Inquiries in Philosophical Pragmatics: Theoretical Developments 9783030564360, 9783030564377

218 75 5MB

English Pages [210] Year 2021

Report DMCA / Copyright


Table of contents :
About the Authors
Introduction: Inquiries in Philosophical Pragmatics – Theoretical Developments
Three Mistakes About Semantic Intentions
Mistake (I): Intending to Refer
Objection 1: Implausible Starting Point
Objection 2: Incomplete
Objection 3: Redundant
Objection 4: Misleading
Mistake (II): Intending to Communicate
Mistake (III): Constraints on Intentions
Common-Knowledge-Based Pragmatics
A Coordination Problem Model of Communication
Two Problems with the Gricean Answer
Common Knowledge Generators
Generating Common Knowledge in Communication Coordination Problems
Social Roles
Common Knowledge-Based Pragmatics
The Primacy of Semantics and How to Understand It
The Territorial Wars: Semantics Versus Pragmatics
The Pragmaticist Proposal
The Semanticists Strike Back: A Radical Proposal
The Guidance Proposal
The Isolation Strategy and Why It Is Misleading
Anaphora and Guided Saturation
‘Few’, ‘A Few’, ‘Only’: Negative Quantifier Noun Phases and Negative Polarity Items – The Horn-Atlas Debate 1991–2018
Historical Introduction
“Only Proper Name” and “Few N” Sentences: Entailments, Implicata, and Monotonicity
Zwarts’s (1996) De Morgan Taxonomy of Negative Noun Phrases
“Only Proper Name”: The Standard View (Geach 1962, Pre-2002 Horn)
“Only Proper Name”: The Stuttgart View (Atlas 1991, 1996)
The Post-Horn (1996) Pragmatic View
The Standard View of Few N: Chomsky (1972), Horn (1989, pp. 244–250), Zwarts (1996)
The Atlas View of Few N
Assertion and the Division of Semantic Labor
Peter Geach’s Account of “Only Proper Name” Sentences: Excluding Force
What Are Geach’s Options?
The Atlas Dis-Solution of the ‘Only’ Problem
Speech-Act-Theoretic Explanations of Problems of Pure Indexicals
Problematic Cases of the Personal Pronouns I and You
The Analyses Given by Goffman (1981) and Austin ([1962]1975)
Analysing the Problematic Cases
Concluding Remarks
Prospective Reference
1. Theories of Prospective Reference
2. The Practice of Prospective Reference
2.1. Commerce and Law
2.2. Abortion Discourse
3. Experimental Data
4. Significance of the Data
4.1. Traditional Lexical Semantics
4.2. Semantic Externalism
Categorization, Memory and Linguistic Uses: What Happens in the Case of Polysemy
Polysemy as a Semiotic Necessity
Perspectives on the Study of Polysemy
Lexical Pragmatics and Attention to Use
Some Proposals for a “Dynamic” Interpretation of Polysemy
Interpretation of Copredicative Sentences: A Rich Underspecification Account of Polysemy
Classification of Underspecification Models
Core-Meaning Approaches
Thin Semantic Approaches
Rich Underspecific Structures
Analysis of Empirical Results
Representation, Interpretation and Copredication in the Activation Package Model
First Person Implicit Indirect Reports in Disguise
The Scope of Pragmatics
First Person Implicit Indirect Reports
Second Person Indirect Reports and Implicit ‘De se’ Attributions
Operation Cases
Extending the Analysis
Assessing Dialectical Relevance Using Argument Distance
The Pragmatic Nature of Dialectical Relevance
Probative and Topical Relevance
The Nuclear Waste Example
The Bank Robbery Example
The Tuna Fish Example
Modeling Probative Distance with Carneades
First Abstract Example
Second Abstract Example
The Communicative Functions of Metaphors Between Explanation and Persuasion
Metaphors: Between Persuading and Explaining
Explanation Versus Persuasion
Persuasive and Explanatory Metaphors
Metaphors and the Blurred Distinction Between Explanation and Persuasion
Describing the Conversational Goals of Metaphorical Utterances
Metaphorical Moves
From Metaphorical Utterances to Metaphorical Moves
Goals of Metaphorical Explanations
Stereotypes Favour Implicatures and Implicatures Smuggle Stereotypes: The Case of Propaganda
The Evolution of Critical Attention: Implicit Strategies as Persuasive Devices
Stereotypes in the Semantics of Advertising
The Combined Effects of Stereotypes and Implicatures in Advertising and Propaganda
Recommend Papers

