual frames, the meaning of icons would be infelicitous. While symbols activate rules and provide a default common ground, indices locate entities thus signifying a particular common ground. It is the interface of symbols and indices where the functional category of lexical marker is allocated to. Lexical markers express relational meanings which are calculated with regard to the marker’s connectedness with a contribution and its proposition, force, and local and global contexts.

5. LEXICAL MARKERS OF COMMON GROUND(S) This volume presents original and up-to-date research on the connectedness between common ground and grounding on the one hand, and between different types of common ground on the other hand. It provides an interdisciplinary perspective on the two primarily socio-pragmatic

10 Lexical Markers of Common Grounds notions of grounding and common ground, and it intends to further our understanding of the complex processes involved in producing and interpreting lexical markers, and to illustrate their relevance and function in the administration of common grounds. The first part, “Grounding and Common Ground”, examines the relationship between grounding and the activation of common ground based on the analysis of the lexical markers ok, before and after, some modal particles and explicit metalinguistic operations in the contexts of face-to-face, computer-mediated, written communication and translations. The studies address the relationship between construing common ground by means of lexical markers, which guide the grounding process, and their references to assumed common ground. ˇ Sherri L. Condon and Claude G. Cech’s “Ok, Next One: Discourse Markers of Common Ground” examines the multifarious functions of the discourse marker ok in the context of ˇ computer-mediated discourse. Condon and Cech show that the discourse marker ok provides a strategy for grounding utterances with a minimal expenditure of resources. Furthermore, they illustrate the different functions of ok with respect to the activation of common ground, ˇ navigating between different phases in decision-making interactions. Condon and Cech’s research demonstrates that ok marks the common understanding that a previous decision routine has been completed and that the expected next decision routine is being initiated. Kerstin Fischer’s contribution “Grounding and Common Ground: Modal Particles and Their Translation Equivalents” investigates the function of German modal particles with respect to displaying common ground and contrasts them with discourse particles which are mostly involved in the grounding process. Her research is based on a translation corpus with professional and non-professional translations, which she uses to compare and contrast the different translations of modal particles. The contrastive perspective allows her not only to identify a broad spectrum of lexical (as well as grammatical) devices that function as markers of common ground; but the results also show that while modal particles are grammaticalized means to activate common ground, many other linguistic devices can fulfill similar functions. Another interesting overlap between discourse particles, generally involved in the grounding process, and modal particles, generally involved in the activation of common ground process, can be noticed: not only can the same items function both as discourse and as modal particles, but with respect to thematic structure, discourse particles also fulfill common ground activating functions. Karin Pittner’s “Common Ground in Interaction: The Functions of Medial Doch in German” analyses the functions of the medial particle doch in German with respect to grounding common grounds. She demonstrates that medial doch functions as a metapragmatic instruction to update common ground. Based on the differentiation between individual and collective common grounds, Pittner illustrates how the meaning of the stressed particle arises from the combination of the particle’s lexical meaning and the meaning of the focus accent. Her solution to the relationship between grounding and the activation of common ground is thus that modal particles function as update operators, activating a previous proposition and modifying it at the same time. Modal particles, in her account, have important functions with respect to both grounding and the retrieval of common ground. Carlos Rodríguez Penagos’s “A Common Ground for Knowledge through Knowledge of Language: A Computational Research of Consensus-based Meaning in Scientific Papers” investigates the grounding of common ground in the context of expert knowledge. He analyses Explicit Metalinguistic Operations whose function it is to introduce a new concept into a common knowledge space. He illustrates how scientific knowledge is constructed through

Introduction 11 metalinguistic speech acts and demonstrates the relevance of context-dependent tools with respect to building non-standard knowledge engineering resources which allow the accommodation of gradual changes. Thora Tenbrink’s “Imposing Common Ground by Using Temporal Connectives: The Pragmatics of Before and After” examines the different functions of before and after in spoken and written corpora. She demonstrates that subclauses introduced by these connectives present information as (if it was) common ground. In case the information is not shared, it is imposed onto the hearer. Thus, before and after may evoke common ground, or present new information as not in need of ratification, thus evading a grounding process. While subclauses are designed to activate common ground, they may contribute or rather impose new information. The author presents the particularized discourse contexts which are specific to individual markers and discourse configurations. The second part, “Common Grounds”, examines the relationship between the contextindependent notion of common ground and the context-dependent notion of particular common grounds for the lexical markers mais (but), maintenant (now), toujours (always), and the lexical strategy of reformulation in the contexts of media communication, literary discourse and face-to-face communication. François Nemo’s “The Pragmatics of Common Ground: From Common Knowledge to Shared Attention and Social Referencing” analyses pragmatic approaches to common ground by Grice, Anscombe and Ducrot, Sperber and Wilson and others. The author argues that speaking is a matter of attracting somebody’s attention to something and asking her/him to take it into account in a particular way. The role of lexical markers of common ground consists in managing attentional and scalar frames and in marking information as “business as usual”. Nemo further distinguishes between interlocutive, discursive and cultural common ground, which result from different kinds of ratification processes. Anita Fetzer’s “Reformulation and Common Grounds” examines the functions of the metapragmatic device of reformulation in the negotiation of intersubjective meaning and in the construction and administration of common grounds. She illustrates the different functions of self- and other-reformulations in media discourse and pays particular attention to the construction and reconstruction of individual dialogue common ground and of collective dialogue common ground. The author demonstrates that the meaning of a reformulation depends on its producer, sequential status and the connectedness between participants, communicative contribution, dialogue common ground and social, sociocultural and linguistic contexts. Moeko Okada’s “Whose Common Ground? A Misunderstanding Caused by Incorrect Interpretations of the Lexical Markers of Common Ground” investigates the construction of common ground in the context of a literary dialogue. Her analysis is based on relevance theory and focuses on misunderstandings. Because of its bounded frame of investigation and the corresponding bounded common grounds, the author can illustrate how individual participants construct their individual representation of what they consider to be common ground, how they strengthen their conceptions of common ground, and how they modify and reconstruct their representations of common grounds. Thanh Nyan’s “Common Ground, Categorization and Decision Making” adopts a co-evolutionary stance and argues for common ground to be constructed in terms of pre-existing means, which are intrinsically connected with encoded schematized situations, categorization and decision making. Her analysis is based on argumentation theory, Searle’s conception of Background and Damasio’s model for decision making. She illustrates the functions of the

12 Lexical Markers of Common Grounds argumentative markers mais, maintenant and toujours with respect to their role in suggesting a categorization of the situation as relevant for the conclusion in argumentative discourse. The author thus distinguishes between imposed common ground, schematized situations and background knowledge.

REFERENCES Akman, Varol et al. (eds), 2001. Modeling and Using Context. Third International and Interdisciplinary Conference, CONTEXT 2001, Dundee, UK, July 2001, Proceedings. Springer, Heidelberg. Austin, John L., 1980. How to Do Things with Words. Cambridge, Cambridge University Press. Bouquet, Paolo et al. (eds), 1999. Modeling and Using Context. Second International and Interdisciplinary Conference on Modeling and Using Context, Context’99, Trento, Italy, September 1999, Proceedings. Heidelberg, Springer. Clark, Herbert H., 1996. Using Language. Cambridge, Cambridge University Press. Duranti, Alessandro and Goodwin, Charles, (eds), 1992. Rethinking Context: Language as an Interactive Phenomenon. Cambridge, Cambridge University Press. Fetzer, Anita, 2002. Communicative intentions in context. In: Fetzer, A. and Meierkord, C. (eds), Rethinking Sequentiality: Linguistics meets Conversational Interaction. Amsterdam, Benjamins, 37–69. Fetzer, Anita, 2004, Recontextualizing Context: Grammaticality Meets Appropriateness. Amsterdam, Benjamins. Gardner, Rod, 2001. When Listeners Talk. Amsterdam, Benjamins. Garfinkel, Harold, 1994. Studies in Ethnomethodology. Cambridge, Polity Press. Givón, Tom, 1989. Mind, Code and Context. Hillsdale, Laurence Erlbaum. Givón, Tom, 1993. English Grammar: A Function-Based Introduction. Amsterdam, Benjamins. Grice, Herbert P., 1975. Logic and conversation. In: Cole, M. and Morgan, J.L. (eds), Syntax and Semantics. Vol. III. New York, Academic Press, 41–58. Gumperz, John J., 1977. Sociocultural knowledge in conversational inference. In: Saville-Troike, M. (ed.), Linguistics and Anthropology. Washington, Georgetown University Press, 191–211. Gumperz, John J., 1992. Contextualization and understanding. In: Duranti, A. and Goodwin, C. (eds), Rethinking Context: Language as an Interactive Phenomenon. Cambridge, Cambridge University Press, 229–252. Hymes, Dell, 1974. Foundations in Sociolinguistics: An Ethnographic Approach. Philadelphia, University of Philadelphia Press. Lambrecht, Knud, 1994. Information Structure and Sentence Form: Topic, Focus and the Mental Representation of Discourse Referents. Cambridge, Cambridge University Press. Lehmann, Christian, 1995. Thoughts on Grammaticalization. München, Lincom. Levinson, Stephen, 1983. Pragmatics. Cambridge, Cambridge University Press. Levinson, Stephen, 1995. Interactional bias in human thinking. In: Goody, E. (ed.), Social Intelligence and Interaction. Cambridge, Cambridge University Press, 221–260.

Introduction 13 Lewis, David, 1979. Scorekeeping in a language game. Journal of Philosophical Logic 8, 339–359. Saville-Troike, Muriel, 1989. The Ethnography of Communication. Blackwell, Oxford. Sbis`a, Marina, 2002. Speech acts in context. Language and Communication 22 (4), 421–436. Searle, John R., 1969. Speech Acts. Cambridge, Cambridge University Press. Searle, John R., 1995. The Construction of Social Reality. New York, The Free Press. Searle, John R., 1999. Mind, Language and Society: Doing Philosophy in the Real World. London, Weidenfeld & Nicolson. Sperber, Dan and Wilson, Deidre, 1996. Relevance: Communication and Cognition. Oxford, Blackwell. Stalnaker, Robert, 1999. Context and Content. Oxford, Oxford University Press. Thomason, Richmond H., 1992. Accommodation, meaning, and implicature: Interdisciplinary foundations for pragmatics. In: Cohen, P.R. et al. (eds), Intentions in Communication. Cambridge, MIT Press, 325–363. Traugott, Elizabeth, 1995. Subjectification in grammaticalization. In: Stein, D. and Wright, S. (eds), Subjectivity and Subjectivisation. Benjamins, Amsterdam, 31–54. Vanderveken, Daniel, 2002. Universal grammar and speech act theory. In: Vanderveken, D. and Kubo, S. (eds), Essays in Speech Act Theory. Amsterdam, Benjamins, 25–62. Wischer, Ilse and Diewald, Gabriele (eds), 2002. New Reflections on Grammaticalization. Amsterdam, Benjamins. Wittgenstein, Ludwig, 1978. Philosophical Investigations. Oxford, Blackwell.

This page intentionally left blank

Part I GROUNDING AND COMMON GROUND

This page intentionally left blank

Lexical Markers of Common Grounds Edited by Anita Fetzer and Kerstin Fischer © 2007 Elsevier Ltd. All rights reserved.

OK, NEXT ONE: DISCOURSE MARKERS OF COMMON GROUND ˇ Sherri L. Condon and Claude G. Cech

1. INTRODUCTION Researchers in discourse and pragmatics recognize that much of what is communicated in verbal interaction is not explicitly formulated in the words that are exchanged. Instead, the linguistic forms are linked to understandings in the cultural, personal, and textual context, to objects and events in the physical setting, and to the joint actions in which interlocutors are engaged. Clark (1996: 12) defines common ground among participants in interaction as the “knowledge, beliefs, and suppositions they believe they share.” Following Stalnaker (1978) and Lewis (1979), Clark asserts that the meanings and understandings associated with each speaker’s contribution to an interaction increase their common ground and provide an essential foundation for subsequent meanings and understandings that they will coordinate. All of the linguistic units in a discourse contribute to the shared understandings that are achieved in the discourse, and therefore to the common ground that participants maintain. However, some units such as deictic expressions seem to function more saliently in the processes by which participants coordinate their talk to accomplish joint activities. While most lexical items are structured in utterances that contribute propositional content to the common ground, some forms foreground additional presuppositions, attitudes, expectations, or relations that are otherwise tacit in the speech event. For example, greetings and terms of address such as “yo, dude!” reflect social relations and attitudes between speaker and recipient. Other forms seem to function primarily in strategies designed to ensure that understandings have been successfully achieved, such as use of “copy” to verify receipt and understanding of a radio message. These expressions can be called lexical markers of common ground. The functions associated with discourse markers (Louwerse and Hite Mitchell, 2003; Schiffrin, 1987) can be construed in terms of common ground, and several papers in this volume explore the effects of expressions that have been identified as discourse markers on the content or management of understandings in interlocutors’ common ground. Fischer (in this volume) suggests that discourse particles like yeah, uh huh, and okay provide evidence of understanding that contributes to the joint verification of what is being said. But as Fischer and others in this volume (e.g. Nemo, Fetzer) point out, understanding what is said includes understanding how it is to be construed in the current activity of the speech event. At the same

18 Grounding and Common Ground time, this current situation is itself jointly established by the participants through what is said. Both Fischer and Nemo (in this volume) use the term framing to describe the process by which a discursive common ground emerges and becomes a resource for interpreting participants’ contributions to the interaction. Fetzer’s notion of collective dialogue common ground (in this volume) seems to describe the result of this process, though we will continue to use the simpler term common ground to refer to participants’ shared understandings of their interaction. We focus on the discourse functions of the marker ok, which resembles other interjections like yeah and uh huh in several ways. All of these forms serve a variety of functions in interaction, they lack specific denotations or propositional content, and they are subject to few syntactic constraints. Yet evidence shows that they occur frequently in conversation and that their functions are significant to the verification and framing processes that determine the success of speech events. The study reported here investigates whether the same discourse functions that have been associated with ok in face-to-face interaction also occur in computer-mediated interaction. Research on computer-mediated interaction often focuses on the many differences between face-to-face and computer-mediated environments (Baym, 1996; Bordia, 1997; Ferrara et al., 1991; Garcia and Jacobs, 1999; Herring, 1996; Oviatt et al., 1998). In computer-mediated interaction, participants do not share a physical context, and the language is written instead of spoken, to mention a few of the many differences that may occur. Consequently, many resources that can be exploited in face-to-face interaction are not available in computermediated interaction, including facial expressions, gestures, body language, prosody, and the timing of turns (see, for example, Clark and Brennan, 1991; Whittaker, 2003). As a result, participants in computer-mediated interaction must rely more on linguistic conventions to establish and manage understandings in their common ground. Though they can be expected to modify their encoding strategies to adapt to the environment (e.g., Clark and Brennan, 1991), we do not expect them to abandon conventions that can be productive in computer-mediated contexts. In previous research with participants solving the same decision-making tasks in faceto-face and computer-mediated contexts, we found identical strategies for decision-making across media. Participants formulated the same kinds of utterances organized into the same routines. However, participants in computer-mediated interaction relied more strongly on the decision routines, with higher proportions of utterances serving routine functions (Condon ˇ and Cech, 1996a, b; Condon et al., 1999). They also relied more extensively on explicit ˇ language to manage the interaction (Condon and Cech, 2001). But do participants in computermediated interactions use discourse markers to signal common ground? This study examines the discourse functions of ok in sixty computer-mediated decision-making interactions to verify that ok is used in the same contexts in both communication environments. In the next section, we describe the discourse functions of ok that have been identified in face-to-face interaction and the role of ok in managing the understandings that are established in the common ground of interlocutors. Then we describe the methods used to record decisionmaking interactions in face-to-face and computer-mediated environments. We statistically compare the occurrences of ok in the two environments by presenting the discourse functions of utterances that immediately precede and follow discourse ok in the interactions. Finally, we compare the results to other quantitative studies of ok in face-to-face interaction and discuss the relatively infrequent use of ok in computer-mediated interactions. Discussion of the results focuses on their implications for grounding.

Discourse Markers of Common Ground 19

2. DISCOURSE OK AND COMMON GROUND According to Clark (1996), the joint activity of successful interaction requires positive evidence that the common ground has been appropriately incremented. He asserts that every contribution to the discourse by one conversational partner requires positive evidence from the other conversational partner that it was understood, and he describes four main classes of positive evidence that participants in interaction can provide to verify that they understand what has been communicated. Presuppositions of understanding occur when participants respond appropriately to their partners’ contributions by taking up the joint project proposed. As Schegloff (1982) observes, each time a conversational partner responds to the previous contribution without taking the opportunity to repair any perceived misunderstandings, it can be inferred that no misunderstandings are apparent. We call this default verification because participants can assume understanding as long as they satisfy each other’s expectations for the current state of the interaction. Understanding is the default assumption in the absence of evidence to the contrary. Assertions of understanding occur when participants use acknowledging signals such as yeah, uh huh, and I see. We use the term backchannel verification to emphasize the backgrounded quality of these contributions. Schegloff (1982) argues that these particles claim understanding, but they do not actually demonstrate it. Nevertheless, backchannel verification is an effective strategy because it is not tied to the content of the communication. Lacking syntactic or semantic constraints, backchannel acknowledging expressions are all-purpose devices for verification that are independent of the particular understandings achieved. Clark also distinguishes displays of understanding and exemplifications of understanding. He explains that answers to questions display how the questions are construed, which is “more valid” (1996: 229) evidence of understanding than asserting or presupposing understanding. Exemplifications of understanding such as repetition and paraphrase also allow participants to check for an acceptable construal. In general, stronger evidence of understanding will be more closely tied to the specific understandings achieved, so that demonstrating understanding is less likely to be accomplished with all-purpose devices like backchannel verifiers. However, Schegloff (1982) suggests that forms like yeah and uh huh do seem to show, rather than merely claim, understanding when they are used as continuers. When these forms are used to pass instead of taking a full turn while the conversational partner is narrating a sequence of events, they show understanding by identifying the current state of the discourse as one at which passing on a turn is appropriate. The discourse use of ok is another example of a minimal form that lacks propositional and syntactic structure and serves to demonstrate understanding in a variety of contexts. Descriptions of ok identify the occurrence of the form at significant, yet expected, transitions in the discourse. Discourse ok marks transitions of topics (Bangerter and Clark, 2003; Bangerter et al., 2004; Beach, 1990; Hopper, 1989; Schegloff, 1986), including the initial topic in phone conversations (Schegloff, 1968, 1979). It also occurs as callers shift to closure of phone conversations (Bangerter and Clark, 2003; Bangerter et al., 2004; Schegloff and Sacks, 1973). Discourse ok marks shifts between stages of service encounters (Merritt, 1978) and transitions to the doctor’s next question in medical interviews (Beach, 1993, 1995). Discourse ok marks mutually verifiable properties of the interaction: the current state of the discourse is identified as a confluence of structural boundaries and as an expected transition. Therefore, ok

20 Grounding and Common Ground is maximally informative about the speaker’s construal of the previous talk, while consuming minimal resources and avoiding an explicit representation of what is understood. Bangerter and Clark (2003; Bangerter et al., 2004) view the functions of discourse ok in terms of navigating joint projects. They observe that joint projects are typically organized hierarchically into subprojects that must be accomplished in sequence to achieve the goal of the joint project. They provide evidence that markers like ok and alright are used to navigate vertically among subprojects and from lower-level subprojects to higher subprojects and projects. In contrast, markers like yeah and uh huh are used to navigate horizontally through the sequences of joint actions within subprojects. Bangerter and Clark’s corpora, like ours, are elicited by asking participants to engage in interactions with specific goals, so that the occurrence of ok in our corpora can also be described within their framework of horizontal and vertical navigation through a hierarchy of goals. We examine occurrences of the discourse marker ok in face-to-face decision-making interactions that require participants to plan events such as a party or getaway weekend. Consequently, completion of the task requires a sequence of decisions to specify the details of the events that are planned. These decisions are organized by a discourse routine in which an utterance orients the decision by formulating a goal for the decision, which is followed by a suggestion and agreement. The excerpt in (1) from Condon (1986) illustrates a decision routine that occurs at the beginning of an interaction and the use of ok to initiate the interaction. (Language in face-to-face interactions was transcribed verbatim, but no attempt has been made to record intonation, stress, or prosodic information except that question marks are used to indicate question intonation and ellipses represent pauses.) (1) a. b. c. d. e. f.

Father: ok (long pause, paper shuffling) ok    two whole weeks Mother: ok Teenager A: Hawaii Mother: Hawaii ok Teenager B: Hawaii Father: for all fourteen days?

Here (1) illustrates how tightly the routine binds the interpretation of the utterances. The father’s orientation in (1a) is for the entire project. An answer sheet that the family was required to complete identified at least twenty eight subprojects because each day of the two-week vacation was divided into a space for a location and a space for activities. When the teenager formulates the suggestion in (1c), the default understanding is that the suggestion is a proposal that satisfies the orientation in (1a), and this relation is so strong that the father seeks to verify it immediately. We have claimed that adjacency pairs are special cases of routines and that utterances linked in routines have the same relations associated with adjacency pairs: each utterance in the sequence projects an expectation for the function of the subsequent utterance, and this function determines the interpretation of the latter. As a result, sequences of utterances in routines require relatively little linguistic form to encode their meanings. In (1) only Hawaii is needed to encode the suggestion and to express the agreements that result in the decision, which is presupposed in (1f). Routines rely on a common ground of expectations and understandings that make it possible to accomplish joint projects with minimal verbal coordination.

Discourse Markers of Common Ground 21 The use of ok in (1a) is clearly different than the use in (1d) to accept the suggestion in (1c). Whereas the latter occurs in response to the suggestion, the former is not preceded by any talk except a sound check for the tape recorder. (1a) is a clear example of the use that we call discourse ok, marking a significant, yet expected, transition in the interaction. The beginning of the interaction is a significant structural boundary for the discourse, and it is expected that the family will initiate the decision-making project. Bangerter and Clark (2003) speculate that when ok occurs with no apparent first pair-part, there is actually an implicit pair-part that is presupposed and that frames the interpretation of the subsequent utterance. In (1), the first pair-part might be something like “let’s do the project we have been asked to complete.” We agree that the understandings associated with discourse ok seem to be the same as the understandings associated with ok when it occurs as a second pair-part expressing consent or compliance. Frequently, ok follows a request for action such as closing a door for which ok alone does not satisfy the request. Instead, ok indicates that the request was understood and that the speaker intends to act in a way that satisfies the expectations of the request by performing the requested action. Similarly, when ok expresses consent, the speaker indicates that the request for permission or the suggestion for future action was understood and that the speaker intends to behave in a way that is expected for someone who has given consent or who has agreed to a suggestion. This logic allows ok to function similarly, even when it responds to an utterance that is not a clear request or suggestion. Goodwin (1987: 211) provides a good example, which is presented as (2). (2)

a. Don: I’ll go get some more water. (Leaves with pitcher) b. John: Okay.

By using ok in (2b), John affirms that (2a) was understood and that his subsequent action will conform to their expectations for appropriate action in Don’s absence, including anticipation that Don will return with water. There does not seem to be a need to hypothesize an implicit first pair-part for ok to signal understanding of the current state of the discourse and the intent to act in a way that satisfies the expectations for that state. We would argue that it is exactly this function of ok that makes it a lexical marker of common ground which is used not simply in response to a previous utterance, but in response to all of the framing and incrementation that has led to the current common ground of the speech event. Consequently, ok provides an effective strategy for vertical navigation among subprojects: as ok initiates the next open subproject, it can be construed to signal that the speaker understands the position of the transition in the hierarchy of subprojects that has emerged. There are many task-oriented interactions presented by Bangerter and Clark (2003) to illustrate the use of ok for vertical navigation among subprojects; other illustrations may be found in Beach’s (1995) examples from medical interviews in which doctors use ok at transitions among questions in the interview and at transitions among larger subprojects, such as movement from the patient’s medical history to details of the patient’s current complaint. The decision-making interactions in our corpora are additional examples of task-oriented speech events that are organized in subprojects. Planning events such as vacations, parties, or award ceremonies requires many individual decisions that participants accomplish using the

22 Grounding and Common Ground decision routine. The frequent occurrence of ok at the beginnings of these routines illustrates the use of ok to navigate subprojects by marking the common understanding that the previous decision routine has been completed and that the expected next decision routine is being initiated. An example of several decision sequences in face-to-face interaction is annotated in (3) with three instances of discourse ok highlighted in boldface type. (3) a. P1: ok then (orients suggestion) we could walk down scenic uh tourist district you know and go and sightsee in the French Quarter area (suggests action) b. P2: ok (agrees with suggestion) c. P1: ok so (discourse marker) we’ll all see Jackson Square (suggests action) d. P2: yeah (agrees with suggestion) e. P1: and then (orients suggestion) we’ll take a walk (suggests action) that’ll be pretty much our morning (suggests action) f. P2: well (discourse marker) g. P1: breakfast at Cafe du Monde (elaborates, repeats) h. P2: yeah (acknowledges only) and then (orients suggestion) go through Jackson Square and the–uh–Riverwalk (suggests action) i. P1: and the Riverwalk right (agrees with suggestion) ok then (orients suggestion) we’ll eat lunch at Mospero’s (suggests action) The annotations illustrate the coding system used to record the functions of utterances in decision-making interactions. In each case, ok follows the agreement that completes a previous routine and precedes the orientation or suggestion that initiates the next routine. In previous work, we have used the coding system to quantify the stability and frequency ˇ of the decision-making routine (Condon and Cech, 1996a, b; Condon et al., 1999) and the occurrence of ok at the beginnings of decision sequences (Condon, 2001). In face-to-face decision-making interactions, discourse ok was frequently followed by utterances that initiate decision routines: utterances coded as Orients Suggestion, as in (3a, i), those coded as Suggests Action, as in (3c), and expressions coded as Discourse Marker, as in (3c). The sequence ok + so + Orients Suggestion was frequent in the data, and more examples are provided in (4). (4)

a. b. c. d.

ok ok ok ok

so so so so

in the afternoon first day we there we    maybe go sightseeing what are we doing in the afternoon? where do you want to go Saturday night? after they’ve eaten what do we do?

This paper addresses the question whether the discourse functions associated with ok in face-to-face interaction can also be found in computer-mediated interaction. We do not anticipate that the function of ok would be different in computer-mediated interaction, but a previous attempt to verify this expectation (Condon, 2001) using an initial corpus of sixteen face-to-face and sixteen computer-mediated interactions was unsuccessful because there were not enough occurrences of discourse ok to allow confidence in the results. For the present study, we analyzed the use of discourse ok in an additional sixty computer-mediated decision-making tasks.

Discourse Markers of Common Ground 23

3. METHODS Data were collected for this study by asking participants to engage in a task that requires cooperative decision-making. Although the interactions are not spontaneous, participants who complete the task achieve all the understandings required to successfully engage in problem solving. Despite the laboratory setting and the “make-believe” nature of the task (participants will not really go on trips, nor produce a TV show), we believe that participants must employ the same processing strategies that they use in other, more spontaneous decision-making events in order to achieve the proper understandings in their joint planning. By focusing on the understandings necessary to accomplish that planning, we are able to observe the variety of communicative strategies that emerge and to compare the influence of factors such as machine mediation on those strategies. As Bangerter et al. (2004) note, this method allows us to focus on the entire encounter rather than short sequences. All participants in the study were students enrolled in psychology courses at the University of Louisiana/Université des Acadiens who received credit for their participation. The face-to-face corpus was generated by sixty mixed-sex pairs who jointly planned a social event (a picnic or barbeque) and an itinerary for a weekend trip. They sat together in the psychology lab at a table with a tape recorder and wrote their plans on an answer sheet that provided spaces for specific decisions: locations and activities for the morning, afternoon, and evening of each day for the weekend trip, and location, time, food, beverages, activities, and entertainment for the social event. They generally finished within an hour. The 120 participants in the computer-mediated interactions included sixty mixed-sex and same-sex pairs. They were asked to plan an MTV Video Awards show in which music videos are nominated for various categories and the winners are announced by celebrities during a ceremony that includes musical performances, clips of the nominated videos, and so on. All participants were provided with the same description of the problem, including a list of the decisions they should include, such as the award categories, nominees, winners, and presenters. We will refer to this task as the MTV task. Participants were allowed two hours to complete the task and completion times ranged from one to two hours. Some general findings from these two tasks have already been published elsewhere ˇ ˇ (e.g., Cech and Condon, 1998; Condon and Cech, 1996a, b). For the past several years, we have used the MTV task to systematically explore discourse management strategies by observing the effects of changes in the computer-mediated interface that might be expected ˇ ˇ to influence various cognitive factors (Cech and Condon, 2004; Condon and Cech, 1996a, b, 2001). Although there is a risk that differences in use of discourse ok across conditions ˇ might be due to task rather than medium, we believe that risk to be minimal: in Cech and Condon (1998), we compared results for computer-mediated versions of the social events tasks and the MTV task, and found very similar proportions of discourse functions, despite the differences in tasks. Hence, we are comfortable examining the use of a specific discourse marker, discourse ok, in a large corpus we have obtained using a different problem than that presented to the face-to-face participants. At issue is the generality of function of this marker. Finding similarities of use across media and problems will provide further evidence for that claimed generality. Partners in the computer-mediated condition were each located in a separate room of the psychology department’s computer lab. They used software designed and implemented by the second author so that the computer-mediated environment would be similar to oral interaction

24 Grounding and Common Ground within the constraints of our methodology. For example, participants sent messages to their partners by pressing the ENTER key after they had composed the text, though a word-by-word transmission would more closely approximate oral processing. We opted for well-defined turns in part because oral exchanges are maximally comprehensible when structured in sequential turns and in part because we anticipated future studies involving a methodology of capturing and changing messages. The computer screens are arranged with a message area in the top portion, a status area in the lower portion, and a brief description of the MTV problem which remains at the bottom of the screen throughout the interaction. To reduce message permanence, only one partner’s message can appear on the screen at a time. Therefore, messages disappear as soon as participants begin to reply, and if a participant sends a message while the partner is composing one, the interrupting message replaces the composer’s message, though the incomplete message remains in a buffer and need not be retyped. To assist participants in managing their turns and to prevent them from thinking that they had lost contact with the partner while the latter was typing, the software includes a status area. As soon as a partner starts typing, the recipient’s status area indicates that the partner is working on a message and will be sending it shortly. The status area text flashes until the message is sent to reassure participants that the system is active. We also systematically varied the size of the message text area in order to investigate the influence of screen size on messages in synchronous computer-mediated interaction. We were concerned that a relatively small message window might bias participants to adopt strategies that resembled the short turns of face-to-face interaction. Consequently, participants in the computer-mediated interaction were randomly assigned to one of three possible interfaces with different text window sizes. If participants initiate a new message immediately after sending one without waiting for the recipient to reply, the earlier message scrolls eventually (and irretrievably) off the screen to make room for the second message. Furthermore, the status area on the recipients’ screen changes color and informs the recipient that there is more of the message coming in. Therefore, though participants were limited to 4-, 10-, or 18-line text windows (a maximum of eighty characters per line), they could easily send as many transmissions in a row as they desired. Text windows were graphically framed, and several warning beeps informed participants when they had only one line left in which to complete their messages. In addition, participants could edit their entire text by backspacing. The software stored all keystrokes (including backspaces) and the times at which a message was started and sent, although the partner only received the corrected message. Finally, we provided a digital clock in the top right corner of the screen, as earlier experience had indicated that participants tended to lose track of time: we did not want to have to interrupt the interaction to warn participants that their time was nearly up. Further details of the interface and the results contingent on the message area ˇ size may be found in Cech and Condon (1998). Face-to-face interactions were transcribed from audio recordings into computer files ˇ (see Condon and Cech [1992] for coding and transcription conventions). All interactions were divided into utterance units, defined as single clauses with all complements and adjuncts, including sentential complements and subordinate clauses. Interjections and discourse markers like yeah, now, well, and ok are treated as separate utterances. Utterance units are annotated to reflect their discourse functions according to a scheme documented in the training manual for

Discourse Markers of Common Ground 25 coders. Coders were regularly tested using standard annotations prepared by the first author throughout training and subsequently using their coded transcripts. The coding scheme is best viewed as a measuring device, and it is not intended to make strong claims about interaction, though some assumptions are necessarily adopted. For example, it is assumed that utterances are multifunctional, which conflicts with the requirement of many statistical analyses that code categories be mutually exclusive. In order to preserve some of the multifunctionality of utterances, they are classified in several broad groups of discourse functions. For example Move functions such as suggests action and requests information are separated from Response functions such as agrees with suggestion and answers request for information. The routine function orients suggestion is grouped in a third set of Other functions. The groupings make it possible to code orientations encoded in questions like (4b–d) as both orients suggestion and requests information. Another example is suggestions in which orientations are formulated as fronted adverbials as in (3a, e, h, i), which are coded as both orients suggestion and suggests action. Because utterances could sometimes be coded in several different categories within each group, categories or functions were arranged in a hierarchy for each of the three broad groups of discourse functions. In keeping with the focus on decision-making, categories that are likely to be more relevant to the decision-making are placed higher in the hierarchies. Coders assigned an utterance to the highest possible function within each of these three hierarchies. Each category is described briefly here. More complete descriptions can be found in Condon ˇ and Cech (1992, 1996b). Move functions are those which invite a Response. In fact, most Move functions are first pair–parts with obligatory second pair–parts. The highest Move function is suggests action and it corresponds to the suggestion/input function in the decision routine, as in (1c) and (3a, c, e, h, i). Suggestions are identified as any utterance that could be interpreted as satisfying the main goal of the task or a subgoal, which is probably expressed in a closely preceding orientation. Below that function is the category requests action for utterances that propose behaviors in the speech event. For example, most face-to-face interactions included requests concerned with recording answers on forms provided to participants, as in (5). (5)

a. write it in activities b. well list your two down there

Utterances coded as requests information seek information not already provided in the discourse, as in (4b–d), while utterances coded as requests validation seek confirmation or verification of information provided in the discourse. The final Move category, elaborates– repeats (3g), serves as a catch-all for utterances with comprehensible content that do not serve any other Move or Response functions. Frequently these are repetitions and utterances that support or comment on suggestions. Response functions generally are the second pair-parts of Move functions. The highest of these is agrees with suggestion (1d, e; 3b, d, i) for utterances that are interpretable as positive evaluations of previous suggestions and as having the consequence that the proposal formulated in the suggestion acquires the status of a decision. The next lower category, disagrees with suggestion, includes utterances that negatively evaluate suggestions and proposals which are incompatible with suggestions. This category also includes a few refusals to comply with requests. The category complies with request identifies utterances that indicate compliance

26 Grounding and Common Ground with any of the three types of requests, and the category acknowledges only (3h) was restricted to forms like yeah that acknowledge previous utterances and to repetitions or completions of a partner’s previous utterance. The Other functions combine categories designed to reflect discourse management strategies as well as two categories included to assess affective functions. As mentioned above the Other group includes orientations in the decision routine, which are coded as orients suggestion (1a; 3a, e, h, i; 4). The category requests/offers personal information identifies utterances in which participants discuss personal information or make other personal comments not required to complete the task. The jokes–exaggerates category includes utterances that inject humor. The highest Other function is discourse marker, which is used for a limited set of forms: ok, well, anyway, so, now, let’s see, and alright are the forms coded as discourse markers when they could not be interpreted as functioning in other ways. For example, coders were instructed to identify ok as a discourse marker only if it could not be seen as agreeing, complying, or acknowledging. The coding system used to annotate the face-to-face interactions differed from the system used to annotate the computer-mediated interactions with respect to functions that we label explicit management. Explicit management refers to meta-pragmatic utterances that manage the interaction by explicitly referring to the interaction, as in the messages in (6), which are taken from the computer-mediated interactions. (Messages from computer-mediated interactions are presented as they appeared to the conversational partner, except for line breaks and fonts: we do not attempt to normalize spelling and punctuation, nor do we call attention to departures from spelling and punctuation conventions.) (6)

a. b. c. d.

Actually we need to start figuring out who the presenters are too. Let’s move on to the opening and performing bands time is short. ok, now we need to decide another band to perform. how about if you pick two categories and I’ll pick two also making it a total of 4.

Explicit management is identified primarily by verbs that represent ongoing mental or verbal processes such as figure out, move on, decide, and pick in (6). Other verbs include mean, type, and say. Utterances like (6) are also coded as requests action. We observed so much explicit management in the computer-mediated interactions that we modified the coding scheme for the analyses of the computer-mediated interactions: Explicit Management became a fourth group of categories that included decision management, as in (6), transmission management, as in (7), explicit repair, and several other categories. (7) a. are you there? b. Do you find this computer is slow in sending the messages, or are we slow in typing??? c. O.k. but try not to type when I type if you are, cause it messes up my screen for some reason. The messages in (7) illustrate how explicit management can be identified by expressions that refer to the talk or the communication environment such as there in (7a), this computer

Discourse Markers of Common Ground 27 and the messages in (7b), and my screen in (7c). Because we only refer to the total numbers of utterances with explicit management functions in this paper, we do not describe each category in the revised group of Explicit Management functions, but complete descriptions can be found ˇ in Condon and Cech (2001). The most important difference between the coding scheme used for the face-to-face interactions and the scheme adopted for the computer-mediated interactions is illustrated in Figure 1 (several additional functions were later added to the Moves and Other categories, but as their use was quite low, we do not include them here). The coding scheme used for the face-to-face interactions employed only the first three columns of categories in Figure 1. For the computer-mediated interactions, the explicit management category was removed from the Other column, and utterances that functioned as explicit management were coded in one of the categories in the fourth Explicit Management column. The major difference between the two coding schemes is that the explicit management and orients suggestion categories are mutually exclusive in annotations of the face-to-face interactions, but not in annotations of the computer-mediated interactions. Consequently, in the coding scheme used for the face-to-face interactions, an utterance like (6c) that explicitly orients a specific suggestion would be coded only as explicit management, whereas in the coding scheme used for the computer-mediated interactions, (6c) was coded as orients suggestion in the Other column and decision management in the Explicit Management column. Therefore, the proportions of utterances coded as orients suggestion in the two types of interaction are not fully comparable, though they are close due to the fact that there were few explicit orientations in the face-to-face interactions. In contrast, because explicit management is higher than orients suggestion in the Other column used for the face-to-face interactions, all explicit management in those interactions is annotated as explicit management and can be compared to the total number of utterances in the computer-mediated interactions that were annotated in the Explicit Management column (excluding No Clear Management). Although we allowed utterances to be multi-functional, we did not require this of them: utterances that could not be assigned a clear function within any given hierarchy were labeled as having no clear function of the sort coded by that hierarchy (e.g., no clear Move; no clear Response, etc.).

Figure 1. Coding schemes for annotation of utterance functions

28 Grounding and Common Ground

4. RESULTS The sixteen face-to-face interactions produced a corpus of 4140 utterances, averaging 259 per discourse, while the sixty computer-mediated interactions consisted of 8096 utterances with an average 135 per discourse. Therefore, participants in computer-mediated interaction completed complex decision-making tasks in about half of the number of utterances used by participants in face-to-face interaction to complete simpler tasks. The decrease in the amount of language used to accomplish tasks in computer-mediated interaction is consistent with the results of the quantitative studies reviewed in Bordia (1997), which document decreases whether frequencies are compared by time or task units. To compensate for the varieties of interaction lengths, the data are presented as proportions. For the average proportions per interaction, the frequency of a function in an interaction is divided by the number of utterances in the interaction and the proportion for each interaction is averaged for all of the interactions in the corpus (face-to-face or computer-mediated). For the proportions per corpus, the frequency of a function in the corpus is divided by the number of utterances in the corpus. We specify which proportions are presented.

4.1. Horizontal and vertical transitions in the decision-making routine First, we observe that participants in computer-mediated interaction clearly adopt the decisionmaking routine that was identified in face-to-face interaction. An example of a sequence with several routines from a computer-mediated interaction is presented in (8). (8) a. b. c. d.

P1: who should win best alternative video. (orients, requests info) P2: Pres. of the United States (suggests, complies with request) P1: ok (agrees) P2: who else should we nominate. (orients, requests info) bush. goo-goodolls and oasis (suggests) e. P1: sounds good, […] (agrees)

Also, in the computer-mediated interactions, discourse ok initiates decision routines as in (9). (9) a. P2: Okay,now who do we want to perform the opening act? (orients, requests info) b. P1: ME (suggests, complies with request) c. P2: Cool! (agrees) Therefore, as anticipated, participants in computer-mediated interaction do adopt the same strategies observed in face-to-face interaction. However, the extent to which these strategies are relied on varies in the two communication environments. Indeed, there is a striking difference in the occurrence of ok coded as discourse marker in the face-to-face (ftf) and computermediated (cmc) communication, replicating the result cited in Condon (2001). Discourse ok appears a total of 227 times in the face-to-face corpus, but only 96 times in the computermediated corpus, despite the fact that the latter includes many more interactions. The average

Discourse Markers of Common Ground 29 frequencies of 14.2 discourse oks per oral interaction and 1.6 per computer-mediated interaction do not take into account the significantly larger sizes of the oral exchanges, but the difference is still large when we assess the average proportion of discourse oks per interaction (5.4% for ftf versus 1.2% for cmc), and the ratio of oks to the total number of words exchanged (1.3% versus 0.2% respectively). Does discourse ok signal the understanding that a decision has been achieved and the transition to the next decision sequence, as claimed? To assess this claim, we will first look at the transitions that discourse ok presumably anchors, and compare the discourse functions immediately prior to discourse ok with those immediately following it. For purposes of comparison, we will present the same type of analysis for ok as a signal of agreement or compliance, and for well as a discourse marker (see Condon 2001 for some preliminary results involving what happens following discourse well versus discourse ok in the oral corpus). Let us first start by presenting some predictions concerning the three relevant lexemes, discourse ok, agreement ok, and discourse well, as we expect these to exhibit quite different patterns. Our initial analyses will focus on four functions that are relevant to the decisionmaking routine. These functions are orientation, suggesting an action, elaborating or repeating, and agreeing with a suggestion. As Condon (2001) points out, orientations ought strongly to be expected following discourse ok because they typically initiate decision sequences. But the contrast between vertical and horizontal transitions (Bangerter and Clark, 2003; Bangerter et al., 2004) suggests an even stronger hypothesis. We claim, along with Bangerter et al., that discourse ok marks vertical transitions, and within this framework, agreement ok should be marking a horizontal transition that closes a decision routine. If these hypotheses are correct, then the functions of utterances preceding and following the two types of ok should differ with respect to orientations. Both types of ok should be preceded by low proportions of utterances functioning as orientations because the routine transition from the orientation function is to the suggestion function. Orienting utterances ought to be more frequent following discourse ok rather than agreement ok, even though it is possible to navigate from an agreeing ok that closes a decision sequence directly to an orienting utterance that initiates the next decision sequence: not all of these vertical transitions are marked by discourse ok. Therefore we predict that if we compare the functions of utterances following ok to the functions of utterances preceding ok, we ought to see a large increase in orientations for discourse ok and a smaller increase for agreeing ok. Schiffrin (1987: 126) observes that discourse well “anchors the speaker into a conversation precisely at those points where upcoming coherence is not guaranteed.” Consequently, the discourse functions of well are very different than those of ok, and the grounding functions are correspondingly distinct. In horizontal transitions, each turn grounds the previous turn by satisfying the expectations projected by the previous turn or by more explicit verification strategies. Both the transition and the successful grounding are expected outcomes of the talk, and this is also true of vertical transitions among subprojects. However, in vertical transitions, a turn grounds entire structures of turns instead of just the preceding turn. When participants shift from one decision routine to the next, they demonstrate their understanding that the previous turn was a horizontal transition which completed the subproject accomplished in the previous sequence of turns. Moreover, in the transition to the next subproject, participants orient to the larger project that governs the sequence of subprojects and creates expectations about the goals of the next decision routine. These properties motivate our claim that discourse ok marks significant, but expected transitions in the talk.

30 Grounding and Common Ground Unlike ok, discourse well typically appears when expectations are not satisfied. In fact, well seems to signal that the transition is a relevant response, even though it is not the expected response at the current state of the discourse. In this view, well performs an important grounding function because it indicates that the previous turn was understood in spite of an apparent lack of coherence. These functions of well do not seem to be limited to horizontal or vertical transitions or to specific positions in the decision routine. Consequently, there is no reason to expect any difference in the frequency of orienting utterances preceding and following the marker well. What about suggesting an action? Agreements are the preferred second pair-parts of suggestions; therefore, we predict high proportions of suggestions preceding agreeing ok. In contrast, the proportion of suggestions preceding discourse ok should be low, and we should be more likely to see suggestions following discourse ok than preceding it. According to the decision routine, the utterance following discourse ok functions as an orientation. However some utterances combine orientation and suggestion functions, as in (3a, e, h, i) and (4a). Also some decision sequences like (3c) skip the orientation function and move directly to a suggestion. Thus, a relatively high number of suggestions are expected following discourse ok. Of course, it is also possible to move from the agreement that closes a decision sequence directly to the next decision sequence, so that agreeing ok may be followed by suggestions, too, though the proportion of suggestions following agreeing ok should be much lower than those preceding it. Consequently, whereas there should be more utterances functioning as suggestions following discourse ok than preceding it, we predict that there will be a large decrease in the proportion of utterances following agreement ok compared to a high proportion preceding it. Though discourse well might function at any point in the discourse, it is often used to introduce dispreferred responses (Schiffrin, 1987: 102), such as suggestions that disagree with a previous suggestion, as in (10). (10)

a. well then we could–I’d rather uh–do that like in the evening (after the partner suggests that they walk around Bourbon St. and the French Quarter in the morning) b. well they can just eat on this all weekend (after the partner suggests that they serve other meals in addition to crawfish if their crawfish boil is going to last the entire weekend)

The speaker who produced (10a) used well before many suggestions, even when they were not counterproposals and his partner did the same. In fact, there were a relatively high number of disagreements in this interaction, and the use of well may reflect some tension between the two participants. Disagreements are so dispreferred that they are frequently preceded by acknowledgments, which results in familiar sequences like yeah, but, and sequences in which the suggestion that elicits disagreement does not immediately precede well, while the suggestion that expresses the disagreement immediately follows well. Therefore, we might expect somewhat higher proportions of suggestions following well than preceding it. After orientations and suggestions, the final component of the decision routine is agreement. Because agreements close decision sequences, they should be rare following discourse ok and relatively frequent preceding it. We would expect very low levels of agreement in the

Discourse Markers of Common Ground 31 utterances immediately adjacent to agreeing ok and discourse well, as they themselves signal an agreement (or its lack). Elaborations are not an explicit part of the default routine, but we include them here because of discourse well. A natural prediction is that discourse following well might include some elaboration of the previous suggestion in order to identify potential dissatisfaction with it. In fact, participants always provide reasons for disagreeing with a suggestion if they do not offer a counterproposal. We can also anticipate that elaborations are more likely to follow rather than precede agreement ok because agreements are often elaborated (e.g. ok that’ll be fun). Finally, a coding convention assigns the elaborates–repeats category to orientations when they are not formulated as requests for information or incorporated in suggestion. Therefore, we might expect that utterances in this category will be more frequent following than preceding discourse ok. However, if elaborations follow agreements, then they are likely to precede the vertical transitions that discourse ok marks, too, so no clear prediction emerges. To provide statistical tests of these hypotheses, we adopt a different approach than Condon (2001), where the proportions of functions that were presented characterized the corpora as a whole. Below, the data are average proportions of functions per interaction. Figures 2 and 3 present the results relevant to these predictions for the face-to-face interactions and the computer-mediated interactions respectively. In those figures, the solid lines and circles present proportions of functions in the utterance following a specific lexeme, whereas the dotted lines and squares present the results for the utterance preceding it. The functions presented are orients suggestion (os), suggests action (sa), elaborates–repeats (er), and agrees with suggestion (as). As can be seen from Figure 2, most (though not all) of the predictions are reasonably well supported. We performed separate analyses of variance on the three lexemes, using order (preceding versus following utterance) and function as our variables. In the analyses on the two forms of ok, the interaction between order and function was significant, F3 45 = 4315, p < 01, and F3 45 = 16483, p < 01, respectively. However, this interaction did not prove significant for well: F = 1481, p > 20. Each analysis also revealed main effects of function (some of which will be discussed later) but not of order. First we will concentrate on the interactions of function and order. 1

PROPORTIONS

0.8

Discourse Marker OK

Agreement/Compliance OK

Discourse Marker WELL Preceding utterance Following utterance

0.6 os sa er as

0.4 0.2

orients suggestion suggests action elaborates-repeats agrees with suggestion

0 os sa

er

as

os sa

er

as

os sa

er

as

SELECTED FUNCTIONS

Figure 2. Functions preceding and following markers in face-to-face interactions

32 Grounding and Common Ground 1

PROPORTIONS

0.8

Discourse Marker OK

Agreement/Compliance OK

Discourse Marker WELL Preceding utterance Following utterance

0.6

os orients suggestion sa suggests action er elaborates-repeats as agrees with suggestion

0.4

0.2 0 os

sa

er

as

os

sa

er

as

os

sa

er

as

SELECTED FUNCTIONS

Figure 3. Functions preceding and following markers in computer-mediated interactions As predicted, when we compare the functions of utterances following discourse ok to those preceding it, orientations increase and agreements decrease, t45 = 410 and −484, respectively, p < 01. There was a predicted increase in suggestions following ok which was significant at a one-tailed .01 level, but not at a two-tailed .01 level, t45 = 261. (We ought to note that we are using a conservative alpha to adjust for false positives due to conduction of multiple t-tests; hence, unless otherwise specified, all t-values given below are significant at the .01 level.) Clearly, the increase in orientations is more robust than the increase in suggestions. Finally, there was a decrease in elaboration, also significant by one-tailed, but not two-tailed, test, t45 = −261. The results were quite different, however, for the utterances surrounding an agreement ok. As predicted, there was a significant decrease in suggestions, t45 = −993, and elaborations increased, t45 = 993. Also as predicted, there were very few agreements either before or after agreeing ok, and they did not differ in proportion from the preceding to the following utterance. Nor was the difference in the proportions of orientations significant at conventional levels, t45 = 177. We had predicted that there would be higher proportions of orientation following rather than preceding agreement ok because orientations initiate decision routines, while agreements close them. The high proportions of elaboration following agreeing ok and the frequent use of discourse markers preceding orientations result in fewer instances in which orientations directly follow agreeing ok. We did not find a large increase in elaborations following well, contrary to expectations. There was a significant increase (by one-tailed test) in suggestions, t45 = 371, but as the overall interaction was non-significant for well, this latter result ought to be viewed with caution. Were there similar sorts of patterns in the computer-mediated environments? The corresponding computer-mediated data are represented in Figure 3. Again, although function proved significant in all three analyses (and order in the analyses of the two discourse markers), we will concentrate on the interactions between the variables. In the data of Figure 3, the interaction of order with function proved significant for each analysis: F3 177 = 4554, p < 005 for

Discourse Markers of Common Ground 33 discourse ok; F3 177 = 6059, p < 001 for agreement ok; and F3 177 = 7214, p < 001 for discourse well. To summarize the results, the change in relative frequency of function in computermediated interactions was similar to that found in the oral interactions, with the exception of the increase in elaboration following agreement ok found earlier. The decrease in elaboration is consistent with participants’ tendency to minimize expenditures of linguistic resources in the computer-mediated interactions. As in the face-to-face interactions, there were higher proportions of orientations and lower proportions of agreements following discourse ok than there were preceding discourse ok, t177 = 428 and −600, respectively. Also like the face-toface interactions, the computer-mediated interactions exhibit considerably higher proportions of suggestions preceding than following an agreement ok, t177 = 884. The increase in suggestions following well again proved significant, t177 = 973, which raises confidence in the earlier result. In part due to the greater statistical power achieved from the analysis of 60 interactions, the decrease of 6% in orientations following well was also reliable, t177 = −265, p < 01.

4.2. Discourse ok and the decision routine In the previous section, we contrasted horizontal transitions in the decision routine represented by agreeing ok with the vertical transitions among decision routines that discourse ok marks by examining the functions of utterances preceding and following the lexemes. In this section, we present additional evidence that discourse ok marks vertical transitions among decision routines in both face-to-face and computer-mediated interactions by contrasting the proportions of functions following discourse ok with the average proportions of those functions in the interactions. Table 1 presents the proportions of all functions that were annotated in the coding scheme used for the face-to-face interactions. We present the results for the computermediated interactions using the same structure of code categories with a note that describes the single difference between the two sets of proportions. (These differences are specified in the discussion of Figure 2 above.) The proportions are all averaged per interaction, rather than over the corpus. Table 1 also presents the 95% confidence intervals, where appropriate, for the mean proportions following discourse ok. For example, the column labeled “Mean” for the faceto-face interactions in Table 1 shows that the average proportion of utterances coded as orients suggestion per interaction was .07, but the proportion of utterances coded as orients suggestion following the discourse marker ok was .19. The 95% confidence interval for the latter mean is plus or minus .07, so that the lower limit of this confidence interval does not include the overall mean for orientations. Similarly, for computer-mediated interactions, the average proportion of utterances coded as orients suggestion per interaction was .10, while the proportion of utterances coded as orients suggestion following the discourse marker ok was .13. The confidence interval surrounding this latter mean includes the average overall mean. The results in Table 1 provide a more fine-grained picture of what happens following discourse ok in the two media. Figures 4 and 5 are based on these data, but show whether there is an increase or a decrease in the proportions of each code category following ok: they graph the difference between proportions for discourse ok and overall proportions per interaction for the face-to-face and computer-mediated environments, respectively. The functions in Figures 4

34 Grounding and Common Ground Table 1. Proportions of functions following ok and average proportion per interaction Face-to-face Function

ok (95% CIs)

Computer-mediated

mean

ok (95% CIs)

mean

Moves Suggests action (sa) Requests action (ra) Requests validation (rv) Requests information (ri) Elaborates-repeats (er)

.16 .07 .04 .07 .41

(.07) (.04) (.04) (.03) (.09)

.18 .03 .04 .06 .29

.11 .05 .01 .16 .16

(.06) (.04) (.02) (.08) (.07)

.30 .07 .02 .12 .24

Responses Agrees with suggestion (as) Disagrees with suggestion (ds) Complies with request (cr) Acknowledges only (ao)

.02 .01 .01 .00

(.04) (.01) (.01) (–)

.11 .02 .07 .09

.00 .00 .03 .00

(–) (–) (.04) (.01)

.11 .00 .10 .03

Other Discourse marker (dm) Explicit management (em) Orients suggestion (os) Personal info (pi) Jokes, exaggerates (je)

.23 .11 .19 .01 .00

(.08) (.07) (.07) (.01) (.01)

.12 .05 .07 .03 .01

.15 (.08) 27∗ (.02) .13 (.06) .00 (–) .00 (–)

∗

.05 19∗ .10 .01 .00

not mutually exclusive with “orients suggestion”.

and 5 are represented according to the abbreviations provided in Table 1. As is evident from these figures, there are some significant differences across medium. Most notably, the media appear to differ on suggestions, elaborations, and orientations, although there is also some apparent difference regarding acknowledgments. To analyze these data, we conducted analyses of variance in which overall proportions versus subsequent-to-ok proportions was one of the factors. We also conducted post-hoc t-tests (with an adjusted .01 alpha level on the differences between the overall and the ok-based frequencies for each of the code functions in the analysis, i.e., a test of the interaction, as before). Tests that show a significant difference in this case correspond to the claim that a data point in Figure 4 or 5 differs significantly from zero. As before, all t-statistics reported below are significant at this level, unless otherwise noted. The interaction of function and environments following discourse ok was significant for the Move group in both the face-to-face condition, F4 60 = 3846, p < 01, and the computer-mediated condition, F4 236 = 7593, p < 001. Similar results were obtained for the Response group, F3 45 = 8955, p < 001, in face-to-face interactions and F3 177 = 19966, p < 001, in computer-mediated interactions. Looking at the results relevant to orientations first, the interaction of code function frequency and whether the function followed a discourse ok proved significant for the face-to-face condition, F4 60 = 4366, p < 005, but was marginal, at best, for the computermediated condition, F4 236 = 2318, p < 058. The proportion of utterances coded as orients

Discourse Markers of Common Ground 35 0.2

PROPORTIONS

0.1

0

–0.1

–0.2 sa ra

rv

ri

er

MOVES

as ds cr ao RESPONSES

dm em os pi

je

OTHER

Figure 4. Difference in proportions of functions following discourse ok vs. corpus averages in face-to-face interaction

0.2

PROPORTIONS

0.1

0

–0.1

–0.2 sa ra

rv

ri

MOVES

er

as ds cr ao RESPONSES

dm em os pi

je

OTHER

Figure 5. Difference in proportions of functions following discourse ok vs. corpus averages in computer-mediated interaction

suggestion following discourse ok was significantly higher than the average proportion of utterances coded as orients suggestion in the face-to-face condition, t60 = 882, but the difference did not prove to be reliable for the computer-mediated condition, t236 = 208. Thus, despite strong evidence in the earlier before-and-after analyses that discourse ok marks transitions by initiating orientations, the evidence is not as strong in the computer-mediated conditions when only proportions following discourse ok are examined.

36 Grounding and Common Ground The claim that discourse ok signals transitions to a new decision routine also predicts that it should be followed by relatively high proportions of utterances coded as suggests action and requests information, because these often co-occur with orientations. Examples of utterances coded as both orients suggestion and suggests action are (3a, e, h, i) and (4a), while (4b–d), (8a, d) and (9a) are utterances coded as both orients suggestion and requests information. Another function that can co-occur with orientation is requests action, which is used to annotate orientations that are accomplished using explicit management, as in (6c). The two request functions do occur in higher proportions following discourse ok than they do overall in the interactions, though the proportions are not significantly higher. Suggestions, on the other hand, were less frequent following discourse ok, although this decrease proved significant only in the computer-mediated condition, t60 = −156 and t236 = −1369. According to our claims about discourse ok and the decision routine, agreements should precede rather than follow discourse ok. In fact, low frequencies of all utterances serving Response functions following discourse ok is anticipated by the claim that ok marks expected transitions at significant structural boundaries, especially projects and subprojects, which are typically decision routines in our data. Response functions are primarily second pair-parts of adjacency pairs, and the routine transition from a first pair-part is to a second pair-part, not to the kind of structure that would be marked by discourse ok. As may be seen in these figures, there were significantly small proportions of utterances coded as agrees with suggestion, complies with request, and acknowledges only following ok both in the face-to-face environment, t45 < −10, and in the computer-mediated environment, t177 < −56. As Bangerter et al. (2004: 2) observe, “[a]djacency pairs can be viewed as carrying out minimal joint projects,” which, as minimal projects, would not include subprojects that would be marked with ok. Of course, adjacency pairs can be separated by exchanges such as insertion sequences, and there are instances in which the expected response to a first pair-part is to take up a joint project. An example from the computer-mediated interactions is (11). (11)

a. P2: what else b. P1: ok so what do c. P1: ok how about best video

P2’s request in (11a) refers to incomplete subprojects that may require action in order to complete the task. Therefore, P1’s response in (11c) both complies with the request by completing the minimal joint project of providing something else and initiates a new subproject with a new decision routine. The aborted message in (11b) suggests other ways that a second pair-part might simultaneously complete and initiate subprojects. If P1 had written “so what do we have,” she would be initiating a joint project of gathering the information needed to complete the project initiated by P2. In other words, (11b) would begin an insertion sequence. Alternatively, if P1 had written “so what do we need,” she could respond to P2 by initiating a project of identifying everything else needed to complete the task. By comparing the proportions of orientations, suggestions, agreements, and related functions following discourse ok to the overall proportions of those functions in the interactions, we have obtained some additional support for the claim that discourse ok marks vertical transitions among decision sequences in both face-to-face and computer-mediated interactions. However, not all of the predicted differences were significant at the conservative .01 alpha

Discourse Markers of Common Ground 37 level we selected, and the proportions of utterances coded as suggests action were lower rather than higher following discourse ok.

4.3. Other discourse functions of ok The transitions among decision sequences are not the only vertical transitions marked by discourse ok in the interactions, and the proportions in Table 1 reflect these additional discourse functions. The most frequent category of utterance that follows discourse ok in the face-to-face interactions is the catch-all category elaborates–repeats, which is also the most frequently occurring category in all the face-to-face interactions. Nevertheless, compared to the corpus average, this function occurred significantly more frequently following discourse ok, t60 = 934. Since utterances coded as elaborates–repeats are also assigned categories in the Other group, some of the utterances in the elaborates–repeats category initiate decision routines because they are also coded as orients suggestion or, if they employ explicit decision management, as explicit management. However, it appears that there is such a high proportion of utterances coded as elaborates–repeats following discourse ok primarily because of other transitions that ok is used to mark in the face-to-face interactions. Discourse ok precedes utterances at the beginning of the interaction when participants initiate the interaction and orient to the task, as in (12). (12) a. ok I’m speaker “A” – I’m 21 and I’m from Jeanerette b. ok I think this is the plan sheet for this one c. ok the barbeque problem As Condon (1986) observes, discourse ok is frequently used to initiate interactions when interlocutors are participating in a research context and have been asked to engage in a joint project. Bangerter and Clark (2003) and Bangerter et al. (2004) report the same function for discourse ok in similar elicited, task-oriented dialogues. The occurrence of discourse ok at the boundaries of high level joint projects in our computer-mediated corpus is illustrated in (13). (13)

a. Okay, hello, My name is Ms. Johnson of BMG records and i was informed to contact you in regards to our upcoming MTV Music Video Awards.? The first thing needed to be discussed however is the date, time and place of said event b. Ok, where do you want to start? c. ok first we have to decide who will be the host

The use of ok before the greeting in (13a) is particularly revealing because ok signals the common ground that frames the entire interaction, including the greeting. Similarly, discourse ok is used to close the computer-mediated interactions, and the association of ok with closings is also reported in Clark and Bangerter (2003), Bangerter et al. (2004), and Schegloff and Sacks (1973). An example from the computer-mediated interactions is (14). (14) a. P2: what do we do now b. P1: I will press escape just watch your screen>

38 Grounding and Common Ground c. d. e. f.

P1: P2: P1: P2:

send me back an ok so i know you got the message ok and by the way don’t you think all USL students should go free? yes! ok

In (14c) P1 sends a second message that specifically requests the use of ok to signal the partner’s readiness to end the interaction by pressing the <escape> key. The final ok signals the sender’s intention to behave in a way that reflects the understanding that the current action is closure. In the face-to-face interactions, discourse ok also marks transitions when participants shift from making decisions to recording their decisions on the answer sheet. In (15) participants record their decisions in a long sequence at the end of the interaction. ([si] and [se] annotate the beginning and end of a sequence of simultaneous speech.) (15)

a. b. c. d. e. f. g.

P1: P2: P1: P2: P1: P2: P1:

ok it’s [si] fourth of July [se] [si] Saturday – [se] fourth of July July fourth – Saturday if – that’s good enough from (writing) ten to six ten a.m. to six p.m. location Chicot park (writing) Chicot State Park Chicot State Park    ok food uh we got the burgers we got the hotdogs it’s true let’s make this some somewhat readable and chicken and steak and potatoe salad I don’t know if I spelled that right but hey and chips salads and ok sodas (writing throughout)

(15a) illustrates the use of discourse ok at the transition from making decisions to recording them, and the two instances of discourse ok in (15g) mark transitions to subprojects of the decision-recording subproject as P1 enters their decisions in the blanks on the answer sheet that are designated for food and beverages respectively. Except for ok, all of the utterances in (15) are coded as elaborates–repeats because they repeat decisions that were previously established. The sequence in (15) also demonstrates that not only does the partner who is writing verbalize the words as they are written, but also the partner who is not writing often verbalizes the words that the partner should write. In (15f) P2 appears to correct P1’s name for the park in which the barbeque will be held. This collaborative writing increases the incidence of utterances coded as elaborates–repeats in the face-to-face interactions and the frequency of utterances coded as elaborates–repeats following ok in the face-to-face interactions. In contrast, the computer-mediated interactions did not involve any collaborative writing like we observe in (15), and the average proportions of utterances coded as elaborates–repeats is much lower. The proportion of elaborates–repeats following discourse ok is even lower: t236 = −577, p < 001. Some of the highest proportions of functions following discourse ok are in the three Other categories that are most closely associated with management of the interaction. We have already reported the results for orientations, which have a significant role in managing the interaction: they initiate decision routines, which determine both the resolution of the joint project and the structure of the talk. Also, we observed that the transitions marked by discourse

Discourse Markers of Common Ground 39 ok are frequently marked with another discourse marker, especially so and sometimes now, and when ok is followed by discourse markers in these interactions, those markers are typically followed by orientations, as in (4), (6c), and (9a). The frequency of discourse markers is higher following discourse ok than the average frequency in both communication environments, t60 = 809, and t236 = 692. In the face-to-face interactions, the category explicit management also includes utterances that orient suggestions, but these were annotated as explicit management because they were formulated with explicit references to the talk, as in (16) (16) ok where first of all we have to decide where um where would be a good place to have a barbeque Table 1 shows that a large proportion (53%) of the utterances following discourse ok are classified as either discourse marker, explicit management, or orients suggestion, which is more than twice the average frequency of utterances serving those functions in the face-to-face corpus. Consequently, there is strong support for the claim that discourse ok is involved in discourse management, for the more specific claim that ok marks significant but expected transitions, and for the even more specific claim that these transitions are typically vertical shifts among subprojects. Consistent with these claims, the increases observed in Figure 4 for these three functions are significant. Table 1 shows that around half of the utterances following discourse ok are classified as either discourse marker, explicit management, or orients suggestion in computer-mediated interactions, too. The increases seen in Figure 5 reach conventional levels of significance for discourse markers and explicit management (the t-value for the latter is 5.54), although there was, at best, a trend in the right direction for orientations, as reported earlier (the t-value for os was significant at the 0.05 level, rather than the alpha level of 0.01 we selected to guard against false positives). The similarities among utterances following discourse ok in face-to-face and computer-mediated interaction occur in spite of differences between interactions in the two communication environments. For example, requests for action were significantly higher in the face-to-face condition, t60 = 311, and requests for information were significantly elevated in the computer-mediated condition, t236 = 288. Many more utterances in the computermediated interactions are organized by the decision routines, as evidenced by higher proportions of suggestions and orientations in Table 1. Condon et al. (1999) examine sequences of two and three utterances and demonstrate that higher proportions of these sequences conform to sequences in the decision routine in computer-mediated interaction. We have claimed that discourse routines are effective tools for managing interaction that allow interlocutors to communicate with minimal linguistic form by taking advantage of a common ground of understandings associated with the routine. In Clark and Brennan’s (1991) terminology, discourse routines incur lower costs because fewer linguistic and interactional resources are required to complete the joint action. Given the well-documented fact that participants in computer-mediated interaction produce less language than comparable interlocutors in face-to-face interaction (Bordia, 1997), it seems likely that the former are motivated to reduce the costs of their communication and will rely on discourse routines to achieve a reduction. Consequently, it is significant to observe that another difference between the

40 Grounding and Common Ground face-to-face and computer-mediated interactions in Table 1 has the opposite effect: participants in computer-mediated interaction produce higher proportions of explicit management. ˇ Condon and Cech (2001) reason that explicit management is dispreferred in any communication environment because talk about talk expends resources and incurs those costs for achievements other than the goals that organize the talk. For example, designers of speech recognition systems assume that implicit verification strategies are preferable to explicit verification strategies, reasoning that the latter require “extra turns, which users may find annoying” (Krahmer et al., 1999: 1423). Many conventional expressions such as greetings can be viewed as strategies for accomplishing interactional work that would otherwise require explicit management. Thus speakers can simply say “hi” instead of “I recognize that you are a potential interlocutor and I hereby initiate joint conversational action with you.” Though explicit management is costly, it is an effective strategy for managing interaction when demands on processing are increased and mutual understanding is threatened. In a study that examined explicit management in face-to-face and computer-mediated interactions, Conˇ don and Cech (2001) found increases in explicit management for the more complex MTV task compared to the simpler party and travel planning tasks in both communication environments. Furthermore, explicit management increased for both types of tasks in the computer-mediated environment. Similarly, Condon and Miller (2003) report very high frequencies of explicit management in a machine-translated instant messaging environment. In each case, greater demands on the interaction create a higher potential for compromising fundamental grounding processes. Therefore, both task complexity and communication environment contribute to the nearly fourfold increase in the proportions of explicit management for computer-mediated interaction in Table 1. As we noted earlier, explicit management was so frequent in the MTV task corpus that we modified the annotation system to include a fourth group of explicit management categories. Because the explicit management category was given a higher priority than the orients suggestion category in the original annotation system used for the face-toface interactions, all utterances that functioned as explicit management were coded in the appropriate category and the frequency can be compared to the frequency of utterances coded in the Explicit Management group in the modified annotation system for the computer-mediated interactions. However, some utterances that also functioned as orientations were coded as explicit management in the original annotation system because they could not be coded in both categories. In contrast, the modified annotation system permits utterances to be coded as both orients suggestion and decision management in the Explicit Management group. Therefore, the frequency of utterances functioning as orientations is slightly underreported for the faceto-face interactions. Also, it is not possible to sum the categories discourse marker, explicit management, and orients suggestion to estimate the proportion of utterances following ok that are associated with discourse management in computer-mediated interaction. Nevertheless the increase in utterances serving managerial functions in computer-mediated interaction is clear, as is the fact that very high proportions of utterances following ok have managerial functions. In fact, summing the proportions in the discourse marker and explicit management categories alone accounts for 42% of the utterances following discourse ok in the computer-mediated interactions. Despite a number of similarities, discourse ok occurs far less frequently in the computer-mediated interactions. We speculate that discourse ok occurs less frequently in

Discourse Markers of Common Ground 41 computer-mediated environments because more explicit management strategies are preferred in those environments. We return to this difference in the next section.

5. CONCLUSIONS The analyses clearly demonstrate that the functions of discourse ok are the same in face-toface and computer-mediated interactions. In both communication environments, ok occurs at transitions between decision sequences and preceding utterances that manage the decisions. In Bangerter and Clark’s (2003) terms, discourse ok marks transitions among the hierarchy of subprojects or plans that must be completed in order to achieve a joint project. Discourse ok demonstrates understanding of the current state of the interaction by verifying that the current state is a significant, but expected, transition. Therefore, discourse ok anchors the talk to the common ground of understandings at the current state of the discourse and signals that the expectations which arise from this state will frame the contribution that follows. Condon (2001) emphasizes the fact that ok signals expected or default transitions and suggests that this reflects a fundamental organization of interaction that is based on the logic of defaults and the contrast between marked and unmarked continuations in the speech event. In this system, discourse ok contrasts with discourse well, which signals unexpected or marked transitions. Moreover, discourse ok seems to link the talk to specific understandings in the common ground: rather than orienting to any previously established propositions or referents, ok attends to the business at hand or, as Beach (1995) generalizes, in medical interviews ok marks transitions to “official” matters. Similarly, Condon (1986) and Bangerter and Clark (2003) note the use of ok to return to the previous topic after digressions. By orienting to the speaker’s perception of “what is going on,” ok naturally links the talk to the current joint activity or subproject. The idea that speakers’ understanding of the talk is organized by default expectations about the current joint activity resembles Nemo’s (in this volume) suggestion that discursive common ground frames events according to whether they are “business as usual.” He argues that an account of shared understandings purely in terms of inferences from a collection of shared propositions fails to allow for the alternative “interpretative grids and prisms through which we see and inseparably judge reality.” We would also argue that inferences, even directed by maxims and interpretive principles, are too underdetermined to account for the powerful constraints that discourse routines impose on language behavior. When they can be relied on, default expectations about “business as usual” provide a common ground for interlocutors to communicate quickly and effectively in spite of an ever-present potential for failure. Forms such as yeah, uh huh, and ok confirm those expectations both at the utterance level and for the larger business at hand. None of the generalizations we have formulated about discourse ok explain why some decision routines are initiated by ok, while others are not. Condon (2001) speculates that a likely motivation for using discourse ok is the need to call into play a considerable amount of shared contextual knowledge that will be required to process subsequent discourse, and decision routines do incorporate a cognitive decision schema plus all of the defaults associated with the schema and the routine. However, these understandings should be uniform for all decision routines and cannot account for the variability in use of discourse ok. Bangerter and Clark (2003) and Bangerter et al. (2004) suggest that differences in use of project markers

42 Grounding and Common Ground result from differences in the ways that participants conceptualize their actions, but this does not explain the systematic difference in the frequency of discourse ok in face-to-face and computer-mediated interactions. One way to view the relatively infrequent use of discourse ok in computer-mediated interactions is in terms of discourse management strategies. We have seen that the decrease in use of discourse ok by participants in computer-mediated interaction is accompanied by an increase in use of explicit management strategies. We observed that discourse ok is useful in part because it is not explicit, which makes ok a versatile, all-purpose verifier of common ground. However, the lack of specificity can be a drawback in contexts where explicitness is preferred. Another factor that is likely to lower the use of discourse ok is the strategy of minimizing the linguistic and interactional resources expended on the speech event by relying on structures such as discourse routines that take advantage of default expectations. Because the transitions that discourse ok marks are expected, they are likely to be successfully navigated without formulation of the marker. An example of the extreme strategies that participants in computer-mediated interaction can adopt in order to reduce the cost of the interaction is in (17), responding to the barbeque planning task. (17)

a. b. c. d. e. f. g.

P1: P2: P1: P2: P1: P2: P1:

WHAT DAY Sunday ACTIVITIES AND ENTERTAINMENT touch football, volleyball, softball ENTERTAINMENT we could hire a magician and comedian MUSIC SHA!

In (17a–e) only the required contents are formulated: the function of each utterance is determined by the defaults of the decision routine. The default agreement function is even eliminated, which is a successful strategy because agreement is entirely predictable from the fact that the partner passes on the opportunity to disagree. Even the reproach and suggestion in (17g) are minimal (sha is a Cajun address term derived from French cher). There is certainly no room for discourse ok in these sequences. The computer-mediated data provide some support for both the idea that ok is used when there is a need to call into play considerable common ground and the idea that the use of discourse ok in computer-mediated interaction is related to the use of explicit management strategies. The utterance function that occurs most frequently following discourse ok in computer-mediated interaction is explicit management. In addition to decision management, explicit management includes other references to the interaction such as discussion of the task (e.g., Are we supposed to write this down?), transmission management, as in (7), and explicit repair. Consequently, it appears that discourse ok is marking explicitly managed vertical transitions at particularly significant boundaries in computer-mediated interaction. Further research on discourse ok promises to be as productive and revealing as previous work has been. As Bangerter and Clark (2003) and Bangerter et al. (2004) observe, there is a need to take into account the effects of prosody on discourse ok. In particular, there is a lengthened discourse ok with a rising intonation on the second syllable that seems to mark something like I intend to act in accordance with the normal, default expectations at this transition even though something unexpected has occurred. Furthermore, the variation in use

Discourse Markers of Common Ground 43 of discourse ok, ok so, and ok now both within and between interlocutors and contexts still requires a satisfactory explanation. We have found comparisons of interaction in different communication environments to be a useful strategy for investigating discourse management strategies, and we hope that these methods will continue to shed light on the issues raised here. Discourse ok is a simple, but powerful means of displaying understanding of the common ground established in an entire structure of utterances. Like other forms such as yeah and uh huh that verify the speaker’s understanding of the current state of the discourse, ok enjoys syntactic and semantic freedoms that make it a versatile, all-purpose marker. Unlike backchannel markers, ok does not simply claim understanding: it provides stronger verification by marking a mutually identifiable property of the current state. By marking the current state of the discourse as a significant structural boundary and a transition that is expected by the business at hand, discourse ok allows speakers to verify their mutual understanding of “what is going on” and to project a common frame for interpreting the next contribution to the interaction. Therefore, like other markers of common ground, discourse ok provides insight into fundamental processes that make interaction successful.

ACKNOWLEDGEMENTS We would like to thank the research assistants who annotated the interactions and the editors of this volume, who organized the IPrA panel where we presented an initial version of this paper. We are also grateful to the editors for comments that helped us improve the manuscript. The research was supported in part by funding from the University of Louisiana at Lafayette and a Louisiana Board of Regents Research award. The co-author’s affiliation with The MITRE Corporation is provided for identification purposes only, and is not intended to convey or imply MITRE’s concurrence with, or support for, the positions, opinions or viewpoints expressed by the authors.

REFERENCES Bangerter, Adrian and Clark, Herbert H., 2003. Navigating joint projects with dialogue. Cognitive Science 27, 195–225. Bangerter, Adrian, Clark, Herbert H., and Katz, Anna R., 2004. Navigating joint projects in telephone conversations. Discourse Processes 37, 1–23. Baym, Nancy, 1996. Agreements and disagreements in a computer-mediated discussion. Research on Language and Social Interaction 29, 315–346. Beach, Wayne, 1990. Language as and in technology: Facilitating topic organization in a videotex focus group meeting. In: Medhurst, M.J., Gonzalez, A., and Peterson, T.R. (eds), Communication and the Culture of Technology. Washington State University Press, Pullman, Washington, 197–219. Beach, Wayne, 1993. Transitional regularities for ‘casual’ “Okay” usages. Journal of Pragmatics 19, 325–352. Beach, Wayne, 1995. Preserving and constraining options: “Okays” and ‘official’ priorities in medical interviews. In: Morris, G.H. and Chenail, Ronald J. (eds), The Talk of the Clinic:

44 Grounding and Common Ground Explorations in the Analysis of Medical and Therapeutic Discourse. Lawrence Erlbaum, Hillsdale, New Jersey, 259–289. Bordia, Prashant, 1997. Face-to-face versus computer-mediated communication: A synthesis of the experimental literature. The Journal of Business Communication 34, 99–121. ˇ Cech, Claude and Condon, Sherri, 1998. Message size constraints on discourse planning in synchronous computer-mediated communication. Behavior Research Methods, Instruments, & Computers 30, 255–263. ˇ Cech, Claude and Condon, Sherri, 2004. Temporal Properties of Turn-Taking and TurnPackaging in Synchronous Computer-Mediated Communication. Proceedings of the Hawaii International Conference on System Sciences, Waikoloa, Hawaii, January, 2004. Clark, Herbert H., 1996. Using Language. Cambridge University Press, Cambridge. Clark, Herbert H. and Brennan, Susan E., 1991. Grounding in communication. In: Resnick, L.B., Levine, J.M., and Teasley, S.D. (eds), Perspectives on Socially Shared Cognition. APA Press, Washington, DC, 127–149. Condon, Sherri, 1986. The discourse functions of OK. Semiotica 60, 73–101. Condon, Sherri, 2001. Discourse ok revisited: Default processes in verbal interaction. Journal of Pragmatics 33, 491–513. ˇ Condon, Sherri and Cech, Claude, 1992. Manual for coding decision-making interactions. Unpublished manuscript, University of Southwestern Louisiana/Université des Acadiens. (Revised May 1995). Available online at http://www.georgetown.edu/ luperfoy/Discourse/Treebank/dri-home.html ˇ Condon, Sherri and Cech, Claude, 1996a. Functional comparisons of face-to-face and computer-mediated decision making interactions. In: Herring, S. (ed.), ComputerMediated Communication: Linguistic, Social and Cross-Cultural Perspectives. John Benjamins, Philadelphia, 65–80. ˇ Condon, Sherri and Cech, Claude, 1996b. Discourse management strategies in face-to-face and computer-mediated decision-making interactions. Electronic Journal of Communication/La Revue Electronique de Communication 6 (3). ˇ Condon, Sherri and Cech, Claude, 2001. Talk about talk in face-to-face and computer-mediated interaction. In: Nemeth, E. (ed.), Cognition in Language Use: Selected Papers from the 7th International Pragmatics Conference, Vol. 1. International Pragmatics Association, Antwerp, 56–69. ˇ Condon, Sherri, Cech, Claude, and Edwards, William, 1999. Measuring Conformity to Discourse Routines in Decision-making Interactions. Proceedings of the 37th Annual Meeting of the Association for Computational Linguistics, 238–245. Condon, Sherri, and Miller, Keith, 2003. “Can you Read this Well?” Error Handling in a Translated Message Environment. Proceedings of the ISCA Workshop on Error Handling in Spoken Dialogue Systems. Chateau-d’Oex-Vaud, Switzerland, August, 2003. Ferrara, Kathleen, Brunner, Hans and Whittemore, Greg. 1991. Interactive written discourse as an emergent register. Written Communication 8, 8–34. Fetzer, Anita, this volume. Reformulation and common grounds. Fischer, Kerstin, this volume. Grounding and common ground: Modal particles and their translation equivalents. Garcia, Angela Cora and Jacobs, Jennifer Baker, 1999. The eyes of the beholder: Understanding the turn-taking system in quasi-synchronous computer-mediated communication. Research on Language and Social Interaction 32, 337–367.

Discourse Markers of Common Ground 45 Goodwin, Charles, 1987. Forgetfulness as an interactive resource. Social Psychology Quarterly 50, 115–130. Herring, Susan, (ed.), 1996. Computer-Mediated Communication: Linguistic, Social and CrossCultural Perspectives. John Benjamins, Philadelphia. Hopper, Robert, 1989. Speech in telephone openings: Emergent interaction v. routines. Western Journal of Speech Communication 53, 178–184. Krahmer, Emil, Swerts, Marc, Theune, Mariet, and Weegels, Mieke, 1999. Problem spotting in human-machine interaction. Proceedings of Eurospeech99, 3, 1423–1426. Lewis, David K., 1979. Scorekeeping in a language game. Journal of Philosophical Logic 8, 339–359. Louwerse, Max M. and Hite Mitchell, Heather, 2003. Towards a taxonomy of a set of discourse markers in dialog: A theoretical and computational linguistic account. Discourse Processes 35, 199–239. Merritt, Marilyn, 1978. On the use of ‘O.K.’ in Service Encounters. Sociolinguistic Working Paper 42, University of Texas at Austin, April. [Also In: Bauman, Richard and Sherzer, Joel (eds), 1980, Language and Speech in American Society: A Compilation of Research Papers on Sociolinguistics. Southwest Educational Development Laboratory, Austin, Texas, 139–147]. Nemo, François, this volume. The pragmatics of common ground. From common knowledge to shared attention and social referencing. Oviatt, Sharon, Bernard, Jon, and Levow, Gina-Anne, 1998. Linguistic adaptations during spoken and multimodal error resolution. Language and Speech 41, 419–422. Schegloff, Emanuel, 1968. Sequencing in conversational openings. American Anthropologist 70, 1075–1095. Schegloff, Emanuel, 1979. Identification and recognition in telephone conversational openings. In: Psathas, G. (ed.), Everyday Language: Studies in Ethnomethodology. Irvington, New York, 23–78. Schegloff, Emanuel, 1982. Discourse as an interactional achievement: Some uses of uh huh and other things that come between sentences. In: Tannen, D. (ed.), Analyzing Discourse: Text and Talk. Georgetown University Press, Washington, DC, 71–93. Schegloff, Emanuel, 1986. The routine as achievement. Human Studies 9, 111–151. Schegloff, Emanuel and Sacks, Harvey, 1973. Opening up closings. Semiotica 7, 289–327. Schiffrin, Deborah, 1987. Discourse markers. Cambridge University Press, Cambridge. Stalnaker, Robert C. 1978. Assertion. In: Cole, P. (ed.), Syntax and Semantics 9: Pragmatics. Academic Press, New York, 315–332. Whittaker, Steven, 2003. Theories and methods in mediated communication. In: Graesser, Arthur, Gernsbacher, M.A. and Goldman, S.R. (eds), Handbook of Discourse Processes. Erlbaum, Mahway, New Jersey, 243–286.

This page intentionally left blank

Lexical Markers of Common Grounds Edited by Anita Fetzer and Kerstin Fischer © 2007 Published by Elsevier Ltd.

GROUNDING AND COMMON GROUND: MODAL PARTICLES AND THEIR TRANSLATION EQUIVALENTS Kerstin Fischer

1. INTRODUCTION In this paper, I analyse the functions of German modal particles and their English translation equivalents in order to discuss the relationship between grounding and common ground. I shall proceed as follows. First, I am going to propose that the main function of modal particles is to relate the current utterance to a particular aspect of common ground, namely a proposition “at hand”. In particular, their main function is to signal one’s understanding of what the situation is all about with respect to the argumentative relations built up in the current situation. Secondly, I am going to illustrate my proposal for the function of German modal particles as lexical markers of common ground on the basis of their English translation equivalents. The hypothesis is that the function of modal particles, to relate the current utterance to the argumentative background of the current situation and thus to indicate what is assumed to be common ground, is usually faithfully translated into the target language English (contrary to previous assumptions). An empirical investigation of data from a translation corpus will show which kinds of functions are translated and what the corresponding markers in English are. Thirdly, the analysis will subsequently be used to distinguish the linguistic labour of modal particles from other lexical markers of common ground, such as discourse particles. This discussion will lead us into distinguishing between common ground and grounding and between different types of common grounds.

2. GROUNDING AND COMMON GROUND Situational context is taken here to be something that is not given but which needs to be interactively established. Because it is impossible to pin down objectively what exactly a particular context consists of (e.g. Schegloff, 1997), participants need to signal what they consider the context to be to each other at all times. While it may be the case that there are externally given objects, states, or events that are salient to all participants, what the situation

48 Grounding and Common Ground consists in depends on the participants’ interpretation of these external factors. Context is therefore not external, but jointly established (Schegloff, 1997; Fischer, 2003). The set of assumptions that result from this process, comprising the implicitly and explicitly negotiated as well as the presupposed that the participants take to be shared, is the common ground between the interactants (Clark, 1996). Two kinds of processes can be distinguished. One has to do with what the speakers say and how they understand each other’s contributions, the grounding process (Clark and Schaefer, 1989). Participants display their understanding of each other’s utterances and the relationship that holds between these utterances to each other. These interpretations may be ratified or challenged by the coparticipants in the following turn (Sacks et al., 1974). Because speakers are in constant need to signal to each other their understanding of their partners’ utterances, the procedures are often implicit. For instance, Clark and Schaefer (1989: 267) list a number of techniques users employ for grounding, most of which are implicit: Continued attention, initiation of the relevant next contribution, acknowledgement (by means of nodding or items like uh-huh or yeah), demonstration, and verbatim display. “These types are grades roughly from the weakest to the strongest” (Clark and Schaefer, 1989: 267), and also from least explicit to most explicit. The result of this process is the conversational record, to use Thomason’s (1992) terminology, or the dialogue common ground in Fetzer’s terms (in this volume). The second major process involved is the signalling of common ground, the speakers’ assumptions about what the situation consists in as it is taken to be shared with their communication partners. Like the contents of the utterances which can only be assumed to be sufficiently grounded if they have been ratified by the communication partner, the communicative situation is not objectively given either. Speakers categorise situations on the basis of previous experience and make sense of what they encounter on the basis of personal, social, and socio-cultural knowledge. Clark (1996) presents a typology of aspects that speakers usually assume to constitute a shared basis, common ground. These aspects of the joint situation are treated as unproblematic, and the participants’ displays of their understanding of the situation are usually implicit (see also the criticism of Clark’s model in Pickering and Garrod, 2004). As Gumperz (1982, 2001) has shown, speakers use these cues to provide their communication partners with indicators as to how their utterances should be understood. That is, speakers display their understanding of the situation as it is necessary for the current utterances to be interpretable. This mechanism has been called framing (Tannen, 1979). Because of the concurrent nature of such displays, Gumperz has focussed on paralinguistic cues. However, as will be shown here, many different lexical and grammatical means can relate the current utterance to an assumed shared basis and thus function as markers of common ground. How are grounding and common ground then related? An obvious relation is that by means of grounding what is being said, common ground is being built up. Many researchers thus look at common ground as a set of propositions resulting from the grounding process. However, as I have argued above, situational context is not objectively and independently given, and thus also the speakers’ displays of their understanding of the situation contributes to the interactive constitution of the dialogue common ground. As Nemo (in this volume) argues, speakers co-define what is business as usual and what is worth attention and how this information should be taken into account with the goal of constituting interlocutive common ground. Moreover, both processes may be managed by the same items, items that can function

Grounding and Common Ground 49 as discourse as well as as modal particles. In the course of this paper, I try to disentangle these two aspects of the joint establishment of the situation further by investigating the role of discourse and modal particles. Discourse particles,1 particularly acknowledgement tokens such as uh-huh, yeah, or okay, serve to provide evidence of understanding to the communication partner. Much recent research has shown how skilfully speakers employ them to provide subtle information about the status of each utterance (cf., most notably, Gardner, 2001; but also Sorjonen, 2001, and Condon ˇ and Cech, in this volume). What is extremely interesting about them is that they can ground information chunks of very different sizes and types; consider the following uses of yeah: (1)

fmjm_3_03: okay, the third sounds good, and, we can get done, just in time so each of us can go to where we need to go for the evening. uh sounds like a date, how “bout you, is that good?” mdrd_3_04: yeah, it’s excellent.

(2) fsma_7_03: well Wednesday I’m busy all day, Tuesday the only time I would have would be at three in the afternoon. flmm_7_04: yeah, I’ll be uh I’ll be busy on Tuesday from two to four thirty, so, maybe we should make it for next week sometime? While the first example constitutes an acknowledgement token, accepting the date for the appointment proposed, in the second example the acknowledgement refers only to the successful transmission of the proposal, yet not to its acceptance. Thus, studying feedback is particularly challenging because acknowledgement tokens can contribute to the grounding process on so many different levels. However, besides directly referring to what is being said, discourse particles may also fulfil other functions, as in the following example: (3) mdkr_5_01: yeah, Cindy, there’s a, couple more things I’d like discuss with you. um can we get together for a couple hours, um this week or next? In this example, yeah is even the first word in the dialogue, so what it ratifies is the speakers’ joint project to sit down and schedule appointments. That is, they also evoke an aspect of the common ground, namely the joint understanding of what the task consists in. Similar uses ˇ of the discourse particle okay can be found in Condon and Cech (in this volume) who show that okay functions basically in two ways: signalling discourse boundaries and thus indicating agreement about the topic boundaries (the discourse marker use), and acknowledging what is being said (the feedback signal use). Similarly, Bangerter and Clark (2003) and Bangerter et al. (2004) argue that there are two uses of acknowledgements in conversation. Some markers are used in correspondence with the boundaries of joint projects and some to give feedback within such a project. However,

1

I take the term discourse particle to be largely synonymous with the term discourse marker, see Fischer (2005).

50 Grounding and Common Ground it is important to keep in mind that there are no two separate sets of items that fulfil functions with respect to the ratification of what is being said and with respect to the marking of discourse boundaries. As we have seen above, feedback signals like yeah also refer to a broad range of different types of information (see also Fischer (2005) for a detailed analysis of okay) and fulfil a broad spectrum of functions, such as topic marking, and signalling perception, understanding and acceptance (cf. also Allwood et al., 1992), even though it may fulfil one function more frequently than another. In these tasks discourse particles always refer to the here and now of the communicative situation. In Fischer (2000) I have argued that the functions discourse particles fulfil should therefore be understood with respect to a communicative background frame, explicating the attended-to domains relevant in verbal face-to-face interaction. In addition to the grounding work that discourse particles/markers may do, contributing to the joint verification of what is being said, their discourse boundary marker use is also related to the activity that the speakers are engaged in. This is related to not only what is being talked about but also about how the situation is being understood. Thus, discourse particles not only mark project boundaries but they constitute attempts at jointly establishing what the current activity consists in. Thus, discourse particles also function with respect to establishing common ground by evoking situational definitions. We can conclude that the distinction between grounding and common ground is not categorical. Moreover, the same items may fulfil functions with respect to both processes, although discourse particles are mostly involved in the grounding process. In the next section we shall see that German modal particles function as framing devices, or contextualisation cues (Gumperz, 1982), providing indirect evidence of what the speaker takes the joint situation to consist in. In this function, modal particles can be distinguished from discourse particles by their reference to the argumentative background of the interaction, which is presented as common ground. In contrast, discourse particles, which are in many cases homophonous with modal particles but which can be distinguished by their occurrence in distinct structural positions, refer to aspects of the here and now of the conversational interaction, including the status of the topic development.

3. MODAL PARTICLES AS LEXICAL MARKERS OF COMMON GROUNDS German modal particles, utterance medial pragmatic particles, such as ja, also, aber, and doch, can be described by means of the following attributes: They are more typical of spoken than of written language (Thurmair, 1989); unlike discourse particles, such as English oh, yes, okay, and well, they are intonationally integrated and occur inside sentence boundaries where their occurrence is restricted to particular syntactic positions (in particular, in the syntactic middle field (Abraham, 1991b)). Furthermore, modal particles are restricted to certain sentence types; for instance, the German modal particle ja, which in its discourse particle uses is often equivalent to English yes, may occur in assertions but not in questions. The meanings of modal particles do not modify particular segments but refer to the whole utterance (Lindner, 1991: 166). The meanings of modal particles are furthermore not truth-functional but indispensable for the “felicity conditions” of an utterance (Abraham, 1991a: 5). Their meanings are also considered to be syncategorematic (e.g. König and Stark, 1991), or to have at least a highly

Grounding and Common Ground 51 abstract semantic content (Lindner, 1991). Finally, many modal particles have a stressed and an unstressed variant whose interpretations are different, and they all have “homonyms” in other word classes, such as discourse particles, conjunctions, and adverbs.2 Modal particles have been proposed to fulfil a number of different functions in dialogues; thus they have been suggested to be illocutionary indicators (Helbig, 1977: 34; Kawashima, 1989: 281). They have also been argued to divide a sentence in given and new information (Krivonosov, 1989: 33–5) and to establish coherence (König and Requardt, 1991). Gelhaus (1995: 371) furthermore proposes that they express speaker attitude (see also Nehls, 1989: 283; Doherty, 1985). Regarding Swedish modal particles, Aijmer (1997) argues that they express epistemic modality. More generally, modal particles have been suggested to anchor an utterance in the communicative interaction (Thurmair, 1989: 2; Helbig and Buscha, 1986: 476), and their indexical nature has often been pointed out (see also Fillmore, 1984; Petric, 1995; Waltereit, 2001). Developing further an account of German modal particles that has been proposed first by Diewald and Fischer (1998) and further detailed in Diewald (2005) and Fischer (2000; 2005), the starting point in this investigation is also the indexical structure of modal particles. The assumption made here is that the general grammatical function of modal particles is to connect the current utterance to a pragmatic pretext, that is, to a proposition “at hand” which is part of the non-verbal argumentative context. In this model, the readings of a modal particle are created by interaction of the invariant meaning of the respective lexical item, the current utterance, and the element indexed: the pragmatic, argumentative, pretext. I take it that speakers need to display to each other in which ways their current utterances relate to the argumentative context (Ducrot, 1996, see also Nemo, this volume, and Nyan, in this volume). That is, with argumentation theory I hold that language is inherently dialogical and that every utterance contains several other “voices” that correspond to that part of the common ground that constitutes the argumentative background of each utterance. The assumed background needs to be conveyed to the recipient, as Nyan (in this volume) puts it: “initiating decision-making in others crucially depends on getting them to view the situation of reference as a member of a category of initial situations to which a category of response-options is paired.” One means to refer implicitly to aspects of the argumentative background, the situation “at hand”, is, I argue, German modal particles. The grammatical function specific to modal particles is thus to relate the current utterance to an aspect of the argumentative common ground, that is, to a propositional aspect of the communicative situation which has not necessarily been verbalised, but that is proposed to be shared. The semantic content of each modal particle lexeme specifies the relation as, for instance, additive or adversative. A schematic description of the meanings of modal particles, which represents the relationship between an utterance containing the modal particle and a pragmatic pretext, is the following: (4) common ground (pragmatic pretext: P) & modal particle (utterance, P)

2 Some researchers exclude the stressed variants from their analyses, for instance, Hentschel (1986). However, her description of the function of stressed doch (1986: 128) corresponds directly to the definition of modal particles employed here, and it is compatible to her own (1986: 2–3, cf. also Weydt, 1986: 401). I conclude that both stressed and unstressed variants need to be considered as modal particles (see also Diewald and Fischer, 1998).

52 Grounding and Common Ground The modal particle thus relates the current utterance to an aspect of the common ground, the pragmatic pretext. This function can be illustrated regarding the following example which constitutes the answer to a suggestion to meet as early as seven o’clock in the morning:3 (5) Natürlich, ich werde aber erst ab neun Uhr da sein. “I certainly won’t be there before nine o’clock.” Aber expresses an adversative relation between the pragmatic pretext and the relevant situation. The pragmatic pretext, the contextual assumption that is taken to be shared by the participants, is thus the negation of the situation described in the current utterance: (6)

pragmatic pretext: at hand: meet before 9 am relevant situation: I won’t be there before 9 am → utterance ich werde aber erst ab neun Uhr da sein

The grammatical function – to relate the current utterance, here, that the speaker will not be there before nine, to an aspect of the situation at hand, namely to meet before nine – is fulfilled by all modal particles, while the respective modal particle lexemes, here aber, specify the relation between pragmatic pretext and the utterance in which they occur, which is in this case adversative. The modal particle aber can thus be seen as an instruction to retrieve (or construe) a pragmatic pretext which is “a logical variant” (Foolen, 1989) of what is being said, in this case its opposite. The pragmatic function which results from the grammatical function to relate the utterance containing the modal particle to the contextual situation, or, more specifically, to the pragmatic pretext, is to mark an utterance as non-initial, that is, as a natural consequence of the communicative situation. At the same time the speaker’s own contribution is played down, and hence possibly offensive acts become less face-threatening (Brown and Levinson, 1987). The main pragmatic function of modal particles is thus interpersonal (see also Weydt, 2005; Fischer, 2005), or, as Nemo (in this volume) puts it, it is justificational. However, this interpersonal function does not exclude functions on other levels of discourse; like discourse particles, modal particles can fulfil several functions at the same time (Fischer, 2000), but they do so in virtue of their grammatical function to relate the current utterance to the argumentative context. While in the previous example the proposition indexed by means of the modal particle had just been at issue, modal particles may also refer to what the current speaker assumes to be part of the situation “at hand” on the basis of previously discussed information. For instance, in example (7), by using the modal particle doch the speaker indexes a proposition that suggests that the meeting should not be postponed to the next week: (7) ich denke wir sollten das Ganze dann doch auf die nächste Woche verschieben. “I think we should rather postpone the whole matter until next week.”

3

The examples are from a corpus of German appointment scheduling dialogues which were translated into English. The corpus will be described in more detail in Section 4.

Grounding and Common Ground 53 Stressed doch has the same function as modal aber: it refers to a proposition “at hand” that does not correspond to the situation expressed. This proposition is claimed to be accessible to both; it is claimed to be common ground (cf. Pittner, in this volume). In fact, in this dialogue it has been uttered by one of the speakers several turns before. Thus, in this example, the pragmatic pretext, the proposition “at hand”, is part of the common ground because it is part of the interactionally grounded dialogue history. However, the pragmatic pretext need not have been mentioned at all. There are also uses of modal particles that do not refer to anything that has been previously said, but to a proposition that is assumed to be held by both communication partners. This is the case for supposedly perceptually available aspects of the shared context and for evaluations that are taken to be shared. For instance, in the following example, the modal particle aber is used to refer to an assumption that has not been talked about before, that the task (to construct a toy airplane) is actually quite complicated (although it is a children’s toy):4 (8) ja. oh, das ist aber schwierig. “yes. oh this is difficult after all”. That is, the use of modal aber proposes that there is a shared evaluation of the task as not difficult “at hand”. Modal particles thus indicate a particular proposition that is claimed to be part of the argumentative context. Common ground is therefore not a fixed set of propositions previously established, but a definition of the situation that is interactively negotiated (see Nemo, Fetzer, in this volume). Modal particles contribute to the negotiation process by indicating such propositions. At the same time, they present an utterance as argumentally related to the supposedly shared situation. An aspect currently under discussion is the question whether there should be a single invariant meaning for each particle or whether a number of different senses should be assumed. König (1997: 59), for instance, criticises that the search for a common meaning is often given up too easily in favour of a polysemic approach. Also Thurmair (1989) and Abraham (1991b) argue for a minimalist approach, which assumes as few distinct readings for each particle as possible. Waltereit (2001: 1414), however, rejects such an account because in his view it has to be traded against very abstract meaning descriptions. This may indeed be true, which is however not problematic in itself. For instance, no one would object if the meaning description of, say, the topicalisation construction would be abstract. The point is instead to provide accounts that are concrete enough to allow the prediction of an interpretation of the respective particle in context. While the morphemic meaning may indeed be an abstract but invariant factor in the interpretation of a modal particle, it may not be the only one. The structural position, that is, the construction in which a morpheme is employed, may contribute further information that allows the interpretation of each particle in its respective context (see also Fischer, 2000, 2005). In our case, tied to the grammatical function of modal particles, there is a particular interpretative mechanism, which is coded in the construction in which the respective morpheme occurs. The morphemic meaning of each particle is indeed taken to be the same 4 This example stems from the toy airplane construction dialogues recorded in the framework of the SFB 360 in Bielefeld (Sagerer et al., 1995, cf. Diewald and Fischer, 1998). The translation is mine.

54 Grounding and Common Ground for its occurrences as modal particle, as well as as adverb, conjunction, or discourse particle, or whatever “homonyms” in other word classes the modal particle may have. However, the invariant meaning tied to a particular lexical item interacts with the constructional meaning of its particular use, for instance, as a modal or discourse particle, in a certain structural context (see Fischer, 2005). The approach presented is thus minimalistic such that an invariant meaning accounts for the relationship between the different readings of a modal particle, as well as between these and uses of the same morpheme in other word classes. It is however specific enough to account for particular occurrences because of the interpretative contribution of the respective constructions. Another important issue in models of modal particles is the role of context. As mentioned above, most approaches hold that modal particles are syncategorematic and thus get their interpretations from the contexts in which they occur, and that they relate in some way or other to the communicative situation. There are, however, different ways of invoking context. Waltereit (2001) presents a recent approach that makes use of the notion of context. His approach is speech–act theoretic such that modal particles are taken to encode information on the preparatory conditions of speech acts by calling up particular situations that are metonymically related to the non-modal counterparts of each modal particle. The mechanism involved may however be very different for different modal particles. For instance, for modal ja, the situation called up is one in which, during an interaction of several turns, the hearer has already agreed on a proposition by using ja as an answer signal (2001: 1413). In contrast, the French particle quand-même points to a situation in which real-world norms hold (2001: 1403f), and the diminutive in Italian points to non-serious situations, such as talking to children (2001: 1409–10). In Waltereit’s model, the common basis of modal particles is thus that they all point to particular situations, which are related by contiguity to the use of their non-modal counterparts. The result is a polysemic approach. This could also mean giving up the search for a common meaning too easily. Furthermore, the heterogeneity of the situations proposed to be involved in the meanings of modal particles indicates the danger of limited predictability and thus restricted learnability of the mechanism proposed. Another way of making use of context in the description of modal particles is to claim that they refer to contextual assumptions. König (1997) proposes a relevance-theoretic approach, which is basically along similar lines to the approach outlined here: the modal particle refers to a contextually available proposition. However, relevance theory suffers from the fact that it does not provide a way for identifying the contextual propositions that are “relevant” (Nemo, 1988). There is no independent evaluation of the “processing costs,” which, according to relevance theory, speakers try to avoid by maximising the informational value available at the same time (Sperber and Wilson, 1986). The selection of contextual propositions is, however, completely open in relevance theory, and thus the criticism Wierzbicka (1990) applied to prototype theory, namely that prototypes are used to save descriptions that make predictions that are too narrow or even wrong, can be applied similarly to relevance theoretic accounts of modal particles regarding the notion of context. In contrast, our approach as proposed here states directly what the pragmatic pretext is – it consists of the proposition expressed in the current utterance minus the contribution of the particle morpheme. The model proposed here is thus not a “context saves” model, but allows a principled way to investigate how speakers make use of, and negotiate at the same time, aspects of common ground.

Grounding and Common Ground 55

4. CORPUS In the following, occurrences of German modal particles and their translations will be analysed, first in order to illustrate the model proposed, and second to identify a set of lexical markers of common ground for English. Since these items serve as translation equivalents for German modal particles, they will, if the model proposed here is correct, also function to relate the current utterance to the argumentative background which is assumed to be common ground by the participants. Three German appointment scheduling dialogues recorded in the Verbmobil project (Wahlster, 2000), for which there are five English translations (by professional translators and students of translation), serve as the data set for this investigation. The translators, who are all, with the exception of (d), native speakers of German, which is the source language in this corpus, were the following: a. b. c. d. e.

graduate student of translation studies Dipl.-Fachübersetzerin (professional translator) graduate student of translation studies, exam candidate lecturer, native speaker of English undergraduate student of translation studies

All five translations will be considered in order to get an overview of different translation possibilities, even though the translations (e) have been created by an undergraduate student who still makes mistakes with respect to the target language. The translations were carried out on the basis of the original spoken dialogues to which the translators could listen as often as they liked. They produced written translations of the original taped speech in the three dialogues. The transcriptions are based on the whole dialogues, not on single turns, and thus the translators and interpreters were involved in a “holistic enterprise” (O’Sullivan and Rösler, 1989).

5. ENGLISH EQUIVALENTS TO GERMAN MODAL PARTICLES Previous contrastive approaches have analysed which lexical and grammatical correspondents modal particles may have; for instance, conjunctions, adverbs, formulaic expressions, auxiliary verbs, and tag questions have been found to be functional equivalents (e.g. Fillmore, 1984; Nehls, 1989; Abraham, 1991b: 206; Fischer and Drescher, 1996: 855–6), that is, to have “essentially the same pragmatic conditions” (Fillmore, 1984: 133). König and Stark (1991: 304), in a lexicological approach to modal particles, argue that “equivalence often has to be established on the level of the next larger unit, the phrase, the clause, or the whole sentence.” However, in studies which also consider the distribution and use of modal particles, most often no equivalents were identified at all (Fillmore, 1984; O’Sullivan and Rösler, 1989; Abraham, 1991a: 2; Fischer and Drescher, 1996); Fillmore (1984: 133), for instance, argues about the translation equivalents of German modal particles that “these expressions, if used in English as often as their counterparts are used in German, would produce very mannered speech.”

56 Grounding and Common Ground There may be, in principle, two reasons for this observation. One reason may be that German speakers attend to meanings or functions to which speakers of other languages such as English do not attend; that is, German modal particles may express something that other languages do not express. This opinion is held, for instance, by Fillmore who explains the differences between the two languages as a matter of “larger” pragmatic practices, such that it is a “‘large’ pragmatic fact about German (  ) that the colloquial language welcomes (one may almost say “requires”) pragmatic particles that reflect choices in which individual utterances can be situated in their discourse context. (  ) the corresponding forms in English cannot be grammatically incorporated into surface clauses” (Fillmore, 1984: 133). The difference is attributed to a different level of formal complexity: “This difference seems to exist by virtue of the fact that the German forms are ‘particles’ while the English forms are ‘formulas’” (1984: 133–4). Some lines below Fillmore adds a second reason for the different distribution of particles and their proposed functional equivalents: “The description of these ‘larger’ patterns of use appears to me to be about as difficult to come by as a description of the two cultures” (1984: 134). Eventually, however, he leaves open what these “larger” facts are. Similarly, O’Sullivan and Rösler (1989) argue that in the translation of English texts into German the use of modal particles should be seen as a holistic enterprise depending on the text type. However, they do not explain what the properties of the respective text types are which make the use of modal particles relevant, and what is so special about German conversation in contrast to English conversation. A more likely reason for the fact that often no translation equivalents have been found could be that it was not clear what the functions of modal particles actually are; none of these previous contrastive approaches described above has started out from a concept of what the grammatical function of the word class modal particle is or what general pragmatic function they actually express. Instead, their search for translation equivalents was guided by local decisions on the current pragmatic function of each modal particle occurrence. This procedure often yields no correspondents at all. However, an attempt to find functional equivalents in another language may be successful if we investigate a more general level than the actual functional interpretation of a modal particle occurrence in a given context. Thus, the determination of translation equivalents may be more successful on the basis of a thorough functional analysis of the grammatical function of the word class. Results from such a functional analysis are also necessary in order to explain what the “larger” pragmatic facts are that determine different distributions in the two languages, if there are any. Therefore the procedure in this investigation is to reanalyse the pragmatic functions of modal particles and thus to pave the ground for a new analysis of their translation equivalents.5 The translation of modal particles into languages that do not have a comparable set of modal particles, such as English (but see, for instance, Aijmer, 1997), allows us to identify a number of other types of linguistic strategies that fulfil the same function. Although it has frequently been argued that no translation equivalents can be found, at least not to a comparable amount (Nehls, 1989; Fillmore, 1981), in particular the function of modal particles to relate

5

It has also been proposed that the linguistic work done by modal particles in German is done by prosody in English (Schubiger, 1965). Prosody indeed has been found to be used as a contextualization cue, that is, to fulfil the function proposed for modal particles in this chapter (Gumperz, 1992, 2001). However, prosody self-evidently also fulfils this function in German (e.g. Krivonosov, 1963: 62). It is therefore taken to be a justified procedure here to concentrate on the lexical and grammatical correspondents of German modal particles which are visible in the transcripts.

Grounding and Common Ground 57 the current utterance to the common ground can usually be found in the translation. That is, if the translation is based not only on individual utterances, but also on the argumentative context, the general grammatical function of modal particles can be demonstrated to be present also in the English target. For the grammatical function of modal particles, of relating the current utterance to the assumed argumentative common ground, it is usually possible to find translation equivalents in the translated dialogue. The relationship between pragmatic pretext and current utterance in English is not only expressed by particular lexical or grammatical means but by the interaction of a large number of factors, several alternatives being possible; for instance, in the example (9) of doch below, discussed in part as example (7) above, we find a number of different possibilities: Dienstags um 10 ist bei mir jetzt wiederum schlecht, weil ich da noch trainieren bin. Ich denke wir sollten das Ganze dann doch auf die nächste Woche verschieben. Geht’s bei Ihnen da? With the current utterance the speaker suggests postponing the appointment to the following week. This is in contrast to his partner’s earlier proposal to meet in the current week, but it turned out that no date could be found without sacrifices on either side. The implicit pragmatic pretext is here that the communication partner would prefer not to postpone the meeting. The speaker now uses doch to relate the current proposal to the pragmatic pretext, that is, the implicit proposition not to postpone the appointment, stressed doch signalling a relationship of contrast. This is rendered by the translators in the following ways: a. Tuesday at 10 doesn’t suit me, at that time I’ll still be, em, doing sports. I think we’d better postpone the meeting till the week after. Does that suit you? b. Tuesday ten o’clock is bad with myself, since I’m still at my sports training then. I think we should put it off until the following week. Would that be fine with you? c. Tuesday at ten is bad on my part since I will be training then. I think we should rather postpone the whole matter until next week. Is that okay with you? d. Tuesdays at ten is inconvenient for me on the other hand because I am still off training. I think in that case we should postpone the whole thing to the week after. Would that be all right with you? e. On Tuesday at ten it would be difficult for me because I still will be training then. In this case I think we should postpone our date to next week. Would that be all right with you? While translation (b) does not reflect the function of the modal particle doch at all, the translator of (a) uses better, which is relational and, like the modal particle doch, compares the current to the previous proposal. In (c) we find the relational item rather discussed above. The translator of (d), who is a native speaker of English, renders the relational structure first by means of on the other hand (although there has been no on the one hand), and secondly by in that case, thus implying a relevant alternative which is in contrast with the current proposal. Likewise, translation (e) uses in this case to refer to the pragmatic pretext. Thus, several of the translation variants activate the propositions “at hand” while they do not necessarily signal the relationship between the current utterance and the argumentative

58 Grounding and Common Ground common ground. Thus, rarely is a relationship specified that would be as strong as the adversative relationship indicated by German doch. Whereas the content of the modal particle lexeme itself is reduced to a weak opposition between two alternatives, one of which is implicit, the grammatical function of the modal particle doch is translated into the target language in most cases. In the following example of German doch, whose interpretation in this utterance can be paraphrased as “contrary to our previous idea to meet later in the day,” its lexical meaning is again not transferred into the target language. Instead, we find better in translations (a) and (d), then in (c), and in (a) and (e) the other hypothetical situation is referred to by means of the subjunctive mood; thus, as in the previous example of doch above, it is the relational structure that is particularly translated into English. That is, while the pragmatic pretext is called up as in the German source, the relationship between pragmatic pretext and current utterance, which is contrastive in German, is not specified in the same way in English. (10) Sie haben recht, das wird etwas knapp. Sollten wir vielleicht doch gleich um 10 Uhr anfangen? a) Well, you’re right. We might not have enough time. Do you think it’s better to start at 10 o’clock sharp? b) You are right – it’s a bit tight – should we start right at ten? c) You’re right. We’ll be short of time then. Maybe we should start right at ten o’clock. d) You are right, that is a bit too short. It would be better to start at 10. e) You are right. This would be a bit short. May be we should already start at 10? The analysis is thus: (11)

pragmatic pretext: at hand: we should start later than 10am relevant situation: we should start right at 10am → utterance Sollten wir vielleicht doch gleich um 10 Uhr anfangen?

In the next example, the speaker reacts to the partner’s proposal to consider a particular week by stating his availability during this week which is restricted to Friday. That is, the work done by the utterance containing allerdings is to state (preliminary) acceptance of the previous proposal but to formulate a restriction: “given that we take this week, I have got the following problem: I can only meet you on Friday’ ” (12)

pragmatic pretext: at hand: we meet this week relevant situation: in this week I only have time on Friday → utterance da ging es bei mir allerdings nur am Freitag

In the translations, translation (b) represents this relation between the current utterance and the proposal made by the communication partner; the work then is doing is very similar to the work done by allerdings: “if we take this week, then I’ll have the following problem: I can only meet you Friday”. Similar work is done by the subjunctive mood in translation (a): “if

Grounding and Common Ground 59 we decide for this week, then only Friday would be possible” (cf. Aijmer, 1997). In translation (c) the proposal to meet next week, which constitutes the pragmatic pretext of the utterance, is referred to by means of next week. Both (d) and (e) use unfortunately which is not relational; instead, translations (d) and (e) express an evaluation, thus showing that the speaker does not reject the partner’s proposal thoughtlessly or without good reason; the translators of (d) and (e) have thus chosen to express the interactive consequences the restriction to Friday may have: (13) Da ging es bei mir allerdings nur am Freitag, da ich die ganze Woche über noch in Washington bin. Wie sieht’s am Freitag aus? a) It would only be possible on Friday since I’ll be in Washington the rest of the week. What about Friday? b) Then, only Friday would be free, since I’ll be in Washington for the whole week. What about Friday? c) The only day I’m free next week is Friday. I’m spending the remainder of the week in Washington. How about Friday then? d) Unfortunately I am only free on Friday that week because I am in Washington the whole week. What about Friday? e) Unfortunately only on Friday because I still will be in Washington on Thursday. What about Friday? Thus, there are a number of choices for translating the function fulfilled by the German modal particle. There are, however, also examples of modal particles whose translation is more difficult to recognise than has been the case with the previous examples because of redundant coding of the same function, as, for instance, in the following example of denn: (14)

Ja, Frau Petz, dann lassen Sie uns doch einen Termin ausmachen. Wann wäre es Ihnen denn recht? a) Well, Mrs Petz, then let’s fix an appointment. When would it suit you? b) Well, Ms./Frau Petz; then let us fix a date. When would you like to meet? c) Let’s make an appointment, Frau Petz. When would it suit you best? d) Good, Frau Petz, let’s arrange a time then. When would it suit you? e) Yes, Mrs Petz, then let’s make a date yet. When would you like it?

The function of denn is to relate the questioning act to the communicative situation. As a conjunction, denn is consecutive. Here, it serves to indicate that the current utterance follows as a consequence from the situation. This is similar to and-prefacing in interview situations; Heritage and Sorjonen (1994) show how the use of and-prefacing by a nurse may present her possibly face-threatening questions as based on a list of questions for which she is not responsible. By presenting the questions as non-initial, the force of the questions can therefore be reduced. Similarly, the modal particle denn in the example here, which constitutes a request to propose a time and as such a possibly face-threatening act (Brown and Levinson, 1987), serves to maintain the interpersonal relationship between the speakers by presenting the utterance as a consequence of the situation. In particular, the claim is that asking the question itself has already been part of the common ground; it has been at hand already.

60 Grounding and Common Ground The subjunctive mood in the German source text may be argued to fulfil the same face-preserving function in a different way; thus, while there are no particular correspondents for denn in the translations, the reason may be that subjunctive mood and modal particle redundantly code the same pragmatic information, indicating something like “if I was allowed to ask, I would ask:”. Thus only one of these means is transferred into the target utterance; note that all five translations use the subjunctive. (15)

pragmatic pretext: at hand: if I may ask: when would it suit you? relevant situation: I ask: when would it suit you? → utterance Wann wäre es Ihnen denn recht?

In another example of denn, two translators employ then, which corresponds to the consecutive meaning of denn as a conjunction. Interestingly, one translator first signals agreement before she continues with the proposal: (16)

Wie wär’s denn um dreiviertel elf? a) What about a quarter to 11? b) How about a quarter to eleven then? c) Okay. How about a quarter to eleven. d) How about a quarter to eleven e) Yes, what about 10.45 then?

Thus, even denn in questions can be shown to be realised in several English translations. Returning to example (14) above, I would like to argue that similarly the relationship to the pragmatic pre-text, whether this is the moment to sit down and schedule appointments, expressed by unstressed doch in example (14) is redundantly coded by dann, which is consistently translated by then, and by the initial discourse particle ja, rendered as yes, well, and even as good by the translators. In this case, the initial discourse particle ja has the same function as the discourse particle yeah in example (3). That is, the contribution of modal doch here is very similar to the other two devices, namely to refer to the participants’ willingness to sit down and to schedule appointments, since this is the argumentative background of the beginning of this dialogue. Similarly, in the next example of unstressed doch, of which we unfortunately only have two translations, one translator refers to the pragmatic pretext indexed by means of as far as I remember, while the other one includes before as a time index for a previous conversation: (17) Guten Tag Herr Müller, wir hatten uns doch schon gesprochen und uns einen Termin Ende April Anfang Mai ausgedacht. Hatten Sie da Zeit? a) Hello, Mr Müller. As far as I remember we have already talked and arranged an appointment for the end of April or beginning of May. Would that suit you? b) Hello Herr Müller we have talked on the phone before and have made an appointment for the end of April or the beginning of May. Would that suit you? Another example is the modal particle also, which is by most translators rendered by means of then, presenting the current utterance as a consequence of the argumentative situation. One translator, however, employs the phrase to recap to account for the meaning of also, thus

Grounding and Common Ground 61 indicating how the modal particle refers to the whole topic, as it was discussed in connection with discourse particles at the beginning of this paper: (18) Schön dann machen wir’s so. Das war also Freitag, 23. Oktober, 14 Uhr 45. Alles klar. Bis dann. Tschüss. a) Okay, then we’ll meet on Friday, 23rd of October, 2.45 pm. See you then, good_bye. b) O.K. We’ll do it this way. Then it is Friday the twenty-third of October – two forty-five p.m.; all right; see you; bye. c) Okay let’s do it like this. It’s Friday, 23rd October, at a quarter to three then. Alright. See you. Bye. d) Right that’s it then, to recap Friday 23rd of October at a quarter to three. Got it, see you then, bye e) Well, we make it this way. On Friday, the 23rd of October at 2.45. Everything okay. See you then. Bye. We can thus identify an interesting overlap between discourse particles and modal particles here. While discourse particles can refer to the thematic organisation of the dialogue because it is part of the current ongoing activity, modal particles may refer to the topic structure because it is part of the shared argumentative background. Thus, without having to give up the functional descriptions of discourse and modal particles respectively, we can account for the functional equivalence of the two types of particles with respect to thematic organisation. To sum up, for some modal particles, their lexical content was not always found to be translated, whereas their relational structure with reference to the pragmatic pretext was usually transferred into the target language. Regarding the translators, although these few examples of no more than five translators are admittedly too few to allow reliable conclusions, some differences became apparent. While especially the native speaker of English translated the German modal particles occurring by means of rendering the relational structure proposed for modal particles, the professional German translators translated the relational structure of modal particles less often. It may be speculated that, resulting from the fact that previously the function of modal particles to relate the current utterance to the pragmatic situation at hand had not been identified and thus was not reported to students of translation, the missing translation equivalents in the professional translators’ translations are actually an artefact of their training. In any case it can be concluded that it is not a (“larger”) fact about English that it would be lacking a certain pragmatic function that is fulfilled by modal particles in German, but that its expression had just not been realised.