Inquiries in Philosophical Pragmatics: Theoretical Developments
 9783030564360, 9783030564377

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

Perspectives in Pragmatics, Philosophy & Psychology 27

Fabrizio Macagno Alessandro Capone  Editors

Inquiries in Philosophical Pragmatics Theoretical Developments

Perspectives in Pragmatics, Philosophy & Psychology Volume 27 Editor-in-Chief Alessandro Capone, University of Messina, Italy Consulting Editors Keith Allan, Monash University, Australia Louise Cummings, The Hong Kong Polytechnic University, Hong Kong Wayne Davis, Georgetown University, Washington, USA Igor Douven, University of Paris-Sorbonne, France Yan Huang, University of Auckland, New Zealand Istvan Kecskes, State University of New York at Albany, USA Franco Lo Piparo, University of Palermo, Italy Antonino Pennisi, University of Messina, Italy Francesca Santulli, Ca’ Foscari University of Venice, Italy Editorial Board Members Noel Burton-Roberts, University of Newcastle, Newcastle, Australia Brian Butler, University of North Carolina, Asheville, USA Marco Carapezza, University of Palermo, Palermo, Italy Felice Cimatti, Università della Calabria, Cosenza, Italy Eros Corazza, Carleton University, Ottawa, Canada Marcelo Dascal, Tel Aviv University, Tel Aviv, Israel Michael Devitt, City University of New York, New York, USA Frans van Eemeren, University of Amsterdam, Amsterdam, The Netherlands Alessandra Falzone, University of Messina, Messina, Italy Neil Feit, State University of New York, Fredonia, USA Alessandra Giorgi, Ca’ Foscari University of Venice, Venice, Italy Larry Horn, Yale University, New Haven, USA Klaus von Heusinger, University of Stuttgart, Stuttgart, Germany Katarzyna Jaszczolt, University of Cambridge, Cambridge, UK Ferenc Kiefer, Hungarian Academy of Sciences, Budapest, Hungary Kepa Korta, ILCLI, Donostia, Spain Ernest Lepore, Rutgers University, New Brunswick, USA Stephen C. Levinson, Max Planck Institute for Psycholinguistics, Nijmegen, The Netherlands Fabrizio Macagno, New University of Lisbon, Lisbon, Portugal Jacob L. Mey, University of Southern Denmark, Odense, Denmark Pietro Perconti, University of Messina, Messina, Italy Francesca Piazza, University of Palermo, Palermo, Italy Roland Posner, Berlin Institute of Technology, Berlin, Germany

Mark Richard, Harvard University, Cambridge, USA Nathan Salmon, University of California, Santa Barbara, USA Stephen R. Schiffer, New York University, New York, USA Michel Seymour, University of Montreal, Montreal, Canada Mandy Simons, Carnegie Mellon University, Pittsburgh, USA Timothy Williamson, University of Oxford, Oxford, UK Anna Wierbizcka, Australian National University, Canberra, Australia Dorota Zielinska, Jesuit University of Philosophy and Education Ignatianum, Kraków, Poland More information about this series at

Fabrizio Macagno  •  Alessandro Capone Editors

Inquiries in Philosophical Pragmatics Theoretical Developments

Editors Fabrizio Macagno Universidade Nova de Lisboa Lisbon, Portugal

Alessandro Capone University of Messina Messina, Italy

ISSN 2214-3807     ISSN 2214-3815 (electronic) Perspectives in Pragmatics, Philosophy & Psychology ISBN 978-3-030-56436-0    ISBN 978-3-030-56437-7 (eBook) © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland


The editors would like to thank the Faculdade de Ciências Sociais e Humanas of the Universidade Nova de Lisboa and the University of Messina for their support. This work was supported by the Fundação para a Ciência e a Tecnologia (research grant no. PTDC/FER‐FIL/28278/2017).