6. PRAGMATIC MARKERS AND THE RELATIONSHIP BETWEEN GROUNDING AND COMMON GROUND The results from this investigation show that the hypothesis that there is a relational semantic basis and grammatical function of modal particles which explains their pragmatic functions is supported by the contrastive analyses; especially their relational structure is rendered in the target language by a number of different lexical items and grammatical constructions which all share the same relational structure. Contrary to discourse particles, which are

62 Grounding and Common Ground intimately connected to the here and now of the actual communicative process, including the current activity in which the speakers are involved, modal particles refer to one particular aspect of the situation: to a proposition that constitutes part of the argumentative background of the current utterance and which is taken to be shared by the communication partners. Particularly contrastive modal particles usually evoke alternative situations, other “voices” that are “at hand”, that serve as implicit shared background for the current discussion. Many English translation equivalents indicate such alternative situations, for example the use of the subjunctive and relational items like better, rather, or on the other hand. While discourse particles thus contribute to establishing the current situation, a big part of which is what is being said, and thus mostly contribute to the grounding process, modal particles claim to refer to “voices” that are implicitly part of the assumed common ground. They therefore also contribute to the grounding of the situation, by evoking propositions that (are claimed to) exist as shared background, but which in fact constitute interpretations of what the situation consists in. In this respect it may be argued that modal particles “update common ground” as Pittner (in this volume) claims.6 The process of grounding and the evoking of common ground are thus highly interrelated and co-dependent. Within this continuum, however, discourse and modal particles are located differently with respect to different kinds of information activated, an overlap, however, occurring with respect to the thematic structure which can be referred to by both kinds of particles. Another result of this investigation is that only a hypothesis about the general function of modal particles allows us to identify the translation equivalents of German modal particles. Thus, not only the relational hypothesis can be verified in this contrastive study, but also several ways by which the function of indicating aspects of common ground may be realised in English could be determined. Candidates for translation equivalents are, for instance, adverbs, such as unfortunately and really, but more often relational concepts,7 such as better, rather, in that case, and even subjunctive mood. Does this mean that all these devices have to be considered as lexical markers of common ground, indicating what the speaker considers to be the context of his/her utterance? Do our results lead to an inflation of the concept of lexical markers of common ground? Our hypothesis was that speakers in interaction need to signal to each other what they consider as the argumentative background of the interaction, and it was proposed that in German this may be done by means of modal particles. If this constitutes, however, a general need of the participants in interaction, English should equally have devices to fulfil this function. Besides extensive literature on the grounding process, which is concerned with the creation of common ground, not much is known about how speakers make their perspective on this common ground available to each other. One influential study is Tannen (1979) who identifies sixteen different surface cues that allow her to infer the contents of the background information that speakers draw upon in their narratives; for instance, negation, generalisation, moral judgement, repetition, false starts, backtracking, hedging, and so on. In this paper, a closer look at two such devices has been taken: discourse and modal particles. For discourse particles it was found that they contribute to common ground as acknowledgements of what is

6

However, the notion of updating presupposes that common ground has been different before, which may or may not be true, depending on the respective modal particle. I would therefore propose to restrict the suggestion to a contribution to the situational definition. 7 See also Okada, this volume, who claims all deictic items to function as lexical markers of common ground.

Grounding and Common Ground 63 being perceived, understood, and accepted and as indicators of the status of the current activity. They thus serve in establishing a joint perspective on the current communicative situation. Modal particles, in contrast, refer to the argumentative common ground. Speakers present their information as contributions to larger argumentative activities such as contrasting, supporting, evidencing, and so on, and these rhetorical relations are indicated by means of modal particles. Thus, modal particles are framing devices par excellence, indicating the current activity, but as their translation equivalents in English have shown, they are by far not the only means to do so. Modal particles are grammaticalised means to refer to the argumentative common ground, as much as feedback signals are grammaticalised means to ground utterances. However, as we have seen, many other means may be used as lexical markers of common grounds.

REFERENCES Abraham, Werner, (ed.) 1991a. Introduction. Discourse Particles: Descriptive and Theoretical Investigations on the Logical, Syntactic, and Pragmatic Properties of Discourse Particles in German. Benjamins, Amsterdam, 1–10. Abraham, Werner, (ed.) 1991b. Discourse particles in German: How does their illocutionary force come about? Discourse Particles: Descriptive and Theoretical Investigations on the Logical, Syntactic, and Pragmatic Properties of Discourse Particles in German. Benjamins, Amsterdam, 203–252. Aijmer, Karin, 1997. “I think” – an English modal particle. In: Swan, T. and Westvik, O.J. (eds), Modality in Germanic Languages: Historical and Comparative Perspectives. Mouton de Gruyter, Berlin, 1–47. Allwood, Jens, Nivre, Joachim, and Ahlsén, Elisabeth, 1992. On the semantics and pragmatics of linguistic feedback. Journal of Semantics 9, 1–26. Bangerter, Adrian and Clark, Herbert H., 2003. Navigating joint projects with dialogue. Cognitive Science 27, 195–225. Bangerter, Adrian, Clark, Herbert H., and Katz, Anna R., 2004. Navigating joint projects in telephone conversations. Discourse Processes 37, 1–23. Brown, Penelope and Levinson, Stephen, 1987. Politeness. Some Universals in Language Usage. 2nd edition. Cambridge University Press, Cambridge. Clark, Herbert H., 1996. Using Language. Cambridge University Press, Cambridge. Clark, Herbert H. and Wilkes-Gibbs, Deanna, 1986. Referring as a collaborative process. Cognition 22, 1–39. Clark, Herbert H. and Schaefer, Edward R., 1989. Contributing to discourse. Cognitive Science 13, 259–294. ˇ Condon, Sherri and Cech, Claude, this volume. Ok, next one: Discourse markers of common ground. Diewald, Gabriele, 2005. Discourse particles and modal particles as grammatical elements. In: Fischer, K. (ed.), Approaches to Discourse Particles. Elsevier, Amsterdam. Diewald, Gabriele and Fischer, Kerstin, 1998. Zur diskursiven und modalen Funktion der Partikeln aber, auch, doch und ja in Instruktionsdialogen. Linguistica 38 (1), 75–99. Doherty, Monika, 1985. Epistemische Bedeutung [Studia Grammatica 23]. Akademie-Verlag, Berlin.

64 Grounding and Common Ground Ducrot, Oswald, 1996. Slowenian Lectures/Conférences Slovènes. Edited by Igor Zagar, ISH, Ljubljana. Fetzer, Anita, this volume. Reformulation and common grounds. Fillmore, Charles J., 1981. Pragmatics and the description of discourse. In: Cole, P. (ed.), Radical Pragmatics. Academic Press, New York. Fillmore, Charles J., 1984. Remarks on contrastive pragmatics. In: Fisiak, J. (ed.), Contrastive Linguistics: Prospects and Problems [Trends in Linguistics: Studies and Monographs 22]. Mouton, Berlin etc., 119–141. Fischer, Kerstin, 2000. From Cognitive Semantics to Lexical Pragmatics. The Functional Polysemy of Discourse Particles. Mouton de Gruyter, Berlin, New York. Fischer, Kerstin, 2003. Notes on analysing context. In: Kühnlein, P., Rieser, H., and Zeevat, H. (eds), Perspectives on Dialogue in the New Millennium. John Benjamins, Amsterdam and Philadelphia. Fischer, Kerstin, 2005. Frames, constructions and morphemic meanings: The functional polysemy of discourse particles. In: Fischer, K. (ed.), Approaches to Discourse Particles. Elsevier, Amsterdam. Fischer, Kerstin and Drescher, Martina, 1996. Methods for the description of discourse particles. Language Sciences 18 (3–4), 853–861. Foolen, Ad, 1989. Beschreibungsebenen für Partikelbedeutungen. In: Weydt, H. (ed.), Sprechen mit Partikeln. Mouton de Gruyter, Berlin, 305–317. Gardner, Ron, 2001. When Listeners Talk. Pragmatics & Beyond New Series 92, John Benjamins, Amsterdam. Gelhaus, Hermann, 1995. Die Wortarten. In Drosdowski, G., et al. (eds), Duden: Grammatik der deutschen Gegenwartssprache. 5., völlig neu bearbeitete und erweiterte Auflage. Dudenverlag, Mannheim etc., 85–398. Gumperz, John, 1982. Discourse Strategies. Cambridge University Press, Cambridge. Gumperz, John, 2001. A discussion with John J. Gumperz, Carlo L. Prevignano and Aldo di Luzio. In: Eerdmans, S.L., Prevignano, C.L., and Thibault, PJ. (eds) Discussing Conversation Analysis. Benjamins, Amsterdam, Philadelphia. Helbig, Gerhard, 1977. Partikeln als illokutive Indikatoren im Dialog. Deutsch als Fremdsprache 14, 30–44. Helbig, Gerhard and Buscha, Joachim 1986. Deutsche Grammatik. Verlag Enzyklopädie, Leipzig. Hentschel, Elke, 1986. Funktion und Geschichte deutscher Partikeln. Ja, doch, halt und eben. Niemeyer: T¨ubingen. Heritage John, and Sorjonen, Marja-Leena, 1994. Constituting and maintaining activities across sequences and prefacing as a feature of question design. Language in Society 23, 1–29. Kawashima, Atsuo, 1989. Textsorte und Partikeln im Japanischen und Deutschen. In: Weydt, H. (ed.), Sprechen mit Partikeln. DeGruyter, Berlin, New York, 276–281. König, Ekkehard, 1997. Zur Bedeutung von Modalpartikeln im Deutschen: Ein Neuansatz im Rahmen der Relevanztheorie. Germanistische Linguistik 136, 57–75. König, Ekkehard and Requardt, Susanne, 1991. A relevance-theoretic approach to the analysis of modal particles in German. Multilingua 10 (1–2), 63–77. König, Ekkehard and Stark, Detlef, 1991. The treatment of function words in a bilingual German-English dictionary. In: Abraham, W. (ed.), Discourse Particles: Descriptive

Grounding and Common Ground 65 and Theoretical Investigations on the Logical, Syntactic, and Pragmatic Properties of Discourse Particles in German. John Benjamins, Amsterdam, 303–328. Krivonosov, Alexej, 1963/1977. Die modalen Partikeln in der deutschen Gegenwartssprache. Alfred Kümmerle, Göppingen. Krivonosov, Alexej, 1989. Zum Problem der Klassifizierung der deutschen Partikeln. In: Weydt, H. (ed.), Sprechen mit Partikeln. DeGruyter, Berlin, New York, 30–38. Lindner, Katrin, 1991. “Wir sind ja doch alte Bekannte.” The use of German ja and doch as modal particles. In: Abraham, W. (ed.), Discourse Particles: Descriptive and Theoretical Investigations on the Logical, Syntactic, and Pragmatic Properties of Discourse Particles in German. John Benjamins, Amsterdam, 303–328. Nehls, Dietrich, 1989. German modal particles rendered by English auxiliary verbs. In: H. Weydt (ed.), Sprechen mit Partikeln. DeGruyter, Berlin, New York, 282–292. Nemo, François, this volume. The pragmatics of common ground: From common knowledge to shared attention and social referencing. Nemo, François, 1988. Book review of Relevance: Communication and cognition: Dan Sperber and Deirdre Wilson, Oxford: Basil Blackwell, 1986. Journal of Pragmatics 12 (5–6), 791–795. Nyan, Thanh, this volume. Common ground, categorization and decision-making. Okada, Moeko, this volume. Whose common ground? A misunderstanding caused by incorrect interpretations of the lexical markers of common ground. O’Sullivan, Emer and Rösler, Dietmar, 1989. Wie kommen Abtönungspartikeln in deutsche Übersetzungen von Texten, deren Ausgangssprachen für diese keine direkten Äquivalente haben? In: Weydt, H. (ed.), Sprechen mit Partikeln. DeGruyter, Berlin, New York, 204–216. Petric, Teodor, 1995. Indexikalische Leistungen der Partikeln und ihre natürlichkeitstheoretische Bewertung. Linguistica 35, 245–259. Pickering, Martin J. and Garrod, Simon, 2004. Toward a mechanistic psychology of dialogue. Behavioral and Brain Sciences 27, 169–226. Pittner, Karin, this volume. Common ground in interaction: The functions of medial doch in German. Sacks, Harvey, Schegloff, Emanuel, and Jefferson, Gail, 1974. A simplest systematics for the organization of turn-taking for conversation. Language 50 (4), 696–735. Sagerer, Gerhard, Rickheit, Gerd, and Eikmeyer, Hans-Jürgen, 1994. Wir bauen jetzt ein Flugzeug. Materialband. SFB 360 “Situierte künstliche Kommunikatoren”. University of Bielefeld. Schegloff, Emanuel A., 1997. Whose text, whose context? Discourse & Society 8 (2), 165–187. Schubiger, Maria, 1965. English intonation and German modal particles: A comparative study. Phonetica 12, 65–84. Sorjonen, Marja-Leena, 2001. Responding in Conversation. A Study of Response Particles in Finnish. Pragmatics & Beyond New Series 70, John Benjamins, Amsterdam. Sperber, Dan and Wilson, Deirdre, 1986. Relevance: Communication and Cognition. Blackwell, Oxford. Thurmair, Maria, 1989. Modalpartikeln und ihre Kombinationen. Niemeyer, Tübingen. Tannen, Deborah, 1979. What’s in a frame? Surface evidence for underlying expectations. In: Freedle, R. (ed.), New Directions in Discourse Processing. Ablex, Norwood, NJ, 137–181.

66 Grounding and Common Ground Thomason, Richmond H., 1992. Accommodation, meaning, and implicature: Interdisciplinary foundations for pragmatics. In: Cohen, P.R., et al. (eds), Intentions in Communication. MIT Press, Cambridge, MA, 325–363. Wahlster, Wolfgang, (ed.) 2000. Verbmobil: Foundations of Speech-to-Speech Translation. Springer, Berlin etc. Waltereit, Richard, 2001. Modal particles and their functional equivalents: A speech-acttheoretic approach. Journal of Pragmatics 33 (9), 1391–1417. Weydt, Harald, 1986. Betonungsdubletten bei deutschen Partikeln. In: Weiss, Walter E., Herbert E. Wiegand and Marga Reis (eds), Textlinguistik contra Stilistik? Wortschatz und Wörterbuch. Akten des VII. Internationalen Germanisten-Kongresses G¨ottingen 1985, Bd. 3 Tübingen, Niemeyer, 393–403. Weydt, Harald, 2005. What are particles good for? In: Fischer, K. (ed.), Approaches to Discourse Particles. Elsevier, Amsterdam. Wierzbicka, Anna, 1990. “Prototypes save”: On the uses and abuses of the notion “prototype” in linguistics and related fields. In: Tsohatzidis, S.L. (ed.), Meanings and Prototypes. Studies in Linguistic Categorization. London, 347–367.

Lexical Markers of Common Grounds Edited by Anita Fetzer and Kerstin Fischer © 2007 Published by Elsevier Ltd.

COMMON GROUND IN INTERACTION: THE FUNCTIONS OF MEDIAL DOCH IN GERMAN Karin Pittner

1. INTRODUCTION Common ground has been viewed as static in monological theories of context, where it is taken to comprise the set of true propositions shared by speaker and hearer. A productoriented conception of common ground accounts for the fact that the true propositions cannot be reduced to a stable set as new ones may be added during interaction. In the field of sociolinguistics and discourse pragmatics, a process-oriented approach to common ground has been established, according to which coparticipants negotiate and reconstruct common ground in interaction. This makes it necessary that speakers display to each other what they consider to be common ground in their interaction. For this purpose, special lexical markers may be used. These markers are derived from lexical items through a process of grammaticalization, which constrains the distribution of the item and reduces the meaning and tranfers it to a more intersubjective and pragmatic level (cf. e.g. Fetzer this volume). This paper takes a closer look at one lexical marker in German, namely medial doch which is usually taken to be a modal particle relating to the context of the utterance. The purpose of this paper is to show that the description of the functions of this particle in interaction may profit from newer developments in the theory of common ground as well as deepen our understanding of the ways common ground is negotiated in interaction. The paper is organized as follows: In Section 2, recent developments in common ground will be pointed out followed by a survey of the functions of doch in Section 3 and previous analyses of the particle in Section 4. An integrated analysis of doch will be developed in Section 5. Finally, a suggestion is made how the analysis can be applied to stressed doch, which is followed by a short summary of the findings.

68 Grounding and Common Ground

2. RECENT DEVELOPMENTS IN COMMON GROUND Drawing a conclusion from the view of various authors, Clark states that “Two people’s common ground is, in effect, the sum of their mutual, common, or joint knowledge, beliefs, suppositions” (1996: 93). Clark (1996) introduces a differentiation between personal and cultural common grounds. Cultural common ground exists between members of certain groups, like people who speak the same language, or belong to the same nation, class or any other kind of group. Personal common ground, on the other hand, is established by interactions between individuals, which means that there is no personal common ground between strangers. For the purposes of this paper, it is important to differentiate between personal common ground, which is the mutual knowledge gained through interactions between individuals, and dialogue common ground, which is the activated mutual knowledge at the time of the current interaction. This part of common ground is what Thomason (1992) called the “conversational record” which is established and constantly modified and updated during discourse. Furthermore, Fetzer introduces a distinction between individual dialogue common ground and collective dialogue common ground (2004 and this volume). Speakers can indicate for instance by reformulation that something that is part of their individual common ground should not be attributed to the collective dialogue common ground (cf. Fetzer, this volume). Common ground is constantly updated and, if necessary, revised by an operation called “grounding”. As Clark (1996: 221) states it: “To ground a thing […] is to establish it as part of common ground well enough for current purposes.” Grounding is anchored to a logic of upward completion and downward evidence, cf. Clark’s explanation (1996: 389): “Levels of action form what I have called “action ladders”, which have the properties of upward causality, upward completion, and downward evidence”. These action ladders, which are vaguely reminiscent of the different acts involved in a speech act as established by Austin, range from the lowest level of executing a behaviour, on to the higher levels of presenting a signal and signalling to the highest level of proposing a joint project. Clark stresses that each of these actions is a joint action in which the recipient is involved, for example by attending to the behaviour at the lowest level up to considering the proposal of the joint project at the highest level. The analysis of doch presented here will make use of the distinction between general common ground and dialogue common ground. It is assumed that the coparticipants in an interaction can draw upon general common ground, but in order to become effective in dialogue common ground, the beliefs, suppositions and so on of general common ground must be activated. If the speaker presents them as her/his individual dialogue common ground, the coparticipants have to ratify them in order for them to become part of collective dialogue common ground. In a process-oriented view of common ground the ways and means of offering, accepting or rejecting the offer to accept something as part of the collective dialogue common ground play an important role. As will be shown, doch is an effective means of grounding in German. The data base consists of 60 examples of doch in written language as well as 60 in spoken language.1 They are complemented by examples found in the literature on the subject.

1

The sample is taken from the corpora provided by the Institut für Deutsche Sprache in Mannheim (www.ids-mannheim.de). I would like to thank Daniela Elsner for her help with the corpus data.

Common Ground in Interaction 69

3. MEDIAL DOCH IN GERMAN Doch is a word which occurs in a number of different functions in German and is accordingly classified as belonging to different word classes. (1) a. Er wollte kommen, doch er hatte keine Zeit. (coordinating conjunction) He wanted to come but he had no time. b. Er wollte kommen, doch hatte er keine Zeit. (conjunctional adverb) c. Kommt er nicht? – Doch. (reaction particle) Won’t he come? Yes, he will. d. Hans hat doch keine Zeit. (modal particle) John has no time, you know. e. Hans hat DOCH keine Zeit. (stressed variant/modal particle). John has no time after all.2 Doch in (a) and (b) functions as a conjunction or conjunctional adverb relating the sentence to the previous, usually translated by but or however.3 In (c) it is a reaction particle which can be used to reply to a question containing a negative bias towards the proposition by the speaker. Moreover, the reaction particle can occur after a negative reply (Kommt er? Nein. Doch., cf. Graefen 2000). Medial doch in German as in (d) is generally assumed to be a modal particle. It has been argued by a number of authors that the whole class of modal particles is due to grammaticalization processes. These particles all have counterparts in other word classes, such as adverbs, adjectives and conjunctions. In contrast to these word classes, modal particles have no lexical meaning in the narrow sense, but operate on a pragmatic level. König (1997) calls them “metapragmatic instructions”. In this way, they are instances of the process of subjectification that Traugott postulates for grammaticalization processes. Moreover, in accordance with the parameters for grammaticalization formulated by Lehmann (1995), they have lost not only semantic but also phonetic substance in a number of cases (cf. Wegener 1997, 2002). These particles also are more constrained in their syntactic distribution, since they can only occur in a position behind the finite verb and the “right bracket” in independent sentences, in the “middle field”. Additionally, modal particles cannot freely occur in all sentence types but have individual restrictions in this respect. Due to these characteristics, modal particles can be assigned the status of a lexical marker operating at a pragmatic level. They can be differentiated from other types of lexical markers as for instance discourse markers. Gohl and Günthner (1999) and Günthner (2000) argue that discourse markers can develop out of conjunctions such as weil (because) or obwohl (although). Neither syntactically nor intonationally integrated into a clause, these discourse markers operate outside the confines of a single sentence indicating the relationship between larger parts of discourse.

2 Stressed doch is translated by after all, nevertheless, yet, still, by an accent on the verb (cf. Section 6 on verum focus) or it may remain untranslated, as a study by Keyser (2002) shows. Compare Fischer (this volume) for the translation of modal particles. 3 Conjunctional adverbs like conjunctions connect sentences, but in contrast to conjunctions they can fill the slot before the finite verb in declarative sentences (“prefield”) alone.

70 Grounding and Common Ground Modal particles, on the other hand, are syntactically and intonationally integrated into a sentence and with a few exceptions they are unstressed (cf. Fischer this volume). These particles situate an utterance in context and can be used to indicate common ground, to hint at assumptions of speaker or hearer thereby relating an utterance to the context. Foolen (1989: 312f.) claims that they hint at implicit, contextually relevant propositions, which are a logical variant of the explicitly expressed proposition. A similar view is held by Diewald, who speaks of a “pragmatic pretext” (Diewald, 1997) or a “pragmatically given unit” (Diewald, 2005) that the modal particle relates to (see also Fischer, this volume). In contrast to the conjunction doch which relates the sentence to the clause immediately preceding it, doch as a modal particle relates the clause to a proposition that is not overtly expressed, but contextually relevant. The development from the conjunction to the modal particle can be seen as an instance of a development from textual functions to meaning on the level of the speaker’s subjective attitude towards the situation which Traugott (1989) postulates as a general tendency for grammaticalization and meaning change. The categorization of medial doch as a modal particle is uncontroversial as long as the particle is not accentuated. For accentuated doch, however, there are controversial views regarding its lexical category. While some authors take it to be a modal particle as well (Meibauer, 1994), others claim that it is a special kind of adverb (“affirmative adverb”, Thurmair, 1989). Meibauer also claims that both uses of doch have a common semantics, to which the semantics of the focus accent is added in the case of accentuated doch. This line of argumentation will be pursued in Section 6. Both the stressed and the non-stressed variant of medial doch will be taken into account, since they are closely related and both have special functions with regard to common ground which shed light on each other. We start by giving a survey of recent analyses of the meaning of unstressed doch.

4. PREVIOUS ANALYSES Generally speaking, there are two basic approaches to the description of the meaning of modal particles. In a minimalistic approach, a core meaning is assumed which occurs in all uses of the particle and may be modified by the sentence type the particle occurs in. In a maximalistic approach, different meanings are postulated depending on the various sentence types the particle can occur in. An example for such a maximalistic approach is Helbig (1990), who defines no less than seven different meanings of doch as a modal particle. The particle has already been widely discussed in the literature on modal particles and a number of proposals have been made to define its meaning and uses in interaction. It is uncontroversial that doch signals the attitude of the speaker towards the utterance. Some authors hold that doch has an affirmative meaning component which groups it together with the modal particle ja (e.g. Borst, 1985). Lütten (1979) speaks of “consensusconstitutive” particles. The fact that ja and doch are the modal particles occurring most often shows that it is an important function of modal particles to establish what is considered to be common ground. Besides its affirmative meaning, however, doch is assumed to have a second contradictory or adversative meaning component (e.g. König, 1997; Weydt, 1986). Doherty (1982, 1985) for instance claims that doch expresses a positive attitude (usually of the speaker) and a negative attitude (usually attributed to the recipient) towards the proposition.

Common Ground in Interaction 71 Thurmair (1989) sees the two elements “known” and “correction” at work in medial doch. “Known” may apply to the knowledge of the hearer or, in the case of questions, to the knowledge of the speaker. She assumes that utterances with doch contain propositions which are uncontroversial to the hearer. In this respect, doch resembles the modal particle ja which has also the function to signal that the proposition is already known to the hearer and in no way controversial. But with medial doch we have an additional assumption of the speaker, namely that the speaker expresses that he/she has reasons to think that the hearer does not take the proposition into account at the moment. It is an instruction to the recipient to correct their assumptions and expectations on the basis of the facts actually known to them (Thurmair, 1989: 112; cf. Lütten, 1979: 36). A similar account is put forward by Ormelius-Sandblom (1997) who sees the two components “affirmation” and “adversativity” at work. In the following formula, she states that the proposition is taken as fact accompanied by a conventional implicature that there is a proposition q in the context which implies ¬p: (2) p [FACT p] implicature [∃q q → ¬p A strictly minimalistic analysis is developed by Lindner (1991). According to her, the common core of all the uses of unstressed doch is the following: (3) (It is necessary that) If the speaker uses MP doch in an illocution type IT referring to  then he/she assumes at the time of speaking that it is not the case that  is being taken into consideration. The variable  represents the proposition in assertive sentences and exclamations. For imperatives the first occurrence of the variable represents the proposition p, the second one “bringing about p”. As Meibauer (1994: 112) points out, there is a problem with (3) as far as optative sentences are concerned, because they express (3) already without the particle, so that the meaning contribution of the particle does not become clear for this sentence type. See the following example (taken from Lindner 1991: 187): (4) (A has to write a paper, the sun is shining, the birds are singing  ) A: Wenn ich doch jetzt in der Sonne liegen könnte! “If only I could go and lie in the sun now!” In optative sentences, the obligatory subjunctive marks the proposition as counterfactual, which means that the proposition is not “being taken into consideration”. For further discussion of this sentence type the reader is referred to Section 5.6. In Lindner’s approach, it is not assumed that “being known” to speaker or hearer is part of the meaning of doch. This minimal analysis of doch avoids meaning elements which would have to be cancelled in certain contexts and thus comes closest to the approach to be developed in this paper. It seems theoretically desirable to define a common core of meaning which holds for all the uses of doch and to develop a compositional view where the meaning contribution is stable and combines with the meaning of the various sentence types the particle occurs in.

72 Grounding and Common Ground An analysis of doch by Foolen (2003) refers to a model by Sekiguchi (1977) who proposed a three-step model, illustrated by the sentence es hat doch geschneit (“it has snowed”): 1. affirmation: it can be assumed that there is snow every year 2. movement towards negation: could it be that there will be no snow this year? 3. return to affirmation: no, it has snowed Foolen subscribes to the model developed by Sekiguchi without assuming that all the steps have to be realized explicitly. An analysis by Graefen (2000) is suggestive of similar steps. In a way, the steps suggested in this model are reflected in the approach presented here, where the meaning of doch is to be explained with relation to common ground using a distinction between general common ground and dialogue common ground. “Affirmation” means that the proposition is presented as part of the general common ground or at least as being compatible with it while at the same time indicating that it is not part of the collective dialogue common ground at the moment (“movement towards negation”). This means for the occurrence in most sentence types that the speaker has reasons to assume that the hearer is not taking the proposition into consideration at the present moment. By means of doch the speaker instructs the hearer to update dialogue common ground and to take the proposition into consideration.

5. TOWARDS AN INTEGRATED ANALYSIS Most authors dealing with the meaning of doch consider only a part of its occurrences. Both Thurmair and Lindner consider unstressed doch in all sentence types it can occur in except causal clauses. Doherty bases her account of the particle on stressed and unstressed doch in declarative sentences and assertive questions. Meibauer tries to explain the difference between stressed and unstressed doch but does not include its unstressed occurrence in causal clauses in his investigation. The aim of this section is to develop an approach to its common core of meaning that covers its functions in all sentence types including its use in causal clauses as well as the stressed variant of the particle. From a theoretical point of view, it is desirable to follow a minimalistic approach which defines the common core of all uses of doch. This does not mean that, especially for the purposes of German as a foreign language, more detailed descriptions of the use of the particle in various contexts may not be desirable (cf. Foolen, 2003). Thurmair (1989) and Ormelius-Sandblom (1997) assume meaning elements which do not occur in all contexts, namely correction/adversativity. The feature “correction” or “adversativity” is a bit misleading and, as we will see, may be more aptly applied to the uses of stressed doch. In my view, if we consider this in the light of recent distinctions made relating to common ground, what the speaker wants the hearer to do is to take the proposition of the utterance into consideration, that is, the speaker has to update common ground but not to correct it in the sense that there is anything wrong with it. This can be construed as the difference between general common ground and dialogue common ground: doch signals that a proposition from the general common ground has to be activated in the dialogue common ground.

Common Ground in Interaction 73 That a feature “correction” or “adversativity” does not occur in all uses of the particle as we will see would mean that these elements would have to be cancelled in certain contexts. Our approach comes close to the one by Lindner which is more reticent in this respect. The purpose of this section is to establish the contribution of the particle in the various sentence types and to define a common core of its meaning. We will consider the particle in all the main clause types it occurs in as well as in dependent types like causal and concessive clauses. Moreover, the stressed variant of doch will be closely looked at. As a first approximation, the following formula characterizes the content of the metapragmatic instruction which is given by doch: (5) replace ¬p by p As we will see, the communicative status of ¬p differs. In the case of unstressed doch it refers to the individual dialogue common ground of either speaker or hearer in which p is not present (and in this weak sense there is non-p). In the case of stressed doch, ¬p is present in the collective dialogue common ground.

5.1. Doch in declarative sentences Doch in declarative sentences signals that the proposition is known to the hearer, that is, part of the general common ground, but the hearer is assumed not to be aware of it at the moment. Consider the following example: (6) [A boy is about to drink a bottle of wine in the presence of a grown up, who says to him]4 Du bist noch nicht groß genug. Du kannst doch nicht eine Flasche Wein allein austrinken. “You are not yet grown up, you cannot drink a bottle of wine all by yourself.” By using doch in the second sentence, the speaker signals to the boy that he knows that the fact he is telling him, but his actions signal clearly that he does not take it into account at the moment. As Thurmair states, “The knowledge of the hearer assumed by the speaker is not so much asserted by the use of doch, as the hearer is instructed to take that knowledge into account” (Thurmair, 1989: 112, translation K.P.). In terms of a differentiated view of common ground, this means that in (6) the boy is required to update his dialogue common ground from his general common ground. The fact that doch signals that the proposition is known and uncontroversial can be exploited, for instance in narratives. Thurmair (1989: 113) observes that, within a narrative, unstressed doch may have a kind of backgrounding function. Although the narrative element containing doch may be new and cannot be considered part of the common ground, it does not contain an important step and serves merely as background for a more interesting narrative element. Thus doch in these cases serves as an instruction to the hearer to treat the respective narrative element as if it were part of the common ground.

4

Example slightly adapted from IDS-corpus (Deutsche Mundarten/German dialects).

74 Grounding and Common Ground

5.2. Doch in imperatives The analysis presented for declarative sentences carries over to imperative sentences: The hearer is not considering to perform the action, that is, he/she is not taking it into consideration. The speaker signals by using doch that there are reasons to assume that the speaker is not considering to perform the action. It could be objected that imperative sentences are used only if the speaker would not carry out the action anyway because otherwise there would be a violation of the Gricean maxim of relation. But the special characteristic of imperatives containing doch becomes clear in comparison to imperatives containing the modal particles eben and halt. The latter particles suggest that the action referred to in the imperative is in a way a consequence of a previous move by the coparticipant: In the case of eben it is presented as the only possible one, in the case of halt as a very plausible one (cf. Thurmair, 1989).5 In imperatives containing doch, however, the speaker signals that he/she is rather in the initiating than responsive position. This differs from situations where eben and halt are used. He/she has reasons to believe that the hearer is not considering to perform the act referred to. The hearer is required to update the dialogue common ground, that is, to realize that there are no reasons why the action should not be performed and why non-p should be replaced by p. Whereas the particles eben or halt signal a compliance on the part of the speaker with an action suggested by the recipient or one which is an obvious one in the situation at hand, doch signals that the action is in some way at variance with the situation. In our sample, doch occurs seven times in a directive context in which the coparticipant is requested to tell more about a certain topic (Erzählen Sie doch mal, “tell us something about it”). Here, it is quite clear that the directive is used to initiate a sequence and the hearer would not tell more about the subject of her/his own accord. It has been observed that doch in negated requests can be used to correct current or continuing behaviour as in (7a) and (7b), but can hardly occur in negated requests relating to actions in the future as in (7c), cf. Thurmair (1989: 118f): (7) a. Mach doch nicht immer so einen Krach! Don’t be so noisy! b. Erzähl doch nicht immer alles der Frau Kling! Don’t always tell everything to Mrs. Kling! c. Und bleib (*doch) nicht zu lange in der Kneipe! Don’t stay in the pub for too long! That imperatives containing doch are in some way at variance with the situation becomes most clear in sequences where the hearer is instructed to change his past or present actions like in the following one: (8) A: Und was ist mit dem Herrn, der stolz wie ein Pfau herumstolziert ist?6 And what about the man who walked about so proudly?

5

It is hardly possible to translate these particles in imperative clauses. In declarative clauses, eben might be translated by just: Das ist eben so. “That’s just the way it is.” 6 Example slightly adapted from IDS-corpus (Belletristik).

Common Ground in Interaction 75 B: Sagen Sie doch so was nicht! Don’t say such a thing! A: Er versteht nichts von Pferden. He knows nothing about horses. In this sequence, where A is instructed to correct his previous utterance, the use of the particle doch supports our assumption that also in imperatives the particle serves as an instruction to replace non-p by p. But, as this sequence shows, the hearer of course does not have to comply with such a request. A does not comply with the request to take back his previous utterance but tries to justify it.

5.3. Doch in questions Doch cannot occur in information questions where there is no bias towards the truth of the proposition. It can only appear in assertive questions, in deliberative questions and in a special kind of rhetorical questions. Doch occurs in assertive questions which have the form of a declarative sentence (indicated by verb-second position). What turns these sentences into questions is that they have a rising intonation contour: (9) Du kommst doch morgen? You will come tomorrow, won’t you? In assertive questions, doch serves as an instruction to the hearer to update individual dialogue common ground of the speaker. The speaker is convinced that the proposition is right (indicated by the form of the declarative sentence), but wants to be sure there is nothing the hearer knows which stands against his or her assumption. In other words, when using assertive questions, the speaker wants to be assured about the truth of the proposition. It is an instruction to the hearer to replace non-p (which occurs to the speaker as a possibility) by p (which the speaker assumes to be correct). In the following interaction, the hearer may or may not react as was expected by the speaker. Consider the following examples from the sample: (10) A: Sie als Jurist haben doch sicher die richtige Einstellung zu Gesetz und Ordnung? You as lawyer certainly have the right attitude towards law and order? B: Ja, hart und konsequent durchgreifen! Yes, rigorous action must be taken! (11) [Die Verteidigerin unternimmt einen Versuch, das Blatt zu wenden:] The defence counsel tries to change the situation: A:  Aber Sie sind doch hauptsächlich zu Frau E. gegangen, um mit ihr über ihre familiären Probleme zu reden? But you mainly went to Mrs. E. to talk with her about her problems with the family? B: Nein. No.

76 Grounding and Common Ground While doch in both cases signals that the speaker is trying to establish the proposition as part of the common ground, the hearer need not comply with this and may reject it, as in (11). Doch may also occur in wh-questions, where it has a somewhat different function. These questions can have two functions: Either they are deliberative questions or they are used as a rhetorical device. In deliberative questions, the speaker has forgotten the answer at the moment and thus is in need to update her/his individual dialogue common ground. Questions of this kind may have an addressee, but they may also be directed at the speaker herself/himself (only in this case they are truly deliberative), as a kind of request to his memory to provide the answer: (12) Wie hieß er doch? Now what was his name? Besides doch, particles occurring typically in these sentences are noch, gleich, wieder, schnell. Their contribution is not exactly the same. Noch (still) indicates a state that persists. In this kind of questions, it signals that the speaker knew the answer and that he/she intends to continue with that status quo. Gleich (soon) indicates that this state will be reached shortly, and schnell (quickly) has a very similar effect. Wieder (again) indicates the restitution of a state, in this case that the speaker knows the answer. By using these particles, the speaker suggests that the hearer can quickly update the individual common ground of the speaker, to whom the answer is at hand, but just at the moment not accessible in his or her memory. This means that the answer is part of general common ground but not part of the speaker’s individual dialogue common ground. The particle suggests that this is a situation which can quickly be fixed by the recipient. Doch may also appear in rhetorical questions. These rhetorical questions prepare the way for presenting the following proposition, which is assumed to be known, that is, part of the general common ground: (13) Wie sagte doch Goethe so treffend? “How did Goethe say so aptly?” Basically, these questions (e.g. occurring in lectures and speeches) suggest that the hearers know the answer and just have to be reminded of it, which means that they have to update their dialogue common ground from non-p to p (present in general common ground). In the case of the wh-questions, it is more correct to say that the proposition with the variable (expressed by the wh-word) is to be replaced by the proposition where the variable is instantiated, that is, where one of the alternatives is chosen.

5.4. Doch in exclamations Exclamation sentences can occur in three forms in German: as verb-second clauses with a strong accent, often on a demonstrative pronoun (14a), as verb-first clauses (14b), as wh-exclamations, which can exhibit verb-second- or verb-end-position (14c), or as sentences introduced by dass (14d):

Common Ground in Interaction 77 (14)

a. DAS ist doch toll! This is great! b. SCHLÄGT der sich doch mit dem Hammer auf den Finger! With a hammer he hits his finger! c. Wie SCHÖN er doch ist! How beautiful he is! d. Dass doch die Jugend immer zwischen den Extremen schwankt! That youth always varies between extremes!

They signal that the state of affairs is in some way unexpected to the speaker, either the fact itself or the degree to which it obtains. Lindner’s formulation that the proposition has “not been taken into consideration” is an apt description of the speaker’s state of mind while uttering these sentences. The speaker replaces non-p by p for himself and for a recipient if any is present. The speaker (and hearer) were not aware of the fact before, which means it is not part of the dialogue common ground although it is taken to be consistent with their general common ground.

5.5. Doch in optative sentences Optative sentences often are called conditionals because they have several features of this clause type: They exhibit the subjunctive, which is used to mark counterfactual propositions. Moreover, optative sentences like conditional clauses in German are either introduced by wenn (if) or exhibit verb-first position. (15)

a. Wenn er doch (nur, bloß) käme! If he only came! b. Käme er doch!

When considering the contribution of doch to optative sentences one has to bear in mind that counterfactuality is already clearly marked by the subjunctive. It is also interesting to note that the optative sentences cannot occur without a modal particle, either doch, nur or bloß. These particles serve to indicate the attitude of the speaker towards the proposition. Whereas the speaker can remain neutral towards the counterfactuality of a conditional clause, optative sentences can only be used if the speaker wants the proposition to come true. This attitude is expressed by means of these particles, albeit in different ways. Nur and bloß are used also as exclusive focus particles (like only), which explicitly exclude the alternatives. This exclusive meaning component gives urgency to the wish expressed with these sentences. Doch expresses the attitude of the speaker towards the proposition in a different way: it signals that the speaker wants to update their individual common ground, he/she wants to follow the instruction expressed by the particle to replace non-p by p. Admittedly, it may seem a bit far-fetched to speak of common ground in this kind of speech act which need not be addressed to anyone but oneself. This is only possible if a notion of individual common ground is employed as proposed by Fetzer (2002).

78 Grounding and Common Ground The contribution of doch is to express the attitude of the speaker towards the state of affairs: He/she would like to see the proposition to come true. That is, the speaker would like to update their individual common ground about that. In this way, a relation to “update common ground” exists.

5.6. Doch in causal verb-first sentences Usually, subordinated verb-first clauses can either be conditional or concessive clauses. A causal interpretation crucially hinges on the presence of doch, without which these sentences cannot occur. These causal sentences have characteristics which causal clauses introduced by a conjunction do not have. As Reis (1985: 285) notes, they can be considered neither coordinated nor subordinated because of their peculiar characteristics. These sentences are not constituents of the preceding sentence because they fulfil none of the constituent tests: Neither they can be pronominalized nor can they function as the answer to a wh-question, be moved or topicalized. Also the fact that no correlates pointing towards the causal clause can occur shows that these sentences are not syntactically integrated in the preceding clause. Moreover, these sentences cannot be focussed and occur only in post-position. We will see how these characteristics correlate with the specific pragmatic functions of these causal clauses and try to shed some light on the important contribution of the modal particle. As will be shown, these sentences can serve as a kind of causals due to the interplay of verb-first position, their postposition and the obligatory particle. Verb-first position has a special function in this context. Önnerfors (1997) argues at length that declarative sentences with verb-first position are closely connected to the previous sentence, which is also the case in the causal sentences treated here. Our approach is inspired by Ford (1994), who takes a close look at the pragmatic functions of postposed causal clauses. She notes a repair function of causal clauses in spoken language which make a non-preferred reaction within a speech act sequence more acceptable. Her hypothesis is that postposed causal clauses in written language can have a similar function. As she observes, these clauses often occur after contrasting and negative sentences which do not fulfil the expectation of the readers or hearers. They serve to make these sentences more acceptable by providing a fact that supports the previous proposition. According to Ford, this is possible because the writer is having a kind of dialogue with a projected recipient. Adapting an explanation of this kind, some of the characteristics of the causals considered here are motivated by their function. These sentences are not focussed, because they do not carry the main information but rather have a subsidiary function. They are postposed because their function is to support a statement already made. It has been observed by several authors that these sentences always assert a fact which is considered to be uncontroversial (Altmann, 1993: 1020; Engel, 1999: 269). Here is where the contribution of doch comes in: It signals that the proposition is part of the general common ground. It would make no sense to make a statement, if one believed the recipient to be aware of that fact at the moment already. Thus, doch in these sentences has the function to instruct the hearer to “update common ground”.

Common Ground in Interaction 79 Sentences with causal doch are primarily, if not exclusively, a phenomenon of written language. Our sample provides thirteen instances of causal doch in written language, but none in spoken language. Consider the following examples: (16) Gerade an der Universität St. Gallen sei es unerlässlich, auf geschlechterspezifische Konflikte hinzuweisen, handle es sich bei der Institution doch immer noch vorwiegend um eine “Männerwelt”, die sich aus 75 Prozent Studenten und nur einem Viertel Frauen zusammensetzt. At the University of St. Gallen of all places it is imperative to point out genderspecific conflicts, since the institution still is a “men’s world” mainly, with 75 percent male students und a quarter female students. (17) Der neue Pächter, Imro Rusnak, ist in St. Margarethen kein Unbekannter, verwöhnte er doch bereits vor 18 Jahren im Falken seine Gäste mit Spezialitäten. The new leaseholder Imro Rusnak is not unknown in St. Margarethen, since he treated his guests to specialities in the “Hawk” 18 years before. (18) Diese Unterführung kenne ich sehr gut, gehört sie doch zu meinem täglichen Arbeitsweg von der Krontalstrasse in den SZB an der Schützengasse 4. I know this subway very well since it is part of my way to work from Krontalstrasse to the SZB in the Schützengasse 4. In each case, the information given in the doch-clause is presented as completely uncontroversial, that is, as part of general common ground (either personal or cultural) or at least as compatible with it. Now the question is what the exact contribution of verb-first position to the interpretation of these sentences is. As Reis (2000: 224) notes, verb-first declarative sentences in German assert true propositions, whereas verb-second declaratives assert propositions as true. Put differently, verb-first declaratives put the truth claim into the background, whereas V-2declaratives foreground it. This supports the view that these sentences contain completely uncontroversial information from general common ground which is brought to the attention of the reader in order to ensure the acceptance of the main clause assertion. As was already mentioned, verb-first position in declaratives signals also the close connection of the sentence to the previous one. But it is the modal particle which provides the instruction to the recipient how to connect this sentence by giving her/him a hint how it relates to the context. Without the modal particle these sentences cannot occur because the recipient would be left at a loss as to the relation of this sentence to the context. These sentences are not so much subordinated sentences but contain own speech acts that have a subsidiary function, which accounts for the impression that they are neither really coordinated nor subordinated. Whereas causal clauses introduced by a conjunction usually give a cause for the proposition of the main clause, these sentences provide an additional fact which makes the acceptance of the proposition of the sentence they relate to more plausible and thus more acceptable to the recipient. The function of these sentences is reflected in their intonational characteristics. These causal sentences cannot be intonationally integrated into the clause they relate to, but have an intonation contour of their own. This means that they contain a separate information unit that is not part of the proposition of the sentence they relate to. Causal clauses introduced by weil

80 Grounding and Common Ground (because), however, usually are integrated into the proposition of the main clause, adding a cause for it.7 Because of their special form verb-first causals cannot be mistaken as part of the previous clause, since they cannot be intonationally integrated into them. Thus, these sentences provide a means of disambiguation in written language between clauses indicating a cause for the proposition and sentences that mention a fact that is merely supportive of the main assertion. In spoken language, this distinction can be made clear by pauses, accents and intonational contours. Any meaning element in the kind of “opposition” or “adversativity” is not present in these sentences, which just signal that they contain a fact which the speaker thinks the recipient is not aware of at the moment. Thus, causal sentences containing doch serve as an instruction to the recipient to update their dialogue common ground with a fact that is part of the general common ground and in this sense to replace non-p by p.

6. STRESSED DOCH The question arises whether the meaning “replace non-p by p” can be found with stressed doch. It is not the main purpose of this section to discuss the syntactic categorization of stressed medial doch but by analyzing its meaning we will find some hints what a useful categorization might be. In this context, it seems worthwhile to pursue the line of argumentation suggested by Meibauer (1994), who explains stressed doch as a combination of the meaning of unstressed doch and the meaning of the focus accent. The basic ingredients of the analysis presented here are the meaning of doch as well as the meaning of the focus accent as developed in alternative semantics. The accent on doch has been described as a contrastive focus accent by Meibauer (1994). Focus on single lexical elements that are neither verbal nor an argument of a verb cannot project, that is, the accent can only be interpreted as indicating a narrow focus on that single element. As a consequence of that, the rest of the sentence is backgrounded, that is, present in the context or presupposed. We adopt the view of focus which has been proposed in alternative semantics as it was introduced by Rooth (1992): Focus establishes a relation between alternatives, where the kind of alternatives are determined by the type of the focussed constituent. In the case of stressed doch, the alternative is non-p. By placing a focus accent on doch the speaker indicates that non-p is present in the context, namely in the collective dialogue common ground. In this way, it does not relate to the individual dialogue common ground of either hearer or speaker in the way unstressed doch does, but indicates that there are objective reasons, which are present in the collective dialogue common ground to assume the contrary. The question arises how this special effect is achieved by stressed doch. The focus accent on doch obviously makes an important contribution. Although focus on doch cannot project to

7

Sometimes clauses introduced by weil may also give a reason for the speech act performed in the main clause. In this case they have an independent intonation contour and are never focussed because of their subsidiary function (cf. Pittner, 1999: 235ff.).

Common Ground in Interaction 81 larger parts of the sentence according to the rules for focus projection, it can be assumed that the focus accent on the particle relates indirectly to the whole proposition because doch has scope over the proposition (Meibauer, 1994: 130). The effect of this is that the presence of an alternative proposition (namely non-p) is signalled, which may be expressed in the context or may remain implicit. Additionally we have the meaning to “replace non-p by p” which makes this alternative more explicit than the so-called “verum focus” (Höhle, 1992) does. Verum focus strengthens the truth of a proposition and is usually placed on the finite verb in declarative sentences or on the complementizer in subordinate clauses. The presence of verum focus indicates that the truth of the proposition has been under debate or at least seems to be questionable in the context. By using verum focus, the speaker stresses the truth of the proposition by repudiating non-p at the same time. Thus, its effect is very similar to stressed doch: compare the following interactions:8 (19) A: Er ist nicht gekommen. He did not come. B: Er IST gekommen. (verum focus) He DID come. (20) A: Er ist gekommen. He has come. B: Er ist NICHT gekommen. He did NOT come. A: Er IST gekommen./Er ist DOCH gekommen. He DID come./He came after all. The second interaction shows that a proposition containing a focussed negation can be contradicted by an utterance containing a verum focus accent either on the verb or, even better, on doch. The closeness of stressed doch to verum focus is illustrated by this example. It is interesting to note that B in (19) could reply by using the responsive particle doch instead, which represents a special use of stressed doch. Stressed doch is more explicit than verum focus in repudiating non-p because in addition to the focus on the particle there is its lexical meaning “replace non-p by p”, which is not there in the case of a verum focus accent on the verb or complementizer. The effect of stressed doch can be compared to the sentence adverbs wirklich (“really”) and tatsächlich (“actually”): These elements may carry a focus accent to the effect that the rest of the sentence is backgrounded. These sentences occur in contexts where the truth of the proposition has been been in doubt or has been controversial (cf. Pittner, 1999: 179f). The mechanism is the same as in the case of stressed doch. Although focus from these elements cannot project because they are adjuncts, the focus accent on them indirectly relates to the proposition via the scope these elements have over it thereby indicating non-p as an alternative. 8

The study by Keyser (2002) shows that stressed doch can be translated by emphatic do or stress on the verb which signals verum focus; see. the following example: (i) Sie hat gelogen und ich bin doch ihr Vater. (Max Frisch, Homo Faber, p. 183) She was lying and I am the father. (English translation, p. 154)

82 Grounding and Common Ground These elements can occur in the prefield (i.e. before the finite verb in declarative sentences). Usually they are stressed and preceded by und (“and”): (21) Und tatsächlich/wirklich/DOCH kam er. And actually/really/yet he came. The difference between stressed doch and the stressed sentence adverbs seems to be that there must be counterevidence in the case of doch whereas this is not necessary for the use of the sentence adverbs tatsächlich (actually) and wirklich (really). The following example illustrates this point: (22) Wir hofften alle, dass Peter kommt und tatsächlich/wirklich/∗ DOCH kam er. We all hoped that Peter would come and he actually/really came. Whereas in the case of tatsächlich and wirklich it simply may have been unclear whether Peter will come or not, doch signals that there are reasons that speak against it. Again, it is the meaning “to replace non-p by p” which is responsible for this effect. The distribution of stressed doch in the sentence shows that a categorization as modal particle obscures rather than elucidates how stressed doch works. Also it is not yet entirely clear whether there is any difference between medial and initial stressed doch (cf. the discussion by Meibauer, 1994). Initial position is a further argument not to count stressed doch among modal particles. Several authors dealing with the meaning of stressed doch assume that the opposite proposition has to be expressed explicitly (e.g. Meibauer, 1994: 118). Brauße, for instance, states that by using stressed doch “one does not refer to implicit inferences, but two opposite opinions are contrasted, that both have been uttered before, namely q and –q” (Brauße, 1988: 99, translation K.P.). As will be shown, it is too strong an assumption that stressed doch always presupposes the explicit presence of the opposite proposition in the context. But first we take a look at portions of a dialogue from the sample where the opposite proposition is expressed explicitly. In this talk the nature of the institution of marriage is discussed at length. Speaker 1 tries to establish as part of the collective dialogue common ground that marriage is not simply a natural and primitive condition but an institution which was formed essentially by Christianity: (23) S1: ich meine, unter Ehe verstehen wir doch nicht nur eine Liebe, so ne besondere Liebe zu einer Frau, sondern Ehe ist eine ganze Institution mit sagenhaft vielen Verpflichtungen die beide Partner haben    und meine These (vielleicht ist sie falsch) ist doch die, dass unsere Vorstellung, der abendländische Begriff der Ehe, wesentlich geprägt ist durch das Christentum und durch    I mean, by marriage we do not understand just love, such a special love for a woman, but marriage is a whole institution with many obligations both partners have    and my thesis (maybe it is wrong) is that our notion, the occidental notion of marriage is essentially formed by Christianity and by   

Common Ground in Interaction 83 By using unstressed doch the speaker indicates that he thinks his opinion is part of the general common ground. After some discussion by two other speakers about the dwindling influence of Christianity today, it occurs to speaker 1 that the Chinese are kind of abolishing marriage and finally he revises his view that Christianity is the prerequisite for marriage as an institution: S1:    wie könnten die Chinesen plötzlich die Ehe abschaffen how could the Chinese suddenly abolish marriage S2:    S1: nein, nein du hast recht, die Chinesen waren nicht christlich und haben DOCH eine Form der Ehe.9 no no you are right, the Chinese were not Christian and nevertheless have a kind of marriage. By using stressed doch in the second part of the last sentence the speaker signals a contrast to the assumption he presented before, namely that marriage is an institution, which goes together with Christianity. He has revised his individual dialogue common ground and it is now established as collective common ground that Christianity is not a prerequisite for marriage. Now we will turn to examples where the contrasting proposition is not explicitly expressed. A case in point is the use of stressed doch in concessive clauses: (24)

a. Obwohl es regnete, ist er DOCH spazieren gegangen. Although it was raining, he went for a walk. b. Wenn er auch berühmt war, war er DOCH nicht eingebildet. Even though he was famous, he was not conceited.

Concessive clauses express a condition under which the main clause proposition normally would not hold, as Brauße (1994: 146) states it: “In concessive clauses the truth of both p and q is asserted and stands in contrast to an expectation that both cannot be true at the same time, but either p or q must be false” (translation K.P.). This background assumption need not be explicitly expressed but is considered to have a presuppositional character (e.g. König, 1991). This means that the opposite proposition (non-p) is usually not openly expressed if doch is used in concessive constructions.10 The effect of stressed doch becomes most clear when it occurs in sentences without a written or spoken context. These sentences are usually introduced by and. The connective indicates that the sentence relates to some discussion generally known. To give just two examples: (25)

10

a. Und And b. Und And

die Bibel hat DOCH recht. (title of a book) the bible is right. sie bewegt sich DOCH. (Galileo Galilei) it moves.

Stressed doch has a merely supportive function in clauses with obwohl, stressing the truth of the main clause proposition although there is a reason that speaks against it. In other types of concessive clauses like in (24b), it occurs more often and contributes to the concessive meaning of the construction (cf. Pittner, 1999: 265 for details).

84 Grounding and Common Ground The reader or hearer can conclude from the presence of stressed doch alone that the opposite view is contextually relevant also without knowing the discussion presupposed to be generally known. This means that the effect of stressed doch is to indicate the presence of the opposite proposition in collective dialogue common ground while at the same time denying its truth. It can be concluded that the formula “to replace non-p by p” applies to stressed doch as well. The focus accent hints at the proposition non-p which is explicitly or implicitly present in the collective dialogue common ground.

7. SUMMARY In a process-oriented approach to common ground where it is negotiated between coparticipants and reconstructed by them in interaction it is necessary that speakers display to each other what they consider to be common ground in their interaction. For this purpose, special lexical markers may be employed. The paper investigated how the medial particle doch in German may be used in this respect. Medial doch is a modal particle functioning as a metapragmatic instruction whose basic meaning was argued to be an instruction to replace non-p by p. In this context, a distinction between dialogue common ground and general common ground proved useful. While p is positively evaluated by the speaker in all uses, the status of non-p differs. Also, p may be assumed to be part of the general common ground but not of dialogue common ground to the effect that the particle in these cases serves as an instruction “to update common ground”, that is to activate p in the collective dialogue common ground. According to the sentence type doch occurs in, the instruction is directed towards the hearer’s or the speaker’s individual common ground. Doch signals that p is not present in either the speaker’s or the hearer’s common ground while at the same time presenting it as part of the general common ground or at least as compatible with it. The particle is obligatory in postposed sentences with verb-first position which have a causal meaning. It was argued that these sentences serve to give a supporting fact for the main clause proposition thereby ensuring the acceptance of the main clause. This is possible because this additional fact is uncontroversial, that is, it is part of the general common ground. The reader is assumed not to be aware of this fact, however, and has to update dialogue common ground. The meaning of the stressed variant of this particle was argued to arise from the combination of the lexical meaning of the particle and the meaning of the focus accent. It comes close to the so-called “verum focus” which stresses the truth of a proposition in contexts where non-p is under debate. The focus accent on the particle cannot be projected to the rest of the sentence but indirectly relates to the proposition via the propositional scope of doch, thereby implying non-p as an alternative. The status of non-p in this case is that of a proposition present in collective dialogue common ground (explicitly expressed or as an inference from other facts). If there is no co-text at all, that is, if a sentence occurs in isolation, the focus accent on doch alone signals the presence of non-p in the dialogue common ground. In each case, doch is used as a means of grounding, anchoring p in the dialogue common ground by simultaneously ruling out non-p. With unstressed doch, p is signalled not to be present in the individual dialogue common ground (and in this weak sense there is non-p). Stressed doch signals that non-p is part of the coparticipants’ collective dialogue common

Common Ground in Interaction 85 ground. Thus the particle serves as an instruction to update (activate p) or revise common ground (remove non-p). The distinction between individual and collective dialogue common grounds has proved useful for the description of the effects of doch. Unstressed doch indicates that p is not present in the individual dialogue common ground, depending on the sentence type either the speaker’s or the hearer’s. Stressed doch indicates that non-p is present in the collective dialogue common ground.