 Introduction: Inquiries in Philosophical Pragmatics – Theoretical Developments����������������������������������������������������������������������������������������������������    1 Fabrizio Macagno and Alessandro Capone  Three Mistakes About Semantic Intentions��������������������������������������������������    9 Michael Devitt Common-Knowledge-Based Pragmatics ������������������������������������������������������   21 Richard Warner  The Primacy of Semantics and How to Understand It��������������������������������   33 Nenad Miscevic  ‘Few’, ‘A Few’, ‘Only’: Negative Quantifier Noun Phases and Negative Polarity Items – The Horn-Atlas Debate 1991–2018������������������������������������   49 Jay David Atlas  Speech-Act-Theoretic Explanations of Problems of Pure Indexicals����������   63 Etsuko Oishi Prospective Reference��������������������������������������������������������������������������������������   79 Paul Saka  Categorization, Memory and Linguistic Uses: What Happens in the Case of Polysemy ����������������������������������������������������������������������������������   95 Grazia Basile  Interpretation of Copredicative Sentences: A Rich Underspecification Account of Polysemy ������������������������������������������������������  111 Marina Ortega-Andrés  First Person Implicit Indirect Reports in Disguise ��������������������������������������  133 Alessandro Capone  Assessing Dialectical Relevance Using Argument Distance ������������������������  149 Douglas Walton vii



 The Communicative Functions of Metaphors Between Explanation and Persuasion������������������������������������������������������������  171 Fabrizio Macagno and Maria Grazia Rossi  Stereotypes Favour Implicatures and Implicatures Smuggle Stereotypes: The Case of Propaganda��������������������������������������������  193 Edoardo Lombardi Vallauri

About the Authors

Fabrizio  Macagno  (Ph.D.  UCSC, Milan, 2008) is Assistant Professor of Philosophy and Communication Sciences at the Faculdade de Ciências Sociais e Humanas of the Universidade Nova de Lisboa. He is author of more than 80 papers on definition, presupposition, argumentation schemes, and dialogue analysis published in major international peer-reviewed journals such as Discourse Processes, Discourse Studies, Lingua, Journal of Pragmatics, Pragmatics and Cognition Intercultural Pragmatics, Argumentation, and Philosophy and Rhetoric. He applied the theoretical insights developed in the fields of argumentation theory and pragmatics to the fields of legal interpretation, classroom dialogue, and medical communication, leading and collaborating in European and Portuguese major research projects. His most important publications include the books Argumentation Schemes (CUP 2008), Emotive Language in Argumentation (CUP 2014), Interpreting Straw Man Argumentation (Springer 2017), and Statutory Interpretation: Pragmatics and Argumentation (CUP 2020). Alessandro Capone  is Full Professor of General Linguistics in the Department of Cognitive Science at the University of Messina. He has a doctorate from the University of Oxford, where he studied with James Higginbotham and Yan Huang, and one from the University of Palermo, where he studied with Franco Lo Piparo. He has two habilitations as Full Professor of Linguistics and Philosophy of Language. He is chief editor of the Springer series Perspectives in Pragmatics, Philosophy and Psychology. He is co-editor for the journal Pragmatics and Philosophy, Mouton De Gruyter. He has published a monograph for Springer, entitled The Pragmatics of Indirect Reports: Socio-philosophical Considerations (2016). He has published papers in Lingua, Linguistics, Pragmatics and Cognition, Pragmatics and Society, Journal of Pragmatics, Intercultural Pragmatics, Australian Journal of Linguistics, La Linguistique, Argumentation, RASK: International Journal of Language and Communication, International Journal of Language Studies, Oxford working papers in Linguistics, Reti Saperi Linguaggi,



About the Authors

and Lingua e Stile. He has published sixteen volumes with CSLI, University of Chicago Press, Springer, Clueb (Bologna), and ETS (Pisa). He is a member of the editorial boards for Lingua, Journal of Pragmatics, Intercultural Pragmatics, Pragmatics and Society, International Journal of Language Studies, and Brill Research Studies in Pragmatics.