REFERENCES Altmann, Hans, 1993. Satzmodus. In: Jacobs, J., et al. (eds), Syntax. Ein internationales Handbuch der zeitgenössischen Forschung. 1. Halbband. De Gruyter, Berlin, 1006–1029. Borst, Dieter, 1985. Die affirmativen Modalpartikeln doch, ja und schon. Niemeyer, Tübingen. Brauße, Ursula, 1988. Modalpartikeln in Fragesätzen. Linguistische Studien Reihe A. Arbeitsbericht 177, 77–113. Brauße, Ursula, 1994. Lexikalische Funktionen der Synsemantika. Narr, Tübingen. Clark, Herbert, 1996. Using Language. Cambridge University Press, Cambridge. Diewald, Gabriele, 1997. Grammatikalisierung. Eine Einführung in Sein und Werden grammatischer Formen. Tübingen, Niemeyer (Germanistische Arbeitshefte 36). Diewald, Gabriele, 2005. Discourse particles and modal particles as grammatical elements. In: Fischer, K. (ed.), Approaches to Discourse Particles. Studies in Pragmatics 1, Elsevier, Amsterdam, 403–426. Doherty, Monika, 1982. Doch. Deutsch als Fremdsprache 19, 174–178. Doherty, Monika, 1985. Epistemische Bedeutung. Akademie-Verlag, Berlin. Fetzer, Anita, 2002. Communicative intentions in context. In: Fetzer, A. and Meierkord, C. (eds) Rethinking Sequentiality: Linguistics meets Conversation Interaction. Benjamins, Amsterdam, 37–69. Fetzer, Anita, 2004. Recontextualizing Context: Grammaticality Meets Appropriateness. Benjamins, Amsterdam. Fetzer, Anita, this volume. Reformulation and common grounds. Fischer, Kerstin, this volume. Grounding and common ground: Modal particles and their translation equivalents. Foolen, Ad, 1989. Beschreibungsebenen für Modalpartikelbedeutungen. In: Weydt, H. (ed.), Sprechen mit Partikeln. De Gruyter, Berlin, New York, 305–317. Foolen, Ad, 2003. Niederländisch toch und Deutsch doch: Gleich oder doch nicht ganz? Linguistik online 13(1) (http://www.linguistik-online.de/2− 00/index.html). Ford, Cecilia E., 1994. Dialogic aspects of talk and writing: Because in the interactive-edited continuum. Text 14 (4), 531–554. Gohl, Christine and Günthner, Susanne, 1999. Grammatikalisierung von weil als Diskursmarker in der gesprochenen Sprache. Zeitschrift für Sprachwissenschaft 18 (1), 39–75. Graefen, Gabriele, 2000. Ein Beitrag zur Partikelanalyse – Beispiel doch. In: Weydt, H. and Poljakova, S. (eds), Linguistik online 6/2000. Günthner, Susanne, 2000. From concessive connector to discourse marker: The use of obwohl in everyday German interaction. In: Couper-Kuhlen, E. and Kortmann, B. (eds), Cause, Condition, Concession, Contrast. De Gruyter, Berlin, 439–468.

86 Grounding and Common Ground Helbig, Gerhard, 1990. Lexikon deutscher Partikeln. Verlag Enzyklopädie, Leipzig. Höhle, Tilman N., 1992. Über Verum-Fokus im Deutschen. In: Jacobs, J. (ed.), Informations struktur und Grammatik. Linguistische Berichte Sonderheft 4, 1991–92, 112–141. Keyser, Thilo, 2002. Die Modalpartikel doch und ihre Entsprechungen im Englischen. Unpublished manuscript University of Stuttgart. König, Ekkehard, 1991. Konzessive Konjunktionen. In: von Stechow, A. and Wunderlich, D. (eds), Handbuch Semantik. (Handbücher zur Sprach- und Kommunikationswissen schaft). De Gruyter, Berlin, 631–639. König, Ekkehard, 1997. Zur Bedeutung von Modalpartikeln im Deutschen: Ein Neuansatz im Rahmen der Relevanztheorie. Germanistische Linguistik 136, 57–75. Lehmann, Christian, 1995. Thoughts on grammaticalization. Unterschleissheim: Lincom Europa (Lincom Studies in Theoretical Linguistics, 1). (Überarbeitete Version: http://www.uni-erfurt.de/sprachwissenschaft/ASSidUE/ASSidUE09.pdf) Lerner, Jean-Yves, 1987. Bedeutung und Struktursensitivität der Modalpartikel “doch”. Linguistische Berichte 109, 203–229. Lindner, Katrin, 1991. Wir sind ja doch alte Bekannte. In: Abraham, W. (ed.), Discourse Particles. Descriptive and Theoretical Investigations on the Logical, Syntactic and Pragmatic Properties of Discourse Particles in German. Benjamins, Amsterdam, 168–201. Lütten, Jutta, 1979. Die Rolle der Partikeln doch, eben und ja als Konsensus-Konstitutiva in gesprochener Sprache. In: Weydt, H. (ed.), Die Partikeln der deutschen Sprache. De Gruyter, Berlin/New York, 30–38. Meibauer, Jörg, 1994. Modaler Kontrast und konzeptuelle Verschiebung. Studien zur Syntax und Semantik deutscher Modalpartikeln. Niemeyer, Tübingen. Önnerfors, Olaf, 1997. Verb-erst-Deklarativsätze: Grammatik und Pragmatik. Almqvist & Wiksell, Stockholm. Ormelius-Sandblom, Elisabeth, 1997. Die Modalpartikeln ja, doch und schon. Zu ihrer Syntax, Semantik und Pragmatik. (Lunder Germanistische Forschungen 61). Almqvist & Wiksell, Stockholm. Pittner, Karin, 1999. Adverbiale im Deutschen. Untersuchungen zu ihrer Stellung und Interpretation. Stauffenburg, Tübingen. Reis, Marga, 1985. Satzeinleitende Strukturen im Deutschen. Über COMP, Haupt- und Nebensätze, w-Bewegung und die Doppelkopfanalyse. In: Abraham, W. (ed.), Erklärende Syntax des Deutschen. Narr, Tübingen, 271–311. Reis, Marga, 2000. Anmerkungen zu Verb-erst-Satz-Typen im Deutschen. In: Thieroff, R., et al. (eds), Deutsche Grammatik in Theorie und Praxis. Niemeyer, Tübingen, 215–227. Rooth, Mats, 1992. A theory of focus interpretation. Natural Language Semantics 1, 75–116. Sekiguchi, Tsgio, 1977. Was heißt “doch”? In: Weydt, H. (ed.), Aspekte der Modalpartikeln. Studien zur deutschen Abtönung. Niemeyer, Tübingen, 3–9. Thomason, Richmond H., 1992. Accommodation, meaning and implicature: Interdisciplinary foundations for pragmatics. In: Cohen, P.R., Morgan, J., and Pollack, M.E. (eds), Intentions in Communication. MIT Press, Cambridge, MA, 325–363. Thurmair, Maria, 1989. Modalpartikeln und ihre Kombinationen. Niemeyer, Tübingen. Traugott, Elizabeth Closs, 1989. On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65, 31–55. Ulrich, Engel, 1999. Deutsche Grammatik. Heidelberg: Groos.

Common Ground in Interaction 87 Wegener, Heide, 1997. Zur Grammatikalisierung von Modalpartikeln. In: Barz, I. and Öhlschläger, G. (eds), Zwischen Grammatik und Lexikon. Niemeyer, Tübingen, 37–55. Wegener, Heide, 2002. The evolution of the German modal particle denn. In: Wischer, I. and Diewald, G. (eds), New Reflections on Grammaticalization. International Symposium, Potsdam, 17–19 June 1999. Amsterdam, Benjamins, 279–394. Weydt, Harald, 1986. Betonungsdubletten bei deutschen Partikeln. In: Weiss, W., et al. (eds), Textlinguistik contra Stilistik? Wortschatz und Wörterbuch. Grammatische oder pragmatische Organisation von Rede? Niemeyer, Tübingen, 393–403. Corpora provided by the Institut für Deutsche Sprache in Mannheim www.ids-mannheim.de (COSMAS II)

This page intentionally left blank

Lexical Markers of Common Grounds Edited by Anita Fetzer and Kerstin Fischer © 2007 Elsevier Ltd. All rights reserved.

A COMMON GROUND FOR KNOWLEDGE THROUGH KNOWLEDGE OF LANGUAGE: A COMPUTATIONAL RESEARCH OF CONSENSUS-BASED MEANING IN SCIENTIFIC PAPERS Carlos Rodríguez Penagos

1. INTRODUCTION “The language of science is, by its nature, a language in which theories are constructed; its special features are exactly those which make theoretical discourse possible. But this clearly means that the language is not passively reflecting some pre-existing conceptual structure; on the contrary, it is actively engaged in bringing such structures into being. They are, in fact, structures of languages (   ) A scientific theory is a linguistic construal of experience.” M. A. K. Halliday (1993: 12) Language is not just an instrument for conveying messages. It serves other social purposes as well, such as regulating and structuring the relationships of individuals and groups. At the same time, language is not a static or neutral medium where ideas and words reside for us to find and use, but is itself shaped constantly through group interaction. In this paper we want to explore how language can be decisive in creating and sharing knowledge about reality as is jointly held by a group of experts in a domain, and how this cognitive common ground is linked to a linguistic common ground. We wish to restrict our discussion to the role that agreeing on what Clark (1998) has called a Communal Lexicon (CL) plays on the advancement of the consensus-based knowledge of an academic field, as evidenced in corpora of papers and articles from peer-review journals and other highly specialized texts. We focus on metalinguistic exchanges, fundamental discourse operations with powerful and lasting consequences not only for the successful continuation of scientific dialogue, but for the very constitution both of technical knowledge as a collective enterprise and of social groups defined by a domain expertise and a shared lexicon. Agreeing on epistemic common ground through negotiation of a shared terminology is possible because metalinguistic exchanges have

90 Grounding and Common Ground cognitive and formal characteristics that include, among other things, powerful lexical markers that at the same time make them prominent in discourse and guide their interpretation. This twofold function, we will argue, conforms to the description of what makes a regular word a lexical marker: the fact that, in addition to its usual semantic meaning and grammatical role, it has a contextually determined pragmatic function that signals (and perhaps enacts) for participants in the exchange a discourse relation or a speech act. What follows is an example from our research corpora of the kind of definition-like metalinguistic utterances to be examined here, although, as in this case, only part of the sentential construction can be considered to be truly about some aspect of language: Integral power results in a fundamental type of social classification which, adapting Bernstein’s terminology, I shall call “frame”. (Bernstein 1971) We will start by reviewing the general concept and function of metalanguage and describe the conditions under which it functions in one kind of learned exchange: printed scientific discussion mediated by peer-review. We will show how such exchanges allow the systematic construction of a shared epistemic common ground, and how they help further the ultimate goal of scientific pursuit: the theories and models that explain and describe reality through consensus, empirical evidence and rational interpretation. We will also present a speech act analysis of metalinguistic sentences, followed by a simple model of such predication in general, in which lexical markers signal, articulate and enact explicit metalinguistic operations (EMOs). Finally, we will describe some experiences in the development of computational tools to exploit the pragmatic and semantic regularities described by our model.

2. FORMAL AND COGNITIVE PROPERTIES OF METALANGUAGE The concept of metalanguage originates with Hilbert and Gödel within the traditions of logic and mathematics, and is the cornerstone of modern formalization of abstract systems, especially symbolic systems like language. It addresses the need to have a formal language of a higher order with which to describe and establish the truth-conditions of another one.1 A metalanguage thus supplies the “conditions of possibility” (to borrow Kantian terminology) of the language to which it refers, defining on the one hand the signifying elements of it, and on the other the combinatorial rules necessary for the creation of propositional sense. In order to function at this very basic semiotic level, natural language has to be split (at least methodologically) into two distinct systems that share the same rules and elements: a metalanguage, and an object language which can refer to and describe objects in the mind or in the physical world. Metalanguages and object languages thus belong to two different semiotic levels, even though they may share the same surface elements and be isomorphic. An important point to make here is that to distinguish between these two semiotic levels is not just a practical convenience but a theoretical prerequisite for its logical workings. From a formal perspective, and even with isomorphism, marking these distinct levels is very basic.

1

In fact, Gödel showed that no formal language is autonomous, or capable of being its own foundation.

Common Ground for Knowledge Through Knowledge of Language 91 Making them prominent in discourse and being able to recognize them, as readers or hearers of technical discourse, is a fundamental property of metalanguages. We will come back to this point when we examine in more detail the role of lexical markers in these exchanges. An interesting aspect of the coexistence of two distinctive systems in the same representational space is that a discontinuity in semiotic levels occurs when linguistic items are mentioned instead of being used normally in an utterance.2 As we will discuss in detail, natural language employs, when necessary, various resources (quotation marks, typography and lexical items) to functionally mark this vital distinction in discourse. Statements about the language that is being used are vital nodes in the interaction between content and form, between knowledge and expression, between personal intuition and intersubjective scientific understanding. Metalanguages are foundational in nature (Lara, 1989), since they have the power not only of describing a communicative code, but of directly enacting and creating it. Strictly speaking, such linguistic operations are not a product of logical derivation or inference but of the axiomatic foundation of a system. Metalinguistic statements actually establish the semiotic systems that constitute the basis of communicative codes. Language conventions are enacted through what Lewis (1969) calls signaling doublets, in which producers and interpreters of utterances link meaning with linguistic representations. As coordinating devices, they allow modeling of intentions, actions and contents as well as alignment of production and interpretation. At the speaker’s side, the intention “when I use X to refer to Y” will be matched by the hearer with “when I encounter X I should take it to refer to Y”. For Lewis, these conventions are valid language-wide (language as a whole is a system of signaling conventions), and this dovetails with Harris’ (1991) metalinguistic hypothesis that implicit linguistic specification devices underlie systemically for all normal utterances. In any case, the information that a metalinguistic statement provides is not derivable inferentially from either the context, a previously standing code, or information publicly available at that point in discourse. This fact ensures its relevance and informativity (de Beaugrande and Dressler, 1981; Sperber and Wilson, 1986). The kind and quantity of information that is deemed relevant in each metalinguistic exchange depends on the context in which it appears (Thoiron and Béjoint, 1991), the computed linguistic common ground that holds for the targeted audience, and the methodological protocols that ensure successful reference in scientific communication. When interpreting text, regular lexical information is applied by default under normal conditions, but more specific pragmatic or discursive information can override it if necessary (Lascarides and Copestake, 1995). But how do these metalinguistic operations materialize in text? Definitions are the kind of metalinguistic statements that readily come to mind, and although they are not the only possible metalinguistic operation, they certainly are the better-studied ones. They have been viewed as, among other things, deep-structure equations that relate a term with its semantic content, or as answers to lexical questions (Bierwisch and Kiefer, 1969; Ogden and Richards, 1929; Riegel, 1987). The classical view of a definition is of Aristotelian vintage: specification of a genus, followed by a statement of differentia that allows unique identification of a referent. But on many occasions conceptual completeness is not the norm in definitions, and what really is being provided is partial information: an additional semantic trait for an entity already

2

This last distinction is credited to Quine (1960).

92 Grounding and Common Ground introduced in discourse, the modification of a pragmatic restriction for an already known item, introduction of a synonym or an evaluation of word usage in a restricted context. In previous work (Rodríguez, 1999, 2004) we have introduced the theoretical notion of EMOs to account for all varieties of metalinguistic modification. In this framework, classical definitions are just specific materializations of EMOs. We will describe and formalize this notion in later sections, but in general we can describe an EMO as a discourse act where specific knowledge about the sublanguage being used is explicitly provided, a semantic or pragmatic knowledge that can be about the items or rules of usage of the sublanguage. This last sentence is in itself an instance of an EMO, as it has served to introduce this very term into our common knowledge space.

2.1. Corpora used in this research In our corpora, we looked at how term introduction and modification was done as well as at how metalinguistic activity was signaled in text, both by lexical and by paralinguistic means. Some of the indicators found included verbs and verbal phrases like called, known as, defined as, termed, coined, dubbed, and descriptors such as term and word. Other non-lexical markers included quotation marks, appositional structures and text layout. The markers we observed broadly overlap with inventories composed by Pearson (1998), Meyer (2001) and others who have described technical definitions in context. We used these patterns as queries to the written portions of the Social Science, Applied Science, Natural and Pure Sciences, Belief and Thought domains of the British National Corpus (BNC). The retrieved 10 937 sentences were then manually classified as metalinguistic or otherwise, with 5407 (49.6 percent of total) found to be truly metalinguistic.3 We have used this extensive collection of sentences (hereafter the EMO Corpus) to analyze metalinguistic discourse operations, as well as for developing an automatic extraction system.

3. NEGOTIATING MEANING AND COMMON GROUND 3.1. Expert-domain communities and the buildup of a common knowledge Scientific discourse is a joint activity (Clark, 1996), at the same time collaborative and adversarial, combining common goals (mutual understanding, lexical coordination) with diverging ones (establishing superiority). Scientific interaction through language can be viewed as an extended conversation among expert peers, and the ongoing production of scientific papers as turn-taking in these conversations regarding both the subject matter of the discipline as well as the correct or acceptable way to convey that message employing a sublanguage’s conventions. These turns are anything but spontaneous, and are carefully crafted within stylistic, argumentative and thematic constraints. There is little room here for improvisation, and positions are expected to be backed by empirical evidence or rational arguments that must be accessible, in principle, to all that have similar training and previous knowledge of the general state of

3

The criteria for classification are described in a later section.

Common Ground for Knowledge Through Knowledge of Language 93 the field. Domain experts form communities of peers4 who do not only control high level knowledge, but for whom producing new knowledge constitutes one of their main activities and goals. Discovery and systematization of new knowledge is a dialogic, interactive and collaborative enterprise. Scientific activity involves constructing intricate conceptual systems that, by virtue of rational consensus, can become communal knowledge spaces, repositories of accepted facts, theories and terminology. In addition to the background facts needed for semantic interpretation of regular, everyday utterances, scientific interlocutors share an enormous amount of assumptions and ontological commitments as domain-specific common ground. In fact, a significant aspect of their own constitution as experts consists of assimilating publicly available empirical evidence, explanations, models and theories that all can refer to, and thus become the contexts for, the dynamical debate that powers scientific advancement. We could call all these, collectively, the conceptual common ground which scientific research is all about, and which in general is part of the dynamically updated collective common ground described in Fetzer (2004, and in this volume), providing shared goals, inference strategies, etc. But these knowledge-based communities are also language-based communities, insofar as one of their foundational conditions, along with a common domain subject, is a shared sublanguage. For Clark (1998: 80), a Communal Lexicon is “the vocabulary associated with a community of people who are distinguished by their common knowledge of a particular field of expertise.” These linguistic conventions can arise in what Clark (1998) calls explicit agreement, when an author stipulates explicitly how he is going to use a term, and what meaning conditions he will attach to them. Clark identifies the following as some of the phrasal and lexical markers used in these operations: “what I shall call”, “let us call this”, “hereafter”, “for short”, “termed”, “named” and “abbreviated”, but states that more elaborate formulas are possible. If we extend the notion of “conversational record” (Thomason, 1992) to include the multiparty subject-specific exchanges of scientific debate, we can imagine that participants keep in mind such a registry of meaning-in-context when they write or read scientific papers, in order to establish the conceptual common ground of the expert community they belong to. Such “score-keeping” would of course extend into conceptual positions and established empirical facts, but it would crucially include a semantic and pragmatic dimension that ensures close alignment of linguistic expression and intended meaning; in short, terminological alignment. It is unlikely that either in scientific settings or in everyday conversation the entire aggregate of all possible data constituting a common ground between two interlocutors is activated and presented when processing language. A more realistic model of common sense and shared technical knowledge would involve selective activation of inferential chains as discourse develops and memory brings into focus, as required, one or another sector of a systematic network of concepts. A situation model (especially one as complex as a scientific theory) presupposes an enormous amount of interlinked facts, events and actors, but does not require such a lattice to actually be present in the mind for adequate interpretation of language. It only needs to be available for lookup when needed, and for some mental and linguistic generative devices to be used in a way that is shared by all. There must also be social group mechanisms for sharing and agreeing upon such background models. A communicative

4 To call the participants of a scientific debate “peers” can be misleading. As in conversation, certain speakers’ utterances carry more weight than others, and the influence of a pecking order in learned discussion is difficult to dismiss.

94 Grounding and Common Ground common ground has to be negotiated between peers, although factors other than rationality or communicative efficacy influence such decisions. The rhetorical aspect of scientific debate and the role that language plays in competitive interaction have been studied by Fowler et al. (1979), Kotschi (1986), Gross (1991), Myers (1992), Hunston (1996) and Liddicoat (1997), among others. Terminology can thus be thought as much a battle ground as a common ground.

3.2. Terminology, scientific discourse and the buildup of knowledge We have argued that a key feature of scientific discussion as a joint activity is the communitywide alignment of conceptual contents and the expressions to be used for them. In fact previous turns in those multiparty dialogues (i.e., prior publications dealing with the same subject or problem) usually explicitly put forward these systematic ensembles of “facts” for consideration by a community of peers, so that they can be identified, referred to and retrieved through standard quotational devices. A standing terminology accepted by a significant portion of the participants in a specialized exchange becomes a veritable “common ground” when it can be jointly and interactively changed by the participants in expert community. It is a framework that, paradoxically, is continuously strengthened by its vulnerability to change. Terminological alignment has system-wide consequences, and it involves modeling the sets of shared beliefs and ontological commitments of the interlocutors, similar in some ways to what Bach and Harnish (1979) would call Mutual Contextual Beliefs, Pickering and Garrod (2004) Situation Models, or Clark (1996) Common Ground. This cumulative common ground has a very stable core (unless amid a process of revolutionary change in the field), and an outer layer that is fluid and in a state of constant change. Both general and specialized knowledge, insofar as abstract conceptual structures, behave like metastable5 systems that undergo localized changes that globally transform the structure of a system. Parallel to the introduction or modification of conceptual structures (and intimately linked to them), sublanguages grow and evolve in a process in which insidious change slowly but completely transforms the theoretical landscape. The acquisition of both everyday language and more formal and technical sublanguages imply a process of assimilation of the rules and lexical components of a communicative code. Knowledge of language is also a special kind of specialized knowledge. Students of a technical domain need instructions about how to use a sublanguage. Specialization is precisely this acquisition of a particular domain of knowledge; it is a gradual and conscious effort developed throughout a considerable amount of time. Directive utterances concerning language use are vital in those learning processes, and play a very important role in both of them. But here similarities end. A deeper background representing the general lexicon of natural language is shared by all researchers due to their own normal linguistic competence as individuals. A technical common ground, however, has to be learned, constantly re-created and discussed, and it serves to sustain a varied set of epistemological assumptions about reality. In the case of learning everyday language, lexical items and grammar rules are part of a collective heritage that is relatively stable historically and culturally, and is retrieved from a repository where there is no personal attribution for meaning. But terms and sublanguages are consciously

5

Gentilhomme (1994) borrows this term from chemistry.

Common Ground for Knowledge Through Knowledge of Language 95 created, evolve rapidly and are linked to specific theories, personalities and texts, to events and to specific functions. They were introduced somewhere, sometime by someone, and then adopted (or rejected) volitionally by a community of speakers that is a subset of the speakers of a language.6 Domain-dependent words or terms have low occurrence statistics in the normal interaction of the general population, but not so in the specialized contexts where they have been spawned and where they evolve. They help define the communities that use them as much as these communities define those very terms. Language and knowledge-based communities are defined by their practices, beliefs and interaction, by their common ground. They are constituted against the wider frame of society in general, and establish with other groups and individuals certain links and relationships. Putman’s hypothesis of the universality of the division of linguistic labor (1975) suggests that some lexical items require interpretative recourse to (and also some form of interaction with) groups of individuals that in some sense control meaning conditions outside the mainstream of normal linguistic competence. Every linguistic community (…) possesses at least some terms whose associated “criteria” are known only to a subset of the speakers who acquired the terms, and whose use by other speakers depends upon a structured cooperation between them and the speakers in the relevant subsets. (Putnam, 1975: 228) When experts interact amongst themselves, they must assume that their audience of peers knows most of the Communal Lexicon (Clark, 1996, 1998) that holds for their field. This audience also assumes that a speaker will use both a common repository of knowledge accepted by all as well as an accepted and unambiguous terminology to refer to it and at the same time will expect that new knowledge will be offered by the speaker at its turn at scientific communication. It will also assume that if new knowledge or new terminologies are introduced in discourse, then evidence, arguments or interpretative conditions will be provided so that these innovations, terms or ideas can be understood and accepted at large by the community of experts whom the speaker is addressing. Of course, expert communities can also form subcommunities defined by areas of specific knowledge in the field, or by acceptance of diverse and competing explanatory theories of the phenomena. These subcommunities would share some, but maybe not all, of the terminology and common ground of the whole field, and must correct accordingly their assumptions about what is commonly accepted as far as linguistic conventions, conceptual inferences and empirical evidence are concerned.

3.3. Lexical markers, communal lexicons and metalinguistic repair Lexically marking an utterance as metalinguistic creates special conditions for its interpretation, and signals the need for a repair in the information flow. Jointly reworking the communicative code is a resource-costly cognitive process that integrates within coherent conceptual structures a considerable amount of new and very complex terms. Metalinguistic verbs and descriptors

6

We could even argue that all terms owe their existence to an original baptismal act where it was introduced to the speech community, although an impossibly exhaustive historical corpus for all disciplines would be needed to fully support such a statement.

96 Grounding and Common Ground make clear that there is an issue with the shared linguistic conventions that needs to be resolved before discussion of normal domain-centered topics can resume. In the background of this process there is the belief that the speech community shares a more or less established linguistic convention (Clark, 1998; Lewis, 1969) that allows them to communicate, and that anyone with the right expertise in the subject matter shares a core vocabulary and rules of usage for the particular sublanguage, even if he or she doesn’t master every possible term in the field. The underlying assumption here is that there is a consensus about what is meant when using shared linguistic expressions, a belief that goes beyond the truth of falsity about what is actually being stated. Metalinguistic markers do more than signal a common ground. They also signal when something is not assumed to belong to the common knowledge state of the participants, something that has to be negotiated or reported in order to become part of the lexical background needed to ensure mutual understanding. Metalanguage brings a new or modified lexical item to the attention of the expert community in order to specify its meaning conditions. What needs to be addressed in a metalinguistic speech act is the presumed with inoperability within a specific context of what Bach and Harnish (1979: 7) call the Linguistic Presumption (LP): “The mutual belief in linguistic community Cl that: 1. the members of Cl share [language] L, and 2. that whenever any member S utters any [expression] e in L to any other member H, H can identify what S is saying, given that H knows the meaning(s) of e in L, and is aware of the appropriate background information.” Metalinguistic repair ensures that expression e is clarified and becomes part of the common communicative code that S and H share, and which in an important way defines them as a linguistic community.7 In the case of technical terminology without wide currency “then special beliefs to the effect that H is acquainted with the vocabulary are required by S” (Bach and Harnish 1979: 8). Metalanguage, then, not only affects linguistic performance per se, but establishes and renews sociocultural links between individuals. In metalinguistic exchanges a communicative misalignment needs to be repaired and agreed upon, if only for the sake of the arguments that might follow. A clarification, qualification (Guelich and Kotschi, 1995) or reformulative procedure is called for which may resolve communication problems such as ambiguity, coherence or reference. The expert community must actively maintain coordination of a usable lexicon that reflects the states, objects, events and causes accounted for in their theories, explanations and empirical descriptions. A cognitive contrast has to be presented between what is assumed to hold for all, and what the speaker will introduce as new information. This is one of the important roles of lexical items used to mark metalanguage, and the reason for its dual nature with regard to meaning and function. Marked metalinguistic topics introduce an issue that has to be dealt with on a semiotic level that is distinct from the purely referential level of explanatory and descriptive discourse. This signaling has, then, a twofold purpose: to call attention to an item that is not yet part of the common vocabulary of the expert community, and to introduce the information needed 7

Bach and Harnish (ibid:8) state that “as a matter of social fact, the LP in a community is so strong that not to know the language is often a sign of nonmembership in the community”.

Common Ground for Knowledge Through Knowledge of Language 97 for its subsequent interpretation. Their intended perlocutionary effect is for members of a language community to adopt that usage for the short- or long-term scope that the speaker intends to be valid.8 Explicitness is, again, paramount for correct interpretation of such repair actions. Reformulations are repair strategies concerned with contextual meaning, but metalinguistic operations can also be concerned with “default” meaning beyond the local scope of ongoing discourse, beyond the immediate “conversational record”. Modification of lexical meaning can work retroactively or prospectively to change or introduce interpretative conditions at the most basic level. Also, the specification of scope for a proposed metalinguistic modification or clarification becomes important to establish a wider or more restricted context for a repair procedure. Against this theoretical background, we will now turn to how these alignment and repair mechanisms function in scientific discourse, and how lexical items, in their dual role as metalinguistic markers and operators, help accomplish these processes.

4. METALINGUISTIC EXCHANGES AS SPEECH ACTS 4.1. Linguistic alignment and the double nature of lexical markers Metalinguistic statements are special kinds of speech/textual acts, and involve a definite performative dimension. We have argued that in theoretical research knowledge is constantly being created and monitored through metalinguistic statements that allow for negotiation of the conceptual frameworks that materialize in complex terminologies. Through learned discussion in specialized journals a terminological proposal is put forward within a textual instance that does not receive immediate feedback or challenge (as in regular face-to-face conversation), and a collective process with two or more interlocutors (and thus with a wider scope than just dialogical) is initiated, resulting in either tacit, implicit or explicit acceptance of the new linguistic representation, or in a counterproposal that reinitiates the interactive alignment process. Acceptance of a new terminology can be made explicit, or it can be inferred from the lexical choices of subsequent utterances in turns of others, in future texts. This acceptance has a strong expression of contextual effect (to borrow Sperber and Wilson’s term) that will create a fairly specific context for the interpretation of follow-up exchanges and the transmission of subject-matter information. In terms of situation models, EMOs aim at alignment of the communicative channel through which a shared world-view can be agreed upon. It is an extended and more complex instance of lexical alignment between interlocutors (Clark and Wilkes-Gibbs, 1986; Garrod and Anderson, 1987; Pickering and Garrod, 2004). In contrast with common conversational devices, most of the negotiating procedures affecting terminological change can’t be implicit, and require explicitness to be minimally effective. Lexical and paralinguistic devices make salient a standing problem in the communicative code, and at the same time provide a procedure or some information to resolve the issue. 8

Another effect is to be recognized by the members of their community as individuals that can exert some kind of control or power over the meanings and content of the jointly established subject matter.

98 Grounding and Common Ground Lexical markers do not operate solely at the level of purely semantic meaning, but in the wider discursive context of the joint construction of sense through procedural devices and categorization (Sperber and Wilson, 1996; Nyan 1996). The inventory of these kinds of lexical markers is finite and regular. They are subject to less expressive variation than other lexical items and we can assume some of them have undergone certain grammaticalization. The metalinguistic lexical markers described here signal a negotiation process, more than just pointing to propositions that are already part of an implicit common ground. They also play a processing role, which is why I have termed them Marker/Operators (Rodríguez, 1999, 2004). Lexical markers, in general, are defined by this dual role of meaning something while at the same time signaling a discourse process or relation. Since the “dialogues” inherent in scientific debates are played out in widely separated turns, the simple priming mechanisms suggested by Garrod and Anderson (1987) are unlikely to be at work here, but nevertheless we can conceive metalinguistic Marker/Operators as “priming” the reader to process sentences or phrasal fragments that deviate from the expected thematic content of the exchange, and which require special processing conditions, long-term storage and the full cooperation of all participants.9 Pickering and Garrod (2004) emphasize dialogic interaction as a prime model for linguistic processing, whose study (they claim) has formerly relied too much on monologue. In their view, interlocutors automatically align their linguistic representations simultaneously at many levels, instead of through interaction only at levels suited for more simplistic and linear accounts. Through simple priming and inferential mechanisms that create routines common to them, interlocutors ease the computational load of producing and interpreting language. The very nature of metalinguistic negotiation makes it difficult to conceive a significant role for low-level alignment mechanisms in these textual contexts. Coordination at a content level requires coordination of meaning. Establishing a linguistic “common ground” requires a high-level alignment that crucially involves modeling the interlocutors’ sets of beliefs and entities. In order for a scientific paper to have any kind of communicative success, its referential expressions must be established intersubjectively through complex protocols. Establishing terminological conventions allows dealing with a central problem of language use: “coordinating what a speaker means and what their addressees understand them to mean” (Clark, 1996: 3). In pragmatic terms, EMOs could be described as two very basic discourse actions: to inform on a linguistic state of affairs (i.e., establish a standing communicative code), or direct either the coding or the decoding of a message. Both can be seen as being equivalent on a higher level of abstraction, as an informative statement can imply that a convention stands, and could be interpreted as a hidden directive to follow when communicating within a group. Using Austin’s (1962) typology of illocutionary acts, EMOs can be equally interpreted as Directives (make speakers adopt the convention), Assertives (assert that a convention is part of a terminological state of affairs), Comissives (make others adopt the convention from hereforth). We could then add a more specific act to Austin’s list, a Metalinguistic speech act that would have a double nature: as an invitation to use a sign with a conventional meaning, and a declaration that certain meaning is attached to a linguistic sign within the accepted code of a linguistic community. This double nature would reflect its effect on both production and interpretation of the joint communicative activity.

9

See, for example, the processing described by Nyan in this volume.

Common Ground for Knowledge Through Knowledge of Language 99

4.2. Relevance, intentionality and felicity conditions As with other speech acts the intentions of metalinguistic utterances have to be recognized in order for them to be effective at all, and their felicity conditions imply the tacit or explicit acceptance by a community (or at least part of it) as a valid report, modification or addition to the standing communicative code. Making those speech acts prominent through the use of certain lexical items is a signal of a reflexive intention (Grice, 1957; Searle, 1969), an intention that is intended to be recognized as such. Further communication that actually uses the conventions would indicate that those speech acts have attained their goal. The producer of such a message won’t find closure evidence during his own dialogic turn, but might infer it by monitoring follow-up turns for either explicit acceptance of his proposal or implicit acceptance showing in language use that follows it. Bach and Harnish (1979: 42) classify the speech act of christening as a conventional illocutionary act that is fulfilled by satisfying a convention, as opposed to other speech acts fulfilled by means of recognition of their intentions. Conventional acts, in their view, affect “institutional states of affairs”, and can fall into the categories of effectives and veridictives. Lexical and phrasal markers, in the case of christening and other metalinguistic operations, activate the convention of modifying or reporting a vocabulary, affecting the state of affairs of the standing communicative code in varying ways and scopes, a state of affairs that is constitutive (if not identified with) joint common grounds. The caveat here is that these acts can only affect such states of affairs if they are accepted as acts of metalinguistic modification or report, not only with regard to their speaker intention, but in their actual communal effect by the hearer’s acceptance of (either new or standing) semantic/pragmatic content for the defined term, even if that acceptance is granted for the sake of the argument or subjected to other constraints. Although the intention may be fully recognized, a metalinguistic speech act that fails to be adopted by the linguistic community to which it is addressed cannot complete its illocutionary goal. Computing what constitutes relevant metalinguistic information to be offered in discourse involves making certain assumptions about what others in the community accept as part of the field’s common ground. It involves a calculus of which information about the sublanguage can be considered new, and which should be considered as given. What participants know about the subject at hand must be taken into consideration (Is this information relevant? Did someone else offer it before me?), as well as their knowledge of the terminology that will be used to enact that knowledge (Do I need to create a new term for this referent or concept? Is there an existing term I could use? Does my audience prefer another equivalent term?). Although this kind of high-level calculus of basic common ground is seldom done for everyday conversation, this terminological alignment is crucial in special domain communication. Some of these presuppositions about what can be taken for granted by all can be very complex, but at the same time they should be explicit and publicly available in a way that normal conversational assumptions may not be. Insofar a speech (textual) act that needs to be interpreted, a metalinguistic statement presupposes, in order to be considered relevant and informative, that (1) an actual problem obtains with regard to the linguistic code itself, (2) there is some information that is new to whom it is addressed, and (3) this information is not inferable from other sentences already uttered or from information that is readily available. Readers generally expect high relevance of statements requiring unusual interpretative conditions. A good reason has to be established for the resource-costly alteration of a common

100 Grounding and Common Ground linguistic code. The very explicitness of metalinguistic discourse is grounded in the fact that relevant information about the communicative code is being provided, in effect making opaque and visible what is usually “transparent” and invisible (language use itself). Altering the meaning conditions of scientific discourse can only be justified if such operations have some impact in the common cognitive state of the expert-domain community, if they allow better understanding, structuring or communication of scientific knowledge. Metalinguistic predication allows grounding of shared knowledge states by establishing conventional meanings for discourse. We could state this by saying that when a speaker puts forward a modification of the standing linguistic code, or he informs of a linguistic convention, he does so to ensure felicity of previous or subsequent communicative acts, and to establish successful reference within a shared cognitive state. We could state two principles of metalinguistic predication as follows: 1. When describing or modeling a state of affairs, it is possible to use terms with interpretative conditions that are not assumed to be known by the linguistic community to whom the utterance is addressed, provided that the producer makes explicit the pragmatic or semantic rules needed for intersubjective interpretation of those utterances. 2. Metalinguistic statements imply an invitation to generalize and adopt, from that point forward and within certain contexts and a stated scope, a convention in the standing communicative code, and to incorporate it into lexical and cognitive common grounds valid for the language-defined community. We will avoid dwelling deeper into the question of whether Markers/Operators function on the basis of conventions, standardization or inference, or whether they enact rules, practices or norms. They may actually share some traits from this diverse menagerie of theoretical concepts. The dual nature of lexical and phrasal markers as flaggers of special interpretative conditions and as illocutionary operators points to the need for a complex explanation of their range and effectiveness, but we believe that our description is sufficient for current purposes.

5. EXPLICIT METALINGUISTIC OPERATIONS: A LEXICAL AXIS FOR PREDICATION We have argued previously that lexicographic definitions used to study metalinguistic predication provide too narrow a descriptive framework. We will propose a model of how EMOs function that attempts to describe these discourse operations, while at the same time allowing for an elementary formalization for their computational treatment. Explicit Metalinguistic Operations are theoretical entities modeling those discourse acts in which specific knowledge about the sublanguage being used is explicitly provided, a knowledge that can be about the items or rules of a sublanguage. We have discussed metalinguistic predication as deep-structure equations that help explain its foundational nature. EMOs can be logically analyzed as constituting complex formulas where the predicates receive their functional value from the semantics of the verbs and lexical markers involved. We can posit that metalinguistic operations are discourse actions involving three distinct elements,

Common Ground for Knowledge Through Knowledge of Language 101 articulated into a predication where those elements play specific roles that make the sentence meaningful and effective at the same time. In metalanguage, an element of a clear semiotic nature (word, sign, term, or however we might refer to it) is specified with regard to a system or code by virtue of a link with some entity or information acting as a predicate for it. This articulation into meaning is done mainly by certain lexical elements that have the dual role we have suggested before: (a) flagging or marking the metalinguistic as well as autonymical10 nature of the sentence as a whole, and (b) functioning as operators performing a concrete modification of the code, supplying specific interpretative cues as to the particular operation being performed. These two different roles can materialize in text through the same surface elements. All metalinguistic operations can be identified by several explicit or implicit roles and elements: a self-referential lexical item, relevant information about it, and a way to convey that information and link it to that item. In order to identify such operations in ongoing text, we specify some minimal requirements, and established basic elements that should always appear in the sentences performing such operations. The features in listing I describe the general requirements that a sentence or a phrase should have in order to be considered an EMO. Listing I. EMO requirements: 1. There has to be a linguistic sign that is the subject (either in logic, discourse or grammar) of a predication. 2. The utterance should count as a contribution of relevant information about the status, coding or interpretation of the linguistic sign. 3. The whole operation should be prominent or conspicuous because of its metalinguistic nature. An EMO, then, should always have at least three basic constitutive elements corresponding to these requirements, which we present in listing II. Following each one, we show how each of those elements can be projected over an example sentence from our exploratory corpus: Listing II. Constitutive elements of EMOs A) a (complex or simple) term or linguistic sign functioning as a selfreferential term or autonym, which stands as the subject of the operation; KENES B) a semantic or pragmatic information to be linked up with the term or lexical unit; The bit sequences representing intentionally similar to “genes”.

10

quanta

of

knowledge

KENES

Any element of linguistic code can become a name unto itself and automatically acquire nominal traits, in what Rey-Debove (1978) terms a “metalinguistic rewriting rule”. Rey-Debove and Carnap (1934) call this self-referential condition autonymy.

102 Grounding and Common Ground C) a metalinguistic or “definitional” verb or verbal phrase, a typographical, punctuation or dispositional mark, or a combination of other semiotic resources that connect (and conceptually articulate) element A and element B, while flagging the extraordinary (non-referential) nature of the segment (Markers/Operators). The bit sequences representing quanta of knowledge will be called “KENES”, a neologism intentionally similar to “genes”. Regardless of the syntactic configuration of an EMO sentence, its deep structure can be considered an articulation of the three theoretical objects described above. In the BNC-based EMO corpus we have described previously, a total of 5407 sentences were identified as EMOs using these requirements. In the next sections we will review each of the EMO elements in more detail, and show how metalinguistic information constituting the terminological common ground of specialized discourse is provided in a predication articulated by an axis of lexical markers.

5.1. Elements and basic features of EMOs For the sake of clarity, I will start by presenting some examples from our corpora in Table 1, where the first full-length row contains a complete EMO, the first column in the following

Table 1. EMO elements This means that they ingest oxygen from the air via fine hollow tubes, known as tracheae. Informational segments

Markers/Operators

Term

fine hollow tubes

known as  apposition

Tracheae

In 1965 the term soliton was coined to describe waves with this remarkable behavior. Informational segments

Markers/Operators

Term

to describe waves with this remarkable behaviour

coined  the term

Solition

Integral power results in a fundamental type of social classification which, adapting Bernstein’s terminology, I shall call frame (Bernstein 1971). Informational segments

Markers/Operators

Term

a fundamental type of social classification  adapting Bernstein’s terminology,

call  Quotation marks

Frame

Common Ground for Knowledge Through Knowledge of Language 103 row shows the lexical item(s) figuring as autonyms (or self-referential elements). The next column contains the lexical, pragmatic or paralinguistic elements that help flag and articulate these discourse operations, while the last column presents the actual informational segments where something is stated about the lexical item.

5.2. Autonyms EMOs essentially present lexical items that stand in autonymical condition, and which regardless of their usual grammatical properties exhibit nominal syntactic features. They constitute the subject about which something is predicated, with regard to value, meaning or interpretation. A definitional sentence might not always contain an item explicitly marked as metalinguistic. In those cases, it can be argued that besides operating on a conceptual level, as opposed to a purely semiotic one, such sentences can receive an optional metalinguistic interpretation. A definitional sentence, such as the following one from a biomedical abstract, would not be considered to be an EMO under the current framework, but could certainly receive a metalinguistic interpretation:11 Amyotrophic lateral sclerosis (ALS) is a neuro-degenerative disorder with both sporadic and familial forms.

5.3. Metalinguistic Markers/Operators Metalinguistic Markers/Operators ensure that autonyms can be successfully interpreted as such, and they provide further clues as to how metalinguistic predication must be performed. We have argued that lexical markers of this sort constitute the predicative axis of the discourse operation. The most common markers in our exploratory corpora were quotation marks and the verbal items “called” and “known as”. Appositional structures are also very common markers in technical texts that usually signal relations of equisignificance, especially when providing synonyms and acronyms. Some markers are linked to very specific kinds of EMO: for example, the combination “where” plus colon usually signals meaning attribution within a formula, as when defining the variables in formalization, and must be considered to have very “local” scope that usually does not affect a linguistic code on a global level. I define the Neighborhood Sorting Index (NSI) as: , where: Y is household income, i indexes households, n indexes neighborhoods (…) In some cases it is very difficult to discriminate between valid metalinguistic uses of a verb lemma such as “call” and other uses that are not metalinguistic, as we shall discuss when we briefly describe a computational implementation of these principles. In general, using certain tenses and auxiliary verbs in the co-presence of other indicators like quotation marks and

11 Many cases of copulative sentences in technical corpora can receive this metalinguistic interpretation, and constitute cases that could be called “Implicit Metalinguistic Operations”.

104 Grounding and Common Ground lexical modifiers will help perform a better job at correctly identifying metalinguistic instances from our corpora. In the following listing (III) we present an inventory of elements that appear in sentences containing EMOs. The nature of each one contributes in its own special way to the overall predicative and cognitive processes. Listing III. Common Markers/Operators in EMOs from our corpora (A) Lexical: 1. Descriptors: term, word, phrase, terminology, vocabulary, name, definition, etc. 2. Metalinguistic verbs: calls (is called), means, termed, name, refers to, use (is used), speak, designates, known as, stands for, defined as, coins, corresponds to, applies to, dubbed, designates, labels, indicate, said, etc. 3. Other lexical indicators: subtitle, oxymoron, where, etc. (B) Syntactic: Apposition, copulative clauses, etc. (C) Pragmatic: Informational structure, hedging, etc. (D) Paralinguistic: • Layout: footnotes, highlighted text, tables, etc. • Typography: bold, italic, different type face, etc. • Punctuation: quotation marks, parenthesis, etc. There are two important observations to make here. First of all, none of these elements by itself is enough for a positive identification of an EMO. It is the whole articulation of items from different dimensions of linguistic and textual structure that bring about a complex, foundational discourse action such as the ones we have been analyzing. These elements and patterns are not always exclusive to metalinguistic statements, but most of the time they can signal that such processes are taking place in text. Unlike the fields of lexical databases or the feature structures we sometimes employ to represent lexical entries, actual language use does not usually segregate semantic, pragmatic or syntactic data for our convenience. Moreover, some lexical markers are polysemic and can perform different functions in different contexts; some otherwise reliable lexical indexes of metalinguistic activity can be misleading and would need disambiguation. An interesting finding of our review of metalinguistic sentences is that usually at least two or more different lexical and paralinguistic elements are present. The constitutive markedness of metalinguistic operations means that most of the times these sentences will have at least two indicators present, for example a verb and a descriptor, or quotation marks, or even have preceding sentences announcing them in some way. Redundancy of markers diminishes ambiguity or confusion. Some elements might just have the function of flagging something as metalinguistic (e.g., quotation marks), while others also contain some processing directive by virtue of their semantic content. Formal as well as cognitive properties of EMOs facilitate the task of locating them accurately in text and interpreting them correctly.

5.4. The variety of informational predicates Referring expressions and pronouns that represent referents for the autonyms are the more common instances of informational segment identified in our exploratory corpora, but they

Common Ground for Knowledge Through Knowledge of Language 105 are by no means the only kinds of linguistic realizations for these EMO components. Besides providing a referent through denotation or supplying conceptual data to identify it (i.e., referent identification through extension or intension), EMOs might offer a judgment (usefulness, validity, etc.) on the use of a term within a context, or an evaluation of word usage. Durkheim’s usage of the terms “psychic” and “affective” seems euphemistic. An EMO might provide a characterization of a term that qualifies it in some way or other. While these kinds of statements do not provide core semantic information, they might be important as instructions about how a term should be or is actually used in context or with regard to its connotation. Sometimes the information provided in metalinguistic contexts is restricted to the fact that some meaning conditions need to be reexamined, without locally presenting a new configuration, as in the next example: At this point we should give some further consideration to what exactly the widely used term identity is supposed to mean. Furthermore, metalinguistic information provided might amount to a restriction on an item’s usage or a specification about what that item does not mean, rather than what it actually means in a specific context: By primacy we do not mean exclusivity Intra- and inter-sentential relationships also play a role in identification of the informational segment that contains what is being linked to the autonym. Anaphoric pronouns and other similar lexical items (like WH-pronouns in relative clauses, or determiners) might point to other sentence constituents, or refer to other preceding or subsequent sentences that express entities, or modifiers and complements for the term. The following example presents an instance where the information about how exactly the term flux pattern is to be used (or more precisely what its referent is) will have to be retrieved from a previous utterance, and that information is connected to the EMO by the use of the pronoun “this” (bolding is mine). This shall be referred to as the flux pattern. Another special case of informational segments is when there is no explicit or superficial textual fragment that could be linked to an autonymical role, and we have to look at a deeper logical level. Some markers like so called flag a denomination where the referent is elided completely and has to be inferred from context: But the so called “opinion” polls have become more frequent. What is being stated here is merely the existence of a referent for the autonym, so that in this last example we know that there is something that is called with the name opinion polls. In

106 Grounding and Common Ground these cases of indirect reference we have to posit a logical form where an existential operator is linked to a variable, or what we may call an existential placeholder, formalized as ∃x Nx y, where x is a term denoting an entity, y is a linguistic sign, and N is a nominative predicate linking y with term x.12 The hidden reference would be paraphrased as “there is something, and that something is referred to with autonym y”. Finally, information about an autonym might not be localized in a single segment, and would have to be retrieved from various constituents, to be reconstructed either as a single piece of information or as two or more distinct pieces of information. Besides core semantic and pragmatic data, EMOs can also offer other potentially useful pieces of information within the metalinguistic operation, like scope of terminological specification, illocutionary force, hedging, participants in the exchange, source attribution and so on. These expressive devices help the speaker position herself/himself with regard to the actual knowledge being evaluated, and signal the consensus processes taking place in the framework of the interaction of the expert group. We have argued previously that metalinguistic exchanges in technical text are subjected to the same complex dynamics that govern dialogic interaction in conversation, and that language provides the devices used by the participants both to create meaning as well as to perform discursive actions that play a decisive role in their non-linguistic interactions. These intentions, strategies and inferences, and the whole discourse production process, leave traces (Guelich and Kotschi, 1995) in the linguistic materializations of EMOs, and although these traces might include paralinguistic traits and punctuation, lexical markers are the main functional elements in these discourse actions, providing a way to flag these sentences as well as supplying information about how to interpret them. We have presented the formal and cognitive properties of metalinguistic statements that allow them to be the foundational cornerstone of symbolic systems. We have also offered a perspective of metalinguistic exchange in scientific and technical sublanguages that constitute the context of these operations within the community of users of those sublanguages, with an extended description of what role metalanguage plays in the creation and structuring of knowledge. Finally, we have elaborated a descriptive model of EMOs that allows the identification of their constitutive elements and the functions each one contributes to the global working of these fundamental discourse operations. These findings suggested that EMOs could be exploited fruitfully as knowledge-rich contexts for extracting terminological information. Although statistically sparse,13 EMOs can be located in a straightforward manner by using lexical markers as indicators for recognition in computer applications that do not need large lexicons or high-level processing like deep syntactic parsing. The predictability and regularity of metalinguistic predication across domains, along with the constitutive markedness of such sentences, allowed development of a computational system that used Information Extraction techniques and heuristics to extract the metalinguistic information that all specialized text contain alongside its regular subject-matter information. The proof-of-concept system (Metalinguistic Operation Processor, or MOP) we briefly describe in the next section shows how this goal was achieved. In a sense, it is a practical demonstration of the powerful theoretical characteristics of the lexical Marker/Operators we have put forward as functional

12

This analysis accomplishes Quine’s goal of translating a linguistic expression into a logical form that supports its ontological commitments better. 13 They constitute between 0.6 and 5 per cent of the corpora we have explored.

Common Ground for Knowledge Through Knowledge of Language 107 axis of EMOs. In the framework of this volume’s issues, it can be considered an attempt to automatically extract a linguistic common ground from a corpus of related texts.

6. DISAMBIGUATING LEXICAL MARKERS IN THE MOP SYSTEM The first issue to tackle when automatically mining metalanguage is how to obtain a reliable set of candidate sentences for input into the next processing stages. Lexical markers for metalanguage (unlike other discourse markers like reformulation flags, e.g. “but”) are highly context-dependent, and their meaning and function need to be inferred. This disambiguation, although trivial for human processing, presents unique problems for computer systems. A key problem with approaches that use lexical-based “triggers” is how to control the amount of “noise”, or non-relevant instances, that might be encountered alongside useful ones. Although we have stated that lexical markers fulfill a “flagging” function, they are not completely reliable since they can serve other purposes in different contexts. The experiments performed during the development of the MOP system, and presented in this section, compare two different approaches for this filtering task that are representative of the main symbolic and statistic techniques in Natural Language Processing. From our initial analysis of various corpora we selected forty-four lexical patterns that showed the best statistical reliability as EMO indicators. Their selection was based on a high degree of reliability as indicators of EMOs in text (with a lower threshold of around 75 percent EMOs of all sentences), as well as other considerations: for example, (a) although some patterns presented a high percentage of EMOs, their number of hits was too low for inclusion; (b) in cases where there was a lower but significant percentage of EMOs, if reliable collocation-based exclusion rules could be formulated it was selected, since metalinguistic instances would be identified with a tolerable margin of error. Using these patterns, we first extracted a set of candidate sentences for filtering out non-metalinguistic instances before further processing. Our filtering strategies distinguished between useful results, for example using the lexical pattern called in (1) from a non-relevant instance in (2) that uses the same lexeme: 1. Since the shame that was elicited by the coding procedure was seldom explicitly mentioned by the patient or the therapist, Lewis called it unacknowledged shame. 2. It was Lewis (1971, 1976) who called attention to emotional elements in what until then had been construed as a perceptual phenomenon. We experimented with two strategies for lexical disambiguation: first, we used collocations found in our corpora as added restrictions (verbal vs. nominal occurrences of our lexical markers, for instance) to discard non-metalinguistic examples, e.g., attention in sentence (2) next to the marker called. We also implemented learning algorithms14 trained on a subset of our EMO corpus, using as vectors either Part-of-Speech tags or word strings, at one, two and three positions adjacent before and after our lexical markers. The details and results of

14

Naive Bayes and Maximum Entropy algorithms that have been reported to work well with sparse data.

108 Grounding and Common Ground these experiments have been reported elsewhere (Rodríguez, 2004), and we won’t repeat them here, except to say that both Language-Engineering and supervised learning approaches were adequate for initial filtering of metalinguistic sentences, although learning algorithms allowed easier transport into new domains. Overall, it seems clear that the lexical patterns (along with their most immediate co-text) that our corpus analysis identified constituted very good indicators of metalinguistic activity in text. Subsequently, the MOP system created Metalinguistic Information Databases (MIDs) that constitute useful sources of information for lexicographers, as well as valuable electronic resources for language processing. Conventional resources like lexicons and dictionaries compile meaning definitions that are considered stable and widely shared by a linguistic community. They can be seen as repositories of the default, core lexical information for terms used by a research community (i.e., the information available to an average, idealized speaker). A MID, on the other hand, might contain the multi-textured real-time data embedded in research papers, and in this sense could be conceptualized as an anti-dictionary: a listing of exceptions, special contexts and specific usage of instances where meaning, value or pragmatic conditions have been spotlighted by discourse for cognitive reasons. Terminological information can be more specific and might be better suited for the interpretation of certain texts. While the resulting MOP “definitions” generally do not present high readability or completeness, these informational segments are not meant to be read by laymen, but used by domain lexicographers updating existing glossaries for neological change, or, in machine-readable form, by other applications.

7. STUDYING COMMON GROUND THROUGH KNOWLEDGE OF LANGUAGE MIDs cannot be viewed as end-user products but as semi-structured resources (midway between raw corpora and full-fledged lexical knowledge bases) that need further processing to convert them into usable data sources. We might better characterize them as auxiliary lexical knowledge resources, more than core lexical references. The MIDs could in principle supply new interpretation rules in Artificial Intelligence applications when inferences won’t succeed because the state of the lexico-conceptual system has changed, when the Communal Lexicon is out of step with the conceptual common ground of the expert community. Changes in the conceptual and terminological configuration of a discipline might be traced and be better understood by the dynamical updates reflected in these databases. Conceptual reformulation is a key process in all disciplines. At first sight, MIDs might look like the database implementations of common ground that the Introduction to this volume describes as frequent in Computer Science. They are a record of sorts of an extended written conversation between peers, allowing lexical and cognitive alignment in the long run. But in Artificial Intelligence this view of common ground as a list of shared elements has been shown to be neither theoretically fruitful nor practical, especially in Natural Language dialogue managers. As Allen et al. (2001: 29) state, “the tasks are too complicated to represent as a series of parameterized contexts. In fact, these tasks require the system to maintain an explicit model of the tasks and/or world and reason about these models. The language and the dialogues become significantly more complicated, and one also needs to start explicitly modeling the collaborative problem solving process that

Common Ground for Knowledge Through Knowledge of Language 109 the system and user engage in.” Being able to reason using these elements involves much more than a monotonic aggregation of information. A rich and fine-grained model of common ground is not a luxury, either for Linguistics proper nor for Human Language technologies. We have explored metalinguistic information that is being put forward in text because it cannot be assumed to be part of the collective expert-domain competence. Our research has focused on the linguistic mechanisms with which theoretical knowledge is created, modified and negotiated within an expert community, and on how that knowledge is enacted through a sublanguage and a terminology. The mechanics of consensus within an expert community is a privileged keyhole for seeing how common ground, any common ground, can be built up cooperatively. Our analyses and their application in the MOP system can prove to be useful tools to do empirical research into the nature of expert knowledge, as is attested in the interaction of scientists and scholars that struggle to communicate theoretical explanations and descriptions to their colleagues. The creation and maintenance of a theoretical common ground for scientific discussion is a complex, interactive process that requires constant intervention by the participants to repair miscommunication and to structure their knowledge linguistically. These vital discourse operations are performed in metalinguistic speech acts with very specific characteristics. The role that lexical markers of the metastable common ground of Science play in the constitution and evolution of technical knowledge cannot be overstated.

REFERENCES Allen, James, Byron, Donna, Dzikovska, Myroslava, and Ferguson, George, et al., 2001. Toward Conversational Human-computer Interaction. Artificial Intelligence Magazine 22 (4), 27–37. Austin, John L., 1962. How to Do Things with Words. Oxford University Press, Oxford. Bach, Kent and Harnish, Robert, 1979. Linguistic Communication and Speech Acts. MIT Press, Cambridge, MA. de Beaugrande, Robert and Dressler, Wolfgang, 1981. Introduction to Text Linguistics. Longman, London. Bierwisch Manfred and Kiefer, Ferenc, 1969. Remarks on Definitions in General Language, Studies in Syntax and Semantics. D. Riedel Humanities Press, Dordrecht, 55–79. Carnap, Robert, 1934. The Logical Syntax of Language. Routledge & Kegan, London (1964). Clark, Herbert H., 1998. Communal lexicons. In: Malmkjoer, K. and Williams, J. (eds), Context in Language Learning and Language Understanding. Cambridge University Press, Cambridge, 63–87. Clark, Herbert H., 1996. Using Language. Cambridge University Press, Cambridge. Clark, Herbert H. and Wilkes-Gibbs, Deanna, 1986. Referring as a collaborative process. Cognition, 22, 1–39. Fetzer, Anita, 2004. Recontextualizing Context: Grammaticality Meets Appropriateness, John Benjamins, Amsterdam. Fetzer, Anita, this volume. Reformulation and common grounds. Fowler, Roger, Hodge, Robert, Kress, Gunter, and Trew, Tony, 1979. Language and Control. Routledge & Kegan Paul, London. Garrod, S. and Anderson, A., 1987. Saying what you mean in dialogue: A study in conceptual and semantic coordination. Cognition 27, 181–218.

110 Grounding and Common Ground Gentilhomme, Y. 1994. L’éclatement du signifié dans les discours techno-scientifiques. Cahiers de Lexicologie LXIV-1, 5–53. Grice, Herbert, 1957. Studies in the Way of Words. Harvard University Press, Cambridge. Gross, Alan, 1991. The Rhetoric of Science. Harvard University Press, Cambridge. Guelich, Elisabeth and Kotschi, Thomas, 1995. Discourse production in oral communication. Aspects of Oral Communication. W. de Gruyter, Berlin, New York, 30–66. Halliday, Michael A. K. and Martin, James R. 1993. Writing Science: Literacy and Discourse Power. The Falmer Press, London. Harris, Zelig, 1991. A Theory of Language and Information. Clarendon Press, Oxford. Hunston, Susan, 1996. Evaluation and Ideology in Scientific Writing, and Register Analysis, Theory and Practice. Printer Publishers, London. Kotschi, Thomas, 1986. Procedes d’evaluation et de commentaire metadiscursif comme estrategies interactives. Cahiers de Lange Francaise 7, 231–258. Lara, Luis. F., 1989. Une critique du concept de metalangage. Folia Linguistica Tomus Mouton deGruyter XXIII (3–4). Lascarides, Alex and Copestake, Ann, 1995. The Pragmatics of Word Meaning. Proceedings of the AAAI Spring Symposium Series: Representation and Acquisition of Lexical Knowledge: Polysemy, Ambiguity and Generativity, Stanford, CA, 75–80. Lewis, David, 1969. Convention: A Philosophical Study. Harvard University Press, Cambridge. Lyons, John, 1977. Semantics. Cambridge University Press, Cambridge. Liddicoat, Anthony, 1997. The function of the conditional in French scientific writing. Linguistics 35 (4), 767–780. Meyer, Ingrid, 2001. Extracting knowledge-rich contexts for terminography. In: Bourigault, D., Jacquemin, Ch. and L’Homme, M.-C. (eds), Recent Advances in Computational Terminology. John Benjamins, Amsterdam, xviii, 279–302. Myers, Greg, 1992. “In this paper we report    ”: Speechs acts and scientific facts. Journal of Pragmatics 17 (4), 295–313. Nyan, Thanh, 1996. Metalinguistic Operators: Why they matter to linguistic theory, Le discourse: cohérence et connexion. Etudes Romanes 35, Museum Tusculanum Press, Copenhagen. Nyan, Thanh, this volume. Common ground, categorization, and decision-making. Ogden, Charles K. and Richards, Ivor A., 1929. The Meaning of Meaning. Routledge & Kegan Paul LTD, London (1969). Pearson, Jennifer. 1998. Terms in Context. Studies in Corpus Linguistics. Vol 1. John Benjamins, Amsterdam. Pickering, Martin and Garrod, Simon, 2004. Toward a mechanistic psychology of dialogue. Behavioral and Brain Sciences 27 (2), 169–190. Putnam, Hillary, 1975. The Meaning of “Meaning”. Mind, Language and Reality, Vol. II. Oxford University Press, Oxford. Quine, Willard V. O., 1960. Word and Object. MIT Press, Cambridge, MA (1992). Rey-Debove, Jossete, 1978. Le Métalangage. Le Robert, Paris. Riegel, Martin, 1987. Definition directe et indirecte dans le langage ordinaire: les énoncés définitoires copulatifs. Langue Francaise 73, 29–53. Rodríguez Penagos, Carlos, 2004. Mining Metalinguistic Activity in Corpora to Create Lexical Resources using Information Extraction Techniques: The MOP System. Proceedings from

Common Ground for Knowledge Through Knowledge of Language 111 the Association for Computational Linguistics. Barcelona, 2004. MIT Press, Cambridge, MA, 215–222. Rodríguez Penagos, Carlos, 1999. Explicit metalinguistic operations in specialized discourse: The construction of lexical meaning in theoretic science. In P. Sandrini (ed.), Terminology and Knowledge Engineering TKE’99, Innsbruck, Austria. TermNet Publications, Vienna, 150–158. Searle, John, 1969. Speech Acts. An Essay in the Philosophy of Language. Cambridge University Press, Cambridge. Sperber, D. and Wilson, D., 1986. Relevance: Communication and Cognition. Harvard University Press. Thoiron, Philippe, Béjoint, Henry, 1991. La place des reformulations dans les textes scientifiques. Meta XXXVI (1), 101–109. Thomason, Richmond H., 1992. Accommodation, meaning, and implicature: Interdisciplinary foundations for pragmatics. In: Cohen, P.R., et al. (eds), Intentions in Communication. MIT Press, Cambridge, MA, 325–363.

This page intentionally left blank

Lexical Markers of Common Grounds Edited by Anita Fetzer and Kerstin Fischer © 2007 Elsevier Ltd. All rights reserved.

IMPOSING COMMON GROUND BY USING TEMPORAL CONNECTIVES: THE PRAGMATICS OF BEFORE AND AFTER Thora Tenbrink

1. INTRODUCTION The semantics of before and after have long been recognized as interesting to linguists as well as cognitive scientists because of the relation to the conceptualization of time. Temporal clauses involving these connectives serve to give a time reference for the main clause, that is, the event described in the main clause is anchored in time in relation to the event described in the subordinate clause. Consequently, the event in the subordinate clause is “taken for granted” to some degree, as it is capable of serving as an anchor for other events (e.g., Talmy, 2000). This observation motivates a pragmatic analysis of these connectives, addressing the circumstances in which they are employed by speakers in natural discourse. In this study, naturally occurring instances of before and after, collected from corpora of spoken dialogues, are analysed with respect to the question to what degree the speaker can assume that the listener takes the event in the temporal clause “for granted” – either because the information conveyed in the temporal clause is already known material from the previous discourse context, or because the speaker “imposes” knowledge of the event on the hearer, in order to be able to anchor the main clause’s event in time without necessarily initiating a grounding process. With this range of possibilities in mind, the folk concept of “taking for granted” is specified for the present analysis of temporal clauses with regard to its relation to the notions of presupposition, givenness, common ground, and other much debated concepts in the field of discourse pragmatics. The analysis addresses the speakers’ range of options in conveying their temporal “anchor” in a transparently coherent way in the discourse, their use of other linguistic devices together with the temporal connectives that serve to suggest common ground, such as definite articles, anaphora, prosodic (de-)emphasis, or pronouns, as well as the range of variety in listeners’ reactions.

114 Grounding and Common Ground

2. BASIC NOTIONS All instances of naturally occurring language are based on the speaker’s (or writer’s) assumptions about many things: for example, the world, the interlocutor (listener or reader), the present situation, the relation to the interlocutor, and so on. This fact is well established in the literature, but treated fairly differently across research directions: for example, Halliday’s (1985) systemic-functional framework is based on the assumption that linguistic utterances can be analysed as representing three kinds of metafunction reflecting three aspects of context (field, tenor, and mode). Clark (1996) emphasizes that the use of natural language is based on the interlocutors’ motivation to carry out joint activities which are part of larger ongoing projects, embedded in a complex cultural background. Formal approaches, on the other hand, have often restricted their analysis to the immediate textual context, and are only gradually attempting to formalize the intricate connections between the real world and the language that is used in it (various papers in this direction can be found, for instance, in Bras and Vieu, 2001). Corresponding to this diversity in treating what is called “pragmatic” aspects of language use, there is also diversity in the employment of relevant terminology. In the following, some concepts often encountered in the literature, and some problems associated with them, are outlined briefly.

2.1. Common ground Clark (1996) treats the notion of common ground as a technical term denoting that which the interactants of a conversation believe to be their shared knowledge basis. This belief needs to be based on some kind of justification, for which there is a great variety of possible sources, for example, shared common knowledge of a community (communal knowledge), joint personal experiences (personal knowledge), or the ongoing discourse itself (the accumulation of common ground in the course of an interaction). Clark (1996: 92) argues that, on entering a conversation, we presuppose certain common ground, and with each joint action – each utterance, for example – we try to add to it. To do that, we need to keep track of our common ground as it accumulates increment by increment. Closely related is the notion of grounding: “To ground a thing (…) is to establish it as part of common ground well enough for current purposes” (Clark, 1996: 221). Interactants try to reach closure on their actions, to receive evidence that their action has been successful. Such evidence is provided, for example, by the listeners’ reactions (acknowledgements, denials, etc.) to an utterance (see Fischer, this volume, for more details). Formal approaches attempt to capture the interactants’ common ground by including relevant items into the formal discourse representation, a model that is gradually updated with each utterance in the discourse (e.g., DRT (Discourse Representation Therapy): Kamp and Reyle, 1993, and its extension SDRT (Segmented Discourse Representation Theory): Asher and Lascarides, 2003). At present, however, it is clear that only a very small portion of the shared background between the discourse participants can be captured in this way;

Imposing Common Ground by Using Temporal Connectives 115 therefore, formal approaches tend to restrict themselves to those elements of common ground which are needed to be able to represent the aspects that are currently focused on (see, for instance, Krause et al., 2001, for a DRT analysis of a spatial localization dialogue). The Hallidayan approach, in contrast, does not make use of the term common ground at all, focusing instead on the linguistic phenomena that reflect the relationships to various kinds of context, such as phoricity, discourse coherence as established by various linguistic devices, or the Given/New distinction (see below). Reference to elements (assumed or presented as) known from the textual context or other resources is indicated by lexical choices and the treatment of those elements as Given. Crucially, in contrast to other approaches Halliday and Matthiessen (1999: 2) point out that linguistic devices reflect speakers’ meaning rather than knowledge. Thus, they treat “information” as meaning rather than as knowledge and interpret language as a semiotic system, and more specifically as a social semiotic, rather than as a system of the human mind. (…) Meaning is a social, intersubjective process. If experience is interpreted as meaning, its construal becomes an act of collaboration, sometimes of conflict, and always of negotiation. In their approach, shared knowledge is therefore less important than shared meaning. The difference has important consequences with regard to the Given/New distinction (see below).

2.2. Foregrounding/backgrounding Linguistic elements can be either foregrounded or backgrounded in the overall discourse or in a single clause. Roughly, this notion captures the intuition that some elements are presented as more important than others. Thus, linguistic means are employed to achieve prominence in discourse: main clauses are generally considered to be foregrounded, while subordinate clauses are backgrounded; additionally, there are specific linguistic devices enabling emphasis or de-emphasis of parts of an utterance, such as left-dislocation or omission. In contrast to some other notions dealt with in this paper, the foregrounding/backgrounding distinction is generally viewed as a relative concept: some elements are relatively more foregrounded than others in a discourse, while they can at the same time be relatively more backgrounded than yet other elements. Halliday and Matthiessen (1999: 239) point out: If the Theme of the clause (realized as the element in first position) is also explicitly located as Given, this has a very strong “backgrounding” effect: the message is “you already know this; now use it as stepping-off point for a further move, to something you don’t know”. By the same token, the remainder of the clause (either the whole of it, or at least the culminating element) is strongly “foregrounded”. The total construction is obviously a powerful device for reasoning and argumentation. Thus, backgrounding is achieved by packaging pieces of information in a linguistic form that signals to the listener that the information is not to be focused upon.

116 Grounding and Common Ground

2.3. Presupposition In a broad sense, presupposition is ubiquitous in language: in producing language, something is always “supposed beforehand”; it is impossible to use language without assuming some kind of background (see Carston, 2002: 28ff. for an argument against the existence of “eternal” – background-less – utterances). In a narrower sense, it has been stated that presuppositions need to be satisfied by their local context. In this view, specific presupposition triggers serve to activate specific expectations about the previous discourse context, as is immediately obvious with regard to expressions such as too, which “marks that the reported event or state is not the only one of its kind reported or assumed in the context” (Zeevat, 2002: 64). Furthermore, explicit markers of common ground such as ok and German doch presuppose some aspects ˇ of the discourse context in order to fulfill their function, as spelled out by Condon and Cech (this volume) and Pittner (this volume). Two kinds of definitions are prevalent in the literature: In the semantic approach, presuppositions are taken to be those parts of the sentence that remain true when the main proposition is negated (negation test; cf. Strawson, 1950). In the pragmatic approach, “presuppositions are what is taken by the speaker to be the common ground of the participants in the conversation” (Stalnaker, 1978: 320). Correspondingly, Clark (1996) takes presuppositions as evidence of common ground, that is, presupposed elements are elements that one speaker assumes to be already grounded. Both views overlap to such a high degree that many authors do not explicitly state which one they adopt, especially since, by now, knowledge about the notion of presupposition is often presupposed by authors in the field. But they have clearly differing consequences with regard to the question whether discourse-new information may be introduced in a clause that is semantically presuppositional (i.e., cannot be negated directly). It may seem that the negation test simply captures on a logical basis which elements speakers assume to be unquestionably true since they are regarded as common ground. However, as Nemo (this volume) points out, there is a potential discrepancy between what is supposed to be shared knowledge and the specific assumptions that utterances make in relation to a given discourse context. A direct consequence of the pragmatic view seems to be that presuppositions should not be new to the discourse, since elements that already belong to the common ground cannot be new. But there is nothing in the negation test that precludes that elements remaining true by negating the main proposition are discourse-new. Thus, a sentence like1 1. She kissed them both before she brushed their hair. in which a sequence of actions that were not mentioned before is presented, contains a semantically presupposed clause that is undoubtedly discourse-new. The previously unmentioned event in the temporal clause remains true when the sentence as a whole is negated. Thus, a consequence of the pragmatic view is that the propositions of clauses introduced by before or after are not necessarily presupposed, while in the semantic view, they are (Heinämäki, 1974; Schilder, 2001).

1

John le Carré, The Russia House.

Imposing Common Ground by Using Temporal Connectives 117 Since a great amount of paper has already been dedicated, mainly by formal theoreticians, to these (and other) approaches to presupposition, I will not pursue this issue further. For present purposes, it is merely important to distinguish clearly between elements assumed to be common ground, discourse-new elements, and elements that remain true when the main proposition is negated. It is precisely this distinction that opens up an opportunity for the speaker to “impose” common ground on the listener by presenting facts in a linguistic construction that is hard to challenge, even if the proposition expressed in it is discourse-new. This phenomenon is also known for other linguistic constructions, as mentioned, for instance, in Biber et al. (1998). They show that, in their example – 2. The fact that in many insect groups the newly emerged adults show a slow gliding type of flight linked to dispersal suggests, however, that the earliest winged insects may have evolved in temporary habitats of small, erect plants. – discourse-new facts are introduced as established facts through the use of a clause starting with the fact that. This way of presenting information exemplifies how linguistic constructions do not merely convey knowledge, but they convey it in a way that reflects the speaker’s stance with regard to the information presented. Moreover, it enables the listener to infer specific assumptions even in the absence of shared common ground prior to the utterance (Nemo, this volume).

2.4. Given/New Among the most controversial distinctions with regard to information conveyance is that between Given and New. In her discussion of approaches up to that date, Prince (1981) notes that Given has variously been taken to mean recoverable, salient (to the hearer), and known (by the hearer). She then redefines the latter one as assumed familiarity and proposes a taxonomy of possible values this can take, based on the status of the discourse entities with respect to the previous and situational context. In addition to the distinctions introduced by Prince, a further distinction can be made between the degree of assumed familiarity, and speakers’ linguistic strategies to communicate elements as New or Given irrespective of the hearer’s assumed state of knowledge. Halliday and Matthiessen (1999: 529) characterize the New part of an utterance as that which construes a piece of information (…) as something having news value – something the listener is being invited to attend to. It may not contain anything the listener has not heard before; a great deal of “news” is totally familiar, being simply contrasted or even reiterated. On the other hand, the entire message may consist of unknown information, for example the first clause of a piece of fiction. But the message is construed along prototypical lines as an equilibrium of the given and the new, with a climax in the form of a focal point of information: “this is to be the focus of your attention”. In this view, New does not in the first place denote discourse-new knowledge (or information assumed not to be familiar) which the speaker wishes to convey, but rather the meaning that

118 Grounding and Common Ground he/she wishes to focus upon. Obviously, these notions overlap to a high degree, the unmarked case being that discourse-new elements are also communicated as “news” to the recipient, that is, linguistically marked and focused upon as New. The distinction between these notions is also blurred through the fact that elements can be known to speaker and hearer not only on the basis of preceding text but also on the basis of world or situation knowledge, or the immediate surroundings. In these cases, the corresponding elements are linguistically marked as Given, as indicated by linguistic features such as definite articles, anaphora, and deictic elements. The employment of such features suggests that the speaker either assumes the elements to be recoverable or inferable or that he/she simply wishes to present them as such. Further complications might be caused by the fact that discourse-new elements can be hearer-old (known to the hearer from other resources not related to the discourse context) – something which the speaker might or might not assume to be the case. Such phenomena arise especially when speakers are acquainted and share some amount of personal knowledge (cf. Clark, 1996). In addition to the linguistic features described so far, which are sometimes termed local newness markings, the clause structure also reflects a Given/New pattern (global marking). In Halliday’s (1985) terms, the Given/New structure is reflected prosodically in spoken language, with each clause possessing its own Given/New structure in terms of prominent and nonprominent elements. The unmarked structure is that the Given part is followed by the New. In these cases the Given part corresponds to the clause’s Theme, representing what the speaker is talking about (something which through longer stretches of the discourse signifies thematic organization). But it is also possible that clauses start with the New part; this is realized by prosodic prominence and, in written language, by marked syntactic structure. Goffman (1983) points out how prosodic structure helps organizing the discourse for the listener by highlighting more important or contrasted elements in relation to the context, emphasizing the contribution of prosody to fulfilling the felicity conditions of presupposed elements. With these notions in place, it becomes clear in which way the semantic and the pragmatic view of presuppositions are closely interconnected. Consider a classical example: 3. The king of France is bald. Here, it is presupposed that there exists a king of France, as shown by the negation test: 4. It is not true that the king of France is bald. (i.e., The king of France is not bald.) where there still exists a king of France even though the main proposition is negated. The sentence has an unmarked information structure: it begins with a Given part, locally marked by the usage of a definite article (the king) and the exophoric reference to France. The New part of the sentence is the claim of baldness, which can be negated. An unmarked prosodic contour would place emphasis on the latter part of the sentence (Halliday, 1985); otherwise, the sentence would be marked prosodically and accordingly receive a marked information structure. Thus, linguistic elements that are clearly Given often remain true in negated sentences, while New elements – in the unmarked case – fall under the scope of negation. In subordinated clauses, the case is different, since subordinated clauses have an information structure of their own. Furthermore, since subordinate clauses are backgrounded, they are clearly not supposed

Imposing Common Ground by Using Temporal Connectives 119 to be in the focus of attention. Presuppositional elements can therefore reflect what speakers wish to present as established even if they do not assume these elements to belong to the common ground. In this way, a direct grounding process (Clark, 1996: 221ff.) may be evaded by introducing new elements as though they were already shared knowledge. Some temporal subordinate clauses clearly belong in this category, while others (e.g., example 1) resemble other kinds of subordinate clauses which are not necessarily presupposed, such as “I think that you are right” (where both clauses seem to be have equal news status).

2.5. Conclusion As this short overview of basic notions shows, a fairly broad range of related issues play a role in the analysis of linguistic elements that point to the “common ground” of discourse participants. In the analysis of natural discourse, it is obviously hard to specify the exact common ground of the interactants, and speakers’ reasons for presenting specific linguistic elements as either Given or New are equally hard to determine. The next section addresses various aspects of speakers’ and listeners’ strategies in actual discourse with regard to the presentation of discourse-new information in presupposed (or otherwise backgrounded) clauses.

3. SPEAKERS’ STRATEGIES AND LISTENERS’ REACTIONS IN ACTUAL DISCOURSE Presuppositional elements that do not already belong to the common ground need to be accommodated, that is, integrated into the listener’s discourse model (or knowledge base) by using appropriate inference processes. In most cases, they then automatically belong to the common ground. But, as Karttunen (1974) points out, this is not necessarily the case:2 If the current conversational context does not suffice, the listener is entitled and expected to extend it as required. He must determine for himself what context he is supposed to be in on the basis of what was said and, if he is willing to go along with it, make the same tacit extension that his interlocuter appears to have made. This is one way in which we communicate indirectly, convey matters without discussing them. Thus, while it seems to be the standard case that listeners tacitly accept the information speakers present in presupposed clauses, alternatives to straightforward accommodation are possible. Listeners can explicitly reject the presupposition, resulting in a cancellation of the presupposition (this possibility is captured in formal approaches by distinguishing between local and global accommodation), or they can ignore it either consciously or without being aware that a presupposition has been implied. Then, the speaker will believe something to belong to the common ground that, in fact, is not represented in the listener’s current representation of the discourse. Such misrepresentations may cause communication problems;

2

My emphasis.

120 Grounding and Common Ground they may be detected later in the discourse, or they may have no effect at all, being irrelevant. The accommodation process itself can happen either consciously or without awareness. These possibilities can result in some facets of speakers’ and listeners’ linguistic behavior that may be regarded as problematic in some respects. For instance, speakers may choose to formulate their utterances such that (in their opinion) irrelevant or unimportant new information is not presented as new, but presupposed, hoping that this will not cause problems. They may also consciously impose some personal opinions on the listener by presenting them as presupposed, in order to avoid discussing potentially controversial topics. In this way, a grounding process (Clark, 1996) may be evaded. Listeners may choose to ignore such “imposed” elements, considering them irrelevant, and focus on the main topic of the discourse. Additionally, the lack of grounding may lead to misunderstandings if listeners’ inferences do not correspond to speakers’ intended meaning, as demonstrated by Okada (this volume). If listeners choose to challenge a backgrounded proposition, this cannot be done by a simple no (as in 6); instead, some effort is needed to refer back to presupposed elements (7): 5. Will you have some more wine before you leave? 6. No (thanks, I will not have any more wine). 7. Oh (you are mistaken), I do not intend to leave for quite some time. Since it is easier (and smoother in the course of an interaction) to follow the “main thread” of foregrounded elements in a discourse than to cancel a presupposition, it is possible that presupposed elements are canceled less often than could otherwise be expected. This has two kinds of consequences for the interaction between speaker and listener: on the one hand, presupposed elements constitute a tool for the speaker which may or may not be used for subtle manipulation of the listener, depending on the listener’s reaction. On the other hand, information that is new to the listener may receive too little attention. This is what is captured by the German saying nur in einem Nebensatz erwähnt (only mentioned in a subordinate clause), which might be interpreted as follows: things that should have been in focus, that is, foregrounded in the discourse and expressed forcefully by main clauses, only appear in a backgrounded construction and are therefore not considered intensively enough by the listener to be able to realize the impact of the information. For example, suppose that a student has missed a session in which the teacher has told the audience about a forthcoming exam. In the next session, she hears the following utterance: 8. You need to hand in some documents before you take the exam. In order to realize that the information in the temporal clause is both new and important for her personally – but not for the rest of the audience – the student needs to be specifically attentive and conscious of the fact that she has missed a session. Otherwise, it is conceivable that she considers the above utterance as irrelevant for herself because she has never heard of any exam and therefore does not see why she should need to hand in any documents. These phenomena are ubiquitous in language and concern not only presupposed propositions but – to various degrees – also other kinds of backgrounded elements, for instance, other kinds of subordinated clauses, lack of stress in intonation, or nominalizations. The present paper, however, touches only on a section of this field, namely, how speakers and listeners use

Imposing Common Ground by Using Temporal Connectives 121 the temporal connectives under analysis to foreground or background propositions, with regard to how they agree on common ground, cancel or accommodate presupposed information that is new to the listener,and so on. Such questions need to be addressed empirically. The next section sets the stage by outlining previous work with regard to the discourse connectives before and after.

4. BEFORE AND AFTER Previous work focusing on the usage of before and after in natural discourse has established several features of these temporal expressions with regard to their semantics and application. Most obviously, they are used to denote temporal relationships between events or states (Heinämäki, 1974; Herweg, 1990; Sánchez Valencia et al., 1994, and others). Lascarides and Oberlander (1993) point out that temporal connectives trigger presuppositions that cannot be accommodated in all contexts, and Schilder (2001) works out in some detail how the discourse relations between the previous discourse, the main clause, and the temporal clause can account for the corresponding accommodation or binding processes. This approach was carried further using natural language data in Tenbrink and Schilder (2003), resulting in a framework specifying the general insight that specific discourse relations are required in the application of before and after in natural discourse. The present analysis starts from this framework. The starting point of the analysis in Tenbrink and Schilder (2003) is the hypothesis that speakers do not “out of the blue” represent events as being in specific temporal relations to each other, but do so for a specific reason, a point already emphasized by Heinämäki (1974). One such reason is that they are in a proximal temporal relationship to each other (cf. Herweg, 1991). This “default case” can only be canceled by explicitly stating the length of time between the events or states. Then, the reason for mentioning the temporal order is precisely to make a statement about the length of a time span. But the notion of proximity per se is not enough to explain why speakers choose the events they do in their linguistic representations. In addition, it has been recognized widely that the juxtaposition of events by temporal markers often involves a conceptualization of an intricate relationship between the events that is not purely temporal. Most uncontroversially, a causal relationship can often be identified. Our own qualitative analysis of naturally occurring instances of the temporal terms before, after, and then showed that these terms are used preferably in situations in which one of several specifiable semantic relationships between the events or states described in the connected elements (usually main and temporal clause)3 can be inferred. One such relationship is described by the notion of Immediacy, where one event happens “next” on a specific level of granularity. Where this is not the case, the nature of the relationship between the two elements connected by before or after can be described via discourse relations: as our data showed, using SDRT categorizations (Lascarides and Asher, 1993) either a causal relationship (Explanation or Result) or Narration is possible, the latter indicating that the discourse is

3

In Tenbrink and Schilder (2003) no distinction was made between temporal markers as conjunctions versus as prepositions. I here adopt the term temporal component for the clause or prepositional phrase headed by before or after, leaving the analysis of possible fine-grained differences for later research.

122 Grounding and Common Ground carried further and the events or states do not overlap.4 The broad concept of Narration is further specified by three subconcepts, namely Insertion, Regulation, and Dependency. The following categories of usage have been identified: • Providing an Explicit length of time. • Immediacy: Where events are described as occurring directly after one another, this can be traced linguistically on either the agent, the patient, or the narrative level, as in “John entered the house. Then, he went into the kitchen. After that, he went upstairs.” where the agent’s actions are described as following one after the other, indicated by the lexical repetition of “he”.5 • Causality: One event causes the other, as in “He fell down after he stumbled over a stone”. This concept is typically expressed using after; where before is used, causal connections do not seem to be as easily inferable: “He stumbled over a stone before he fell down” seems to imply that the events are not causally connected. • Dependency: Before and after can both be used to express the fact that the later event is based (non-causally) on the former. Then, the former event is a precondition for the latter event to occur, as in “He entered after opening the door.” • Insertion: New information about an event that happens earlier than another expected event is provided. The latter event may in some cases not occur at all, being precluded by the inserted event, as in “He died before he could apologize.” This effect is wellknown in the literature as the counterfactual effect of before. In other cases, based on the insertion of the new event it is indeterminate whether the expected event happened after all. Thirdly, it may be presupposed that the expected event occurred. After cannot express the concept of Insertion. • Regulation: Before and after can be used to point to the order of known events. This is often the case when scheduling future actions (as in: “The excursion will take place before the movie”), and also in other cases where the order of two events is considered newsworthy. This is supported by prosodic prominence on the temporal term and by linguistic indicators such as already or even. As this overview shows, some of the identified categories of usage necessitate a specific kind of information status, while others do not: for instance, if Regulation is inferred then both connected elements should be Given, as the most important information concerns not the connected events but the order itself; and for an interpretation in terms of Insertion the event described in the temporal component must be Given. But for Immediacy, Causality, and Dependency, no such requirements have been identified. This variability will be specifically relevant for the present analysis. A further distinction needs to be made between the two kinds of ordering that can occur with the temporal terms. As described above, in the standard case a sentence departs from what is Given in the discourse context. Thus, the ordering of the elements of a sentence has implications with regard to the information flow. In Schilder and Tenbrink (2001) we show that sentences beginning with a temporal connective differ from those in which the

4 5

Immediacy can also be viewed as a semantically weak kind of Narration. Mostly, this concept is identified with then rather than after or before, but the latter two also support it in some cases.

Imposing Common Ground by Using Temporal Connectives 123 connective appears only later in that these sentences comprise two kinds of discourse relations instead of only one: preposed elements need to be connected in specific ways to the previous context, while postposed constructions have a weaker, and more unrestricted, relationship to the context. Based on this previous work, it seems that not all components introduced by before or after can be analysed as common ground for both speaker and hearer. Instead, intricate relationships between the temporal component, the main clause, the previous context, and other sources of knowledge account for the employment of a temporal term. Since temporal components serve as an anchor for specifying the time of a further event, the question needs to be addressed as to which discoursal requirements can be identified that account for the employment of such an anchor even if the event or state described in the temporal component is discourse-new. A corpus investigation was carried out to deal with this issue. It is presented next.

5. CORPUS INVESTIGATION In this study, instances of natural language in spoken discourse were analysed. They were taken from the Switchboard corpus distributed by the Linguistic Data Consortium (LDC),6 a large corpus of conversational speech over long-distance telephone lines which involves numerous speakers representing a great variety of sociological and dialectal backgrounds. To collect the linguistic data, speakers (paid volunteers recruited from different companies and agencies) chose a topic about which they wished to talk, out of a repertory of 70 topics. The participants are considered by the collectors to be broadly representative of adult speakers of American English between 20 and 60 years of age, though (according to the accompanying information on data collection) a bias toward higher socioeconomic and educational levels was considered inevitable due to the task requirements. The LDC provides online access to transcriptions (without prosodic information) as well as to the original sound files of all utterances. Thus, the present analysis could be carried out by first searching for occurrences of before and after in the transcriptions and then listening to the original utterances in order to specify the prosodic organization. This corpus is advantageous for present purposes since the speakers communicating with each other are not acquainted, and it can therefore be safely assumed that they do not draw upon information from previous encounters. Thus, personal knowledge (Clark, 1996: 100ff.) which would otherwise be hard to control is ruled out. Furthermore, since the speakers are talking on the phone they cannot refer exophorically to entities that are perceptually accessible to both partners. The only drawback is that the speakers were allowed to “warm up” before they started the recording by pressing a button, resulting in a certain degree of previous knowledge about each other that is not available to the analyst. Additionally, of course, they share a great amount of knowledge about the world and the community they live in; it will be seen that such communal knowledge (Clark, 1996: 100ff.) is indeed used continuously as shared background. As the present analysis is intended to be qualitative rather than quantitative, no statistical methods are employed. Qualitative exploration is a necessary prerequisite for follow-up studies

6

http://www.ldc.upenn.edu; last visited: June 19, 2005.

124 Grounding and Common Ground aiming at statistical significance, which require hard criteria for coding semantic properties to enable inter-annotator agreement (Carletta et al., 1997). Previous research on encoding semantic properties has shown that this is not easy to achieve (Spenader, 2001). To identify regularities in the circumstances of presenting Given or discourse-old, New or discourse-new information in temporal components, naturally occurring instances of before and after in spoken dialogue were analysed with respect to the following issues: • Common ground (based on textual context, communal knowledge, etc.) • Information conveyed in the temporal component: discourse-new, known, or inferable based on the previous discourse context • Linguistic indicators of information structure including prosodic organization (Given/New distinction based on speakers’ meaning rather than knowledge) • Relationship between the two connected elements and the previous discourse (if relevant) • Listener’s reaction with regard to the information provided in the temporal component, be it Given, inferable, discourse-old or discourse-new, or (presented as) New. These notions were analysed for each of the categories identified in Tenbrink and Schilder (2003), working out regular patterns in speakers’ and listeners’ linguistic behavior.

5.1. Results The qualitative analysis of spontaneous occurrences of the connectives supports the hypothesis that the categories of usage of the temporal connectives involve different degrees of Givenness in a discourse. In the following, a few exemplary analyses of short dialogues are presented in some detail in order to illustrate how speakers employ the connectives in context. Further analyses can be accessed in the appendix in a more sketchy form, for further reference and illustration. However, the main body of dialogues that serves as the basis for the conclusions drawn below cannot be represented for reasons of space. In the following dialogue excerpt (interpreted as a case of Immediacy: one TV program is presented as following another) the information given in the temporal component is presented as New by clear prosodic prominence7 on Who, followed by a pause and elaborated further by a prepositional phrase: Corpus example 1 (Context: sitcoms) B: I wish they’d put those, that’s why I say, did you stay up late to watch this Red Dwarf. It came on after Doctor Who (pause 1 sec) A: Uh-huh. B: on Saturday nights, here at least. And it should, of course I li-really it was, I don’t think they should show it during prime time, but still 7

Items with obvious intonational prominence in the original utterance are set in CAPS to provide an impression of the utterances’ prosodic organization. A more fine-grained analysis would certainly reveal further differences in relative prominence that go beyond the scope of.

Imposing Common Ground by Using Temporal Connectives 125 The pause (along with a rising intonation) seems to indicate the speaker’s wish for confirmation of the new information, which is immediately given. By backchannelling, the listener informs the speaker that he has managed to accommodate the information. In spite of the speaker’s obvious aim of having the information safely in place, it is nevertheless not presented as belonging to the main thematic thread of the discourse. It is presented only in a backgrounded prepositional phrase introduced by after, notably not a full temporal clause. Thematically, the speaker is talking about Red Dwarf, and anchors this temporally. To be able to do this, she needs the listener to confirm that he is indeed familiar with this background information, that is, the interlocutors share common ground on the basis of communal knowledge. Corpus example 2 (here interpreted in terms of Causality) illustrates the special impact of preposing a temporal clause. Corpus example 2 (Context: painting) A: (…)8 and then you can just paint right over the dark. (…) After about, uh, half dozen coats it looks reasonable but it kept, the, the dark it was, uh, just a real dark, uh, wood grain type paneling – B: Uh-huh. Here, the temporal component provides discourse-new information, introducing new linguistic material (“about half dozen coats”) in a prepositional phrase, where both actor and process are omitted. This informationally sparse construction is sufficient for the introduction of new information for two reasons: First, prosody supports information transfer by the speaker’s hesitation (uh) before half dozen and by prominence on coats. Second, the information is inferable because its relationship to the previous discourse is transparent: As is common for preposed temporal components, there is an Elaboration relation to a part of the previous discourse. In this case, half dozen coats elaborate painting over the dark mentioned in speaker A’s previous utterance (cf. Schilder’s considerations of subpart relations in Schilder, 2001). Accordingly, the listener obviously has no problems accommodating the information. In corpus example 3, the expected event is precluded by the main clause, yielding a counterfactual (Insertion) reading of the before-clause (in which temporal precedence is less important than deontic precedence, i.e., desirability): Corpus example 3 (Context: extending school years) A: (…) I understand there’s a lot of talk now about extending the school year which seems inordinately short anyway, (…) well how do you think that would work in grammar schools and B.: I think, uh, in, in some sense, um, [lipsmack] what I would do before I would extend a school year, is I would make it better. Before I would continue the agony, (trailing off) A: Yeah, yeah, that’s right (laughter). 8

Unfortunately for reasons of space some stretches of text have to be omitted. They do not contain information relevant to the present analysis.

126 Grounding and Common Ground Here, the speaker apparently wishes to emphasize that the event referred to in the temporal clause should be precluded. This is indicated by the double construction, where, in the first version, the speaker simply repeats the wording of previous utterances to point to the expected event. The second version, in contrast, elaborates the first statement by a clearly negative assessment of the event (agony). Accordingly, the first version only carries some weak prominence on before and better, while the second version shows a fairly strong prominence on agony. Then, the speaker does not finish the sentence, since the main proposition (the inserted event) has already been given and does not need to be modified. The listener explicitly acknowledges the speaker’s evaluation. Schilder (2001) describes the capability of before to indicate a Termination relation between the main and the temporal clause; for some reason (possibly related to text type), no example for this category was found in the data analysed in Tenbrink and Schilder (2003). In the Switchboard data analysed here, a Termination relation was found to be fairly frequent, especially with preposed temporal components. Termination only occurs with before, and is indicated by the use of duratives (descriptions of states rather than events) in the main clause. The concept supported in this construction can be specified as follows: either • the event in the temporal component terminates the state of affairs as described in the main clause where the temporal component describes an event, or • the state described in the temporal component was different to the state of affairs as described in the main clause where the temporal component describes a state. To support this concept, it is not necessary that the event or state in the temporal component is known beforehand, as corpus example 4 illustrates. Corpus example 4 (Context: gardening) B.: Yeah. Before I moved down here, when I was up, uh, north in Wisconsin living with my folks, they lived on, uh, three and a half acres of land, and, uh, we started mowing that with a push mower. A.: Oh. The context of this piece of conversation indicates that the speakers know vaguely about each other where they live, although there was no mention of moving house. Thus, the information given in the temporal clause is discourse-new but can be accommodated easily. Prosodically, there is almost no prominence at all in the temporal clause, the main emphasis clearly being on land in the main clause. The speaker is talking very fast, giving the impression that he wishes to be finished with providing the necessary background information in order to be able to get to his main point, where he considerably slows down his speech. He is, however, obviously fully aware that the listener needs the information, as he gives a further evaluation of it in the paralleled when-clause. The listener does not react to this information directly, but backchannels on the later main proposition. The data contain frequent occurrences like the previous one where discourse-new information is presented as Given in a temporal component. Usually, such information is accommodated on the basis of communal knowledge: the events or states mentioned are simply not unusual, though the interlocutor could not know them to be the case prior to mention. Corpus

Imposing Common Ground by Using Temporal Connectives 127 example 5 (interpreted as a case of Regulation: the temporal order itself is focused upon, while the described situations are treated as Given) is similar in this respect. Corpus example 5 (Context: pets) B.: I had one cat for eighteen years. I got her when I lived in Hawaii. (…) Before I even got married I had my cat. A.: She’s old. From the previous context, the listener cannot know that the speaker is married. However, this information can be easily accommodated based on communal knowledge (many people get married). The information is not foregrounded even though it is discourse-new; the strongest prominence is on cat.9 The importance of the temporal information is indicated by even. The speaker’s intention in this context seems to be to convey the length of time that has passed since she got the cat. This is immediately supported by the listener’s subsequent remark. Thus, the speaker has successfully “imposed” knowledge on the listener that is not being questioned. However, a different reaction is possible, as in the following example: Corpus example 6 (Context: studying) A.: So, are, are you going to go for your doctorate? B.: Probably not. See I’m in the Navy. A.: Yeah. B.: And, uh, you know, unless the Navy funds it I probably won’t do it until I, if I do do it, it will be after I get out of the Navy. A.: When do you get out? B.: Oh, probably be another twelve years. A.: Oh, okay (laughter). Again, the information in the temporal clause is treated as Given – carrying only very little prosodic stress on after and Navy – even though it is discourse-new. On the basis of communal knowledge, the information that the speaker will get out of the Navy is inferable, though when that will be is not. Therefore, the listener asks for this information, which is then provided by the speaker. Interestingly, this information is indeed important for understanding why speaker B assumes that he will not do his doctorate, since the time span is very long.

5.2. Discussion The analysis of natural language data has shown that the information in the temporal component need not in all cases be known beforehand, in spite of its being classified (in some accounts in the literature) as presuppositional. Thus, before and after cannot be said to mark common ground. However, discourse-new information presented in temporal components is usually

9

Since this prosodic contour seems to be untypical, a graphic depiction of this utterance is provided in the Appendix.

128 Grounding and Common Ground easily accommodated on the grounds of communal knowledge, or because of its relation (e.g., an Elaboration discourse relation) to the previous discourse context, as is specifically often the case in preposed temporal components. In some cases, however, the speaker uses temporal components to refer to information that is only treated as Given although it is actually discourse-new, so that it can be viewed as being “imposed” on the hearer (see also Nemo, this volume). The hearer then has the choice to accept the new information explicitly or inexplicitly, to ignore it or to challenge it. Assessing the speaker’s expectations about the information status of the hearer cannot be achieved in an uncontroversial way. Linguistic devices only reflect the speaker’s presentation of elements as Given or New. Since the data source used for the analysis was particularly suitable for determining how much the interactants know about each other, it could be shown that – in contrast to some previous approaches that do not recognize the necessary distinction between New and discourse-new elements – the hearer’s information status (or the speaker’s assessment of it) is not straightforwardly mirrored by the use of linguistic expressions or prosodic prominence. Instead, some pieces of information are completely new in the discourse and still backgrounded linguistically and prosodically. But information that has been mentioned before can equally well receive prosodic emphasis and linguistic foregrounding. Thus, these devices may or may not be used to indicate the speaker’s expectations with regard to knowledge on the part of the hearer; rather, they seem to reflect the speaker’s intention to place specific emphasis on parts of the linguistic material (in accord with the Hallidayan analysis of Given/New structure). Emphasis can be placed on elements of the temporal component, but the thematic thread of the discourse is mostly carried by the main clause. The various kinds of concepts supported by before and after were found to exhibit the following features. I. Explicit length of time: Explicit statements about length of time do not require the information in the temporal component to be Given, but it should be inferable as it is not presented as important to the main proposition. The main point of the clause complex is either the time span itself or the proposition in the main clause which is supplemented by the information about the length of time. Listeners can challenge or evaluate the information in the temporal component but usually do not do so, following the main thread of the ongoing discourse. II. Immediacy: Immediate succession reflects the speaker’s conceptualization of events following one after the other on one level of granularity. The event in the temporal component may be known beforehand, inferable or discourse-new. If it is known, then the utterance anchors a “previous” or “next” event in time in relation to the known event. If it is inferable or discourse-new, then the speaker anchors an event he/she is talking about in time, establishing the reference time referred to as predecessor or successor “on the spot”. III. Causality: Where after is used to express a causal relation between two events (or states), the main emphasis is on the result rather than the cause, which is reflected via the fact that the cause is packaged into a backgrounded component. This might be one reason why before cannot be used to express this concept: in that case, the result would appear in a backgrounded component, while the cause would be foregrounded. While this structure is theoretically possible, it seems pragmatically unnatural. Most linguistic constructions expressing cause foreground the result rather than the cause,

Imposing Common Ground by Using Temporal Connectives 129

IV.

V.

VI.

VII.

as is the case in clauses introduced by because. Where causes appear in a main clause, the result is presented in a coordinated, rather than subordinated, clause, as with therefore, as a result,and so on. Thus, the cause introduced by after does not seem to receive the interactants’ attention. It is possible that discourse-new information is introduced as a cause, but the result is more important. In the data examined in the present analysis, discourse-new information given in the temporal component could be accommodated without problems and was not questioned by the listener. Notice that the causal relation itself is even less challengeable for the listener, since it is not expressed explicitly but needs to be inferred by the juxtaposition of two otherwise unrelated events. Accordingly, no cases were found in which the listener questioned a causal relationship implied by the employment of after. Dependency: Dependency relationships between the elements connected by temporal terms occur whenever a precondition between events can be identified. Since the precondition can be either of the events, either the precondition or the event that is based on it can be foregrounded. In either case, the backgrounded event may be known, discourse-new, or inferable; but since the speaker focuses on the foregrounded event, the listener naturally follows their train of thought without challenging inferred information. Insertion: In cases of Insertion, the event in the main clause is inserted between present and the event referred to in the temporal component, which is either expected based on the previous context, or else treated as expected, that is, linguistically presented as Given. Usually, if the information itself is not explicitly stated in the discourse history, it can be accommodated easily. The event in the main clause may prevent the event in the temporal component: in counterfactual instances, this is obviously the case, while in indeterminate cases, the expected event may or may not take place, this becomes unimportant because of the inserted event. Regulation: In Regulation, the temporal order of two expected or known events is focused upon. Usually, the information of both connected elements is discourse-old or easily inferable. In any case, it is linguistically treated as Given. If it is in fact new to the listener, he/she can either accommodate the presupposition or question it. Termination: Instances of Termination are indicated by duratives in the main clause; they support the notion that some kind of change has occurred at some point in time. The dependent component provides the temporal information relevant to the change. It can consist of discourse-new, old, or inferable information; since it is clearly backgrounded linguistically, a congruent listener’s reaction is not to react to it at all. In the data, no listener gave any comment on a before-component indicating Termination.

In accord with previous accounts regarding subordinate clauses, it is common to all concepts supported by before and after that the situation described by a temporal clause (or reduced construction) is backgrounded, that is, seldom focused upon in the discourse. This is mirrored by a nearly consistent lack of reaction on the part of the listener with regard to the information packaged in a temporal component, as well as a general tendency for the information to be inferable on the basis of the given context or communal knowledge. However, the degree to which New information can be conveyed in a backgrounded temporal component differs systematically, reflecting a continuum of inferability (or accessibility) in the discourse context.

130 Grounding and Common Ground On one end of the scale, the two connected elements are nearly equivalent with regard to the “newness” of information: this is the case in Immediacy, where in some cases events are listed as happening one after the other without there necessarily being any previous knowledge with regard to either event. Similarly, in cases of Explicit length of time, Termination, Causality, and Dependency the main emphasis lies on the semantic relation between the events or the expressed time span, but there is no need for the backgrounded event to be previously known to the hearer. Here, however, the focus is more clearly on the main clause. On the other end of the scale, in Insertion and Regulation the information in the temporal component is regularly presented as Given. The two categories differ in that in Regulation, also the information in the main clause is treated as Given. It seems, then, that the information in a temporal component should not in all cases be classified as presupposed, since at least in cases of Immediacy information that does not previously belong to the common ground can be conveyed straightforwardly in the temporal component, as in example 1 above (which – as indicated – is taken from a novel). However, such extremely balanced (with respect to newness of information in the two elements) examples turned out to be rare in the conversational examples analysed here. Overwhelmingly, people took care to present only easily inferable information even in cases of Immediacy. The following example (interpreted in terms of Immediacy as indicated lexically by right) illustrates this point: Corpus example 7 (Context: credit cards) B.: Yeah, we’ve tried to do that, we’ve paid ours off, you know, all the way down to where we had everything down to zero and especially right before I, I quit work two years ago, to stay home with the kids and right, and before I left we had everything paid off, we were in great shape, we were putting money, you know, because we were both working, A.: Uh-huh. The speaker starts introducing information in a temporal clause but stops in mid-utterance to start anew. It is clear from the prosodic curve of this reformulation that the speaker now starts a new clause in order to introduce new information (I receives prosodic stress after the restart), rather than continuing the temporal clause. The temporal clause is then formulated only later after having made sure that the listener knows what the speaker is talking about. This example supports Fetzer’s (this volume) account of reformulations as being highly relevant to the establishment of common ground, though from a different perspective. These observations lead to a reconsideration of the fact that the present analysis is restricted to one particular type of discourse, namely telephone conversation between strangers. As outlined above, this restriction has several invaluable advantages which have allowed for this kind of analysis in the first place. However, in terms of Halliday’s threefold distinction between field, tenor, and mode, it is obvious that only the field is varied at all throughout the corpus as a whole. Tenor is restricted via the fact that the interlocutors are strangers, so that not much variety of familiarity or social status and distance can be expected; the mode of conversation (talking on the phone) also remains the same. Field is varied through the relatively broad variety of topics; however, since the speakers were not supposed to invent a further topic, and naturally did not feel free to talk about something completely different while

Imposing Common Ground by Using Temporal Connectives 131 on the phone, there is some restriction here as well. A consequence of these observations may be that speakers in this particular setting do not have any reason for imposing common ground on their listeners in a way that could be considered unfair. There is no reason, for example, to suggest in a covert fashion that the other person should do something, as could be the case, for instance, in a question from host to guest such as the constructed example 5 in Section 3 above. Likewise, there is no reason to accuse the interlocutor of having done something (in fact, mostly the speakers talk about themselves rather than their interaction partner), as could be the case in a courtroom discourse in which a question like 9 might have been asked even before the accused has pleaded guilty at all: 9. What did you do after you robbed the bank? That imputations such as these do in fact occur has been claimed time and again by authors, sometimes assuming unfair behavior on the part of the speaker10 and sometimes claiming unawareness. The following example taken from a novel11 illustrates the phenomenon: 10. “Anything you want to know before you sign?” I am using my special voice after all, for self-protection. He is at firstpuzzled, then amused. “Why? Have you got more answers you want to tell me?” (…) [Not much later, he signs] In this short excerpt – which may be interpreted as a case of Insertion – the narrator suggests by using a presupposed temporal clause that his interlocutor will sign, although it is clear from the previous text that the man has not yet agreed to that. The common ground shared by both, however, is that it is expected of him to sign. In the narrator’s question, something is inserted between the moment of speaking and an expected event. The listener’s reaction is puzzlement: he does not accept the presupposition directly, nor does he question it. Furthermore, since the interactants in the Switchboard corpus probably have no further aims apart from conversing on the phone as they are expected to do – they do not wish to achieve anything, neither buy nor sell, and the like – many pragmatic phenomena that have been pointed out in the literature, such as those related to performative speech acts, do not occur at all. Thus, many possible linguistic phenomena are quite naturally ruled out by the Switchboard setting. It remains open for future analysis of language use in more natural settings to work out whether speakers indeed use linguistic devices to impose common ground on their interlocutor in a blunter way than has been identified so far.

6. CONCLUSION The present analysis of naturally occurring spoken discourse has shown that discourse-new information can unproblematically be presented in a backgrounded temporal component if it is easily inferable. However, the degree to which the information needs to be accessible 10

Dan Scorpio’s online essay on “Presupposition” (http://www.angelfire.com/nd/danscorpio/presup.html) deals in considerable detail with such “unfair” usage of presuppositions. 11 John le Carré, The Russia House.

132 Grounding and Common Ground in order to support a particular concept differs systematically between the different concepts supported by before and after. In particular, Insertion and Regulation require high accessibility of the information in the temporal component, while Immediacy is much less restricted. Thus, “imposing” common ground is a matter of degree, since it cannot be assumed that common ground is being imposed unfairly where the information is not required to be known beforehand. Insertion and Regulation are specifically good candidates for “imposing” common ground since they involve treating the events in the temporal component linguistically as Given to a higher degree than the other conceptual categories do. The interactants in the Switchboard telephone scenario did not seem to make unfair use of this possibility. Further research in other text types is necessary to identify the impact of variation in tenor and mode.

ACKNOWLEDGEMENTS I thank the participants of the workshop “Lexical Markers of Common Grounds” at the Pragmatics conference in Toronto 2003 for lively discussions, John Bateman and two anonymous reviewers for valuable comments on earlier versions of this paper, and Frank Schilder for earlier fruitful cooperation.

BIBLIOGRAPHY Asher, Nicholas and Lascarides, Alex, 2003. Logics of Conversation. Cambridge University Press, Cambridge. Avrahami, Judith and Kareev, Yaakov, 1994. The emergence of events. Cognition 53, 239–261. Biber, Douglas, Conrad, Susan, and Reppen, Randi, 1998. Corpus Linguistics: Investigating Language Structure and Use. Cambridge University Press, Cambridge. Bras, Myriam and Vieu, Laure (eds), 2001. Semantic and Pragmatic Issues in Discourse and Dialogue: Experimenting with Current Dynamic Theories. Elsevier, Amsterdam. Carletta, Jean, Isard, Amy, Isard, Stephen, Kowtko, Jacqueline C, Doherty-Sneddon, Gwyneth, and Anderson, Anne H., 1997. The reliability of a dialogue structure coding scheme. Computational Linguistics 23 (1), 13–31. Carston, Robyn, 2002. Thoughts and Utterances: The Pragmatics of Explicit Communication. Blackwell, Cambridge. Clark, Herbert H., 1996. Using Language. Cambridge University Press, Cambridge. ˇ Condon, Sherri L. and Cech, Claude, this volume. Ok, next one: Discourse markers of common ground. Fetzer, Anita, this volume. Reformulation and common grounds. Fischer, Kerstin, this volume. Grounding and common ground: Modal particles and their translation equivalents. Goffman, Erving, 1983. Felicity’s condition. American Journal of Sociology 89 (1), 1–53. Halliday, Michael A.K., 1985. An Introduction to Functional Grammar. Edward Arnold, London. Halliday, Michael A.K. and Matthiessen, Christian M.I.M., 1999. Construing Experience: A Language-Based Approach to Cognition. Continuum, London, New York.

Imposing Common Ground by Using Temporal Connectives 133 Heinämäki, Orvokki, 1974. Semantics of English Temporal Connectives. Dissertation, Department of Linguistics, University of Texas, Austin. Herweg, Michael, 1990. Zeitaspekte: Die Bedeutung von Tempus, Aspekt und temporalen Konjunktionen im Deutschen. Universitätsverlag, Wiesbaden. Herweg, Michael, 1991. Temporale Konjunktionen und Aspekt. Der sprachliche Ausdruck von Zeitrelationen zwischen Situationen. Kognitionswissenschaft 2, 51–90. Kamp, Hans and Reyle, Uwe, 1993. From Discourse to Logic: Introduction to Modeltheoretic Semantis of Natural Language, Formal Logic and Discourse Representation Theory. Kluwer, Dordrecht. Karttunen, Lauri, 1974. Presupposition and linguistic context. Theoretical Linguistics 1, 181–193. Krause, Peter, Reyle, Uwe, and Schiehlen, Michael, 2001. Spatial inferences in a localization dialogue. In: Bras, M. and Vieu, L. (eds), Semantic and Pragmatic Issues in Discourse and Dialogue: Experimenting with Current Dynamic Theories. Elsevier, Amsterdam. Lascarides, Alex and Asher, Nicholas, 1993. Temporal interpretation, discourse relations and common sense entailment. Linguistics and Philosophy 16, 437–493. Lascarides, Alex and Oberlander, Jon, 1993. Temporal coherence and defeasible knowledge. Theoretical Linguistics 19 (1), 1–35. Nemo, François, this volume. The pragmatics of common ground: From common knowledge to shared attention and social referencing. Okada, Moeko, this volume. Whose common ground? A misunderstanding caused by incorrect interpretations of the lexical markers of common ground. Pittner, Karin, this volume. Common ground in interaction: The functions of medial doch in German. Prince, Ellen F, 1981. Toward a taxonomy of given-new information. In: Cole, P. (ed.), Syntax and Semantics: Vol. 14. Radical Pragmatics. Academic Press, New York, 223–255. Sánchez Valencia, Víctor, van der Wouden, Ton, and Zwarts, Frans, 1994. Polarity, veridicality, and temporal connectives. In: Dekker, P. and Stokhof, M. (eds), Proceedings of the 9th Amsterdam Colloquium. ILLC-University of Amsterdam, Amsterdam, 587–606. Schilder, Frank, 2001. Presupposition triggered by temporal connectives. In: Bras, M. and Vieu, L. (eds), Semantic and Pragmatic Issues in Discourse and Dialogue: Experimenting with Current Dynamic Theories. Elsevier, Amsterdam, 85–108. Schilder, Frank and Tenbrink, Thora, 2001. Before and After: Sentence-internal and -external Discourse relations. Workshop from Sentence Processing to Discourse Interpretation: Crossing the Borders. Utrecht, The Netherlands, 2–3 July 2001. Spenader, Jennifer, 2001. Between binding and accommodation. In: Kühnlein, P., Rieser, H., and Zeevat, H. (eds), Proceedings of BI-Dialog 2001, June 14–16 at ZiF, Bielefeld, 162–173. Stalnaker, Robert C., 1978. Assertion. In: Cole, P. (ed.), Syntax and Semantics 9: Pragmatics. Academic Press, New York, 315–332. Strawson, Peter F. 1950. On referring. Mind 52, 320–344. Talmy, Leonard, 2000. Toward a Cognitive Semantics, Vol. 1. MIT Press, Cambridge, MA. Tenbrink, Thora and Schilder, Frank. 2003. (Non)temporal concepts conveyed by before, after, and then in dialogue. In: Kühnlein, P., Rieser, H. and Zeevat, H. (eds), Perspectives on Dialogue in the New Millennium. John Benjamins, Amsterdam and Philadelphia, 353–380.

134 Grounding and Common Ground Zeevat, Henk, 2002. Explaining presupposition triggers. In: van Deemter, K. and Kibble, R. (eds), Information Sharing: Reference and Presupposition in Language Generation and Interpretation. CSLI Publications, Stanford, 61–88.

APPENDIX This appendix presents further examples of dialogue analyses according to the criteria motivated in the body of the paper. Corpus example 8 (Context: death penalty) B: (…) I just saw on the news last night, that they said the average time a sentenced murderer, you know, is in jail is two years before he’s paroled, and a rapist’s is like six months, (…) A: That’s pathetic. (…5 turns…) and the way the law reads, uh, if they sentence you to life in prison, then he’s available for parole. If it’s life and a day, then he’s not eligible for parole. Category: Explicit length of time Common ground: communal knowledge (convicts can be paroled) Information: inferable: close semantic relation to previous context of legal procedures Given/New: strong prosodic prominence is on years, weaker stress on paroled: the information in the temporal clause is news to the hearer, but not as prominent as that in the main clause. Listener’s reaction: at first no comment on the temporal component; several turns later, evaluation (conceivably the backgrounded information took more time to react to) Corpus example 9 (Context: movies) A: (…) I must admit I did look at my watch after about an hour. B: Did you? Category: Explicit length of time Common ground: communal knowledge on time Information: discourse-new Given/New: presented as Given; no prosodic prominence on the temporal component Listener’s reaction: no comment Corpus example 10 (Context: owning houses) A: We just bought our first home (…) after being in school and so, we’re just first time homeowners. We’ve only been living here about six months.

Imposing Common Ground by Using Temporal Connectives 135 B: Okay, uh, well, I’ve lived in the, in the Dallas and Richardson area for about, uh, twenty-five and a half years, twenty-six years. Category: Immediacy (covering large time spans) Common ground: communal knowledge (people go to school) Information: discourse-new Given/New: presented as Given: prosodically backgrounded, Agent omitted Listener’s reaction: no comment on the temporal component Corpus example 11 (Context: pollution) A: (…) I don’t know if you ever saw that sixty minutes thing where they showed a chicken processing plant. (…) I couldn’t bear to think about eating a chicken after seeing that. And, but the government says it’s safe. Uh-huh. B: Uh, yeah. Category: Causality Common ground: discourse context Information: discourse-old, anaphorical construction Given/New: prominence on that, indicating the speaker’s intention to point to the importance of the event referred to (enough to serve as a reason for the speaker’s aversion to eating chicken) Listener’s reaction: no comment Corpus example 12 (Context: politics) A: The only, the only thing I see about Cuba though, is, uh, after Fidel Castro dies, I don’t think they’ll be a communist power anymore. I, I can’t see communism in that country carrying on past him. B: Well, when you take a, a situation where I think in particular in Salvador (…) Category: Causality Common ground: world knowledge (people die) Information: discourse-new Given/New: presented as Given, no prosodic prominence; the emphasis is on the result, not the cause Listener’s reaction: no comment on the temporal component Corpus example 13 (Context: Puerto Rico as a state of the US or independent) B: And I think, uh, having listened to you relative to the economy thing, I think if I were being forced to make a decision I would plead ignorance and wait to do more research before picking one of these. So I’m ul-, I guess I’m ultimately in favor of status quo also at this point, A: Well that’s interesting.

136 Grounding and Common Ground Category: Dependency (doing more research as a precondition for picking) Common ground: discourse context Information: inferable from the previous discourse (close semantic relation) Given/New: prosodic prominence on picking, thus emphasizing the point that picking is a major process that needs preparation Listener’s reaction: no comment on the temporal component Corpus example 14 (Context: camping) A: And we managed to, uh, fortunately the rain didn’t last long but we managed to get them all back up and, uh, got some sleep in that night. And after we got, uh, back from the weekend, we checked with one of the parents of one of the other college kids and they were a little concerned. They had heard on the radio that the winds on the dam there anyway, had gotten up to ninety-seven miles an hour, B: Good grief. Category: Dependency Common ground: returning from a holiday; inferable from the discourse context of camping Information: discourse-new; informationally sparse because it is not stated where back is supposed to be, nor was there any mention of a weekend in the previous discourse Given/New: presented as Given by a definite article; prosodic prominence on back, indicating that the most important part of the temporal clause is that we needed to come back in order to be able to check with one of the parents Listener’s reaction: no comment Corpus example 15 (Context: automatic phone calls advertising goods or services) A.: I’ve even had some of them, the, they’re voice activated and you’ve got to say hello twice before they’ll do anything. B.: Uh-huh, uh-huh. Category: Dependency (indicated linguistically by ’ve got to) Common ground: discourse context, communal knowledge Information: inferable (elliptical/anaphorical) Given/New: presented as New: prominence on anything Listener’s reaction: no comment Corpus example 16 (Context: Puerto Rico as a state of the US or independent) A: So, if they’ve just taken such action, it would seem to indicate to me either they’re doing it because they’re afraid they might become a state and want to declare this before they become a state, or maybe because they don’t want to become a state for fear of losing the Spanish, or Hispanic heritage. B: Uh-huh, well that’s interesting.