Introduction: Inquiries in Philosophical Pragmatics – Theoretical Developments Fabrizio Macagno and Alessandro Capone

Together with the volume “Inquiries in philosophical pragmatics: Linguistic and theoretical issues,” this book collects selected contributions to the conference Pragmasophia II held in Lisbon in 2018. The purpose of this twofold selection of essays is primarily to gather the state of the art on the interconnection between pragmatics and philosophy. While the field of pragmatics has developed noticeably in the last years, leading to excellent and famous journals in the field of linguistics and communication such as Pragmatics, Journal of Pragmatics, Intercultural Pragmatics, Pragmatics and Society, and Lingua, the philosophical dimension of this crucial discipline seems to be left in the background. The aforementioned journals aim at promoting and publishing interdisciplinary research on topics socially important, such as language learning, intercultural communication, political discourse, and multimedia communication. However, among the great number of contributions focusing on traditionally “core” linguistic issues (phonology, morphology, syntax and semantics) and more empirical communication studies, the philosophical dimension of pragmatics almost disappears. In the world of peer-reviewed journals, the strongest effort to revendicate the central role of philosophy and theoretical developments in pragmatics has been made by Intercultural Pragmatics, whose editor, Istvan Kecskes, launched in 2018 the Thematic Issue in Pragmatics and Philosophy, managed by Alessandro Capone.

F. Macagno (*) Universidade Nova de Lisboa, Lisbon, Portugal e-mail: [email protected] A. Capone University of Messina, Messina, Italy e-mail: [email protected] © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 F. Macagno, A. Capone (eds.), Inquiries in Philosophical Pragmatics, Perspectives in Pragmatics, Philosophy & Psychology 27,



F. Macagno and A. Capone

The relative underrepresentation of theoretical contributions in the arena of journals does not mirror the importance of philosophy, and theoretical and methodological research, in the field of pragmatics, and the attention that it deserves and draws. This role has been made manifest by the success of the Springer book series Perspectives in Pragmatics, Philosophy and Psychology, edited by Alessandro Capone. This book series, which counts at present 23 titles, has reached impressive distribution figures in 7 years (almost 300000 downloads for the digital editions). By displaying the most recent works by the most important philosophers working on theoretical and methodological pragmatic issues, this book series intends to offer a constant update on this almost hidden but essential dimension of pragmatics. The two volumes intend to explore from distinct perspectives the role of pragmatics in social sciences. Pragmatics is a complex discipline, commonly defined as the study of meaning by virtue of, or dependent on, the use of language (Huang 2014, p. 2; Jaszczolt 2018, p. 134). This generic definition is in fact very specific, as it addresses in particular two crucial dimensions of meaning, which are: 1) the linguistic acts (how words can be used to do things, using Austin’s terminology, Austin 1962, p. 2), and 2) the ways in which the linguistic context determines the proposition expressed by a given sentence in that context (Stalnaker 1970, p. 287), namely how we interpret and comprehend utterances (Kecskes 2013, p.  21). These two objects of study have been addressed from two distinct and interrelated approaches, a philosophical and a linguistic one. The first one takes a top-down perspective, focusing on the conditions and the principles underlying meaning and interpretation, illustrating how they can explain the phenomena we can observe. The second moves from the analysis of linguistic data – which include texts and discourses in different languages, different types of dialogues, different types of interactions, and different modes for expressing meaning – looking for the regularities that govern our production and processing. The philosophical and the linguistic perspectives are captured in these two volumes, each devoted to the two natures of pragmatics. They both follow a similar path of inquiry, moving from the classical topics and methods to the newest challenges of the discipline, namely new theoretical viewpoints or new domains of study. The first volume is devoted to the theoretical developments, namely the distinct philosophical perspectives that are laying the grounds for the growth of this field of study in different directions. Thus, the more explored themes are progressively joined by new methodological proposals that point at the interconnections between pragmatics and discourse studies or conversational analysis. The second volume pursues a symmetrical endeavor, starting from the classical linguistic analyses and more established methods and moving to pragmatic accounts of specific activity types and institutional contexts of language use, until proposing explorations in the new areas of schizophrenic discourse and multimodal communication. This first volume intends to contribute to the dialogue between philosophers and linguists, trying to broaden the boundaries of this discipline defined by the crucial notions of context and verbal action. To this purpose, the contributions are collected in an order that reflects the core and the frontiers of pragmatics, the former constituted by the classical and more philosophical topics as quantifiers, semantic

Introduction: Inquiries in Philosophical Pragmatics – Theoretical Developments


intentions and semantics, and common knowledge, and the latter exploring areas such as the relationship between pragmatics and other fields, such as argumentation and discourse analysis. Between these two poles of theoretical developments, we can find new theoretical challenges to some basic pragmatic problems, such as pure indexicals, deferred reference, polysemy, explicatures and indirect reports. The first broad category of papers is characterized by a purely philosophical approach, which establishes the foundations of the field of pragmatics. The first three papers discuss the role of intentions, semantics, and common knowledge in pragmatics, while the fourth analyzes a very specific issue, the semantics and pragmatics of “only,” through a logical investigation. In “Three Mistakes About Semantic Intentions,” Devitt argues that in semantics three commonly accepted paradigms need to be discussed and corrected. First, intentions to refer are considered to determine reference, but this view is shown to be implausible, incomplete, redundant, and misleading. Second, the common view on Gricean “intention-based-semantics” is that speaker meanings are constituted by the speakers’ communicative intentions. However, this claim cannot be taken to mean that it is essential to the meaningful use of language that it involve an intention to communicate. Rather, Devitt claims, meaning is constituted by the content of the thought expressed. The last misconception concerns the belief constraints that Griceans normally maintain to exist on what a speaker can intend to communicate by an utterance (such as, “X cannot intend to A unless X believes that she will A”). However, Devitt shows how these constraints cannot be accepted neither if taken as constitutive or normative. Intentions lead to another fundamental topic of pragmatics, namely the relationship between reasoning and common knowledge. As Warner underscores in “Common-Knowledge-Based Pragmatics,”, pragmatics involves a kind of ampliative inference, a complex type of reasoning commonly described as inference to the best explanation, application of general communication principles, or induction. This sophisticated type of reasoning is shown by the author to incur a coordination issue: how can speakers and audiences reason in the required ways? This chapter explores the role of common knowledge in the speaker-­hearer coordination game building on the insights from game theories, which define common knowledge as a recursive belief state in which people are know something, know they know it, know they know they know it, ad infinitum. According to Warner, common knowledge arises from the communication roles that an interaction creates: by perceiving their mutual roles in an interactions, the interlocutors commonly know that the conditions for the implicatures hold, and that reasoning is attributed to speakers and hearers. One of the central challenges of philosophical approaches to pragmatics is to define its place within the “realm of meaning.” Two opposite views characterize this discussion. On the one hand, semanticists consider semantics the primary vehicle of meaning, maintaining that the encoded semantic content is sufficient or almost sufficient for determining truth conditions. On the other hand, pragmaticists claim that truth conditions can be attributed only to an enriched semantic representation, namely a development of the logical form that results from the encoded (semantic)