Imposing Common Ground by Using Temporal Connectives 137 Category: Regulation Common ground: discourse context Information: discourse-old Given/New: the information in both the main and the temporal clauses is Given: declare this points back to taken such action, while become a state is an exact repetition of the wording used before. The temporal term is prosodically prominent Listener’s reaction: no comment on the temporal component Corpus example 17 (Context: money/credit) B.: (…) and then these kids start charging and, and they get themselves into some deep trouble. (…) Before they even [laughter] start making money, they’re already in debt, you know. A.: No kidding. Category: Regulation supported by the linguistic indicators even and already that point to the peculiarity of the temporal order indicated by the speaker Common ground: communal knowledge Information: discourse-new Given/New: prosodic prominence on money; weaker prominence on before to further support the importance of the temporal ordering Listener’s reaction: no comment on the temporal component Corpus example 18 (Context: ice storms) B.: when you get, I mean, here they start even to forecast the weather, you know, it’s going to be an ice storm tomorrow, they’ll come, before the ten clock news they’ve already got schools calling in saying they’ve canceled. A.: Really? Category: Regulation (indicated by already) Common ground: communal knowledge on news Information: discourse-new Given/New: weak prominence on before, prominence on news: important information on something that happens even earlier than the (known) ten o’clock news (definite article) Listener’s reaction: no comment

Corpus example 19 (Context: cross-stitching) A: Uh-huh (throat_clearing). There’s also a couple of large projects, baby afghans, and I got one done and I think the other one’s gonna the child is gonna be in high school before I get – B: (Laughter) On that, uh –

138 Grounding and Common Ground Category: Indeterminate Insertion Common ground: discourse context Information: discourse-old Given/New: the expected event (probably, finishing the project) is so obvious that it is not even expressed by the speaker Listener’s reaction: this idea invokes laughter on the part of the hearer: the project will take much too long to make sense, that is, the inserted event renders the project meaningless

Corpus example 20 (Context: reading newspapers) A.: And I live in kind of a bad area where if I have the paper delivered it’sstolen before I can get out and get it (laughter). B.: Gosh that’s, that sounds pretty bad to me. I mean who would steal a newspaper? Category: Counterfactual Insertion Common ground: communal knowledge (papers get delivered outside) Information: inferable from the discourse context; anaphoric (it) and elliptical construction (get out) Given/New: presented as Given, weak prosodic prominence on the first get Listener’s reaction: no comment on the temporal component

Corpus example 21 (Context: credit cards) A.: Do you use credit cards a good bit? (…) B.: Before I got married, uh, about the only credit cards I used were gas cards, because I didn’t like carrying, you know – A.: Uh-huh. Category: Termination Common ground: world knowledge Information: discourse-new Given/New: prosodic prominence on married Listener’s reaction: no comment

Corpus example 22 (Context: movies) A.: Because that, that just reminds me of, you know, like back in school and, you know, the girlfriends just taking off for a day (laughter) or something. (…) Just getting in trouble and probably not that much trouble but, (…) Before we found the shopping malls that’s (laughter) what we did. B.: Uh-huh.

Imposing Common Ground by Using Temporal Connectives 139 Category: Termination Common ground: communal knowledge Information: discourse-new Given/New: presented as Given, definite article; weak prominence on before Listener’s reaction: no comment

Figure 1. Prosodic curve of Corpus example 5 (Section 5.1)

This page intentionally left blank

Part II COMMON GROUNDS

This page intentionally left blank

Lexical Markers of Common Grounds Edited by Anita Fetzer and Kerstin Fischer © 2007 Published by Elsevier Ltd.

THE PRAGMATICS OF COMMON GROUND: FROM COMMON KNOWLEDGE TO SHARED ATTENTION AND SOCIAL REFERENCING François Nemo

Lexical markers of common ground cannot be studied without a theory of what exactly common ground in utterances and interactions is. The aim of this text is to present the various ways common ground is and has been represented in general pragmatic theories since the early 1970s and to show how the study of lexical markers may change our views about the role of common ground in discourse and conversation. From the verbal-logical perspective advocated by Paul Grice, in which common ground plays almost no role, to the conversational-sociological perspectives of Erwing Goffman and ethnomethodologists, in which the existence and/or construction of shared belief and representations is central to the description of conversation, I shall detail mainstream approaches to common ground with a special focus on Anscombre and Ducrot’s description of the role of shared scalar views (called topoï), contrasting it with Sperber and Wilson’s descriptions of ostensive-inferential communication and mutually manifest (but nevertheless chosen) cognitive environments. I shall discuss the capacity of such models to grasp and to allow the ˇ representation of what is at stake (see for instance Condon and Cech, Fischer, Pittner, this volume) in the diversity of uses of Lexical Markers of Common Ground (LMCG). I shall also show that we need to go further and that if speaking is indeed a matter of attracting somebody’s attention on something and asking him or her to take it into account, then most of what LMCG are doing is making sure that this process is successful, so that they basically are common ground builders, a reality which cannot be understood outside of the attentional and scalar frames in which they are inserted and which they help to build.

1. DEFINING COMMON GROUND In pragmatics, defining what common ground is and the role it plays in the construction of meaning or in the interpretation of utterances and discourse is not an easy task. The reasons are

144 Common Grounds the co-existence of wide and narrow uses of the notion and the fact that many authors who are using the term do not define it precisely, while others who offer rather sophisticated descriptions of common ground do not use the term. My first goal here shall thus be to clarify what can be considered as common ground and to describe how it has been described in various models in the recent history of pragmatics. I shall distinguish between models that treat common ground as input to the description of linguistic interaction and approaches that describe it as emerging from the linguistic interaction and hence as an output of the interpretative process.

1.1. No common ground? – Grice In Grice’s central model of communication, common ground plays apparently no role. It is assumed that implications are generated by the encounter of what is said with the knowledge of shared conversational principles. Thus, combining knowledge of the linguistic code with knowledge of the cooperation principle and conversational maxims is all that is needed to communicate efficiently. But this minimalist approach to the issue of common ground is not sufficient for the description of specific examples. For instance, consider “he has a good handwriting” as the reply to the question “What kind of student is he?”; in order to draw the correct implicatures, the hearer needs something to jump from the sentence “He has a good handwriting” to the implicature “It is not a good student”; one needs to know that writing well is not the most important thing for defining the qualities of a student, and that this kind of background knowledge being precisely what many authors have called common ground. It is thus fair to say that in Gricean and post-Gricean models (notably Levinson, 2000), common ground is not technically absent but theoretically unfocused. And even though common ground may seem to be absent from Grice’s own work, it is thus perfectly possible and legitimate to reconsider the relationship between common ground and conversational maxims, for instance by discussing the way (see Fetzer, this volume) a contribution is or is not “such as is required, at the stage at which it occurs, by the accepted purpose of direction of the talk exchange” (Grice, 1989 [1975: 45]).

1.2. Common ground as background A powerful tradition in pragmatics basically considers common ground as background information, and views its role in a ground/figure perspective in which utterances are the figure and common ground is the background. Within this tradition, it is possible to distinguish between maximalist definitions according to which background is made of “everything which a speaker takes for granted when he speaks”, and minimalist approaches, such as Grice’s, which limit this background to the knowledge of the linguistic code on the one hand and the knowledge of pragmatic principles on the other hand.

1.2.1. Culture as a common background? It is also, and perhaps more importantly, necessary to distinguish between defining this background as some kind of a “non-contextual context” and defining background as selected

The Pragmatics of Common Ground 145 by the utterance itself. Some important authors, noticeably Sperber and Wilson, hold that the selected background is a portion of what the interlocutors share about the world (the non-contextual context), while other important authors, noticeably Ducrot and Anscombre, consider the selected background to be imposed onto the hearer. By calling it “non-contextual context”, my aim is to stress the fact that it is possible to consider the problem from two opposite perspectives: • one is to follow the idea that there is a large number of things which the interlocutors share about the world, which could be called a common culture, and that it must be the case at some point that what is said and communicated is related with this shared culture; • the second is to start from the utterance and to look for the kind of assumptions it presupposes. It is indeed the case that in the first case, what we have is not context-specific, since culture is not context-specific but shared by all contexts, and at the same time contextually activated since such or such cultural element may be mobilized by such or such utterance.1 It is thus not only part of the context of a given utterance, but also deeply and fundamentally non-contextual. The other way round, if one starts from the utterance itself, and considers the assumptions which this utterance takes for granted, the question will be to know whether these assumptions are shared by a community or not. As we shall see, it is the discrepancy between these two perspectives, in other words the fact that what is taken for granted at the utterance level is indeed common ground or not at the community level, which makes the question of the status of common ground in pragmatics so problematic. But it is also the fact of ignoring this discrepancy which has allowed the formulation of syncretic pragmatic models, and thus allowed the description of common ground in terms of a set of shared assumptions.

1.2.2. Common ground as cognitive environment In what Levinson (2000) has called “utterance-token” pragmatics, the general idea that the production of an utterance should be described within the space in which it takes place can be found under the following form: a specific utterance-token is produced within a cognitive environment which may be described as a set of mutually manifest assumptions. And inferences are what is produced by the encounter of the new information with the set of already shared information (Sperber, 1974; Sperber and Wilson, 1986: 1–119). Thus, we have something like Figure 1. An important point here is that the former equation “common ground = common knowledge” is being replaced by the equation “common ground = what is mutually manifest”, which allows the utterance of something not because it would be new, strictly speaking, but because it allows the speaker to make it more manifest for the addressee. But as far as defining common ground is concerned, and in order to allow comparisons with Grice and Argumentation in Language Theory (ALT), it is clear that in Relevance Theory (RT) (Episode 1), common

1

Sperber’s work as an anthropologist, in his attempts to describe rituals for instance, was completely embedded in this kind of perspective, and it is clear that for the best and the worst, the fundamental perspective of Relevance Theory has been inherited from this work. See for instance, Sperber’s (1974) description of the way the encounter of a new assumption with a set of old assumptions produces new assumptions.

146 Common Grounds

assumptions a1, a2, a3, a4, a5, a6, a7, a8, a9, a10 a11, a12, a13, a14 cognitive environment

+ +

an

= ai1 ai2 (inferences),

new assumption = utterance

cognitive effect

Figure 1. Relevance Theory (Episode/Perspective 1) ground in the wider sense is described as formed of (a) the linguistic code; (b) a principle of relevance; (c) cognitive environment; (d) something like production rules and an inference engine. And in the narrower sense, it is equated with cognitive environment. Communication, within such a model, works exactly like an expert system in artificial intelligence, with the cognitive environment being the equivalent of the knowledge base of an expert system, but for its attentional rather than informational definition.2 This narrow definition of common ground as a set of shared assumptions seems to have been adopted well beyond RT, as problematic as it may prove to be.

1.2.3. Common ground as argumentative backing In the model developed from the 1970s to the 1990s by Ducrot and Anscombre – and thus to some extent well before RT – which is called Argumentation in Language Theory3 , the perspective is similar to RT as downstream implicit4 conclusions (rather than inferences) are drawn not only from what is said but from an upstream implicit. At the same time, it is different as the model does not require neither something as a cognitive environment nor any principles. Instead, it describes the kind of non-implicational mechanisms which allow for drawing conclusions as in Figure 2 (Anscombre and Ducrot, 1983).

2 A limit of the RT model is that even though it postulates the existence of something like production rules and an inference engine, it has never been able to describe precisely the kind of production rules and inference engine which would be at work, claiming only that they were not implicational but demonstrative. 3 Presentations in English of the ALT framework are available (Nyan, 1998, this volume; Plantin, 2003; Van Eemeren et al., 2003; Malrieu, 1999) including a specific issue of the Journal of Pragmatics (1995, 24, 1/2). Argumentation in Language Theory is not a theory whose object would be argumentation but a theory which uses argumentation, in a technical sense, to account for semantic and discursive observations about the linguistic behaviour of connectives, operators, interrogatives, adjectives, etc. Its basic claims are that (i) “uttering a sentence [for instance P1 “the weather is fine”] is pointing to something else, a possible continuation or conclusion “outside of the sentence”, a second utterance P2 [for instance “let’s go to the beach”]. In other words, in this theory, there is an equivalence between “P2 is the meaning of P1” and “P1 is an argument for P2” (Plantin, 2003: 182); and (ii) that the argumentative force or direction/orientation of an utterance is semantically built by linguistic expressions or words inserted in the sentence independently of any informational content. The model is thus a model, which deals with utterance-type semantics in the levinsonian sense (2000) rather than with utterance-token pragmatics. 4 Making a distinction between downstream implicit and upstream implicit is a necessity in pragmatics especially when it comes to understanding common ground. A distinctive feature is that downstream implicit is dependent on the reality of p, whereas upstream implicit is not: saying the weather is bad or the weather is nice leads to opposite downstream conclusions/inferences but leaves unaffected the shared idea, an upstream implicit inference, that “the better (worse) the weather, the more (less) pleasant it is to go outside.”

The Pragmatics of Common Ground 147

+p+r

sentence p convocation of argumentative warrants (scales, topoï, blocks)

conclusion r

Figure 2. Argumentation in Language Theory

In ALT models, it is the words themselves which are encoding specific topoï (or scales, or blocks), that is, for instance formulas with a “the more p the more r  or “the less p the less r  +p + r structure, which behave as production rules and inference engines. The ALT models are thus non-contextual and do not use pragmatic principles. They consider downstream implicit as produced from an upstream implicit which is produced from the words. As for common ground, the argumentative warrants (i.e. topoï, etc) which form the upstream implicit are defined as general (i.e. not context-specific), gradual and shared by a community (Ducrot, 1996). This does not mean that they have to be shared by the interlocutors. In Ducrot’s early work on presupposition (1972), it is even assumed that this kind of upstream implicit is imposed on the addressee as an interlocutive frame rather than shared with the addressee prior to the utterance itself. Such a distinction, and the manipulatory view of communication it embraces, is in fact extremely important for understanding that the common ground an utterance may build is not necessarily consensual, so that neither the “common ground = consensus” equation and the idea that common ground would be given prior to communication (or even after), both are not valid.

1.2.4. Is common ground given or chosen? In the formulation of RT (Sperber and Wilson, 1986), until page 119, the theory is based (see Figure 1) on the idea of an encounter of a new assumption with a stock of shared assumptions. However, after page 119, it is stated that this cognitive background is not given prior to the uttering of the new assumption but chosen and inferred from the assumption of optimal relevance, which allows the hearer to select the most relevant upstream implicit/cognitive background: instead of starting from both the new assumption and the cognitive environment, we have a situation which departs clearly from the expert-system metaphor used in Figure 1, as illustrated by Figure 3. It turns out to be problematic to define the relationship

assumptions a1, a2, a3, a4, a5, a6, a7, a8 chosen context

Principle of relevance

⇐

an

Principle of relevance ⇒ ai1 ai2 (inferences),

new assumption

Figure 3. Relevance Theory (Episode/Perspective 2)

cognitive effect

148 Common Grounds between common ground, mutually manifest assumptions, the cognitive environment and the newly introduced, chosen context, mainly because it is impossible to reverse the initial perspective in such a way without loosing/destroying the “common ground = cognitive environment = what is mutually manifest” equation as in Figure 1 (Sperber and Wilson, 1986: 119–321). The exact relationship between the chosen context and the cognitive environment has indeed remained implicit and problematic: Sperber and Wilson seem to suggest that the chosen context is necessarily a subset of the initial cognitive environment, therefore avoiding the issue of the interlocutive status of this chosen context: Is the chosen context indeed and always mutually manifest to the interlocutors? In the very few examples which are provided, it might seem to be the case, and indeed choosing a context could be considered as a process of foregrounding within the cognitive environment such or such assumption because it would be especially relevant for what the new assumption is about. But obviously even if this situation may occur, it seems clear that there are many cases in which the relevant background is in fact not mutually manifest at all, because it takes for granted something the hearer disagrees with. Reversing the mechanisms of selection as in Figure 3 is thus not problematic by itself, as there is little doubt that things do occur this way most of the time (Nemo, 1988), but because if the speaker chooses the context, it cannot be taken for granted that this chosen context will be mutually manifest.

1.2.5. Common ground and the modal frames of speech In post-Ducrotian models such as the model presented in Nemo (1988, 1992, 1999), common ground becomes less directly involved in the generation of downstream implicit, whether this implicit is called conclusions, inferences, or implicatures, because it becomes accessible only through a new semantic path, which consists of a modal frame built from the sentence which is uttered. So what we have is shown in Figure 4. If we consider here the status of shared warrants, it becomes apparent that the direct upstream implicit of a sentence is a modal frame, i.e. a set of alternatives, and that this modal frame which is proposed to the hearer as an interlocutive frame, if it may often be backed by shared warrants through relation 2, may introduce an information which was not mutually manifest nor shared by a community, and thus the relationship between the interlocutive background and common ground becomes loose at best. What does that mean exactly? First of all that the simplest upstream pragmatic inferences do not necessarily lead to shared information. For instance, if I hear someone saying

Shared warrants

sentence

Modal frame 2

1 3 address

4

difference it makes 5 attentional-indexical consequence

Figure 4. Modal frame utterance semantics (Nemo, 1988, 1992, 1999)

The Pragmatics of Common Ground 149 “Kofi Annan is alive”, the first inference I shall make is that something has happened to Kofi Annan (an attack, a heart attack, an accident, etc.) which makes it worth saying (i.e. relevant) that he is alive. In other words, I will infer that “Kofi Annan might not be alive” is strongly the case (Nemo, 1988). But this information that something happened to Kofi Annan, without which the utterance would not be relevant, need not be common ground at all. The equation “utterance background = common ground” is thus wrong, the ordinary background of an utterance or a set of utterances being a modal frame instead. If we then compare for instance the following examples – (1) C’était loin donc j’ai pris la bicyclette It was far so I took the bicycle (2) C’était pas loin donc j’ai pris la bicyclette It wasn’t far so I took the bicycle – it can be shown that they do not introduce the same set of alternatives (i.e. the same modal frame), since example (1) introduces the two alternatives (far and bicycle) versus (not-far and not-bicycle) in which not-bicycle is interpreted as meaning “by foot” whereas the example (2) introduces the alternatives (far and not-bicycle) versus (not-far and bicycle) in which not-bicycle is interpreted as meaning “by car, bus, taxi,and so on”. But the fact that the choice was between biking and walking or between biking and using a motor vehicle, which is a semantic condition of relevance for (1) and (2), does not have to be part of any mutually manifest environment; even though the exact determination of the set of alternatives is backed by a general knowledge about human locomotion. Common ground, it seems, may play a role in the definition of the interlocutive modal background, but the common ground itself is not the background. Moreover, if the alternatives which are introduced are the semantic conditions of relevance of the utterance, the addressee does not need to accept their semantic content and may refuse the set of alternatives which has been proposed to himself/herself. But in order to do so he/she must propose another alternative set, and hence the ratification/rebuttal process of the interlocutive background becomes a dialogical process. To illustrate this with a simple example, we can consider the following exchange: (3) A: Paul a peu travaillé (Paul hasn’t worked much/has worked little) B: Paul a travaillé (Paul has worked) The answer by B in (3) seems problematic because it states something which seems to be common ground, being at least presupposed by the first assertion, so that it clearly cannot be predicted in terms of informational cooperation or in terms of effect of a new assumption on a set of old assumptions. But if the answer is fully relevant, as “Paul did little work but he worked” would also be, and even seems to contradict what has been said, it is because of the alternative sets which are introduced: the interest and effectiveness of saying Paul worked being simply that it reminds the first speaker that Paul might not have worked at all, thus refusing the alternative frame proposed by the first speaker (doing little work versus doing a lot of work). In any case, whether the interlocutors will finally agree or not on either frame is unpredictable, but what is sure is that if common ground is finally achieved it will be through a process of ratification of a perspective about what has happened.

150 Common Grounds

1.3. Deconstructing common ground Consequently, there is a general problem with the classical view that common ground could be described in terms of a set of assumptions which would either form the cognitive stage in which further communication would take place or be the outcome of the communication process. It seems indeed that common ground: • is not a set; • does not consist of assumptions; and that its role cannot be framed into a ground/figure perspective in which utterances are the figure and common ground is the background, nor into the perspective of informational convergence or coherence. As for the first assertion, it is quite clear that common ground emerges from the utterance (or discourse) itself, and plays no stable role. It should hence be considered as a discursive (rather than communicational) background whose role is unstable, and which does not have the collective and temporal status of a set of assumptions. If we consider for instance a phenomenon such as bridging, and the kind of examples treated by Wilson and Matsui (1998): (4) Sara left [location A] for [location B]. She hates sandy beaches. with 100 percent of the interpreters understanding sandy beaches to be in location A. And if we remark that changing (4) into (5): (5) Sara left [loc.A] for [loc. B]. She hates sandy beaches. How stupid of her! sandy beaches are immediately relocated in B, it becomes possible to understand that in accordance with Ducrotian ALT models, which basically were not designed to account for this kind of data, what is necessary to account for the two interpretations is only a semantic topos (rather than a pragmatic principle) according to which “the more you hate a place the more you leave it”, with its variant “the more you like a place the less you leave it”. In other words, what is needed at the start is the creation of a common ground semantic gradual relationship between the predicates leaving and hating, leading to the correct interpretation of (4) and then, because there is no way to create the same kind of common ground semantic gradual relationship between leaving a place one hates and being stupid, since improving one’s situation is not being stupid, it becomes necessary to imagine that since she is said to be stupid and since there is a semantic topos which says that “the more one worsens one’s situation the more stupid one is”, then it must be the case that leaving location A for location B is making things worse and it must be the case that the sandy beaches she hates are in B. Since all this starts from the explicit conclusion she is stupid, the chain of reasoning in (5) moves upstream through modal frame integration to the topoï, producing as a by-product the truth conditional interpretation of the second sentence. This shows once again that trying to promote such or such interpretative path as the standard path of utterance interpretation is meaningless because there is nothing such as a standard path of utterance interpretation. More interestingly, what deserves to be observed here is not that ultimately the upstream reasoning ends up in extremely general and partly cultural conceptions of life such as the idea that one

The Pragmatics of Common Ground 151 should avoid what one does not like, since it can be predicted by Ducrotian and post-Ducrotian models, but how this shared “knowledge” (or whatever it could be named) is actually used. In other words, it seems that all the pragmatic models have been too focused on accounting for downstream implicit implicatures, conclusions, inferences, scalarisations, and so on to consider the nature of what exactly was going on. If we consider all examples from (1) to (5), there is indeed a general pattern of behavior, which deserves some consideration, according to which whatever is taking place (taking my bike, working, moving) at a very specific and local level is to be related axiologically with a whole class of attitudes/behaviors which they either contradict or illustrate. And it must be noticed that such a process has little or no connection at all with the notion of adding a new fact (assumption) to a knowledge base (a set of shared assumptions). This means that Sperber’s (1974) attempt to capture the nature of the relationship between culture (as the most global form of common ground) and ordinary utterances has failed, because he treats this relationship as a relationship between autonomous assumptions. In Wittgensteinian terms, the problem is that the language game we are actually playing and the language game played in the new assumption/old assumptions classical model (cf. Figure 1) have in fact very little in common: adding a new information A to a set of old information B, even if some extra inferred information is generated by combining A and parts of B is neither what is at stake in utterances such as (1), (2), (4), and (5) nor what lexical markers of common ground are dealing with.

1.4. Joint attention, controlled attention, and social referencing The most interesting convergence in contemporary pragmatics is probably the convergence between ALT and RT in the characterization of the nature of the communicative game associated with ordinary utterances: both ALT and RT assume that speaking (or communicating) is basically a matter of attracting somebody’s attention on something and asking him/her to take it into account (Nemo, 1999, 2005), a cognitive game which has been called ostensive-inferential communication in RT (Sperber and Wilson, 1986). Here, however, the similarity ends, since ALT holds that it is impossible to attract somebody’s attention on something without showing what you feel about it (or what you want your addressee to feel about it), which is precisely what ALT is using the term argumentative for, whereas RT seems to believe that it is possible to be neutral when you attract somebody’s attention on something, and hence possible neither to show nor to promote any reaction to what is pointed at. The most general justification of ALT and RT’s claim about the attentional nature of communication is to be found outside of pragmatics and linguistics in cognitive psychology and cognitive ethology (Eilan et al., 2005), which are concerned with the types of attentional processes available in interpersonal communication, namely: • mutual attention (the baby looks at his/her mother who looks at him/her), which is the first form of interpersonal communication in humans; • deictic attention (the baby looks at his/her mom looking at something), which develops a few months later; • joint attention (the baby looks at what his/her mom is looking at), also developing a few months later; • controlled attention (the mother uses the capacity of the baby to follow her attentional focus to attract his/her attention on something).

152 Common Grounds Cognitive ethologists have described further not only the dimensions of these attentional processes, in particular the use of controlled attention to divert the other’s attention from something by pointing to something else, but also the mind-reading capacities directly related with deictic and joint attention, and the consequent existence of social referencing, for example of the fact that babies can observe the reaction of their parents (or any other humans) when they look at something and learn from it (for instance if they can see fear, anger, delight, joy, etc.). It could be added that empathy, as a capacity to see and feel the world from somebody else’s point of view, is also part of the same set of cognitive mechanisms available to human beings. So if the first metalanguage of pragmatics was the language of action, and if the second metalanguage of pragmatics was the language of interaction, the third metalanguage of pragmatics, or at least the only metalanguage that may unify pragmatics from face work to information structure and the interpretation of tenses and moods, should be attentional.5 As for here, it is obvious that the diversity of attentional contexts and attentional moves is of considerable interest for pragmatics, and allows us to go well beyond the ostensive-inferential mechanism, which is an example of controlled attention, into the diversity of attentional mechanisms, including social referencing and “argumentation” (see also Nyan, this volume), which are associated with communication. The importance of controlled attention and social referencing for pragmatics should indeed not be underestimated. Even though these mechanisms do not require language, it is clear that linguistic interactions are deeply embedded in such mechanisms and that linguistic communication has inherited some of its most decisive characters from them. An active use of the social referencing mechanism by the speaker is, for instance, the cognitive mechanism on which the argumentative uses of language are based. To move back to the starting point of this paragraph, if social referencing exists and is important, we need to follow ALT in its belief that it is impossible to point at something without expressing or promoting a relation or an attitude toward the pointee, hence focusing on social referencing. In this way we can overcome RT’s complete ignorance of social referencing and its belief that it is possible to express no attitude toward the pointee in the communicative act. In any case, if we consider our problematic language game (cf. paragraph 1.2.6), what is important and interesting is that it is possible to describe it in terms of projection of attitudes, and hence to understand both the role of the common ground semantic gradual relationship we have encountered and the space in which the lexical markers of common ground are being produced.

1.5. Common ground as “business as usual” We have seen above that communication is not only a matter of sharing the same attentional focus but also a matter of sharing the same attitude toward what is focused on. Attitudes may basically fall into two different categories: expressing a “business as usual” attitude toward what happens and expressing concern or fascination about what is happening. We can now

5

On the same ground, some linguists (e.g. Tomasello, 1995) have shown in great detail that the joint attention and controlled attention mechanisms were actually playing an important role in the acquisition of the vocabulary, for instance in the determination of the denotation of words: looking at what somebody is looking at when he/she uses a word is a much more efficient and practical way to catch what it refers to than learning by direct ostentation.

The Pragmatics of Common Ground 153 investigate the role of common ground in the kind of language game discussed above. Before we can classify something as either “business as usual” or as “concerned with/fascinated by”, it is however necessary to define first what “business as usual” is or why we should be concerned about something. The fact that we have modal frame integration backed by general tendencies (and topoï) such as “increasing distance with things you dislike is normal” or “reducing distance with things you dislike is problematic” shows that what is at stake in the most ordinary utterances is not a matter of adding a fact (an information) to other facts (other informations) in a world of facts6 (a world of information) but instead a matter of building reactions to the world, by considering what happens in terms of its conformity or non-conformity to a certain definition of “business as usual”. But then of course defining what is “business as usual” may become problematic, and it can be the case that there is no common ground on common ground. But if all utterances are inserted in an attentional language game which frames facts either as “business as usual” or not, to understand an utterance such as a sign as (6) stuck on a shop’s (or an office) window: (6) Opened on Monday. we first need to answer the question of whether to be opened on Monday is “business as usual” or not, so that in a country where shops are often closed on Mondays, such as France, we are led to the interpretation “opened also on Monday”, whereas in countries in which opening on Monday is business as usual, it would be interpreted as “opened only on Monday”. The role of our knowledge of the fact that shops are or are not opened on Mondays is hence completely dependent on the attentional game we are playing, so that the interpretation of (6) and the inferences we can draw from this interpretation (“also opened the other days” or “closed the other days”) are also pragmatically driven. Moreover, even if the idea that we could describe normal communication in terms of transmission of information, rather than in terms of attentional moves and expression of attitudes, may seem to be empirically backed by the “business as usual” contexts, we must realize that these contexts are not “normal communication” but only one of the alternatives of normal communication. The conclusion of all this is that we cannot understand common ground as a discrete phenomenon, a “to be or not to be common ground” reality, and that we should in most cases describe it in gradual terms, that is, as something which is more or less acceptable (ratifiable). The kind of general semantic relationships we have encountered in Figures 2 and 4 are indeed used as measuring devices and measuring tools, and not as a stock. Consequently, common background, and more generally culture, cannot be described in terms of representations but as a (possibly incoherent) set of interpretative grids and prisms through which we see and inseparably judge reality,7 and hence as a set of measuring tools that allows its holders to grasp reality in terms of the measurement scale they include (see also Nyan, this volume).

6

Human languages are not concerned with describing the world per se. What they are concerned with is in which ways what happens may affect us or should generate reactions, and hence what they point at defines the way we can see the world (Cadiot and Nemo, 1997a, 1997b). 7 As being business as usual or not.

154 Common Grounds

1.6. From common ground to common ground It would be easy to separate common ground as background and as a goal of interaction. Furthermore, it could be useful to distinguish between: • interlocutive common ground (ICG); • discursive common ground (DCG); • cultural common ground (CCG); where interlocutive common ground is the outcome of the ratification process, discursive common ground is the warrant of the successive utterances, and cultural common ground is what is beyond ratification in a given community. This analytical distinction might avoid confusion and vagueness: it may be the case that a large part of what occurs as discursive common ground could be considered as being part of the cultural background, but since discursive and cultural in principle may also be different or the case may simply be undecidable, it is preferable not to conflate the three notions and not to over-generalize specific configurations. But such a distinction should not lead us to consider them separately, for it seems clear that DCG and to some extent CCG cannot be considered as external to the process of building an interlocutive common ground. If indeed communication is not only a matter of sharing the same attentional focus but also a matter of sharing the same attitude toward what is focused on, conversation will deal with central issues such as: • co-defining whether something is worth the attention or not; • co-defining what must be attended to; • co-defining how to take it into account. It is obvious that these central processes must be ruled by certain norms, in the same way as having the right to speak is ruled by conversational mechanisms (Sacks et al., 1974; Schegloff, 1997). Consequently, we should study what the norms are for defining what is worth the attention or not, what the norms are when it comes to defining the attentional set (i.e. what will be taken into account simultaneously), and what the norms are for defining which attitude should be adopted regarding what is being considered. It is of course well beyond the scope of this paper, but it can be predicted that the shortest way to achieve interlocutive common ground is probably to back one’s utterances by a perspective which is already part of common ground. But if this is the case, and if the best way to reach consensus is to start from consensus, thus facilitating the ratification process considerably, this sheds some light on the status of DCG: it is not only the case that the discursive common ground is not independent of what is being said, as it would be in a knowledge base, but also that it mainly has a justificational role. The DCG, to a large degree, and this explains its semantic nature of defining “business as usual”, is used to justify an interlocutive position and to make it easily ratifiable. Hence, we do not need to choose between common ground as background and common ground as an interlocutive goal, because in any case common ground as background is but a means to reach interlocutive common ground.

The Pragmatics of Common Ground 155

1.7. Diverging grounds Although achieving interlocutive common ground is probably one of the axiomatic goals of conversation and verbal interaction, the management of diverging ground is also an essential part of our pragmatic competence. Interlocutors need not only have a representation of what does not need to be ratified and what is disputable, but also a representation of what cannot be ratified. And this not only in order to choose what they will or will not try to achieve, but also in order to integrate diverging ground in the context. The idea that context or context selection should take place only within common ground is falsified by the most simple data, such as ironic utterances. In irony, speakers not only select an interlocutive background that asserts something at least one of the speakers does not agree with, but even make fun of what is being said. Similarly, an efficient path to achieve interlocutive common ground is to start from diverging ground as a discursive context, or to mark it minimally (Berrendonner, 1981) and to move forward to reduce or overcome the divergence. In both cases, utterances or discourses readily become polyphonic, in the sense that they introduce various voices responding one with another. Such a situation occurs not only when the interlocutors adopt different perspectives but also when the divergence is with a third person or when somebody switches between perspectives. The linguistic study of polyphony, which was promoted by Ducrot (see Ducrot et al., 1980: 38–56; Ducrot, 1985) to account for the dialogic dimension of many ordinary utterances, has become an important field of study, noticeably with the ScaPoline school (Nølke et al., 2004), which deals with the grammaticalized polyphony of moods and tenses, the discursive polyphony of discourse relations, and more generally with the integration of viewpoints into structure (Nølke, 1993, 2001a). In doing so, the linguistics of polyphony provides us with extremely detailed analyses of the way linguistic expressions are used to make it possible for the addressee to adopt the different perspectives and standpoints available. Empathy, mindreading, and polyphony all show that building a common perspective about something is not a matter of imposing such or such perspective or attitude, but depends on our capacity to share temporally a perspective and an attitude. Trying to look at things from somebody else’s perspective, adopting this perspective, or finally sticking to one’s own perspective is conversational routine. Yet in all cases, enhancing common ground and reducing diverging ground is a basic conversational task, in which lexical markers of common ground are deeply embedded.

2. LEXICAL MARKERS Possibly because written data was until recently the main source of corpus, which led to an underestimation of the existence and importance of lexical markers of common ground, it has been the case that in the study of LMCGs, empirical work, that is, sign-token pragmatics, has been well ahead of the capacity of general pragmatics to integrate empirical observations in a general theory that would be able to do justice to them. And as always in pragmatics, the only way to do justice to empirical observations is to describe the interlocutive space in which they take place. Progress in pragmatics is thus a matter of constantly enriching our representation of the interlocutive space and understanding of the set of constraints which defines it, by taking

156 Common Grounds into account phenomena which at first sight may seem marginal or extremely local but which constitute in fact traces of what the real constraints on communication and interlocution are. Lexical markers of common ground are all those items that play a role in the process of co-defining the attentional ground and co-defining how it should be taken into account (“attitude”). It includes hence attentional shifters (“we should also take this or that into account”, “we should now focus on something else”); joint attention builders (“could you pay attention and consider this”); empathic experiencers (you know); evidentials, contribution builders and contribution ratifiers (see Nemo, 2005); divergence reducers (“How do you want me to say something else than this”); acknowledgers or rebutters, to mention a few types of LMCGs other than those described in detail, for example by K. Fischer (this volume). And if we are still far from a typology of the different types of LMCGs, it is mainly because in the process of co-definition of an interlocutive ground, the diversity of interlocutive paths and dimensions of the process and the exact nature of the interlocutive norms which make them possible have not been systematically studied. The study of LMCGs is thus an open window on the nature of these mechanisms, and probably one of the most promising ways to allow general pragmatics to swing to a fully interlocutive perspective.

REFERENCES Anscombre, Jean-Claude and Ducrot, Oswald, 1976. L’argumentation dans la langue. Langages 42, 5–27. Anscombre, Jean-Claude and Ducrot, Oswald, 1983. L’argumentation dans la langue. Mardaga, Bruxelles. Blakemore, Diane, 1992. Understanding Utterances: An Introduction to Pragmatics. Blackwell Publishers, Oxford. Berrendonner, Alain, 1981. Eléments de pragmatique linguistique. Editions de Minuit, Paris. Cadiot, Pierre and Nemo, François, 1997a. Propriétés extrinsèques en sémantique lexicale. Journal of French Language Studies 7, 1–19. Cadiot, Pierre and Nemo, François, 1997b. Pour une sémiogénèse du nom. Langue française 113, 24–34. Cadiot, Anne, et al., 1979. Oui mais non mais ou: Il y a dialogue et dialogue. Langue française 42, 94–102. Cadiot, Anne, et al., 1985. Enfin marqueur métalinguistique. Journal of Pragmatics 9, 199–239. Clark, Herbert H. and Schaefer, Edward, 1989. Contributing to discourse. Cognitive Science 13, 259–294. ˇ Condon, Sherri and Cech, Claude, this volume. Ok, next one: Discourse markers of common ground. Ducrot, Oswald, 1972. Dire et ne pas dire. Hermann, Paris. Ducrot, Oswald, 1984. Le dit et le dire. Editions de Minuit, Paris. Ducrot, Oswald, 1985. Le Dire et le dit, Editions de Minuit. Ducrot, Oswald, 1996. Slovenian Lectures/Conférences Slovènes ISH.I. Zagar (ed.), Ljubljana. Ducrot, Oswald, et al., 1980. Les mots du discours. Editions de Minuit, Paris. Eilan, Naomi, Hoerl, Christoph, McCormack, Teresa, and Roessler, Johannes, 2005. Joint Attention: Communication and Other Minds. Issues in Philosophy and Psychology. Oxford University Press, Oxford.

The Pragmatics of Common Ground 157 Fernandez-Vest, Jocelyne, 1994. Les particules énonciatives dans la construction du discours. PUF, Paris. Fetzer, Anita, this volume. Reformulation and common grounds. Fischer, Kerstin, this volume. Grounding and common ground: Modal particles and their translation equivalents. Grice, Paul, 1989. Studies in the Way of Words. Harvard University Press, Cambridge, MA. Levinson, Stephen, 2000. Presumptive Meanings: The Theory of Generalized Conversational Implicature. The MIT Press, Cambridge, MA. Malrieu, Jean-Pierre, 1999. Evaluative Semantics. Cognition, Language and Ideology. Routledge, Oxford. Nemo, François, 1988. Relevance (D. Sperber et D. Wilson), Book review. Journal of Pragmatics 12, 5–6. Nemo, François, 1992. Contraintes de pertinence et compétence énonciative: l’image du possible dans l’interlocution. PhD Thesis. EHESS, Paris. Nemo, François, 1997. The modal frames of speech. In: P. Weingartner, G. Schurz, and G. Dorn (eds), The Role of Pragmatics in Contemporary Philosophy, Vol. 2. Kirchberg am Wechsel, Autriche, 700–707. Nemo, François, 1999. The pragmatics of signs, the semantics of relevance, and the semantic/pragmatic interface. In: K. Turner (ed.), The Semantics-Pragmatics Interface from Different Points of View. CRiSPI Series, Elsevier Science, Amsterdam, 343–417. Nemo, François, 2001. Contributions, énoncés, constructions, morphèmes. Eléments pour une linguistique de la signification et de l’interprétation. Thèse d’habilitation. Université Paris 8. Nemo, François, 2002. But (and mais) as morpheme(s). Delta 18 (2), 87–114. Nemo, François, 2006. Discourse words as morphemes and as constructions. In: K. Fischer (ed.), Approaches to Discourse Particles. Studies in Pragmatics Series, Vol. 1. Elsevier, Amsterdam, 415–448. Nemo, François and Cadiot, Pierre, 1997. Un problème insoluble? (2). Revue de Sémantique et Pragmatique 2, 9–40. Nølke, Henning, 1993. Le regard du locuteur. Pour une linguistique des traces énonciatives. Kimé, Paris. Nølke, Henning, 2001a. Le regard du locuteur 2. Pour une linguistique des traces énonciatives. Kimé, Paris. Nølke, Henning, 2001b. Polyphonie. In: P. Charaudeau and D. Maingueneau (eds), Dictionnaire d’analyse du discours. Seuil, Paris, 444–448. Nølke, Henning, Fløttum, Kjersti and Norén, Coco, 2004. ScaPoLine. La théorie scandinave de la polyphonie linguistique. Kimé, Paris. Nyan, Than, 1998. Metalinguistic Operators with Reference to French. Peter Lang, Bern. Nyan, Than, this volume. Common ground, categorization, and decision-making. Pittner, Karin, this volume. Common ground in interaction: The functions of medial doch in German. Plantin, Christian, 2003, Argumentation studies in France: A new legitimacy. In: F. Van Eemeren, A. Blair, C. Willard, and F. Snoeck Henkemans (eds), Anyone Who Has a View – Theoretical Contributions to the Study of Argumentation. Kluwer Academic Publishers, Dordrecht, 173–188.

158 Common Grounds Rouchota, Villy, 1996. Discourse Connectives: What do They Link? UCL Working Papers in Linguistics 8, 199–212. Sacks, Harvey, Schegloff, Emanuel, and Jefferson, Gail, 1974. A simplest systematics for the organization of turn-taking for conversation. Language 50, 4, 696–735. Schegloff, Emanuel A., 1997. Whose text, whose context? Discourse & Society 8 (2), 165–187. Schiffrin, Deborah, 1987. Discourse Markers. Cambridge University Press, Cambridge. Sellevold, Kjersti, 2001. L’expression linguistique du doute dans les Essais de Montaigne: une approche polyphonique. PhD Thesis, Université de Poitiers. Sperber, Dan, 1974. Le symbolisme en général. Hermann, Paris. Sperber, Dan and Wilson, Deirdre, 1986. Relevance: Communication and Cognition. Blackwell, Oxford. Tomasello, Michael, 1995. Joint attention as social cognition. In: C. Moore and P. Dunham (eds), Joint Attention: Its Origins and Role in Development. Lawrence Erlbaum, Hillsdale, NJ, 103–130. Van Eemeren, Frans, Grootendorst, R. and Snoeck Henkemans, Francisca, 2003. Fundamentals of Argumentation Theory – A Handbook of Historical Backgrounds and Contemporary. Lawrence Erlbaum Associates, Hillsdale, NJ. Wilson, Deirdre and Matsui, Tomoko, 1998. Recent Approaches to Bridging: Truth, Coherence, Relevance. UCL Working Papers in Linguistics 10.

Lexical Markers of Common Grounds Edited by Anita Fetzer and Kerstin Fischer © 2007 Published by Elsevier Ltd.

REFORMULATION AND COMMON GROUNDS Anita Fetzer

1. INTRODUCTION In a discourse-based investigation of natural-language communication coparticipants do more than produce and interpret communicative contributions. They also negotiate the communicative status of their utterance by jointly constructing meaning in context. In the process of constructing intersubjective meaning, coparticipants do not always interpret contributions in the speaker-intended way. For this reason, they may reformulate their own contribution and they may reformulate other’s contribution in order to display their interpretation of other’s contribution. Moreover, coparticipants can never be completely explicit: they can say what they mean, but should they be requested to make all of their indexical expressions explicit, they would not be able to do so (Bar-Hillel, 1998). This is because context cannot be delimitated, or to use Penco’s own words: “there is no ultimate outer context, but it is always possible to transcend the context in which we are. This gives rise to the possibility of an infinite regress, but it seems that this infinite regress is harmless, and represents our condition of limited humans” (Penco, 1999: 280). Since indexicals are context-dependent by definition, there is no ultimate outer referential domain because their delimitated contexts can always be transcended. From a dialogue perspective, contributions are anchored to local and global contexts: they express a connectedness with a prior contribution and its embeddedness in context thus reconstructing context. At the same time they pave the ground for the context that is to follow. Coparticipants do not only rely on context when interpreting their contribution, but also presuppose the validity of a subset of background information, that is, a commonly shared set of propositions and beliefs, which is generally referred to as common ground. The premise of context-dependency holds for both the interpretation of a contribution and its production with respect to force, content and formulation. To produce an appropriate contribution, a coparticipant makes assessments about how explicit he/she intends to be. This includes an assessment about the questions of (1) what kind of context is constructed in and through the process of communication, (2) what kind of background information is presupposed and (3) what propositions and beliefs are contained in the coparticipants’ common ground. Against this background, reformulations are a necessary constituent of natural-language communication,

160 Common Grounds where not only content, force and formulation, but also context and background information with its constitutive propositions, beliefs and interpersonal and interactional presuppositions can be reformulated. In the following, the nature of the connectedness between reformulation and common ground is examined. The next section analyses the concept of common ground by focussing on Clark’s differentiation between personal and communal common ground (Clark, 1996), Thomason’s notion of conversational record (Thomason, 1992), Sperber and Wilson’s concept of mutual manifestness (Sperber and Wilson, 1995) and Fetzer’s context-dependent notions of individual dialogue common ground and collective dialogue common ground (Fetzer, 2002, 2004). The third section investigates the connectedness between reformulation and dialogue. It focuses on the act of ratification and on default and non-default scenarios by making explicit their theoretical backgrounds, premises and contextual requirements. The section also examines the canonical form and function of reformulation with respect to the administration of dialogue common ground. The fourth section illustrates the form and function of reformulation in context, and gives particular references to its function in the administration of collective dialogue common ground. The final section concludes by summarizing where reformulation, dialogue common ground and common ground meet, and where they depart.

2. COMMON GROUND IN CONTEXT In the research paradigms of pragmatics and discourse analysis, the information required for the retrieval of implicit information is referred to as background knowledge, background assumptions, common knowledge, mutual knowledge, presuppositions, common ground and context. The fundamental pragmatic differentiation between what is said and what is meant is based on the distinction between information represented in the explicit mode and information represented in the implicit mode. While the interpretation of a contribution communicated in the explicit mode does not generally rely on the coparticipants’ background assumptions or common ground and does not generally require any of its presuppositions to be spelled out, the interpretation of a contribution in the implicit mode requires the explicit accommodation of a shared set of premises to which the implicit representation of information is anchored. The communication of implicitly represented information is only possible if common ground is assigned the status of a basic foundation whose content is accessible to coparticipants and retrievable for coparticipants. Furthermore, the premises and sets of propositions indexed by the implicit representation have to be a constitutive part of the common ground and they have to be accessible to all coparticipants of a communicative encounter. But what exactly is common ground, what does it contain and how is it represented? Common ground is one of those complex notions whose existence is presupposed but whose contents are hardly made explicit. This may be due to the fact that common ground contains both linguistic and encyclopaedic knowledge represented by an infinite number of assumptions and propositions, such as first-order assumptions, second-order assumptions and multi-order assumptions, and that it contains first-order propositions, second-order propositions and multi-order propositions. It may also be due to the fact that the concept of common ground is inherently dynamic and thus both process and product. This is reflected in the fact that common ground contains already known information, but at the same time deals with

Reformulation and Common Grounds 161 the accommodation of new information, and for this reason is supplemented by some kind of administrator and some kind of update mechanism. Thus, there is the product of common ground, whose contents can be captured, to some extent at least, and there is the process of common ground anchored to the act of grounding (Clark, 1996), which adds further complexity to the examination of the contents of common ground and of their connectedness. So far, the contents of common ground have been approached in a context-independent manner, where they are represented by objective (or true) information whose validity obtains in all possible scenarios. This is reflected in the differentiation between the traditional conceptions of linguistic knowledge as context-independent and of encyclopaedic knowledge as contextdependent, and it is also inherent in Searle’s distinction between declarative knowledge, namely knowledge that, and procedural knowledge, that is knowledge how (Searle, 1969: 14). The latter is connected with the speech-act theoretic concept of felicity condition which is required for the retrieval of implicitly represented information. An allocation of felicity conditions to common ground, however, would demand the complex phenomenon to be conceived of in a context-dependent manner. In the following, personal common ground and communal common ground (Clark, 1996), conversational record (Thomason, 1992), mutual manifestness (Sperber and Wilson, 1995), and individual dialogue common ground and collective dialogue common ground (Fetzer, 2002, 2004) are compared and contrasted, and then connected with context.

2.1. Personal common ground and communal common ground Clark’s approach to communication is both social and cognitive, and as a consequence of that, his conception of common ground is social and cognitive, or to use his own words: “two people’s common ground is, in effect, the sum of their mutual, common, or joint knowledge, beliefs and suppositions” (Clark, 1996: 93). Common ground is generally represented as shared basis, as reflexive and iterated propositions, and as beliefs and assumptions. What is of relevance to the investigation of natural-language communication is the fact that without presupposed common ground, viz. a common domain to which symbols, indices, icons and contextual constructions are anchored, communication as signaling meaning for the addressee would not be possible. Moreover, joint utterances, the minimal joint project of an adjacency pair and other joint projects are only possible against the premise that there exists common ground and that coparticipants have access to their common grounds: The suggestion is that people are ordinarily able to justify their common ground. They believe or assume each piece has a basis that meets the requirements for CG-shared: The principle of justification. In practice, people take a proposition to be common ground in a community only when they believe they have a proper shared basis for the proposition in that community (Clark, 1996: 96). Implicit in the principle of justification is Clark’s context-dependent differentiation between an individual’s personal common ground and their communal common ground: the former is primarily intrapersonally oriented and the latter is primarily interpersonally oriented. The two types of common ground are interconnected and constantly updated and, if necessary, revised. For instance, the felicity of referring and the allocation of the referent to the communal

162 Common Grounds common ground through the act of grounding is a necessary condition for a contribution to be felicitous: “to become a contribution, the utterance has to be mutually accepted before the initiation of the next contribution, and that process may require repairs, expansions, and replacements of all or part of the initial presentation” (Clark and Wilkes-Gibbs, 1992: 490). As has been shown, common ground is an extremely general notion, which, if applied to actual language use, may cause a lot of difficulties because of its dual status as cognitive and social, and as individual-oriented and collective-oriented. Regarding their connectedness, the intrapersonal and individual-oriented conception of common ground manifests itself in the performance of joint action through language, as is convincingly argued for by Goody: Another way of putting this social reality of language is to note that a Secondary Representational Stem does far more than multiply the power of the individual’s primary representational system; it gives every individual the power to enter into the linguistic representation of others, and to use these shared secondary representations to model cognitively the understandings as well as the intentions of others. (Goody, 1995: 6) The dual status of common ground as cognitive and social is not restricted to the domains of knowledge, beliefs and assumptions. It is also reflected in the domain of cognitive operations, namely in individual intelligence and in social intelligence. The cognitive and social domains of common ground depart with regard to cognition and they meet in the performance of coordinated verbal and non-verbal social actions.

2.2. Mutual manifestness In line with those approaches which look at a contextualized notion of common ground, relevance theory (Sperber and Wilson, 1995) refers to its contextualized version as mutual manifestness. It adapts the philosophical construct of common ground to the constraints and requirements of a cognitive framework, where the domain of validity of infinitely regressive knowledge or common knowledge (Lewis, 1969) and of mutual knowledge (Schiffer, 1972) is given a more restricted interpretation and assigned the status of mutual manifestness which is context-dependent by definition. In relevance theory, context is a psychological construct which contains a subset of the hearer’s assumptions about the world which are used as a set of premises for the interpretation of an utterance. Unlike the philosophical construct of common knowledge, the relevance-theoretic construct of manifestness is defined as something that can be represented mentally and which is accepted as true or accepted as probably true by an individual. For this reason, the relevance-theoretic concept of ‘manifest’ is weaker than the philosophy-based notions of ‘assume’ and ‘know’. Moreover, manifestness is conceived of as a scalar notion. This is the reason why degrees of manifestness need to be accommodated as well. Manifestness as such is an individual-anchored concept and therefore of no immediate relevance to communication. In order to be felicitously integrated into a theory of communication, however, manifestness requires a collective interpretation. Individual-centred manifestness is thus supplemented with higher-order propositions and higher-order assumptions which are

Reformulation and Common Grounds 163 intended to represent a subset of collective-oriented mutual manifestness anchored to mutual cognitive environments.

2.3. Conversational record A conception of communication as coordinating action is also implicit in Thomason’s approach to communicative intentionality (Thomason, 1992). According to him, speaker meaning is a sort of coordination-oriented intention. It is intrinsically connected with Lewis’s conception of accommodation (Lewis, 1969) which “consists in acting to remove obstacles to the achievement of desires or goals that we attribute to others” (Thomason, 1992: 332). To account for a coparticipant’s communicative act, he/she is attributed a communicative intention which does not only make his/her goals and attitudes explicit, but also fit into the collective communicative intention. The attribution of intentions, goals and desires to other is intrinsically connected with Garfinkel’s practical-reasoning strategies of ‘unless’ and ‘ad hocing’ (Garfinkel, 1994: 21). The former regulates the acceptance of a social action under some specified conditions, and the latter makes allowances for the odd-one-out scenario so that it can also be accepted. Against this background, Thomason draws the conclusion “that to accommodate accommodation in pragmatics, we must relax the conception that prevails in syntax of rules or generalizations. If there is such a thing as accommodation, genuine pragmatic rules will be routinely violated, and these violations will not seem anomalous; they will achieve an effect that seems natural and normal” (Thomason, 1992: 333). Hence the production and interpretation of a communicative contribution is a context-dependent endeavor par excellence. Not only are contributions compared and contrasted with prior contributions, they are also attributed to a public conversational record, which contains information that is public and, for this reason, must have been ratified by the coparticipants. The conversational record additionally contains presumptions and an update operator. If communication is looked upon from that perspective, coparticipants do not simply exchange contributions and information but rather “are attempting to work on each other. I have said that I would like to replace this with a slightly different picture, according to which they are working together to build a shared data structure” (Thomason, 1992: 339). Naturally, coparticipants are not fully aware of their joint construction of a conversational record, or to use Thomason’s own words: I grant that in most cases that we’d want to describe as “conversation”, there won’t be anything like a literally shared memory. Normally, there will be no shared conversational record; each conversant will have its own representation of the record. But it is normal for conversants to assume that the representation is shared, for each conversant to treat its representation as of the record as if it were a public object. Unless danger signals are perceived, a conversant won’t distinguish between its representation of the record and its interlocutors’ representation. (Thomason, 1992: 339) It is at that ‘danger-signal’ stage where conversational records and reformulation meet, and where the individual coparticipants’ public conversational records depart. The following context-dependent conception of individual and collective dialogue common grounds accommodates that differentiation.

164 Common Grounds

2.4. Individual dialogue common ground and collective dialogue common ground The concepts of conversational record and of personal common ground and communal common ground can be further refined with regard to their participant orientation and their embeddedness in context. Fetzer (2002, 2004) examines particular subsets of common ground, namely context-dependent dialogue common grounds. They capture and administer information which is exchanged in one encounter, communicative project or dialogue by explicitly connecting context-dependent dialogue common ground with the administration of other dialogue common grounds and with the administration of common ground in general. From a functional perspective, dialogue common ground can be equated with Thomason’s concept of a public conversational record (Thomason, 1992), where the dialogue-specific information is stored. Regarding its form, however, it subcategorizes into an individual dialogue common ground and a collective dialogue common ground. Dialogue common ground is a dynamic sociocognitive construct, which administers different types of information exchanged in a communicative encounter. An individual dialogue common ground contains and administers the information of one individual’s dialogue with respect to force, content, formulation, context, background information and its propositions, beliefs and interpersonal and interactional presuppositions thus capturing an individual’s processing and interpretation of information. The information processed is connected with an individual’s possible side activities and may interact with the processing of that type of information. For this reason, the coparticipants’ individual dialogue common grounds may differ, sometimes even to a large extent. In addition to each coparticipant’s construction of their individual dialogue common ground, they also construct a collective dialogue common ground which is anchored to the set of coparticipants and to the set of their ratified contributions. The collective dialogue common ground contains and administers set-specific information, such as collective we-intentions, collective communicative goals, collective inferencing strategies and collective coherence (Fetzer, 2002, 2004). In communication, the set-specific values function as a filter, which constrains and canalizes what-has-been-said accordingly by guiding the interpretation and production of intersubjective meaning in the prespecified manner. Collective dialogue common ground intersects with individual dialogue common ground, with other individual dialogue common grounds, and with common ground. In natural-language communication, the individual and collective dialogue common grounds are permanently updated, and all of the postulated contributions and their presuppositions are allocated to the individual dialogue common grounds, and all of the ratified contributions and their ratified presuppositions are allocated to the collective dialogue common grounds. The latter are expected to be almost identical for the coparticipants because of the necessary condition of ratification. Naturally, the collective dialogue common grounds are permanently updated. In the case of a ratified, but negotiated contribution, they are revised. Furthermore, the collective dialogue common grounds serve as a foundation for the coparticipants’ inferencing processes for both assertional implicatures and background implicatures, to employ Thomason’s terms (Thomason, 1992: 352). Because of their set-specific status, ratified presuppositions are assigned a co-suppositional status. Analogously to individual and collective dialogue common grounds, presuppositions and co-suppositions subcategorize along similar lines. As regards the validity of dialogue common ground, it is only valid in the restricted domain of one particular dialogue, but may, if strengthened through ratification in

Reformulation and Common Grounds 165 other dialogues, become valid in a more extended frame of reference. Unlike presuppositions, co-suppositions administer information about something which is the case and about something which is not the case. Against this background, co-suppositions are only valid in the restricted domain of one particular dialogue, but may, if strengthened through ratification in other dialogues, become valid in a more extended frame of reference. Because of the dynamic conception of common ground and its subcategorization into individual and collective dialogue common grounds, the internal organization of common ground and its subcategories is anchored to a network-based scenario with parallel distributed processing (Searle, 1995, 1999). Common ground, individual and collective dialogue common grounds, personal and communal common grounds, conversational records and mutual manifestness seem to capture almost all of the information exchanged in communication. So, are they functionally equivalent to context?

2.5. Common ground and context In spite of the fact that common ground and context are frequently compared and contrasted (Akman et al., 2001; Bouquet et al., 1999), both are extremely hard to come to terms with. The above-explicated inherently dynamic conception of common ground is also reflected in the notion of context, which subcategorizes into cognitive, linguistic, social and sociocultural contexts (Fetzer, 2004). Against this background, common ground and context meet in the process of administering common ground, and they depart once the product of that process has been reached. They also depart with regard to the administration of sociocultural context, which in the domain of common ground is conceived of in a sociocognitive manner. While social context is seen as given, as is reflected in the presuppositional approach to context, it is also conceived of as interactionally organized, which manifests itself in the joint construction of social contexts (Schegloff, 1991). Here, the connectedness between context and lexical expression and object is explicated by the process of contextualization, in and through which meaning is constructed. A more precise account of social context is offered by the research paradigm of ethnomethodology. It is based on the premise of indexicality of social action and examines the questions of how separate individuals are able to know or act within a common world, and of how members negotiate or achieve a common context. That type of common context is synonymous with social context, which is further distinguished with regard to micro- (or local) social context and macro- (or global) social context. Against this background, common ground is also a context-dependent phenomenon. In the following, the connectedness between reformulation and dialogue is examined.

3. REFORMULATION AND DIALOGUE Natural-language communication comprises the constitutive parts of coparticipants and their presuppositions, such as intentionality and rationality, it comprises the constitutive parts of language and its presuppositions, such as linguistic, communicative and sociocultural competence, and it comprises the constitutive parts of context, for instance a situation in which language is produced and interpreted. Yet there is more to natural-language communication than a purely

166 Common Grounds additive perspective in which a speaker produces a contribution which the hearer interprets. From a holistic viewpoint, natural-language communication is interaction and dialogue, and that is the reason why an appropriate definition of dialogue needs to go beyond the simplistic addition of the necessary and sufficient conditions required for the definition of individual coparticipants. Rather, dialogue requires a frame of reference which accounts for the individual coparticipants and their presuppositions, and for the set of coparticipants and their set-specific presuppositions. The parts-whole perspective has the consequence that the set-specific presuppositions can never be identical with the sum of the individual-anchored presuppositions. This is because the set-specific presuppositions additionally contain intersection-specific presuppositions anchored to the set of coparticipants who have particular set-specific beliefs. Against this background, the linguistic system and its subsystems of phonology, morphology, syntax, semantics and the lexicon, and pragmatics are connected with each other and with other cognitive systems, such as individual cognition and social cognition, and emotional intelligence and social intelligence, and all of them are connected with a social world (Habermas, 1987). In that type of scenario, lexical expressions and grammatical constructions are intrinsically connected with the linguistic system, with the cognitive system and a social world, to which they refer and which they reconstruct. As a necessary consequence of the triadic connectedness, the linguistic system and the cognitive system have access to a social-world notion of common ground, and both contribute to it by adding new information of the world and new information about the world while at the same time strengthening and restructuring the taken-for-granted information. So what does a reformulation do in the network-based scenario? Reformulations tend to occur primarily in spoken language, where they are looked upon as prototypical representatives for a joint (or distributed) production and joint (or distributed) interpretation of dialogue. Since the production and interpretation of intersubjective meaning are cognitive operations, reformulations are intrinsically connected with the domain of social intelligence, such as Brown’s (1995) interactive thinking and Goody’s (1995) anticipatory interactive planning. They signify that a contribution is ratified, but not seen “such as is required, at the stage at which it occurs, by the accepted purpose of direction of the talk exchange” (Grice, 1975: 45). The contribution’s status as not being fully in accordance with the constraints and requirements of the Gricean cooperative principle may be due to one or more of its constitutive parts. From a sequential-organization viewpoint, a reformulation signifies that coparticipants should orient themselves backward, and for this reason, reformulation is an anaphoric device par excellence. But there is more to reformulation than a simple backwardpointing function. From an information-management perspective, it indicates a disruption in the process of attributing information to the collective dialogue common ground, and from a communicative-action viewpoint, a reformulation requests the coparticipant, should he/she intend to have his/her contribution and all of its constitutive parts accepted, to provide reasons why the original contribution and all of its constitutive parts should be accepted, or to accept the modified reformulated contribution. Only then is it possible to repair the disruption and to continue with the default scenario.

3.1. Reformulation and ratification The holistic nature of dialogue has been under investigation in different research paradigms, for instance dialogue analysis (Linell, 1998), cognitive linguistics and psychology (Clark, 1996;

Reformulation and Common Grounds 167 Trognan, 2002), artificial intelligence (Litman and Allen, 1992) and philosophy (Habermas, 1987; Searle, 1999), to name but a few. All of the approaches to dialogue accommodate a higher-level goal, if not a macro-oriented perspective, and all of them explicitly accommodate an in-between move allocated to the transition between a first move postulated by coparticipant 1 and a second move postulated by coparticipant2 . The in-between move is referred to as ratification (Habermas, 1987), as enactment of ratification (Trognan, 2002) and as principle of mutual responsibility (Clark and Wilkes-Gibbs, 1992). In his theory of communicative action (Habermas, 1987), Habermas examines the necessary and sufficient conditions for communication1 . He looks upon communication in an idealistic framework, in which rational coparticipants postulate validity claims and direct them at their fellow coparticipants, who ratify them through an acceptance, rejection or neutral stance. Unlike the then monological conception of a speech act (Austin, 1980, Searle, 1969), Habermas’s validity claim is dynamic, truly dialogical and both process and product. In his frame of reference, a postulated validity claim does not constitute a validity claim per se because the postulated claim requires another coparticipant’s ratification through an acceptance, rejection or neutral stance. Hence the act of ratification is a necessary condition for the postulated claim to be assigned the status of a validity claim. The triadic conception of communication is of immediate relevance to the investigation of reformulation. It is at the stage of ratification where a reformulation is anchored to. At that stage, a coparticipant signifies that they consider the prior contribution not as is required, which may be due to minimally one of its constituent parts of force, content, formulation, context or background information and propositions, beliefs or interpersonal and interactional presuppositions. It is the attested incongruence between coparticipant1 ’s postulated validity claim and coparticipant2 ’s ratified validity claim, which is a necessary condition for a reformulation. Of course, a reformulation requires ratification in dialogue, and for this reason, it initiates a reformulation sequence in which some kind of agreement about the status of the incongruent parts needs to be reached. Like Habermas, Trognan (2002) employs a tripartite scenario. He focuses on personal understanding and mutual understanding, and connects the two with interactional processes. He captures the sequence of a first move, an in-between move, which he calls enactment of ratification, and a second move as follows: The conversational mechanism of mutual understanding (  ) is made up of at least two speakers (L1 and L2), respectively first and second speakers, and the three speech turns: T1, T2 and T3, successively distributed in the following manner: T1: L1 T2: L2 T 3: L3 (Trognan, 2002: 128) Trognan supplements the structural analysis with a functional interpretation in which he explicitly connects his notion of enactment of ratification with the process and product of interpretation and with the process and product of assessment:

1

Fetzer (2002, 2004) has recontextualized Habermas’s macro framework and adapted it to the contextual constraints and requirements of the micro domain.

168 Common Grounds (T1, T2) forms an interpretation relationship. Its second element enacts L2’s interpretation of the action performed by L1 in T1. L2 thus makes this interpretation mutually obvious (…) ((T1, T2), T3) forms an assessment relationship. L2’s interpretation of L1 being available to L1 in T2, he can compare it with his own interpretation and enact a ratification (…) if the two interpretations correspond, that is to say, if L1’s interpretation of this initial utterance is equivalent to that enacted by L2 in T2. He can also reformulate T1 (…) if the two interpretations diverge, that is to say if L1’s interpretation of his initial utterance is not equivalent to that enacted by L2 in T2. However, in both cases, L2’s interpretation of T1 has been shared between L1 and L2. L1 knows it in T2 and L2 knows that L1 knows it in T3. (Trognan, 2002: 129) In his analysis of mutual understanding, Trognan refers to the possibility of reformulation, which, according to him, is anchored to two diverging interpretations, which may be due to their content, force or formulation. Both Habermas and Trognan go beyond an individualcentred conception of speech act, communicative contribution and validity claim by explicitly accommodating a collective-oriented category which Trognan calls assessment relationship and which Habermas calls validity claim proper. Both notions are collective categories because they presuppose a tripartite setting anchored to individual coparticipants as well as to the set of coparticipants. A collective-oriented category anchored to a tripartite setting is also implicit in Clark and Wilkes-Gibbs’ principle of mutual responsibility, according to which “the participants in a conversation try to establish, roughly by the initiation of each new contribution, the mutual belief that the listeners have understood what the speaker meant in the last utterance to be a criterion for current purposes” (Clark and Wilkes-Gibbs, 1992: 488). The principle of mutual responsibility does not only obtain for the contribution as a whole, but also for the constitutive act of referring: “(1) The reference is expressed linguistically with one of the three standard types of noun phrase – a proper noun (…), a definite description (…), or a pronoun (…). (2) The speaker uses the noun phrase intending the addressee to be able to identify the referent uniquely against their common ground. (3) The speaker satisfies her intention simply by issuing that noun phrase. And (4) the course of the process is controlled by the speaker alone” (Clark and Wilkes-Gibbs, 1992: 465). Unlike the product- and speaker-oriented interpretation of referring, Clark and Wilkes-Gibbs look upon the act of referring as a threefold endeavor. For this reason, referring cannot consist of one single act anchored to one single coparticipant. Instead, the process of referring is seen as initiating a reference by the speaker and passing a judgment on presentation by the hearer, which can be an acceptance, a rejection or a postponement of judgment. The principle of mutual responsibility represents the weakest version of the act of ratification because it only refers to the act of understanding and therefore does not explicitly accommodate the contribution’s action-based consequences for the ongoing discourse. In all of the frameworks examined, the act of ratification, the enactment of ratification and the principle of mutual responsibility entail neither the acceptance nor the rejection of the prior move formulated and performed by coparticipant1 . Rather, they signify that the move has been processed and that an uptake is to follow, but what kind of uptake is to follow? Extending Sbis`a’s (2002) claim that speech acts are assumed to be felicitous by default, it is assumed that the default scenario does not only apply to the formulation and postulation

Reformulation and Common Grounds 169 of a validity claim, but also to its ratification. Thus, a contribution is assumed to be felicitous by default, and it is assumed to be ratified by default. But not only is a felicitous contribution ratified by default, its ratification is also assumed to be felicitous. To take the argumentation one move further, it is not only assumed that the contribution is felicitous and felicitously ratified by default, but it is also assumed that the contribution is ratified through an acceptance unless indicated otherwise. An acceptance manifests itself when a discourse topic is continued, when an assessment is agreed with or disagreed with, when an offer or invitation is accepted or rejected, for instance. Thus, a felicitous ratification through an acceptance is not equivalent to the speech act of acceptance as it realizes both an agreement with the contribution’s content, force and formulation, and with its presuppositions and presupposed common ground. Against this background, a reformulation is allocated to a non-default scenario, in which a contribution is felicitously ratified, but not ratified through an acceptance. It signifies that the contribution is not “such as is required” and for this reason needs to be modified and reformulated if the coparticipant intends to have it ratified with an acceptance. But what requires modification and reformulation: the force, the content, the formulation, context, or background information and propositions, beliefs, interpersonal and interactional presuppositions? Who is requested to reformulate the contribution: self or other? And who performs the act of reformulation: self or other?

3.2. Reformulation and common ground The metapragmatic device of reformulation has been examined in the fields of pragmatics and sociopragmatics with regard to the interactional organization of status, power and social control (Bührig, 1996; Sarangi and Slembrouck, 1996; Thomas, 1984, 1985). Depending on its coparticipant-orientation, reformulation differentiates into self-reformulation, in and through which self reformulates what they have been saying, and other-reformulation, where other reformulates what self was saying. While self-reformulation is of no immediate relevance to self for gaining status and social control, other-reformulation has the function of enhancing self’s status in interaction. Reformulation is also distinguished with regard to the question of who initiates the reformulation, viz. self-initiated reformulation and other-initiated reformulation. Again, self-initiated reformulation is of no immediate relevance to social control, whereas other-initiated reformulation has the function of enhancing the status of other. Furthermore, reformulation is distinguished with regard to the reformulation of a contribution’s content and with regard to the reformulation of a contribution’s force (Fetzer, 1994; Jucker, 1986). Since the majority of investigations has concentrated on the perlocutionary effects of a reformulation, they have not explicitly accounted for the connectedness between a reformulation and common ground. In natural-language communication the production and interpretation of a contribution are performed in an intentional manner, which does, however, not mean that coparticipants are fully aware of every single process and of every single operation involved (Searle, 1999). Rather, the processes are highly automatized operations (Givón, 1999; Goody, 1995), and automatization is assigned the status of default. As a consequence of that, automatization is only queried if coordination and communication problems arise. As has been discussed above, the production and interpretation of a contribution in dialogue entails the coparticipants’ retrieval of information from common ground, and it entails the coparticipants’ attribution of

170 Common Grounds information to dialogue common ground. Naturally, common ground and its subset dialogue common ground are not represented as unstructured wholes but rather as dynamic networkbased scenarios, which are administered by the coparticipants in a highly automatized manner. Since reformulation is attributed to a non-default scenario, it disrupts a dialogue’s inherent automatized operations and brings the process of producing, interpreting and administrating contributions to a halt. The disruption may be due to an inappropriateness of a contribution’s force, content, formulation, context, background information and propositions, beliefs, interpersonal or interactional presuppositions. In spite of all the different sources of incongruence, contributions which require reformulation are composed of grammatical constructions and lexical expressions. And it is one or more of the lexical expressions or grammatical constructions, which are seen as inappropriate devices for an appropriate representation of one or more of its constitutive parts. Against this background, the act of reformulation is triggered by one or more lexical expressions or by one or more grammatical constructions whose connectedness with force, content, formulation, context, background information and propositions, beliefs, interpersonal or interactional presuppositions is evaluated as incongruent. Because of its incongruent status, the information communicated by the expression or construction cannot be attributed to the collective dialogue common ground unless it is reformulated in an acceptable manner. Through the explicit accommodation of a lexical form and its function, the prevailing conception of the metapragmatic device of reformulation does not only need to be examined with respect to perlocutionary effects, but also with respect to information management, where reformulation has a direct impact on the administration of dialogue common ground and where it is assigned the status of a reformulatory intervention with the goal of reorganizing the collective dialogue common ground. At the same time, a reformulation can have an indirect impact on the superordinate set of common ground, which might also require some modification if the reformulatory intervention’s force is strong enough. For the administration of common ground and of dialogue common ground it is of relevance that a reformulation is not primarily anchored to the domain of objective information in the propositional format. Rather, it also needs to be anchored to the transmission of sociocultural and social domains of communication so that it can ensure that a coparticipant’s interpretation of a contribution is not idiosyncratic, but rather in accordance with its sociocultural communicative constraints. The distinction between the context-independent notion of common ground and the context-dependent notion of dialogue common ground, which categorizes into the sociocognitive categories of collective dialogue common ground and individual dialogue common ground, allows for a more precise account of the process and product of the negotiation of intersubjective meaning, and of the process and product of reformulating meaning. It illustrates the administration of common ground with regard to the dialogical interactions within an individual and their individual dialogue common ground, and it illustrates the dialogical interactions between the coparticipants, their individual common grounds and their collective dialogue common grounds. Regarding the status of reformulation in the refined frame of reference, a self-reformulation indexes the individual dialogue common ground and signifies that self cannot attribute their contribution to the collective dialogue common ground. An other-reformulation, by contrast, indicates that self does not intend to attribute the other’s contribution to the collective dialogue common ground. Naturally, the notions of individual and collective dialogue common grounds are connected with higher-order types of common ground, such as knowledge- and domain-specific common grounds (cf. Okada, this volume,

Reformulation and Common Grounds 171 Rodríguez, this volume). The different categories of common ground, which are fuzzy rather than discrete in nature, are structured in a network scenario, where they interact with each other in a permanent manner by adding new information, by strengthening or weakening the status of already known information or by restructuring already-taken-for-granted information. To accommodate that change of perspective, a reformulation is used not only as a metapragmatic device but also as a lexical strategy, which is examined in the following.

3.3. Reformulation as a lexical strategy In communication, a coparticipant selects lexical expressions and grammatical constructions to construct a contribution whose function is to transport their communicative intentions. In spite of the fact that selection and construction are highly automatized operations, coparticipants can account for their selections, should the communicative need arise and their choice be queried, as is implicit in the ethnomethodological principle of accountability of social action (Garfinkel, 1994). As contributions are not only constructed and postulated but also ratified through an acceptance or rejection in dialogue, they may need to be repaired, modified or reformulated by self or by other. With a self-reformulation, self presents a refined formulation of their contribution’s content or force. The argumentation also holds for an other-reformulation in which self reformulates the other’s contribution and at the same time requests them to either accept the other-reformulation or to present a further reformulation in which content or force are reformulated more appropriately. Against this background, a felicitous reformulation is only possible if language and language use are no longer conceived of as autonomous entities, but rather as network-anchored faculties which are connected with the cognitive system and its constitutive parts (Allen and Seidenberg, 1999; Givón, 1999): language and language use are connected through coparticipants who use language strategically to realize their communicative intentions in context and to achieve one or more communicative goals in context (Fetzer, 2002, 2004). The integrated approach requires a rethinking of the basic premises of grammatical construction and lexical form in which their relational statuses are explicitly accounted for. The path adopted here argues for a rethinking of lexical form as lexical marker, and of communicative strategy as lexical strategy in particularized contexts. So, what are the necessary and sufficient conditions for a lexical form to be assigned the status of a lexical marker, and what are the necessary and sufficient conditions for one or more syntagma-anchored lexical forms to be assigned the status of a lexical strategy? A network-scenario-based conception of the linguistic system connected with the cognitive system and with the social world is a necessary prerequisite for the linguistic-surface phenomenon of a lexical form to be assigned the status of a lexical marker, and for a number of syntagma-anchored lexical forms to be assigned the status of a lexical strategy. For a lexical form to be assigned the status of a lexical marker, it must have undergone a process of grammaticalization (Brinton, 1996) or a process of subjectification (Traugott, 1995), in which the number of possible inferences, possible semantic collocations, possible morpho-syntactic variations and possible syntactic and discursive positions have been constrained. To be more precise, a lexical marker is more constrained than a lexical form with regard to distribution and therefore allows less variability. Because of the fixation of syntactic and discursive positions,

172 Common Grounds a lexical marker does no longer solely operate on the semantic level, but rather on the intersubjective level of meaning. In spite of its status as a marker of multiple functions, a lexical marker still needs to carry some of its original meaning. In relevance-theoretic terminology (Sperber and Wilson, 1995), it carries conceptual meaning and procedural meaning, as is the case with the prototypical lexical marker of common ground you know or German doch (‘but’) (cf. Fischer, this volume, Pittner, this volume) whose function is to index the domain of taken-for-granted, shared knowledge. The same line of reasoning is true for a number of syntagma-anchored lexical forms which are assigned the status of a lexical strategy. Unlike a communicative strategy, such as the communicative strategy of requesting someone to pass the salt, which can be realized by numerous linguistic surfaces, a lexical strategy, such as reformulation is anchored to a relatively fixed set of lexical forms which are syntagmatically connected in a relatively fixed order. The communicative meaning of that syntagmatic structure is fairly explicit, and that is the reason why a reformulation can be assigned neither the status of a routine formula, nor the status of an idiomatic expression. It is, however, a prefabricated structure (Erman and Warren, 2000) which signifies that the reformulated contribution is more in accordance with the Gricean cooperative principle and its maxims than the original one. Against this background, the syntagmatic structure I mean, which is considered as a self-reformulation, may count as the metapragmatic device of reformulation and it may count as a communicative strategy of reformulation, but it does not count as a lexical strategy of reformulation. This is due to the fact that the syntagma-anchored lexical forms do not only have the communicative goal of introducing the reformulated content and/or the reformulated force. I mean also signifies and introduces an explication-of-pragmatic-presupposition sequence which provides an account to make the contribution more acceptable and more appropriate for the fellow coparticipants. The canonical form of a reformulation depends on its participant-orientation. A prototypical self-reformulation, which is self-centred, is what I am saying is or I am saying that   , and a prototypical other-reformulation, which is other-centred, is what you are saying is or are you saying that   . Their communicative meanings with respect to the interactional organization of status, power and topic control is connected with a further distinction, namely whether the reformulation is self-initiated or whether it is other-initiated: other-initiated reformulations support self’s topic control and self’s status and power management, while self-initiated reformulations are generally employed in the actual process of negotiating meaning. In order to be felicitous, a reformulation requires the following constitutive parts: 1. An explicit reference to a contribution’s informational content which is reformulated, for instance an explicit quotation of the content, the indexical expressions what or that, or other lexical expressions from the semantic field of informational content, such as the point, the position or the issue. If the force of a contribution is reformulated, the content marker is replaced by an explicit reference to force, such as the explicit quotation of the force, the indexical expression how or other lexical expressions from the semantic field of illocutionary force, such as what manner. 2. A reference to the coparticipant who is reformulated, for instance the indexical expressions you or I, or the nominal references Mr Blair or the spokesperson. 3. A lexicalized verb denoting the activity of doing or saying, such as say, mean, assume or claim.

Reformulation and Common Grounds 173 Due to the overarching principle of sociolinguistic variation (Brown, 1995), the lexical strategy of reformulation can employ a repertoire of lexical forms denoting the abstract notions of informational content and activity of speaking. But it is not only informational content and illocutionary force which are reformulated. The discursive and pragmatic presuppositions of a contribution, its sincerity or truth, and the speaker-intended meaning can also be reformulated. These reformulations display domain-specific references to how the informational content has been formulated and how it has been speaker-intended or meant. The lexical strategy of reformulation has only one illocutionary goal, namely to have a contribution formulated in accordance with the Gricean cooperative principle and its maxims so that it can be attributed to the collective dialogue common ground. Its perlocutionary effects, however, are multidimensional and generally refer to the social dimensions of language use, namely status and power management and topic control. Reformulations are a constitutive part of natural-language communication because, following Levinson, “linguistic communication is fundamentally parasitic (   ): no-one says what they mean, and indeed they couldn’t – the specificity and detail of ordinary communicated contents lies beyond the capabilities of the linguistic channel: speech is a much too slow and semantically undifferentiated medium to fill that role alone” (Levinson, 1995: 238). Thus, communication is a context-dependent activity which is not only anchored to the local and global contexts of an exchange, but also to the implicit taken-for-granted common ground. In the following, the functions of reformulation are examined in the corpus of twelve dyadic political interviews, which have been selected because the information exchanged and reformulated is public knowledge and therefore should be accessible to the general public. They all stem from the program on the record (BBC1) and were recorded and transcribed in 1990.

4. REFORMULATIONS IN DIALOGUE In natural-language communication reformulation is of great importance to the coparticipants’ administration of individual dialogue common ground and collective dialogue common ground. Reformulations occur in institutional and non-institutional contexts, but tend to be restricted to spoken interactions, such as dyadic and multiparty face-to-face encounters, telephone conversations or interviews. From an information-exchange perspective, a reformulation is a necessary condition for securing felicitous communication, while from an interpersonal perspective, it carries some face-threatening potential. Regarding their perlocutionary effects, reformulations are key to social control, where they serve as a means of control with respect to local and global discourse topics thus enhancing a coparticipant’s local social status in order to gain global power. Unlike a lexical marker of common ground, such as you know, which indexes common ground, the lexical strategy of reformulation signifies that a prior contribution is assigned only a preliminary status and therefore is not attributed to the collective dialogue common ground. In the following, the form and function of reformulation are examined in twelve dyadic political interviews, in which the lexical strategy of self- and other-reformulation is used to secure a common conversational implicature, to secure the collective dialogue common ground and to secure a common perlocutionary effect.

174 Common Grounds

4.1. Reformulating to secure a common conversational implicature In the context of a dyadic political interview, the majority of self- and other-reformulations are used to secure a common conversational implicature. This is of particular importance because the conversational implicatures are not only calculated by the direct coparticipants of interviewer (IR) and interviewee, but also by the audience in front of their TV sets at home (Fetzer, 2000) who are the main target of the information exchanged. Regarding their distribution, self- and other-reformulations occur in the initial, medial and final positions. The initial position seems to be the preferred variant for realizing both a self- and an otherreformulation, in and through which the conversationally implicated meaning is made explicit. Since a reformulation signifies a non-default scenario, the implicitly realized information is not ratified through an acceptance. However, in order to reject implicit information, the implicit has to be made explicit. Only then is it possible to negotiate its communicative status and to reject it (Fetzer, 1999). The form and function of a reformulation is illustrated with the following six examples. Examples (1) and (2) stem from an interview between Jonathan Dimbleby (IR) and John Gummer (JG: 20/05/1990). Examples (3) and (4) are adopted from an interview between Jonathan Dimbleby and Michael Heseltine (MH: 13/05/1990). Example (5) is from an interview between Jonathan Dimbleby and Collin Moynihan (CM: 03/06/1990), examples (6) and (7) stem from an interview between Jonathan Dimbleby and Tony Blair (TB: 27/05/1990), and examples (8) and (9) are adopted from an interview between Jonathan Dimbleby and David Trippier (DT: 17/06/1990). All of them use the canonical form of a reformulation, which is constructed by a lexical verb denoting the activity of speaking, that is say, ask, do, make a point and mean, which collocates with a marker of information, viz. that, a reference to the direct coparticipant you, we or I, and a reference to the informational content, that is what, that and the point, or the force, which is realized by the progressive aspect and a direct reference to other, namely you: (1) IR are you saying therefore that erm the meat is safe is not the same as saying I can absolutely guarantee unequivocally   (2) JG what I am saying is that I so strongly believe and my scientists and medical people have made it clear that beef is safe to eat   (3) IR now does that mean that if in the weeks and months ahead erm conservative MPs were to come to you and say Michael we don’t think we can win the next election without you as our leader you’ve got to run do you politely show them the door (4) MH oh are you asking a backbencher to accept collective responsibility there’s no backbencher who does that (5) CM the point I was making at that stage was that I did believe and I still do and it’s the FA who enters the team and it’s the FIFA who decide with the Italian authorities where ever they want erm the countries to continue to play   (6) TB what I’m saying to you is that the policy behind it is   In all of the examples, the reformulation is realized in the initial position and refers to the content of a prior contribution, which has not been presented explicitly. In (1), the IR spells out JG’s implicature that the BSE crisis does not have any impact on the safety of beef, and asks him to ratify that version of the implicature. In (2), JG modifies the IR’s version

Reformulation and Common Grounds 175 and explicates his speaker-intended version, namely that beef is safe. Analogously to (1), the IR requests MH in (3) to ratify the interviewer’s inferred implicature, namely that he, the interviewee, does not intend to run against the prime minister. In (4), which stems from a rather controversial interview, MH spells out the IR’s conversational implicature while at the same time rejecting its implicatum. In that particular context, the other-reformulation is assigned the status of a rhetorical question. In (5) and (6), the politicians explicate their speaker-intended meanings, namely that it is not Britain who decides whether a team will stay or leave in (5), and that the policy behind a particular issue is different in (6). While the interviewer’s other-reformulations are primarily directed toward the audience and their calculation of a conversational implicature thus signifying his responsibility for the felicitous construction of a collective dialogue common ground, the politicians’ self-reformulations are used to make explicit their speaker-intended meanings by reformulating their previous contributions in order to make them more appropriate. By making the relevant presuppositions and implicatures explicit, the interviewer requests the politicians to ratify his reformulation of their contributions. While the interviewer does not generally reformulate his own contributions, the politicians use self-reformulations to counter implicit accusations for having been evasive, such as JG in (2) and MH in (4), or to strengthen their argumentation, such as CM in (5) and TB in (6). In spite of the fact that those reformulations are employed in order to secure a common conversational implicature and thus a collective dialogue common ground, their prime function in media discourse seems to be anchored to the perlocutionary domain of topic management and social control. Reformulations to secure a common conversational implicature are predominantly realized in the initial positions. They occur in those contexts in which the implicit which has been taken for granted and is now reformulated in order to secure a common interpretation of the conversational implicature. If a reformulation is realized in the medial position, as is the case in (7), it is also used to secure a common conversational implicature. Because of its embeddedness in a contribution, the processes required for the calculation of the implicature are also made explicit. Thus, more common ground is made explicit in a medial reformulation than in an initial one. In the following example, TB does not only make explicit the underlying premises of his argumentation, but also disambiguates a possible indeterminateness anchored to his prior contributions: (7) TB of course they can’t legitimately come out you might as well say at the moment under the law if employees think they’ve got a primary dispute with their employer and in fact they’ve got a secondary dispute they come out first and then they decide afterwards the court decides afterwards whether it was a primary dispute or not what I’m saying to you is that they cannot come out in those circumstances   In the data investigated, other-reformulations are also realized in the final position, which may be due to genre-specific constraints and requirements of a political interview and to the fact that an implicit request for ratification, which is a constitutive part of an other-reformulation, seems to carry a stronger force when realized finally: (8) IR is that the price that would have to be paid to meet the 2000 target is that what you’re saying

176 Common Grounds (9)

IR yes but yes but let me put back to you two of the points that you’ve made first of all the conditionality of the commitment if says the prime minister other countries do it now the Germans have made their commitment unequivocal are we saying that we will not unless the Americans do it

In (8) and (9), the IR requests the politician to ratify his inferred implicatures, namely that there is a particular price to be paid for reducing the emissions of CO2 in (8), and that Britain’s decision depends on the US in (9). Because of the reformulation’s final position, the politician’s ratification of the reformulation is seen as a direct consequence of being challenged. Therefore, it seems to have a stronger impact on social control than on information management. Self- and other-reformulations to secure a common conversational implicature are based on the content of a communicative contribution. They represent the most common device for securing felicitous communication. While a reformulation to secure a common conversational implicature is implicitly connected with securing the collective dialogue common ground, its perlocutionary effects may be quite different.

4.2. Reformulating to secure the collective dialogue common ground Reformulations have the function to signify that a contribution cannot be attributed to the collective dialogue common ground. They thus initiate a disruption in the structured interplay of postulating, ratifying, accepting and attributing contributions and their contents to the collective dialogue common ground. At the same time, reformulations express a request to secure the collective dialogue common ground. In order to achieve that goal, the implicit, taken-for granted common ground is made explicit so that coparticipants can check whether they share a common conception of the collective dialogue common ground. While a reformulation to secure a common conversational implicature has the prime function of securing a collective understanding on a local level, a reformulation to secure the collective dialogue common ground is realized in a context where a common interpretation of an implicature has already been negotiated, but not yet been achieved. This is reflected in the reformulation’s co-occurrence with an account which provides reasons for the reformulations, as in (10), (11) and (12), or in its co-occurrence with an explicit evaluation, as in (13) and (14). Example (10) stems from an interview between Jonathan Dimbleby and David Trippier (17/06/1990), example (11) is adopted from an interview between Jonathan Dimbleby and Tony Blair (27/05/1990), example (12) stems from an interview between Jonathan Dimbleby and Norman Lamont (13/05/1990), and example (13) is from an interview between Jonathan Dimbleby and John Martin and Danny Finkelstein (03/06/1990): (10) IR what you are saying as to be clear I don’t want to ask you to (11) IR what you’re saying is what you’re saying is then by way of clarification of what probably means is that the employer concerned has to say we will probably do this if it goes that way (12) IR when you say you resist those inflationary pay claims what you will in fact be saying if I understand you correctly even if you take into account some efficiency savings is that people will have to accept pay increases below the going rate of inflation otherwise you will be putting more new money into the economy

Reformulation and Common Grounds 177 Reformulations with an explicit account are performed by the interviewer only. In (10) and (11), the interviewer’s accounts as to be clear and by way of clarification of what probably means index the Gricean maxim of manner thus providing a reason for his unwillingness to attribute the politician’s contributions to the collective dialogue common ground, while at the same time signifying that there cannot be any further felicitous communication unless the disruption in the collective dialogue common ground has been negotiated and a common understanding been reached. In (12), the interviewer’s reformulation presents an unambiguous formulation of the politician’s content, which is not in the interest of the politician. This is reflected in the interviewer’s account if I understand you correctly, which allows the politician to ratify the interviewer’s reformulation through a rejection without the risk of losing face. A reformulation with an explicit evaluation is also used by the interviewer only, and it co-occurs with a negative evaluation: (13) IR are you seriously saying that if David Owen goes and leaves you alone that the SDP will become a more credible force In (13), the interviewer reformulates the politician’s contribution by indexing the Gricean maxim of quality, thereby implicitly referring to the politician’s credibility, which is assigned a rather strong force in that context. In the following, the lexical strategy of reformulating in order to secure a common perlocutionary effect is examined.

4.3. Reformulating to secure a common perlocutionary effect Reformulating a contribution in order to secure a common perlocutionary effect has only been identified once in the data examined. For this reason, it is assigned a marked status in the context of a political interview, where the perlocutionary effect of a contribution is generally left implicit and not spelled out. Because of its marked status, that type of reformulation tends to be restricted to escalating and confrontational sequences, where it makes explicit the other’s, that is the interviewer’s, implicated perlocutionary effect. By challenging the unspoken it is assigned a high face-threatening potential, as in (14), which stems from an interview between Jonathan Dimbleby and Michael Heseltine (13/05/1990): (14)

IR all you’re doing is to try to hike the issue which is an issue which is an issue in which I haven’t anything to add so let’s just take the words for what they are and leave them where they ought to be left

Here MH spells out the perlocutionary effect of the interviewer’s negotiation-of-reformulation sequence. In the context of a medialized political interview, the reformulation of the rather negative and face-threatening perlocutionary effect is made explicit not only for the direct coparticipant, that is the interviewer, but also for the audience. Moreover, the negative perlocutionary effect is presented as the result of the interviewer’s inappropriate performance, his inappropriate interviewing skills and his not acting in accordance with the genre-specific constraints and requirements (Fetzer, 2000).

178 Common Grounds In the context of a political interview, the lexical strategy of self- and other-reformulation is used by the interviewer and by the politician. Its most frequent function is to secure a common conversational implicature between the direct coparticipants, and between the direct coparticipants and the audience. While the securing of a common conversational implicature can be realized as an initial move in a reformulation sequence to secure the collective dialogue common ground, a reference to the actual goal of securing the collective dialogue common ground is used less frequently in the context of a political interview. In that scenario, a reformulation is used by the interviewer only, and it is accompanied by an account and by an explicit evaluation. Reformulations are also used to secure a common perlocutionary effect. This function is not very frequent in the context of a political interview, which may be due to the contextual constraints and requirements of media communication.

5. CONCLUSION Common ground is one of the fundamental premises of natural-language communication. In order to be integrated into a dynamic, negotiation-of-meaning-based framework, the primarily context-independent notion has been given a context-dependent interpretation and thus has been differentiated with respect to a higher-level concept of common ground, and coparticipantand dialogue-centred concepts of individual dialogue common ground and collective dialogue common ground. The latter administer the information exchanged in one dialogue. Individual dialogue common ground administers the information produced and interpreted by one coparticipant, and collective dialogue common ground administers the collectively ratified information, whose status has been agreed upon by the coparticipants. Reformulation plays an important role in the negotiation of intersubjective meaning, and its prime function in the domain of information management is to secure a common conversational implicature, to secure the coparticipants’ collective dialogue common grounds and to secure a common perlocutionary effect. The latter generally contributes to the interactional organization of topic control and power. Regarding its function in the sequential organization of discourse, it signifies a disruptive setting, which is defined by a halt in the flow of information. The communicative meaning of a reformulation depends on the question of who reformulates whom: other-reformulations negatively frame an event and signify that its communicative status is seen as inappropriate, untrue or insincere, and that for one or more of those reasons the contribution cannot be attributed to the collective dialogue common ground. Self-reformulations indicate that the speaker is in the process of negotiating intersubjective meaning, and that their contribution should not yet be attributed to the collective dialogue common ground. With respect to information management, reformulation is anchored to the interface between utterance meaning and speaker-intended meaning, and is, for this reason, looked upon as a lexical strategy. This is due to the fact that a reformulation is constructed of a rather fixed syntagmatic structure whose linguistic representation is highly conventionalized. In order to be felicitous, a reformulation requires (1) a lexicalized verb which denotes the activity of saying or doing, (2) a direct reference to speaker or addressee and (3) a reference to the communicative contribution’s content or force which is being reformulated. The canonical form of an other-reformulation is what you’re saying is ‘x’, and a self-reformulation’s canonical form is what I am saying is ‘x’. Due to the premise of sociolinguistic variation, the lexical verb say can be substituted by different verbs of saying and doing, such as suggest,

Reformulation and Common Grounds 179 mean, do or make, and the content marker what can be substituted by more specific content markers, such as the point or the issue. The key function of a reformulation lies in the coparticipants’ administration of the flow of information by signaling in what manner the information is to be attributed to the collective dialogue common ground and in what way the collective dialogue common ground is to be related to common ground. In dialogue, reformulations index the coparticipants’ collective dialogue common ground and signify that a contribution cannot be attributed to the dialogue common ground in the postulated manner. Instead, one or more of its constitutive parts of content, force, formulation, context or background information and its propositions, beliefs and interpersonal and interactional presuppositions require reformulation and further negotiation. Its communicative meaning depends on its sequential position and source and thus is calculated with regard to the nature of the connectedness between coparticipants, contribution, individual and collective dialogue common grounds, and social, sociocultural and linguistic contexts.

REFERENCES Akman, Varol, Bouquet, Paolo, Thomason, Richmond and Roger Y. Yong (eds), 2001. 3rd International and Interdisciplinary Conference on Modeling and Using Context (Context’01). Springer, Heidelberg. Allen, Joseph and Seidenberg, Mark S., 1999. The emergence of grammaticality in connectionist networks. In: Mac Whinney, B. (ed.), The Emergence of Language. Lawrence Erlbaum, Mahwah, 115–152. Austin, John L., 1980. How to Do Things with Words. Cambridge University Press, Cambridge. Bar-Hillel, Yehoshua, 1998. Indexical expressions. In: Kasher, A. (ed.), Pragmatics: Critical Concepts. Routledge, London, 23–40. Bouquet, Paolo, Serafini, Luciano, Berézillon, Patrick, Benerecetti, Massimop and Francesca, Castellani (eds), 1999. 2nd International and Interdisciplinary Conference on Modeling and Using Context (Context’99). Springer, Heidelberg. Brinton, Laurel, 1996. Pragmatic Markers in English. Grammaticalization and Discourse Functions. De Gruyter, Berlin/New York. Brown, Penelope, 1995. Politeness strategies and the attribution of intentions: The case of Tzeltal irony. In: Goody, E. (ed.), Social Intelligence and Interaction. Cambridge University Press, Cambridge, 153–174. Bührig, Kristin, 1996. Reformulierende Handlungen. Zur Analyse sprachlicher Adaptierungsprozesse in institutioneller Kommunikation. Narr, Tübingen. Clark, Herbert H., 1996. Using Language. Cambridge University Press, Cambridge. Clark, Herbert H. and Wilkes-Gibbs, Deanna, 1992. Referring as a collaborative process. In: Cohen, P.R., et al. (eds), Intentions in Communication. MIT Press, Cambridge, 463–493. Erman, Britt and Warren, Beatrice, 2000. The idiom principle and the open choice principle. Text 1, 29–62. Fetzer, Anita, 1994. Negative Interaktionen: kommunikative Strategien im britischen Englisch und interkulturelle Inferenzen. Peter Lang, Frankfurt/Main. Fetzer, Anita, 1999. Non-acceptances: Re- or un-creating context. In: Bouquet, P., et al. (eds), 2nd International and Interdisciplinary Conference on Modeling and Using Context (Context’99). Springer, Heidelberg, 133–144.

180 Common Grounds Fetzer, Anita, 2000. Negotiating validity claims in political interviews. Text 20 (4), 1–46. Fetzer, Anita, 2002. Communicative intentions in context. In: Fetzer, A. and Meierkord, C. (eds), Rethinking Sequentiality: Linguistics meets Conversational Interaction. Benjamins, Amsterdam, 37–69. Fetzer, Anita, 2004. Recontextualizing Context: Grammaticality Meets Appropriateness. Benjamins, Amsterdam. Fischer, Kerstin, this volume. Grounding and common ground: Modal particles and their translation equivalents. Garfinkel, Harold, 1994. Studies in Ethnomethodology. Polity Press, Cambridge. Givón, Tom, 1999. Generativity and variation: The notion ‘rule of grammar’ revisited. In: Mac Whinney, B. (ed.), The Emergence of Language. Lawrence Erlbaum, Mahwah, 81–114. Goody, Esther, 1995. Introduction: Some implications of a social origin of intelligence. In: Goody, E. (ed.), Social Intelligence and Interaction: Expressions and Implications of the Social Bias in Human Intelligence. Cambridge University Press, Cambridge, 1–36. Grice, Herbert P., 1975. Logic and conversation. In: Cole, M. and Morgan, J.L. (eds), Syntax and Semantics. Vol. III. Academic Press, New York, 41–58. Habermas, Jürgen, 1987. Theorie des kommunikativen Handelns. Suhrkamp, Frankfurt. Jucker, Andreas, 1986. News Interviews: A Pragmalinguistic Analysis. Benjamins, Amsterdam. Levinson, Stephen, 1995. Interactional bias in human thinking. In: Goody, E. (ed.), Social Intelligence and Interaction. Cambridge University Press, Cambridge, 221–260. Lewis, David, 1969. Convention. Harvard University Press, Cambridge. Linell, Per, 1998. Approaching Dialogue. Benjamins, Amsterdam. Litman, Diane. J. and Allen, James F., 1992. Discourse processing and common sense. In: Cohen, P.R., et al. (ed.), Intentions in Communication. MIT Press, Cambridge, 365–388. Okada, Moeko, this volume. Whose common ground? A misunderstanding caused by incorrect interpretations of the lexical markers of common ground. Penco, Carlo, 1999. Objective and cognitive context. In: Bouquet, P., et al. (eds), 2nd International and Interdisciplinary Conference on Modeling and Using Context (Context’99). Springer, Heidelberg, 270–283. Pittner, Karin, this volume. Common ground in interaction: The functions of medial doch in German. Rodríguez Penagos, Carlos, this volume. A common ground for knowledge through knowledge of language. Sarangi, Srikant and Slembrouck, Stefaan, 1996. Language, Bureaucracy & Social Control. Longman, London. Sbisà, Marina, 2002. Speech acts in context. Language and Communication 22 (4), 421–436. Schegloff, Emanuel, 1991. To Searle on conversation: A note in return. In: Searle, J.R., et al. (eds), (On) Searle on Conversation. Benjamins, Amsterdam, 133–128. Schiffer, Stephen, 1972. Meaning. Oxford, Clarendon Press. Searle, John R., 1969. Speech Acts. Cambridge University Press, Cambridge. Searle, John R., 1995. The Construction of Social Reality. The Free Press, New York. Searle, John R., 1999. Mind, Language and Society. Doing Philosophy in the Real World. Weidenfeld & Nicolson, London. Sperber, Dan and Wilson, Deidre, 1995. Relevance: Communication and Cognition. Blackwell, Oxford.

Reformulation and Common Grounds 181 Thomas, Jenny, 1984. Cross-cultural discourse as ‘unequal encounter’: Towards a pragmatic analysis. Applied Linguistics 5 (3), 91–112. Thomas, Jenny, 1985. The language of power. Towards a dynamic pragmatics. Journal of Pragmatics 9, 765–783. Thomason, Richmond H., 1992. Accommodation, meaning, and implicature: Interdisciplinary foundations for pragmatics. In: Cohen, P.R., et al. (ed.), Intentions in Communication. MIT Press, Cambridge, MA, 325–363. Traugott, Elizabeth, 1995. Subjectification in grammaticalization. In: Stein, D. and Wright, S. (eds), Subjectivity and Subjectivisation. Benjamins, Amsterdam, 31–54. Trognan, Alain, 2002. Speech acts and the logic of mutual understanding. In: Vanderveken, D. and Kubo, S. (eds), Essays in Speech Act Theory. Benjamins, Amsterdam, 121–133.

This page intentionally left blank

Lexical Markers of Common Grounds Edited by Anita Fetzer and Kerstin Fischer © 2007 Elsevier Ltd. All rights reserved.

WHOSE COMMON GROUND? A MISUNDERSTANDING CAUSED BY INCORRECT INTERPRETATIONS OF THE LEXICAL MARKERS OF COMMON GROUND Moeko Okada

1. INTRODUCTION In this paper I will show the complexity and difficulty that participants of conversations face while constructing common grounds. I will explore this matter by illustrating a misunderstanding caused by the hearer’s incorrect interpretations of lexical markers of common grounds uttered by the speaker. By lexical markers, I mean the words and phrases that a speaker uses to indicate certain features, such as common ground. A common ground is knowledge, a fact or an assumption that two or more people share. Lexical markers of common grounds are therefore those words and phrases that the speaker uses to signal the fact that the speaker assumes that the hearer shares a certain knowledge, fact or assumption with the speaker. The lexical markers of common grounds include the definite article the, pronouns (also person deixis) such as I, you, he, she, they, this, that and it; place deixis such as here and there; time deixis now and then; and proper nouns without introductions or explanations. Since these utterances presuppose the hearer’s knowledge, expressions such as “what I bought” or “the way you did it” can be included if no clear explanation is given. Therefore, the expression “what I bought” in the utterance “Could you bring what I bought today?” is a lexical marker of a common ground. However, the same expression “what I bought” in “What I bought today is a pretty shirt” is not a lexical marker of common ground, because the speaker does not assume that the hearer has knowledge about what the speaker bought, hence the expression is explanatory. Since the speaker feels no need for clarity, the speaker’s utterances which contain a lexical marker of common ground tend to become ambiguous as a whole. An ambiguity of this kind sometimes generates a misunderstanding between speaker and hearer, and my focus will be on this aspect. More specifically, I will point out that there are cases whereby presupposed common grounds are in fact not at all “common” to speaker and hearer.

184 Common Grounds This discrepancy could happen because the production of a lexical marker of common ground is speaker-oriented, while the processing of a lexical marker of common ground, which is uttered by the speaker, is hearer-oriented. When the hearer processes the lexical marker of common ground as intended by the speaker, the speaker and hearer share the common ground. However, in reality it is difficult to determine whether speaker and hearer really share a common ground or whether they do not. People are in a permanent process of negotiating meaning to make sure that they understand each other’s intended meaning correctly. My attempt is to illustrate a possible case of failure in constructing a common ground between a speaker and a hearer, using dialogues from a drama text.

2. LATENT MISUNDERSTANDING AND DRAMATIC IRONY Linell (1995: 187) classifies cases of misunderstanding or miscommunication events into “overt”, “covert” and “latent” miscommunication events. Using Linell’s notions, Hinennkamp (1999) divides the three categories further. According to him, overt misunderstandings, M1 and M2, are characterised by the immediate recognition of misunderstandings, which are indicated by “repairs”. Covert misunderstandings, M3, M4 and M5, exhibit a gradual recognition of misunderstandings, indicated by incoherence or disturbances in the flow of discourse. Latent misunderstandings, M6 and M7, are not recognised in an explicit mode. Of these, I am especially interested in M6, that is: There is no obvious recognition of a misunderstanding, although an outside observer regards it as a misunderstanding; or one of the participants may have received particular information afterwards (even a long time after) that leads her to reassess the interaction (or parts of it) as a misunderstanding. The interaction in question remains, however, untouched by this discovery or reinterpretation. (Hinennkamp, 1999: 7; italics added). To show the difference between M6 and M7, I will also quote his definition of M7: To an outside observer there is no manifestation and no indication that a misunderstanding has occurred, yet one interlocutor (or even both interlocutors) may have the feeling that either she has or was or they have or were misunderstood. So the misunderstanding may have been noticed but remained unnegotiated. (Hinnenkamp, 1999: 7) In both cases, the flow of discourse does not display any obvious recognition of a misunderstanding on the surface of the interactions. Hence, the interactions themselves are not affected by the misunderstandings. The difference is that in M6 (a) the outside observer recognises a misunderstanding which has occurred, or (b) one of the participants is aware that a misunderstanding has occurred at a later stage. In M7, on the other hand, (a) the outside observer does not recognise the misunderstanding, and (b) one of the participants may wonder whether a misunderstanding has occurred, but does not get any confirmation about this. Therefore, M6 has slightly more concrete signs of a misunderstanding than M7.

Whose Common Ground? 185 The data used in this paper is an extract from a dramatic comedy called The Rise and Fall of Little Voice (1992) by an award-winning Lancashire playwright, Jim Cartwright. The reason why I use dramatic dialogues instead of naturally occurring conversational data is that the dramatic texts sometimes shed more light on aspects that are difficult to point out in real conversations. Among the events of miscommunication, cases of latent misunderstandings in real conversational data are particularly hard to observe. In fact, Hinnenkamp (1999: 3), as an empirical linguist, opts out of discussing latent misunderstandings in his analysis because of the difficulties in accessing the minds of participants in real-world situations. By contrast, the use of a dramatic text enables us to analyse how a case of latent misunderstanding might occur. In a play where the world is complete in itself, where all the situations, the relationships between people, and the characters’ thoughts are revealed to the reader/audience at the end of the play, such an analysis does not cause any real difficulties. The merit is that we, as the reader/audience, observe a situation in which neither the speaker nor the hearer realises that a misunderstanding is taking place or has occurred. Let us return to the definition of M6 above. The “outside observer” represents the audience/reader of a play, and “one of the participants” can be one of the characters in a play; in the case of my data, the character is a woman called Mari. Thus, in spite of the fact that the utterances and the situations are fabricated, we can see a model of how a latent misunderstanding may arise, and from this we can consider, for example, which aspects are taken for granted when the conversationalists speak and listen to one another. Regarding its dramatic structure, a prototypical drama consists of two levels of discourse: the playwright–audience level and the character–character level (Short, 1996: 169).1 The playwright (Addresser 1) writes a play (Message 1) and the reader/audience (Addressee 1) reads/watches it. In the play world (Message 1), a character (Addresser 2) speaks lines (Message 2) to another character (Addressee 2). Due to this unique discourse structure, “dramatic irony” arises, which Short (1996: 169) explains as follows: It is this “doubled” structure which gives rise to the notion of dramatic irony, which typically occurs when the knowledge of some of the characters is less than that of the author and audience at level 1, producing tension for the audience as they wonder what will happen when that knowledge is revealed to the character. This tension can, in turn, give rise to a range of effects from the tragic to the comic. Dramatic irony occurs when Addressee 2 (one of the characters, Mari, in my data) misunderstands the Message at level 2 (utterances) of Addresser 2 (another character, Ray, in my data) and when Addresser 1 (playwright) and Addressee 1 (the audience/reader) know about it. This notion is close to that of the latent misunderstanding M6.

3. CONTEXT AND CONTEXTUAL EFFECTS As a concept, the term “context” is very broad. Halliday (1977), amongst others, distinguishes “co-text”, which is concerned with the location of a particular word in a particular sentence or 1

Hess-Lüttich (1985: 200) offers a three-level discourse structure in drama in which he includes the implied author and the implied reader in the middle.

186 Common Grounds discourse, from “context”, which is concerned with a particular situation in which a particular sentence or utterance is made. In many cases, however, the word context is used to refer to both; therefore, I do not distinguish between the two, unless otherwise stated. For situational context, further categorisation is possible. Summarising Wales (1989: 93–95), Simpson (1998: 37) and Short, Culpeper and Semino (2000: 137–141), the following descriptions of context are possible (my examples). • Physical context comprises the actual setting or environment in which the interaction takes place, such as a house-warming party or a hospital. • Personal context comprises the social and personal relationships amongst the interactants, for instance the relationships between intimate friends or between employer and employees. • Cognitive context comprises the shared and background knowledge held by participants in the interaction, including social and cultural knowledge. It is sometimes referred to schemata. For example, knowledge about how an interview, a wedding or a lecture is conducted. • Textual context comprises the world which the text constructs, that is the textual world. For instance, in Neil Simon’s Fools, the textual context is the situation in which a young teacher struggles to lift a curse brought to the village. Towards the end, he addresses the audience a few times, in which case the textual world is directly connected to the real-world audience or reader. Similarly, Sperber and Wilson’s account of context (1995 [1986]) includes the above features and I will use their notion as a framework for my analysis. They see context as a “psychological” phenomenon, to use their own words: A context is a psychological construct, a subset of the hearer’s assumptions about the world. It is these assumptions, of course, rather than the actual state of the world, that affect the interpretation of an utterance. A context in this sense is not limited to information about the immediate physical environment or the immediately preceding utterances: expectations about the future, scientific hypotheses or religious beliefs, anecdotal memories, general cultural assumptions, beliefs about the mental state of the speaker, may all play a role in interpretation. (Sperber and Wilson, 1995: 15–16) They also claim that context is not given but “chosen” in the search for “relevance” (Sperber and Wilson, 1995: 141). This view is interesting and very useful for the explanation of why people often interpret the same event in different ways.2 Sperber and Wilson propose two terms in relation to context, namely “contextual implication” and “contextual effect” (Sperber and Wilson, 1995: 107–109). Contextual implications are new conclusions, which are generated by the input of new information into old information, but not from old or new information alone. When the contextualisation of the new information in the old information produces effects like these, they are called “contextual effects”. Quoting their example: if the premise about a party is “if Peter, Paul and Mary came to the party,

2

Nemo in this volume, however, argues against Relevance Theory’s conception of context.

Whose Common Ground? 187 the party was a success” (old information), and it turns out that “Peter came to the party”, “Paul came to the party” and “Mary came to the party” (new information), the conclusion (contextual implication) is that “the party was a success” (Sperber and Wilson, 1995: 109). This conclusion cannot be drawn from either the premise or the new inputs alone. Contextual effects can be any of the following: 1. Gaining new assumptions (contextual implications); 2. Strengthening old assumptions; 3. Eliminating old assumptions and replacing them with new stronger assumptions when they contradict each other (summarised from Sperber and Wilson, 1995: 108–117). When new information is rejected and old assumptions remain, there is no contextual effect. If we add new information such as “if the party broke up late, then it was a success” and “the party broke up late” to the above example, then the assumption “the party was a success” is strengthened. On the other hand, if we get more new information which contradicts the old assumption, such as “if Bill came, the party was not a success” and “Bill came”, the old assumption becomes weaker and is finally eliminated. In real-world situations, however, it is extremely difficult to observe a person’s psychological construct by examining the set of assumptions he/she has as its constituent. In the first place, access to a person’s inner state of mind is impossible, as mentioned above. Secondly, we could hardly observe the process of a contextual effect happening in a particular person’s context, because it is difficult to tell which assumptions are old, how much certainty a particular assumption has, which new information he/she has obtained and when he/she has obtained the new information. People do not utter every assumption they have. Therefore, when we observe whether two interlocutors have a common ground or not, we tend to judge it from whether negotiations are taking place in interactions, but not from the assumptions in their contexts. In a dramatic text, which is fictional by definition, all the important assumptions necessary for the development of the story are explicitly provided to the reader/audience through the characters’ utterances and/or kinetic expressions. Therefore, we can easily observe and safely determine which kind of assumptions a particular character has in his/her context or which assumption is held by this character, but not by another character. Therefore, discussing a common ground between characters in terms of sets of assumptions is relatively straightforward.

4. THE ANALYSIS With these notions, I will now show how the interpretations of some lexical markers are affected by context. The Rise and Fall of Little Voice (1996 [1992]) won the Best Comedy for the Laurence Olivier Award and the Evening Standard Award in 1993 in Britain. Before presenting the dialogues, I will briefly introduce the characters. Mari is a workingclass middle-aged widow who hopes to escape from her poverty and misery by finding a rich man. Ray is a middle-aged man in show business who has met Mari at a bar. Mari has a daughter, Little Voice (hence, LV). LV is very shy, rarely talks, and has a very low voice, but she loves the old records her father has left her and she has the ability to sing well and

188 Common Grounds to imitate her favourite stars, such as Judy Garland. Mari dominates her, shouts at her all the time, and places no value on her. One night, Ray visits Mari at home and is impressed when he hears LV singing. The next morning Ray is overjoyed to discover LV’s talent, which he has been seeking for a long time, and he is thinking of making her a star. Billy, another character, is a shy young man from a phone company who is interested in LV, and Sadie is a neighbour of Mari’s. The numbering of the following extract is for convenience.

(1) (2)

(3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)

Mari: Darling, how are you? Goes to embrace him. Ray is dressing as he speaks. Ray: I’ve got to dash Mari. But what can I say? It’s happened at last, eh! I’m so excited, it’s like at the races when you’ve found yourself a little nag no one’s noticed but you know you’re onto a certainty and you’re feeling, this is it! She is the one. Do you know what I mean? A knock on the door. Mari: Go on, ::: yes! :::: Yes! ::: I’m:::: with:::: you ::: lad. :::: Yes    Ray: It’s like… Mari: :::: Yeah. :::: Yeah. (Desperate) Ignore that. It’s too late, Ray is opening it. It’s Billy. Billy: It’s, er, me again. Er, just come to see if your phone’s still alright. Mari snatches handset up. Holds it out to him. Mari: Is that its sound? Billy: Yes. Mari: It’s reet! Slams door. Mari: On Ray, on. Ray: (forgetting where he was) Er… Mari: You were under starters orders and you were off! Ray: Yeah. It’s not just my future, it’s yours. Mari: :: Oh:::: my :::: God.

(15) Ray: I still can’t believe it. It’s what I’ve been looking for for ages. And here it is under this roof, under me very nose. All I can say is… Knock on door. Ray opens it. (16) Mari: No, leave it! For God sake leave it! Sadie is standing there. (17) Ray: (as he leaves) I’ll be back soon. I can’t leave it. Not this. Best to act fast when you’re this sure eh? Mari nods frantically. ::::::::::::::::: (18) Ray: It’s just one o’them once in a lifetime things. He passes Sadie and leaves. Mari takes her by the sleeve and silently draws her in the house. (19) Mari: Sadie, did you hear that utterance? Did you? Did you hear what Ray sayeth. It were almost on the tip there of his raspberry tongue, he wants me. I can’t believe it Sade. The bastard wants me. Get Jackson 5 on, Sadie. We always play it when we’ve something to celebrate, don’t we? Sadie, happy, runs to radiogram and gets it on.

Whose Common Ground? 189 (20)

Mari: At last. At last. Saved, secured. I shall go to the ball. Oh darlings from the sky. ____ Ray’s utterance that contain lexical markers of common ground Mari is showing that she is sharing a ground with him :::: clear indication of Mari’s misinterpretation (Jim Cartwright, The Rise and Fall of Little Voice, Act I; 1996 [1992]: 199–201)

The underlined parts are Ray’s utterances, which contain several ambiguous elements which create the possibility of inviting the hearer’s misunderstanding. Wavy underlines are added to the parts which show Mari’s assumption and appeal that she understands him perfectly and that she shares a common ground with him. The double-underlined parts indicate that Mari has interpreted Ray’s utterances as a declaration of his love, and they therefore are clear indications of her misunderstanding. To reiterate Sperber and Wilson’s definition of context with my emphases in italic type: A context is a psychological construct, a subset of the hearer’s assumptions about the world. It is these assumptions, of course, rather than the actual state of the world, that affect the interpretation of an utterance. A context in this sense is not limited to information about the immediate physical environment or the immediately preceding utterances: expectations about the future, scientific hypotheses or religious beliefs, anecdotal memories, general cultural assumptions, beliefs about the mental state of the speaker, may all play a role in interpretation (Sperber and Wilson, 1995: 15–16; italics added). In the data, we find two contexts which overlap, that is, Mari’s and Ray’s. Mari’s context, that is, her psychological construct includes the following assumptions: – – – – – – – – –

Someday she could escape from her poor, miserable life. The easiest and the best way to achieve this is to find a rich man. She is having an affair with a rich man. The man stayed a night at her house. She has a daughter, who is very quiet and strange. Her daughter is useless. Her daughter loves to sing but her singing is nothing special. The man happened to hear her singing. It is highly unlikely that the singing impressed the man.

On the other hand, Ray’s context would include the following assumptions: – – – – –

Someday he could become a famous music producer. To achieve this goal, he must find a new musician who could be turned into a star. He is having an affair with a woman whom he met at a bar. This woman is not particularly attractive (this is revealed in a later scene). He stayed a night at the woman’s house.

190 Common Grounds – – – – –

The woman has a daughter, who is very quiet and strange. He heard the daughter singing. The daughter has a remarkable singing ability. The daughter is useful. It is highly possible that he will succeed if he can make the daughter a real star.

These two contexts overlap only partially (Figure 1). For example, they share the assumption that Ray heard the daughter singing. However, its impact on both contexts differs immensely. Consequently, it yields contrasting assumptions. In Mari’s context, the assumptions are that “Ray would not be impressed by the daughter’s singing because it is nothing special”, and “the daughter is useless”. In Ray’s context, the assumptions are that “the daughter’s singing is marvellous”, and thus “the daughter is useful”. Bearing in mind that there are both coincidences and discrepancies of assumptions in their contexts, let us consider how Ray’s utterances led Mari to misunderstand Ray. Ray’s utterances in turn (2) bear a major responsibility. I’ve got to dash Mari. But what can I say? It ’s happened at last, eh! I’m so excited, it ’s like at the races when you’ve found yourself a little nag no one’s noticed but you know you’re onto a certainty and you’re feeling, this is it ! She is the one . Do you know what I mean? The lexical marker “it” is used in “it’s happened at last” and “it’s like at the races…”. However, Ray does not specify distinctively what “it” refers to. What can be understood from these utterances is that something significant has happened to him and that he is excited about the incident. What cannot be understood is what exactly it is that makes him so excited. The utterances “this is it” and “she is the one” are similar in the sense that the subjects “this” and “she” refer to “the promising but still unknown horse he has discovered”, and the complements “it” and “the one” refer to something that brings him excitement and happiness. In Mari’s context, an assumption that “Ray gets impressed by the daughter’s singing” does not exist. Hence, she can only connect his sudden excitement to his love affair with her. Therefore, she interprets the reference “it” as being “his discovery of true love for her”. For Ray, “it” means his discovery of “a talented female singer”, namely the daughter LV. In his context, an assumption that “LV has a remarkable singing ability” has a strong impact. The audience/the reader knows this as well, but Mari does not. An interesting aspect is that Ray uses the

Mari’s context

Ray’s context

Figure 1. Relationship between Mari’s and Ray’s contexts

Whose Common Ground? 191 pronoun “she” instead of “he” to refer to the talented horse. He was possibly thinking of LV and this resulted in his selecting the feminine pronoun “she”. It is also possible that because of that feminine deictic expression, Mari imagines that Ray refers to her and the horse in a metaphoric manner. The lexical items which bear some ambiguity, such as the pronouns “she” and “it,” are what Schegloff (1987) calls “problematic references” in his identification of sources of misunderstanding. The final utterance in turn (2) “do you know what I mean?” is interesting and significant. Ray wants to make sure that he is making himself understood. This is an act of checking whether there really is a common ground between them. Unwittingly, he states an ambiguous lexical marker “what I mean” in order to clarify his point. What he wants to say by “what I mean” is that he is excited about the discovery of a talented singer. He should have clarified that by adding, for example, “What I mean is…”, some sort of reformulated expression as Fetzer investigates in this volume. Since no clarification about the incident is provided, it is not possible for Mari to realise that an interpretation of his utterance different to her own, namely “what I mean” means that he is excited about the discovery of his love for her, is also plausible. Thus, the way of his checking whether there is common ground is insufficient. As for Mari, she answers his question with great enthusiasm in turns (3) and (5). She utters “yes” three times and “yeah” twice. Especially the utterance “I’m with you lad” shows an emphasis of her sharing the common ground with him, which in fact she is not. After some interruptions by the visitors, Ray continues his talk with excitement and seriousness in turns (13), (15) and (17). It ’s not just my future , it ’s yours . I still can’t believe it . It ’s what I’ve been looking for for ages. And here it is under this roof , under me very nose . I can’t leave it . Not this . Best to act fast when you’re this sure eh? Similar to the previous phenomenon, the ambiguous lexical marker “it” appears six times and is not interpreted in the speaker-intended manner. The deictic meanings expressed through the other markers, that is “my future”, “yours”, “here”, “under this roof” and “under me very nose” are correctly understood by Mari. So, the speaker could, in principle, convey the intended referents to the hearer in a successful manner by these lexical markers of common grounds. However, since the reference “it” plays a crucial role in the correct understanding which affects the whole sentences and utterances, Mari continues to misunderstand Ray, and she fails to recognise her wrong interpretation. There is another lexical marker which claims a common ground and which also fails to be understood correctly, that is, “this”. It functions in the same way as “it” does. Ray rephrases “it” with the more emotionally proximal form “this”, which shows that Ray regards the incident as genuinely important. The utterance “best to act fast…” is also ambiguous and does not specify “‘what” kind of action he means, and what the action stands for. This ellipsis also results from his assumption that Mari shares his context. Mari utters “oh my God” after hearing Ray’s utterance “it’s not just my future, it’s yours”. As mentioned several times, she mistakenly thinks that she impressed him greatly. Now that he mentions her future and his future, she starts to think that he probably wants her as a long-term partner. Here, Ray’s utterance acts as new information in Mari’s context and

192 Common Grounds yields a new assumption (or a contextual implication). Hearing his utterances up to “best to act fast…”, she “nods frantically” as requested by the stage directions. This is a sign of her joy. Each time she hears him, his utterance acts as new information, and this time it “strengthens” her newly gained assumption that he wants her as a partner. This phenomenon is repeated several times and she becomes more and more confident of the assumption. With the help of her strong wish to lead a better life with a rich man, Mari misinterprets Ray’s utterances in this manner and is now certain about the way she interprets them. Her “expectations about the future” certainly affects her interpretation, as Sperber and Wilson suggest. Ray’s utterances are like a riddle or pun, which has two or more possible answers, as is the case with the question of what is so exciting and affects our future, which I found in this house after so many years. For Mari, the answer is their partnership. The correct answer is LV’s debut as a singer. In that vein, Bergson notes that puns and misunderstandings could be under the same umbrella of the notion “reciprocal interference of series” (1911 [1900]). A similar concept is postulated by Koestler in his “bisociation” theory (1964). Both are situations where one thing is simultaneously belongs to two independent events and, for this reason, can be interpreted in two completely different ways. The ambiguity in Ray’s utterances, which is caused by the unspecified reference “it” which results from his assumption that Mari already shares his psychological construct, falls into the situation of the reciprocal interference of series and the bisociation. “It” belongs to two independent contexts simultaneously, and thus can be interpreted in two different meanings. We can see how a lexical marker of common ground can be interpreted differently in two sets of different assumptions, that is, two different contexts. Ray’s last utterance in the data is, It ’s just one o’them once in a lifetime things. “It” remains unspecified, and it is needless to say, that something which happens once in a lifetime strengthens Mari’s assumption about their partnership. Mari’s misunderstanding is confirmed by her utterances in turns (19) and (20), as in “he wants me”, “The bastard wants me”, “celebrate”, “At last. At last. Saved, secured” and “Oh darlings from the sky”. Especially, the words “saved, secured” reflect her future expectations well. This sad misunderstanding continues throughout the scene, and it does not get resolved until the end of the play when LV abandons the stage, which upsets Ray and leads him to tell Mari that he has never been attracted to her. Hence, Mari realises much later that she has misunderstood Ray’s utterances, and, for this reason, this is a case of latent misunderstanding M6.

5. CONCLUSION Lexical markers of common ground include expressions which indicate the speaker’s belief that the hearer must be sharing certain knowledge or assumption with the speaker. A definite article and deictic words fall into this category. Proper nouns and some expressions such as “what I bought”, if uttered with no explanation as to what the person actually bought, can also be lexical markers of common ground. Therefore, these expressions bear ambiguity and have the risk of not being properly understood. My attempt was to show the difficulty of two conversationalists constructing a common ground. They may believe they share a common ground but they do not share a common

Whose Common Ground? 193 ground in actual terms. Mari’s misunderstanding can be investigated in terms of both co-text and context. On the one hand, it is a product of the playwright’s careful wording, including the lexical markers of common grounds. Hence, this is a matter of co-text. Cartwright could have made Ray utter much clearer lines, such as “I have been looking for a real star for so long and I’ve found LV here in this house. Best act fast to prepare a stage for her.” Instead, the playwright chose to use the problematic reference “ it” and elliptical expressions to invite a crucial misinterpretation on the part of the hearer. He manipulates the use of lexical markers of common grounds to prevent his characters from sharing a common ground. On the other hand, Mari’s misunderstanding is a product of the playwright’s careful setting up of the whole situation, in which Ray can mean one thing and Mari can interpret it in a different way. Therefore, we find two different co-existing sets of assumptions. Hence, this is a matter of context. In both cases, the playwright–audience level of discourse makes clear that Mari has misunderstood Ray’s utterances. The extract shows one possible model of how a latent misunderstanding might occur. Mari’s misunderstanding is an appeal to the audience by adding humorous and possibly sad effects to the play. In fact, from the perspective of drama, this latent misunderstanding is quite significant in the sense that it is one of the comical elements that triggers laughter from the audience. It is almost central to this comedy, as it would be to numerous others, which involve incidents of misunderstandings as a base of the story (Okada, 2001). The lexical markers of common grounds found in Ray’s utterances are the result of his strong belief or assumption that Mari would understand them correctly and with ease. Therefore, for him, no explanations are necessary between them. Mari also is confident that she understands his lexical markers correctly. As the analysis shows, some of them are understood as intended, but the most crucial one is not. There is a discrepancy between the common ground which Ray believes to have and the common ground which Mari believes to have. So, of whose common ground are we talking when we use “common” ground? Sometimes, a common ground may not be common at all. It is only supposed to be common.

REFERENCES Bergson, Henri, 1911 [1900]. Laughter: An Essay on the Meaning of the Comic. Macmillan, London. Cartwright, Jim, 1996 [1992]. The Rise and Fall of Little Voice. Methuen, London. Fetzer, Anita, this volume. Reformulation and common grounds. Halliday, M.A.K., 1977. Text as semantic choice in social contexts. In: van Dijk, T.A. and Petöfi, J.S. (eds), Grammars and Descriptions. Walter de Gruyter, Berlin, 176–226. Hess-Lüttich, Ernest W.B., 1985. Dramatic discourse. In: van Dijk, T.A. (ed.), Discourse and Literature: New Approaches to the Analysis of Literary Genres. Benjamins, Amsterdam, 199–214. Hinnenkamp, Volker, 1999. The notion of misunderstanding in intercultural communication. Journal of Intercultural Communication. 1, 1–24. http://www.immi.se/intercultural/nr1/ hinnenkamp.htm Koestler, Arthur, 1964. The Act of Creation. Hutchinson, London.

194 Common Grounds Linell, Per, 1995. Troubles with mutualities: Towards a dialogical theory of misunderstanding and miscommunication. In: Markova, I., Graumann, C., and Foppa, K. (eds), Mutualities in Dialogue. Cambridge University Press, Cambridge, 176–213. Nemo, François, this volume. The pragmatics of common ground: From common knowledge to shared attention and social referencing. Okada, Moeko, 2001. Linguistic Approaches to the Analysis of Humour in Modern English Dramatic Comedy. Unpublished PhD Dissertation, University of Lancaster. Schegloff, Emanuel A., 1987. Some sources of misunderstanding in talk-in-interaction. Linguistics 25, 201–218. Sperber, Dan and Wilson, Deirdre, 1995. Relevance: Communication and Cognition. Blackwell, Oxford. Short, Mick, 1996. Exploring the Language of Poems, Plays and Prose. Longman, London. Short, Mick, Culpeper, Jonathan and Semino, Elena, 2000. Language and context: Jane Gardam’s Bilgewater. In: Bex, T., Burke, M., and Stockwell, P. (eds), Contextualized Stylistics: In Honour of Peter Verdonk. Rodopi, Amsterdam, 131–151. Simpson, Paul, 1998. Odd talk: Studying discourses of incongruity. In: Culpeper, J., Short, M. and Verdonk, P. (eds), Exploring the Language of Drama: From Text to Context. Routledge, London, 34–53. Wales, Katie, 1989. A Dictionary of Stylistics. Longman, London.

Lexical Markers of Common Grounds Edited by Anita Fetzer and Kerstin Fischer © 2007 Elsevier Ltd. All rights reserved.

COMMON GROUND, CATEGORIZATION, AND DECISION-MAKING Thanh Nyan

1. INTRODUCTION What underlies common ground construction as triggered by argumentative markers such as mais, maintenant, and toujours? While the answer would have to be a matter for empirical testing, a co-evolutionary perspective allows for certain hypotheses to be framed. Rather than go into the details of what such a perspective entails, which can be found in Deacon (1997), I shall confine myself to two assumptions that have direct bearing on current purposes: • Language and the brain are co-evolved. • In addressing its own problems, language, as a latecomer, will normally turn to previously evolved solutions, those that have proved successful in dealing with similar problems, irrespective of the context involved. Unless otherwise specified, “common ground” will be referring to an idealized validation context which the Speaker (S) seeks to impose on the Respondent (R) – or rather his processing system – through the use of the above markers. In terms of organization, I begin by specifying which aspect of the meaning of the above markers I shall be concerned with. Next I propose that the capacity for argumentative discourse is similar to the capacity for higher-order consciousness (as defined by Edelman, 1989) in that both mediate a goal-oriented activity, which is to ensure survival. Following this I set out the consequences of this proposal with regard to supporting structures and processes: on the basis of the above co-evolutionary assumptions, both capacities – I argue – can be expected to share at least some of the same supporting structures and processes, most notably: • a value-dominated conceptual memory consisting of categories of initial situations; • paired with categories of response options;

196 Common Grounds • mechanisms that bring past values to bear on ongoing perceptions, and provide the basis for anticipating sequences of actions necessary to achieve a predetermined goal; • a decision-making system, whereby the appropriate course of action can be selected on the basis of a categorization of initial situations. The next section goes into the details of how argumentative discourse may be viewed in terms of Damasio’s model of decision-making (Damasio, 1994). Inasmuch as the goal of argumentative discourse is one of inducing a decision-making process in R, this section also asks what further requirements this places on the model based on Damasio. The final section provides an overview of how common ground construction would be triggered by the use of the above argumentative markers.

1.1. Data Following Anscombre and Ducrot’s Argumentation Theory1 (AT), the core meaning of the above markers (or rather a key aspect of it) can be construed in terms of encoded “schematized situations” (see also Fischer, this volume). These account for the speaker’s verbalized contribution in respect of appropriateness and validation by providing a model of prior interaction in which S’s utterance is seen to be relevant, together with assumptions that validate it. Consider the following utterances (overheard at a departmental party): 1. 2.

3. 4.

(1) Mike aime les crocodiles, mais de loin. Mike is fond of crocodiles, but from a distance. (2) Ce n’est pas une mauvaise idée. Maintenant [Octavio] a peut-être quelque chose de prévu. That’s not a bad idea. Now, [Octavio] may have made other plans. (3) Si on le descendait, ce sera toujours ça de fait. (4) How about taking him out? At least we’ll be able to cross him off our list.

Here (1) is a response to a prior suggestion that a crocodile would make a suitable birthday present for Mike. The schematized situation encoded in mais consists of interacting viewpoints as shown below: (1s) V1: Mike is fond of crocodiles (P), so let’s get him one for his birthday (C). V2: P is certainly the case, but it does not follow that C. (In other words, the agreement, which is over P in a representational capacity, entails no commitment on S’s part as to whether the situation as described by P is conducive to C. The disagreement concerns the aspect under which the situation of reference is described by P.)

1

As a theory of integrated pragmatics, AT defines utterance meaning in terms of linguistic constraints on the interpretation of the co-text and context of utterance. This entails a “deep structure” (or “phrase”) whose primary component is argumentational (i.e., comprises constraints such as the above, including constraints on continuation). The other component, which is informational, is regarded as derivative. The relationship between these two components may be seen as AT’s central claim, and all major conceptual changes it underwent over the years can be put down to attempts to uphold this idea (Anscombre, 1995).

Common Ground, Categorization, and Decision-Making 197 V2 : The reason for V2 – one arguably more pertinent to the issue at hand – is to do with another aspect of the situation of reference (namely that Mike is fond of crocodiles as long as they are kept at bay = P ). V2 : In view of P  C is the correct conclusion to draw (where C is opposed to C). Of these, V1 is imputed to the Respondent (R), while V2, V2 , and V2 detail various stages of the stance taken by S. Each of the main viewpoints is validated by an assumption, which within AT takes the form of a topos T. Thus, underpinning V1 and V2 one may conceivably have T1 and T2, respectively. (Where T1 = [the more someone is fond of something, the more likely they are to want one of their own]; and T2 = [the more dangerous something is, the more one wants to stay away from it]) The schematized situation for maintenant (as based on Nyan, 1998) can be similarly described: (2s) V1: Inasmuch as the situation of reference is construable in terms of P (i.e., Octavio is fun to have around), C should follow (i.e., Octavio is to be be asked to join us on a certain occasion) V2: I’ll go along with V1. V3: Together with the aspect provided by P of the situation of reference, another aspect, construable in terms of P , should also be taken on board (where P = the possibility that Octavio has other plans). V3 : If P turned out to be the case, this would call for a review of C. Here V1 stands for a view originating from R. V2 is V1 as sanctioned by S. V3 is an alternate view of the situation of reference, one taking into consideration the eventuality of P . V3 is the view both R and S are expected to take, should P turn out to be the case. The validation context in this case would include T1-2 – [the more fun someone is, the more reason to have them around] – and T3 [the busier someone is, the less likely they are to be available]. The schematized situation assigned to (3) is a pared-down version of the one proposed by Cadiot et al. (1985): (3s) V1: Let’s do C (i.e., take X out), because of P (we’ll be able to cross him off our list). V2: P is hardly worth the trouble, so – C (there is no point in carrying out the deed). V3: That may be so. Nonetheless, no matter how small an incentive, P is not to be sneezed at. V4: In view of V3, C is the correct decision to reach. Of these V1, V3, and V4 reflect S’s views. V2 is imputed to R. In terms of validation context, the topoï being referred to would be T1 [the more one has to do, the more reason there is to get on with it]; and T2 [the fewer the incentives, the less reason to do something]. The validation of V3 and V4 is provided by T3 [the stronger the incentives, the more reason to do something], with no lower threshold below which T3 would cease to apply. Alternatively one might have an assumption to the effect that [an

198 Common Grounds incentive, no matter how negligible, is still an incentive], which effectively counteracts the gradability inherent in T2 to justify undertaking a job regardless of the level of incentive involved. In Clark’s terminology (Clark, 1996), these schematized situations correspond, at the lower level, to pre-sequences whose function is to satisfy a preparatory condition, in this case, a condition about relevance. The “joint project” requiring a context of relevance and of validation is that of reaching an agreement over a certain conclusion. Now, inasmuch as these schematized situations are linguistically encoded some comments are in order: • Though featuring sequences of proposals and uptakes, these schematized situations may not reflect any actual prior negotiating process. • Where a schematized situation is not a genuine record of prior interactions, the use of the corresponding marker constitutes an attempt on S’s part to fix her intentions unilaterally (see also Nemo’s manipulatory view of communication in this volume). • While it is, in principle, possible for R to challenge S’s (purported) construal of the situation at the earliest opportunity, this may prove an arduous task if one is dealing with procedural meaning, as is arguably the case here. Indeed, if the meaning of argumentative markers is, by and large, construable in procedural terms (i.e., in terms of instructions or constraints targeting the processing system), speech participants are expected to have a know-how of these markers, rather than a know-that, as in the case of conceptual meaning, hence the difficulty they may experience in having immediate access to aspects that might be challenged.

In other words, if one’s primary concern is the core meaning of argumentative markers: • The role of the actual respondent is not an immediate issue; • the processing system takes precedence as a key “respondent” at this early stage in the communicative process (it is, after all, responsible for initiating the process whereby common ground, as a joint intentional object, may be established). The possibility for reformulations (see Fetzer, this volume) or accomodations (Thomason, 1992) belongs to a later stage.

In the following, I will then focus on this early common ground construction, and how it might be implemented. On the basis of the co-evolutionary assumptions previously introduced, the processes involved should be similar to those sub-serving the same types of function in other cognitive systems. Further specification of the above hypothesis in terms of supporting structures and mechanisms requires that we ask what pre-linguistic capacity presents a clear parallel with the capacity for discourse.

Common Ground, Categorization, and Decision-Making 199

2. DISCOURSE, HIGHER-ORDER CONSCIOUSNESS, AND GOAL-ORIENTED ACTIVITIES From an interactive standpoint, argumentative discourse is construable as a goal-oriented activity, the point of which is to secure R’s agreement over a certain proposition. As such it also involves providing access to the processes leading to the identification of this proposition and of the grounds that validate it. By virtue of these two features, the above linguistic capacity presents a clear parallel with a capacity, which, according to Edelman (1989), arises from higher-order consciousness. An animal endowed with higher-order consciousness is one capable of constructing a coherent picture of past, present, and future, in the sense that it can set itself goals connected to its own survival, keep these goals in mind for the required length of time, and plan sequences of behaviours necessary to their implementation. Without this capacity, the animal would be too susceptible to the demands and distraction of ongoing perceptions to plan for the future. Such a parallel is consistent with the view that language is a major clue to the existence of a fully developed higher-order consciousness, and its possession adds to the adaptive value of an emerging higher-order consciousness (Edelman, 1989: 173–185). Indeed, not only does language reinforce the modelling of internal states, it also provides external representations of those states, with the consequence that others can be manipulated into taking an active part in the implementation of one’s plans.

3. SUPPORTING STRUCTURES AND MECHANISMS On the basis of the co-evolutionary assumptions set out earlier, the following hypothesis can be put forward: the same supporting structures and mechanisms which in Edelman’s Theory of Neuronal Group Selection (henceforth TNGS) account for the emergence of consciousness – both primary and higher-order – can be expected to mediate language in the above representational capacity.2 Irrespective of whether we are dealing with language or higher-order consciousness, the ability to construct a coherent picture of past, present, and future in a way that makes adaptive sense requires two basic things: a. a value-dominated conceptual memory, whose role is to ensure the dominance of values connected to survival; b. mechanisms whereby this value-dominated memory – the past – can be brought to bear on the present, that is, on ongoing perceptual categorizations. These mechanisms must also account for how this memory provides the basis for anticipating the future, in terms of sequences of actions necessary to carry out a predetermined goal. In other words, such mechanisms are about ensuring continuity and ordering.

2

Clearly, further specialized developments would also be required to sub-serve the symbolic aspect.

200 Common Grounds

3.1. Conceptual memory Based on the TNGS, this is roughly how this memory came about. The nervous system may be functionally construed in terms of “non-self” and “self”. “Non-self” refers to the activity of the cerebral cortex, the thalamus and the cerebellum, which are dedicated to categorizing signals from the external world. These include peripheral, voluntary, motor, proprioceptive, and polymodal sensory signals. “Self” is determined by the activity of areas that include “Brain stem and pontine nuclei, mesencephalic reticular formation, hypothalamus, amygdala, septum and fornix and their various connections to prelimbic forebrain areas” (Edelman, 1989: 98). These areas are involved in categorizing signals driven from inner events, whose “adaptive, homeostatic and endocrine functions […] relate to [the individual animal’s] immediate needs for survival and reflect evolutionarily selected values that have contributed to fitness” (Edelman, 1989: 94). So, in respect of categories, there is a basic distinction to be made between non-self, or value-free, perceptual categories relating to the external world, and self-, or value-dominated categories, which result from innate adaptive values acting on perceptual categorization. Innate values determine not only which situations and actions the species as a whole perceives as relevant, but also what situations and actions individuals, on the basis of their own experience, will see as advantageous. In other words some self-categories will be shared by the species, while others will be specific to individuals. Self and non-self categories interact to give rise to and update conceptual memory. This interaction, which involves a comparison of the two categorical systems, occurs, for example, “at the hippocampus, septum and cingulate gyri” and provides a basis for “the conceptual recategorization of this comparison, which takes place in the cingulate gyri, temporal lobe and parietal and frontal cortex” (Edelman, 1989: 98–100, 151–162). The concepts held in this memory are thus composite categories since they result from the interaction between inner values, past categorizations of the world, and current categorizations. In terms of background knowledge, we are seeing the emergence of a distinction between wired-in knowledge and acquired knowledge. Wired-in knowledge corresponds to inner values and arises from the genome. Acquired knowledge subdivides into two kinds: one characteristic of the species (e.g., escape strategies), and one specific to individuals (e.g., a novel way of accessing a food source prompted by circumstances). Note that new adaptive behaviour can conceivably spread from individual to individual and thus become shared by all members of the species. In other words, we have something akin to Searle’s Background3 (1983, 1995), with Deep Background skills at one end, and Superficial Background skills at the other. In light of what precedes, past, present, and future, as featured in the unified picture produced by higher-order consciousness, can now be further specified: “past” stands for the above category-value matching; “present” is shorthand for immediate perceptual categorizations, which are value-free; “future” corresponds to the generation of goal-oriented sequences

3

“Background” (Searle, 1979 and 1983) refers to a network of enabling conditions that makes it possible for particular forms of intentionality – such as meaning or action – to function. In other words, it is a set of practices, skills, habits, and stances that human beings possess, in virtue of both their biological make-up and sharing the same physical and sociocultural environment. Deep Background corresponds to a know-how common to the species, Superficial Background to one that is socio-culturally derived.

Common Ground, Categorization, and Decision-Making 201 of behaviour and depends mainly on the prefrontal cortex of the frontal lobe, with its strong “reentrant” connections with other cortical regions and its capacity for anticipatory set arising from a short-term working memory. The prefrontal cortex is also believed to have a key role in generating motor plans and converting them, via the basal ganglia, into motor programs (Edelman, 1989: 159–160).

3.2. Mechanisms for implementing linkages The basic mechanism is one occurring at the level of neuronal groups. It is known as “global mapping”. “Global mapping” ensures the linkage of maps of neurons from many parts of the brain. These linkages are implemented by a system of “temporally ongoing parallel signaling”, or “reentry” (Edelman, 1989: 93–102). Reentrant connections allow “cross modal” constructions (e.g., signals arising from different sensory features of an object to be brought together to yield a category corresponding to that object4 ). Reentry also allows resolution of conflicts (as when conflicting sensory signals are received), and, most importantly, “recursive synthesis”, a function whereby outputs from higher-order maps can be recycled to lowerorder maps for use5 in the generation of further neuronal responses. This, incidentally, bears some clear resemblance to what one would expect the interpretative process to be able to do, especially at the text level. Through reentry, the brain areas involved in concept formation, that is, the frontal, parietal, and temporal cortex, can stimulate portions of past global mappings, and combine and compare them. Alongside reentry there are convergence centres where these signals come together. Thus, the matching of values and (significant) perceptual categories appears to take place in the hippocampus; as for the linkage between sensory responses and motor responses, it would be mediated by the basal ganglia. Perceptual experience of objects and events – or primary consciousness – results from the reentrant interaction between a value-dominated memory and ongoing perceptual categorizations and their short-term succession. Insofar as there is a conceptual memory that ensures the dominance of internal values, a primary consciousness will give the animal a certain freedom from the present. This freedom, however, still falls short of allowing it to make plans. In order to have this capacity, the animal must be “conscious of being conscious”, that is, possess the beginnings of a higher-order consciousness. In its fully developed form, higher-order consciousness presupposes the ability to symbolize the self- and non-self-distinction, or rather to produce symbolic representations of the self6 acting on the world, or vice versa. We have now reached the stage where we have an animal capable of: • constructing a coherent internal model of past, present, and future, which enables it to make plans; • representing it symbolically.

4 Thus, signals corresponding to the shape and movements of a cat’s tail, which are processed separately, would be brought together by reentrant connections to enable its identification. 5 This would account for the fact that the membership criteria for a category can be both general and specific. 6 The notion of “self”, as used here, exceeds biological individuality and can only arise through social interactions (Edelman, 1989: 187).

202 Common Grounds By now, it has also developed a capacity for attentional selection that allows it to remain focussed on its goals. What is missing in this scenario is how the appropriate course of action is decided. This is where Damasio’s model of decision-making comes in.

4. ARGUMENTATIVE DISCOURSE AND DECISION-MAKING 4.1. Damasio’s model of decision-making A basic model of decision making would include the following two features: • a repository of knowledge consisting, among other things, of categories7 of situations requiring a decision paired with categories of response options (I will refer to them as “complex categories”). These, in turn, are linked with attendant categories of outcomes. This knowledge includes “…facts about objects, persons and situations in the external world, [but] also facts and mechanisms concerning the regulation of the organism as a whole” (Damasio, 1994: 83–84); • a reasoning process, which evaluates and ranks the available options prior to reaching a decision. As it stands, this model lacks efficiency in view of the number of response options to be evaluated within a given time frame, hence the need for a biasing system. Such a system allows response-options to be pre-selected. On Damasio’s account, it relies on the operations of “somatic markers”, or somatic states, with which positive and negative outcomes are tagged. Thus, as a result of past encounters with dogs, you may experience a positive or negative emotion at their sight, and this emotion will enhance or inhibit your tendency to be friendly or to give them a wide berth.8 When faced with a situation requiring a decision (in terms of adaptive behaviour), the very first thing an animal does is categorize this situation. This categorizing marks the initial step in the process whereby appropriate response options are selected by activating relevant complex categories in memory. Somatic markers then intervene to provide a shortlist of options, which serves as input to the reasoning process. The overall picture now in place is one in which, in order to be able to make plans, an animal has to be able to have an internal model of past, present, and future. In terms of supporting systems, this requires a conceptual memory, systems providing linkages of various kinds, and a decision-making system that mediates the interaction between this conceptual memory and the selection of appropriate actions. We can now turn to argumentative discourse as external representation and ask what additional requirements it places on the above apparatus.

7

We need to distinguish between acquired and innate knowledge. The former stems from experience and may vary across individuals; the latter is provided by the genome. 8 For further reference, note that the way somatic markers function involves two alternative mechanisms: one in which the somatic states are experienced by the subject; the other, where the body is bypassed.

Common Ground, Categorization, and Decision-Making 203

4.2. Argumentative discourse and induced decision-making As external representation, argumentative discourse is not about facilitating the speaker’s decision-making, but about inducing a certain decision-making process in others (if only at the interpretative level, in the first instance). This raises the question of how an interpretative and validation context would be put in place by argumentative discourse. In light of what precedes, the answer is, through categorization. When S puts forward an argument in support of a conclusion (e.g., “Mary does not know very many people”, in support of “Let’s invite her to our party”), the pertinent commitment, on her part, is not to the fact that the proposition/Mary does not know very many people/conforms to a state of affairs in the world; rather it is to the fact that the situation of reference falls into a category of situations that calls for the type of action described in the conclusion. In expressing such a commitment, S thus makes reference to a complex category, thereby causing this category, or close alternatives, to be represented in R’s working memory. Thus, initiating decision-making in others crucially depends on getting them to view the situation of reference as a member of a category of initial situations to which a category of response-options is paired, as determined by a complex category in S’s memory. Thus, “common” ground arising from argumentative discourse depends on a categorization process, which, in turn, is part of a decision-making apparatus. Categorization, in other words, is the process whereby background knowledge is accessed (in a selective fashion) to yield an outcome, which is imposed (or proposed) common ground. Common ground – like background knowledge – is thus to be construed in terms of complex categories. As part of induced decision-making, common ground, on account of its implicitness, is imposed rather than proposed. This raises a major difficulty: how can R, with her own valuedominated complex categories, be diverted from constructing her own validation context, carrying out her own evaluation, reaching her own conclusion? In other words: (a) How can R be prevented from making her own category judgements regarding the situation of reference?9 (b) How can she be prevented – for a period of time sufficient for current purposes – from carrying out her own evaluation process (i.e., from comparing and ranking different category judgements of the situation of reference according to her own criteria), and in the order of her choice? From the point of view of S, given what it entails in terms of serial linkage of the succession of complex categories being referred to, this would give rise to an undesirable validation context. (c) How can R’s own somatic markers be inhibited in their biasing action during the selection process? This question arises when the outcome under consideration is known by S to have been assigned different values by R. The above problems are, in fact, twofold, inasmuch as the implementation of S’s goal requires both the inhibition of the processes in question and their replacement with processes which more actively further this goal. Thus (a), (b), and (c) are best thought of as preparing the way for a , b , and c , respectively:

9

Such category judgements would presuppose that appropriate complex categories have been accessed from the respondent’s own background knowledge to provide the basis for her own validation context.

204 Common Grounds a  How can R be prevailed upon to consider certain category judgements of the situation of reference (which she is assumed not to have formed of her own accord), thereby allowing for certain complex categories to be represented in her working memory, and to provide a basis for a validation context similar to S’s validation context? b  How can R be prevailed upon to consider evaluating the category judgements present in her working memory in a certain order, thereby serially linking these category judgements, and hence the corresponding complex categories into the above validation context? As part of this evaluation process, actual and/or potential conflicts also have to be resolved to S’s satisfaction, that is, in a way consistent with her stances and the complex categories that make up her validation context. c  How can R’s selectional process be biased towards a certain outcome? To achieve all of the above, however, a further problem needs to be addressed, namely: (d) How can the stability of the various linkages involved be secured over a certain period of time? Without this kind of stability, no memory10 is possible. Stability, in this sense, is a matter of preserving the order in which categorizations occur, so that they are kept from shifting into unconnected sequences (Edelman, 1989: 115). This is also important for the functioning of the evaluation process, which requires that each category judgement takes place against the background of the previous category judgements (see also Maclaury (1997) on vantage construction and category development). Insofar as a  falls into the category of “selectional problems in the control of action”, the pre-existing means available to handle it lie in the attentional mechanisms (Allport, 1987: 407– 415; Edelman, 1989: 197–207), which select actions in accordance with a predetermined goal. Incidentally, somatic markers appear to have a crucial role in boosting attention (Damasio, 1994: 198). These same means would also address (a), as the ability to direct attention11 to a single action or goal necessarily implies excluding alternatives from its focus (Neumann, 1987: 373–382). To the extent that attention is inherent in all forms of representation, a  would be achieved by the sheer fact that the situation of reference is symbolically represented under those features that highlight its membership to a certain category of situations. By inhibiting, as a matter of course, representations of alternative category judgements, this representation also caters for (a). b , and (b) are also a matter for representation, except that overt representations (as occur in actual exchanges), which invite discussion at every turn, may prove less than satisfactory in terms of processing efficiency. Indeed, S would not have sufficient control

10

Memory in the TNGS arises from the biomechanical basis provided by alterations of synaptic strengths. These alterations lead to refinement in perceptual categorization, which, in turn, leads to a memory that is not replicative but “recategorical”. What is meant for memory to be recategorical is that it is a process involving “facilitated pathways, not a fixed replica or code” (Edelman, 1989: 111). Within this conception of memory, a particular categorical response can be recalled in forms other than the one in which it first occurred. A dynamic view of common ground is compatible with this procedural view of memory. 11 Attentional selection is adaptive in enabling the animal to select actions appropriate to a predetermined goal, and also in preventing any distraction from interfering with the achievement of this goal.

Common Ground, Categorization, and Decision-Making 205 over which category judgements are being considered, the order in which they are considered, and how actual or potential conflicts are resolved. This may be a contributing factor in the emergence of schematized situations as built-in prior context. The fact that these situations are conventionally linked to argumentative markers makes it possible for them to be automatically decoded, especially in their entirety, which means that individual stances cannot be readily singled out for discussion. Whether or not these situations constitute a truthful representation of the actual context of relevance, their conventional status will ensure that they are represented in R’s working memory as the prior context to bring to bear on the interpretation of S’s intention. In addition to the fact that they are automatically decoded in their entirety, their implicitness, which, to a large extent, allows them to remain outside the focus of attention, also makes them less vulnerable to any immediate challenge. In terms of imposed validation context, the complex categories provided for R’s consideration are the ones being referred to by the stances featured in the schematized situations; these stances (which are – to varying degrees – pre-assigned to specific interlocutors) occur in a more or less fixed sequence, thereby determining the order in which the categories at hand intervene. Schematized situations, as we have seen, can also include conflicting stances. The ones I am familiar with, however, are always resolved in a predictable fashion, as each schematized situation – albeit to varying degrees – imposes a certain argumentative orientation (or type of constraint on sequencing), one usually derived from S’s stances. Thus, in addition to an evaluation process, schematized situations also tend to impose a fixed outcome. This is achieved via an appropriate pattern of weighting privileging some viewpoints over others. In other words, we are dealing with a form of tagging germane to somatic marking. Schematized situations may not trigger somatic states, but they usually include preferred stances.12 Once instantiated these stances would determine which complex categories should be brought to bear on the selection of a response option. This takes care of c  and (c). Now, what about (d)? If the validation context that S requires cannot be stabilized for the necessary time period, it is unlikely to become shared common ground. This is where encoded situations come in. Each time an argumentative marker is used, the corresponding schematized situation is automatically decoded, with what it entails in terms of a set sequence of stances attributed to interlocutors, a fixed pattern of weighting assigned to these stances, and a predetermined type of outcome. As a result, a predictable type of validation context is also triggered, with complex categories occurring in a certain sequence, and standing in certain relations, as determined by the sequence of stances and their relationships. Which fully instantiated complex categories are actually triggered depends on the propositional contents being provided. These, however, may not always be sufficient to ensure the relevant degree of control over what gets

12

The usage of toujours exemplified in (3) conveys a measure of dismissiveness in respect of the argument being proposed. Pourtant, as in (4), expresses surprise to varying degrees, depending on the context: (4) A: Jean a passé son UV. Jean has passed his module. B: Pourtant, il n’a rien fichu du semestre. I am surprised/How come? He did precious little over the past semester. Toujours-est-il, as in (5), conveys a mental state ranging from smugness to defiance, depending on the tone and the context. (5) Je veux bien lui en toucher un mot; toujours-est-il qu’il est assez têtu. I don’t mind having a word with him; the fact still remains that he is rather stubborn.

206 Common Grounds represented in R’s working memory. After all, S can wrongly assume that she and R both share a given complex category and fail to provide enough contextual assumptions to ensure the representation of that category. Schematized situations prompt at least two questions: (1) How is one to account for the fact that their encoding came about in the first place? (2) Assuming that schematized situations give rise to an evaluation process, or rather to a representation of such a process, how is this supposed to bring R to a given conclusion? The encoding of schematized situations, as contexts of relevance, may have arisen as part of a context encoding tendency13 that would have emerged in response to problems of processing efficiency specific to the text level. In the case at hand, processing efficiency, which depends on a quick identification of S’s conclusion, is served by the immediate availability of a schematized context of relevance (and validation), which may be seen as contributing to a top–down14 mode of processing). Processes leading up to this encoding include the perceptual system – in particular the process15 whereby stimuli are encoded with their context of occurrence, processes underlying the recategorical conceptual memory, and the automatization process. The latter involves a strengthening of synaptic connections in frequently used pathways, which thus become established. Turning now to (2), how much influence schematized situations have on R is an open question. The likelihood of an automatic causal link between the decoding of a schematized situation and the acceptance of S’s conclusion is rather dubious, despite the fact that the level of mental representation being triggered in R’s working memory can be expected to bias the outcome.

4.3. Schematized situations, categorization, and validation context Consider again the schematized situations assigned to (1), (2), and (3), but this time in terms of categorization and validation context. (1) Mike aime les crocodiles, mais de loin. Mike is fond of crocodiles, but from a distance. The two main opposing viewpoints – V1 (imputed to R) and V2 (corresponding to S’s) – translate into two category judgements of the situation of reference, each presenting it under a different set of features, which highlight its membership to a different complex category: V1 (Mike should be given a crocodile (C), on grounds that he is fond of that species (P)) corresponds to a category judgement to the effect that the situation of reference is a member

13

See Nyan (2004, Chapter 2). An example of a top–down mode of processing is provided by constraints on sequencing, another by schematized situations, although the latter require instantiation. A bottom–up mode of processing would lack efficiency, as it involves constructing contexts from constituent elements. 15 This process is the one responsible for explicit memory encoding. Cf. Kosslyn and Koenig (1995: 317). 14

Common Ground, Categorization, and Decision-Making 207 of a certain category of situations describable in terms of someone’s fondness for something, which, in turn, is part of a complex category consisting in a set of initial situations calling for a decision, and a set of response options. This complex category – which may be seen as a neuro-cognitively grounded version of the topos proposed earlier – would, similarly to that topos, involve a linkage between situations featuring someone’s fondness for something and responses involving procuring it for them. V2 (Mike should not be given a crocodile C  because his fondness for the species stops short of a desire for close interactions P ) consists in viewing the situation of reference under a different aspect, one highlighting its membership to a different category of initial situations altogether: one requiring a more circumspect type of response. The complex category being referred to by V2 involves situations featuring someone’s consistent avoidance of close interactions with something, and responses reflecting a policy of non-interference on the part of those around him. Thus, viewing schematized situations in terms of category judgements referring to complex categories helps explain how such categories would come to be represented in R’s working memory to provide a validation context. In the case of mais the fact that the stances under evaluation stand in a relation of opposition entails16 that the complex categories being referred to also stand in a similar relation. This opposition is resolved through a form of tagging, whereby the second category judgement is marked as preferred. Such tagging may be seen as a further level of automatization to the one mediated by somatic markers: in this case, not only is the somatic state bypassed, but this “bypassing” has also become conventionalized.17 Linkage stability is ensured by the fact that the above linkages and tagging are encoded in the argumentative marker as integral aspects of its schematized situation. We next turn to: (2) Ce n’est pas une mauvaise idée. Maintenant [Octavio] a peut-être quelque chose de prévu. That’s not a bad idea. Now, [Octavio] may have made other plans. The schematized situation assigned to (2) includes: V1, a view emanating from R (to the effect that since Octavio is fun to have around, he should be asked to join them on a certain occasion); V2, which is V1 as sanctioned by S; V3, an alternative view of the situation of reference, one taking on board eventuality P (Octavio may have other plans). Should P turn out to be the case, R and S would have to amend their guest list.

16 Whether or not these complex categories, once instantiated, can actually be regarded as standing in such a relation is a different matter. Having a top–down mode of interpretation means that the schematized situation should prevail in the event of a clash with the contents available for instantiation. 17 Incidentally, concepts for objects, events and properties can be negatively or positively tagged (e.g., casualties/collateral damages and stingy/parsimonious), as has been noted by AT since the early 1990S.

208 Common Grounds Here we start out with a shared category judgement of the initial situation and a shared validation context, consisting in a complex category (CC1) linking a category of initial situations (featuring list A guests) to a category of actions (e.g., inviting such individuals to one’s parties). The situation of reference, as categorized in the prior context, belongs to the former category, and the action S concurs with (in V2) is subsumed under the latter. The appropriateness of this shared common ground is called into question by “maintenant”, which introduces an eventuality = V3 not previously envisaged. Such an eventuality (Octavio has other plans), if proved to be the case, would lead to a different category judgement of the situation of reference, and hence to a new complex category (CC2) being introduced as point of reference (where CC2 is a pairing of a category of initial situations featuring potential guests with prior commitments and a category of response-options involving the cancellation of prior plans in respect of those individuals). Taking CC2 on board would, accordingly, require a different course of action. Unlike what happens in (1), S stops short of imposing an alternative category judgement, leaving it to R to work out the implication of what she is saying. In terms of common ground, we are dealing with something conceivably about to shift. Consider finally: (3) Si on le descendait, ce sera toujours ça de fait. What about taking him out? At least we’ll be able to cross him off our list. In (3) the situation (as presented by the schematized situation) is one of assessing the merits of a proposed undertaking in response to a given situation. Unlike the previous cases, where the decision was about selecting a response option to match a certain categorization of the situation of reference, this case is concerned with selecting a response option on the basis of its predicted outcome. The first option (taking X out) is in the first instance associated with an incentive which is not ranked (they would be able to cross X off their list). Underpinning V1 is a complex category CC1 linking a set of job-incentive pairings with positive responses to them. V1 is the view that Option 1 should be chosen. In V2, Option 1 is recategorized as one associated with few incentives, which leads to a positive course of action being ruled out. This points to a complex category CC2, which associates jobs where incentives are few with a low level of response. Option 2 (not taking X out) belongs to this category. V3 agrees with the ranking proposed for Option 1, but not with the fact that its rejection should ensue. V3 is, accordingly, validated by a complex category CC3, which, irrespective of the level of incentive involved, pairs any jobs with positive attitudes. Inasmuch as V3 comes tagged as the preferred stance, CC3 emerges as the prevailing complex category, and Option 1 as the outcome of the evaluation process.

5. CONCLUSION As we have seen, a co-evolutionary approach allows us to define common ground construction – as mediated by argumentative markers – in terms of pre–existing means. These means include a recategorical conceptual memory (consisting of complex categories), mechanisms responsible for implementing linkages, and a decision-making apparatus.

Common Ground, Categorization, and Decision-Making 209 The type of common ground we have been concerned with is one imposed by schematized situations forming part of the meaning of the above markers. As such it is best thought of as what S wishes to impose as common ground, rather than common ground, as the outcome of a joint decision on the part of S and R. Imposed common ground, as envisaged here, is also distinct from background knowledge as held in conceptual memory: although the complex categories in the former constitute a subset of those from the latter, they are activated in working memory, where they may be retained for a required time period. These categories are accessed through categorizations of the situation of reference, as imposed by instantiated schematized situations. In other words, the construction of this imposed common ground is mediated by the representation in working memory of a decision-making process triggered by schematized situations. The extent to which at the end of this process a genuinely shared common ground can emerge is difficult to ascertain.

ACKNOWLEDGEMENTS This article stems from the body of research I was able to undertake with the support of the Rockefeller Foundation. I also wish to acknowledge the aid I received from the British Academy in its preparation. Finally I am grateful to my anonymous reviewers for their helpful comments.

REFERENCES Allport, D.A., 1987. Selection for action: Some behavioral and neurophysiological considerations of attention and action. In: Heuer, H. and Sanders, A.F. (eds), Perspectives on Perception and Action. Lawrence Elbaum Associates, Hillsdale, NJ, 395–419. Anscombre, Jean Claude (ed.), 1995. La Théorie des Topoi. Kimé, Paris. Cadiot, Anne, Ducrot, Oswald and Vicher, Anne, 1985. Sous un mot une controverse: les emplois pragmatiques de toujours. Modèles Linguistiques 7 (2), 106–124. Clark, H. Herbert, 1996. Using Language. Cambridge University Press, Cambridge. Damasio, Antonio, 1994. Descartes’ Error. The Free Press, New York. Deacon, W. Terrence, 1997. The Symbolic Species. Norton, New York. Edelman, D. Gerald, 1989. The Remembered Present. Basic Books, New York. Fetzer, Anita, this volume. Reformulation and common grounds. Fischer, Kerstin, this volume. Grounding and common ground: Modal particles and their translation equivalents. Kosslyn, Stephen, M. and Koenig, Olivier, 1995. Wet Mind. The Free Press, New York. Maclaury, Robert E., 1997. Color and Cognition in Mesoamerica. Constructing Categories as Vantages. University of Texas Press, Austin. Nemo, François, this volume. The pragmatics of common ground: From common knowledge to shared attention and social referencing. Neumann, Odmar, 1987. Beyond capacity: A functional view of attention. In: H. Heuer and Sanders, A.F. (eds), Perspectives on Perception and Action. Lawrence Erlbaum Associates, Hillsdale, NJ, 361–94.

210 Common Grounds Nyan, Thanh, 1998. Metalinguistic Operators, with Reference to French. Peter Lang, Bern. Nyan, Thanh, 2004. Meanings at the Text Level: A Co-evolutionary Approach. Peter Lang, Oxford. Searle, John, R., 1979. Expressions and Meaning. Cambridge University Press, Cambridge. Searle, John, R., 1983. Intentionality. Cambridge University Press, Cambridge. Searle, John, R., 1992. The Rediscovery of the Mind. The MIT Press, Cambridge. Searle, John, R., 1995. The Construction of Social Reality. The Free Press, New York. Searle, John, R., 2001. Rationality in Action. The MIT Press, Cambridge. Thomason, Richmond, H., 1992. Accommodation, meaning and implicatures: Interdisciplinary foundations of pragmatics. In: Cohen, P.P., et al. (eds), Communication. The MIT Press, Cambridge, 325–263.

INDEX

Acceptance, 49, 58, 79, 97, 99, 169 Accommodation, 5, 119, 120, 163 Accountability, 171 Acknowledgement, 48, 49, 62, 114 Adaptive behaviour, 200, 202 Adjacency pair, 20, 36, 161 Administration, 4, 7, 11, 160, 164, 165, 170, 173 Adverb, 51, 54, 55, 69, 70, 81, 82 after, 10, 11, 113, 116, 121–3, 125, 127, 129, 132, 184 Agreement, 20, 29, 31, 93 allerdings (German), 58 also (German), 50, 60 Ambiguity, 96, 104, 183, 191, 192 and-prefacing, 59 Anscombre, J.-C. and Ducrot, O., 143, 145, 146, 196 Appropriate/inappropriate, 170 Argumentation: argumentative background, 47, 50, 51, 55, 60, 61, 62 argumentative context, 51, 52, 53, 57 argumentative discourse, 195, 196, 199, 202–209 argumentative marker, 195, 196, 198, 205, 207, 208 argumentative situation, 60 in language theory (ALT), 145, 146, 147 Assessment, 8, 126, 128, 159, 169 Assumption: first-order, 160 second-order, 160 Attention, 11, 26, 48, 96, 107, 119, 120, 143–56, 204 Automatization, 169, 206, 207 Autonyms, 103 Backchannel verification, 19 Background/backgrounding:

information, 62, 125, 126, 144, 160, 164, 167, 169, 170 knowledge, 144, 160, 186, 200, 203, 209 before, 10, 11, 60, 113, 116, 121–3, 124, 125, 126, 127, 128, 129, 132, 137, 139 better, 57, 58, 62, 126 Bridging, 150 Capacity: for argumentative discourse, 195 for higher-order consciousness, 195 representational, 196, 199 Categorization: current, 200 of initial situations, 51, 195, 196, 203, 207, 208 of response options, 51, 195, 202, 203, 208 of supporting structures, 199–202 of the world, 200 Category: complex, 202, 203, 204, 205, 206, 207, 208 value-dominated, 200 value free, 200 Clark, H.H., 6, 7, 8, 9, 17, 18, 19, 20, 21, 29, 37, 39, 41, 42, 48, 49, 68, 89, 92, 93, 94, 95, 96, 97, 98, 114, 116, 118, 119, 120, 123, 160, 161, 162, 166, 167, 168, 198 Co-evolution, 11, 195, 198, 199 Co-supposition, 6, 7, 164, 165 Collective: intentionality, 99–100 we-intention, 6, 164 Common ground: collective dialogue common ground, 6, 7, 11, 18, 68, 72, 73, 80, 82, 84, 160, 164–5, 170, 176 common ground construction, 196, 198 communal common ground, 6, 9, 160, 161–2, 164, 165 cultural common ground, 11, 68, 154 discursive common ground, 18, 41, 154

212 Index Common ground (Continued) general common ground, 68, 72, 73, 76, 77, 79, 80 individual dialogue common ground, 6, 8, 11, 68, 73, 75, 76, 80, 84, 85, 160, 164–5, 170 interlocutive common ground, 48, 154, 155 personal common ground, 6, 68, 161–2, 164 signalling common ground, 48 Communal lexicon, 9, 89, 93, 95, 108 Communicative: contribution, 3, 4, 5, 7, 8, 11, 159, 163, 168, 176, 178 project, 164 situation, 48, 50, 51, 52, 54, 59 Computer mediated interaction, 18, 22, 23, 24, 26, 27, 28, 32, 33, 36, 37, 38, 39, 40 Concept formation, 201 Conceptual memory, 195, 199, 200–201, 202 Concessive, 73, 78, 83 Conditional, 77, 78, 150 Conjunction, 51, 54, 55, 59, 69, 79 Connective, 113–32 Consensus, 89, 90, 93, 106, 154 Context/co-text: cognitive, 3, 4, 186 linguistic, 3, 4, 11 personal, 186 physical, 18, 186 prior, 205, 208 social, 2, 3, 4, 165 sociocultural, 3, 4, 165 textual, 17, 98, 114, 115, 124, 186 validation, 195, 197, 203, 204, 205, 206 Contextual: effect, 97, 185–7 knowledge, 41 Contextualization cue, 1, 8, 50 Continuer, 19 Contribution, 3, 6–7, 48, 60, 78, 159, 164, 166, 168, 169, 170, 171, 172, 175 Control, 151–2, 169 Conversational maxim, 144 Cooperative principle, 166, 172, 173 Coparticipant, 48, 68, 74, 159, 163, 164, 166, 167, 169, 171 Corpora, 11, 20, 21, 31, 92, 104 Decision: making routine, 22, 28, 29 routine, 10, 18, 20, 22, 25, 29, 30, 33–7, 39, 41, 42

Declarative, 73, 79, 81, 161 Default: expectation, 41, 42 scenario, 166, 168 verification, 19 denn (German), 59, 60 Dialogue, 2, 7, 8, 11, 37, 51, 53, 60, 78, 82, 165–78 Discourse: boundaries, 49, 50 function, 18, 22, 23, 24, 25, 29, 37 marker, 10, 17–43, 49, 69 particle, 10, 49–50, 61 pragmatics, 67, 113 relation, 90, 121, 123, 128, 155 routine, 20, 39, 41, 42 topic, 169, 173 doch (German), 10, 50, 52, 53, 57, 58, 60, 67–85, 116, 172 Elaboration, 31, 32, 33, 125, 128 Ellipsis, 191 Ethnomethodology, 5, 165 Evaluation, 25, 53, 105, 176, 177 Exclamation, 76 Existential placeholder, 106 Expansion, 162 Expert community, 93, 94, 96, 108, 109 Explicit: language, 18 management, 26–7, 36, 37, 39, 40 metalinguistic operator (EMO), 10, 90, 100–107 Face, 18, 20, 22, 23, 25, 26, 28, 33, 37, 152 Felicity condition, 2, 50, 99–100, 118, 161 First pair part, 21, 25, 36 Focus: accent, 10, 70, 80–1, 84 verum, 81, 84 Foregrounding, 8, 115 Frame/framing: attentional, 11, 143 modal, 148–9, 150 scalar, 11, 143 Garfinkel, H., 5, 7, 163, 171 Given/new, 115, 117–19, 124 Global mapping, 201 Goal: goal-oriented activity, 195, 199 higher-level, 167

Index 213 Goffman, E., 118, 143 Goody, E., 162, 166, 169 Grammaticalization, 69, 70, 98, 171 Grice, H.P., 5, 11, 74, 99, 143, 144, 145, 166, 172, 173, 177 Grounding, 1, 5–12, 47–63 Gumperz, J., 1, 5, 8, 48, 50, 56 Habermas, J., 166, 167, 168 Halliday, M.A.K., 89, 114, 115, 117, 118, 128, 130, 185 Hedge/hedging, 62, 104, 106 Higher-order consciousness, 195, 199, 200, 201 Illocutionary force, 106, 172, 173 Imperative, 74–5 Implication, contextual, 186–7 Implicature: conventional, 71 conversational, 173, 174–6, 178 Incongruence, 167, 170 Indexical, 2, 51, 159, 165, 172 Information: extraction, 106 management, 4, 166, 170, 176 status, 122, 128 structure, 118, 124, 152 unit, 79 Informativity, 91 Inhibition, 203 Intentionality, 5, 99–100, 163, 165, 200 Interactional work, 40 in this case, 57 Irony, 184–5 Joint: action, 6, 17, 20, 39, 68, 114, 162 project, 19, 20, 36, 37, 38, 49, 68, 161, 198 Justification, 52, 114, 151, 154, 161 Knowledge: acquired, 200 base, 5, 93, 95, 108, 146, 151, 154 declarative, 8, 161 encyclopaedic, 160, 161 know-how, 198 know-that, 198 lexical, 108 linguistic, 161

procedural, 8, 161 wired-in, 200 Language: engineering, 108 technology, 109 Learning algorithm, 107, 108 Levinson, S., 2, 7, 52, 59, 144, 145, 173 Linell, P., 166, 184 Linkage/linking: serially, 204 maintenant (French), 11, 12, 195, 196, 197, 208 mais (French), 11, 12, 195, 196, 207 Manifestness, mutual, 3, 162–3 Meaning: adversative, 70 intersubjective, 6, 11, 159, 164, 166, 170 invariant, 51, 53, 54 procedural, 172, 198 Memory: conceptual, 200–201 value-dominated, 195, 199, 200 Metalanguage, 90–2, 96, 106, 107, 152 Metalinguistic: information database (MID), 108 marker, 96, 98, 103–104 operator processor (MOP), 106, 107–108 predication, 100–107 speech act, 11, 96, 98, 99, 109 Metapragmatic: instruction, 10, 69, 73, 84 Metastable system, 94 Miscommunication, 109, 184, 185 Misunderstanding, 11, 19, 183–93 Move function, 25 Mutual: contextual belief, 94 responsibility, 167, 168 Natural language processing, 2, 107 Navigation: horizontal, 20, 28 vertical, 20, 28 Negotiation process, 53, 98 Network, 93, 171 now, 11, 24, 39, 43, 183 ok, 10, 18, 19–20, 21, 22, 23, 24, 26, 28–41, 116 on the one hand, 57 on the other hand, 57, 62

214 Index Ontological commitment, 93, 94, 106 Operator, 5, 97, 98, 100, 101, 102, 106 Optative, 71, 77 Orientation, 6, 20, 25, 29, 30, 32, 33, 34, 36, 39, 205 Ostensive-inferential communication, 143, 151 Particle: discourse, 10, 17, 47, 49, 50, 54, 60, 61, 62 modal, 10, 47–63, 69–70, 71, 79 Perlocutionary effect, 97, 169, 170, 173, 177–8 Pragmatic pretext, 51, 52, 53, 54, 57, 58, 70 Pre-sequence, 198 Presupposition: interactional, 6, 160, 164, 167, 169, 170, 179 interpersonal, 6, 151, 160, 164, 167, 169, 170 pragmatic, 2, 172 semantic, 2 Principle of justification, 161 Processing: efficiency, 204, 206 parallel distributed, 165 Question: deliberative, 75, 76 rhetorical, 75, 76, 175 wh-, 76, 78 rather, 57, 62, 189 Ratification, 6, 7, 11, 154, 164, 166–9 Rationality, 94, 165 Reasoning: practical, 163 process, 202 Record: conversational, 5, 6, 8, 68, 93, 97, 163 of prior interaction, 198 Reentrant connections, 201 Reference, 2, 3, 7, 91, 100, 128, 165, 167, 168, 170, 174, 203 Reformulation: other-initiated, 169, 172 other-reformulation, 11, 169, 171, 174, 175, 178 self-initiated, 169, 172 self-reformulation, 169, 171, 172, 175, 178 Rejection, 167, 168, 171, 177, 208 Relevance: relevance theory, 3, 11, 54, 145, 146, 147, 162, 186

Repair, 26, 95–7, 184 Response, 21, 27, 36, 201, 207 Sbisà, M., 2, 168 Schegloff, E., 19, 37, 47, 48, 154, 165, 191 Searle, J.R., 2, 5, 11, 99, 161, 165, 167, 169, 200 Second pair-part, 21, 30, 36 Situation: model, 93, 94, 97 of reference, 51, 196, 197, 203, 204, 206, 207, 208 schematized, 11, 196, 197, 198, 205, 206–209 so, 22, 26, 39, 43, 105 Somatic: marker, 202, 203, 204, 207 state, 202, 205, 207 Speech act: metalinguistic, 11, 96, 98, 99, 109 Sperber, D. and Wilson, D., 3, 11, 54, 91, 97, 98, 143, 145, 147, 148, 151, 160, 161, 162, 172, 186, 187, 189, 192 Strategy: communicative, 171, 172 lexical, 11, 171–3 Subjectification, 69, 171 Subjunctive, 58, 60, 62, 71, 77 Sublanguage, 92, 93, 94, 96, 100, 106 Suggestion, 20, 21, 22, 25–6, 27, 29, 30, 31, 32, 36, 39, 40 Terminology, 39, 48, 94–7 then, 58, 60, 121, 183 Thomason, R.H., 5, 6, 8, 48, 68, 93, 160, 161, 163, 164, 198 Topos, 150, 197, 207 toujours (French), 11, 12, 195 Transition, 19, 21, 28–33, 38 Translation, 47–63 Trognan, A., 167, 168 Update/updating: common ground, 10, 62, 72, 78 operation, 7 operator, 5, 10 Utterance unit, 24 Validity claim, 167, 168, 169 well, 24, 26, 29, 30, 31, 32, 33, 50, 60 yeah, 17, 18, 19, 20, 24, 26, 30, 41, 43, 48, 49, 50, 60