F. Macagno and A. Capone

meaning. However, in “The Primacy of Semantics and How to Understand it,,” Miscevic underscores that according to this latter view the relationship between semantics and pragmatics becomes unclear, as no “firm guidance” can be derived from the encoded meaning. The author addresses this conflict by proposing an intermediate position, in which the semantic content – like an anaphora – provides guidance to truth-conditions, partially determining them. Unlike semanticists, he acknowledges the role of pragmatics in fully determining the propositional content. These first three contributions offer discussions on the general foundations of pragmatics. The fourth chapter focuses instead on a very specific issue, but offers at the same time an illustration of the applications of logical methodologies to the analysis of semantic and pragmatic puzzles. In “Negative Quantifier Noun Phases and Negative Polarity Items – the Horn-Atlas Debate 1991 – 2018,” Atlas develops an overview of the modern developments and a solution to the problem of the most studied and discussed “exponibile,” namely the quantifier “only.” Exponibilia are defined as propositions, such as “Only Socrates is wise,” which require “further analysis in order to lay bare their underlying logical form and to make clear under what conditions they can be said to be true or false” (Ashworth 1973, p.  137). These propositions need to be expounded normally in a conjunction of two or more simpler propositions; for example, the aforementioned statement can be expressed as “Socrates is wise and no one other than Socrates is wise.” Exponible is commonly used to refer to specific terms such as “only,” “every,” “cease,” characterized by a complex semantic analysis (an unpacking) (Horn 2011). Atlas reviews the modern treatment of this problem, and highlights the weaknesses of the existing proposals related to treatment of exponible propositions in terms of assertion or excluding force. Based on the logical analysis of “few N,” Atlas shows how the treatment of “only” in terms of negation or exclusion can be considered as a mistake. These theoretical researches in pragmatics are followed by a second group of papers that, by proposing new theoretical explorations and new methods of inquiry, bring the core and the frontier of pragmatic research closers. Thus, observing some excerpts from an American drama film, in “Speech-Act-Theoretic Explanations of Problems of Pure Indexicals,” Oishi observes that the standard theory of reference concerning the pronouns I and you (classified as pure indexicals) cannot account for the complexity of their use. Drawing on Goffman’s (1981) distinction among an animator, a principal, and an author and correlating them with Austin’s speech act theory, she shows how these two theories can be integrated leading to an innovative interpretative method. She shows three types of  correspondences:  1)  between “a body engaged in acoustic activity” and the performer of a locutionary act; 2) between the discourse entity who is committed to what the words say and the addresser (the utterer of certain words to produce a certain conventional effect); and 3) between the entity who has selected what and how needs to be expressed and the speaker. Based on this parallelism, she argues that the use of “pure indexicals” can provide an instrument for identifying the interlocutors’ commitments and communicative intentions  – and more specifically the types of speech acts identified by Austin as “expositives.”

Introduction: Inquiries in Philosophical Pragmatics – Theoretical Developments


Reference is also the topic of Saka’s analysis of a specific type of reference, pervasive of our language but neglected by the literature, called “Perspective Reference.” Perspective reference characterizes semantically defective sentences such as “I will deliver my baby in 3 months,” which are regarded as not literally or technically true – due to the fact that the implication that the referent exists fails. Based on experimental studies, Saka observes how such sentences loosely convey a truth based on the perceived similarity or causal relation between their intensional meaning and reference. Polysemy is the subject matter of the following two chapters. In her  chapter “Categorization, Memory and Linguistic Uses: What Happens in the Case of Polysemy,” Basile combines the semiotics and cognitive aspects of polysemy with the recent developments of lexical pragmatics. Polysemy is regarded as a communicative phenomenon, in which the sematic dimension – related to the encoded lexical concepts that underdetermine the speaker’s intended concepts – necessarily interacts with pragmatic inferences based on the interlocutors’ capacity to attribute mental states to each other. The semantic constraints and the mutual attribution of intentions results, according to the author, in a negotiation of multiple meanings. The problem of underspecification of meaning is an essential aspect of the pragmatic analysis of polysemy, to which the chapter entitled “Interpretation of Copredicative Sentences: A Rich Underspecification Account of Polysemy” is devoted. In her study, Ortega-Andrés addresses the analysis and interpretation of words associated with several related senses considering a specific and problematic category, copredicative words. Copredication refers to the use of one polysemous word to express simultaneously two or more related senses in a sentence, such as “book” in “The books are thick and interesting.” While most of the  open-class words we use are polysemous, one a limited number thereof is copredicative. However, they constitute a challenge to the existing theories of underspecification, which the author classifies in three groups – core meaning approaches, thin semantic theories, and rich semantic models – according to the kind of information lexically encoded and the access to the specific senses of polysemous words. The theoretical insights are combined in this paper with a strong empirical orientation, supporting her philosophical insights with the assessment of experimental studies. In this endeavor, she reviews the existing experiments conducted and the results drawn to support a novel theoretical approach, based on a richer account of semantic structure, in which the general knowledge about the world is included inside the “lexical meaning.” The following essay, “First Person Implicit Indirect Reports in Disguise,” concerns one of the crucial phenomena involved in the pragmatic enrichment of the semantic representation of an utterance, commonly referred to as “explicatures.” Capone addresses the boundaries of explicatures, namely cases in which the semantic representation of an utterance needs to be enriched in ways that cannot be predicted considering the superficial structure. For examples, first-person utterances expressing events that the speaker could not experience in first person or cannot possibly remember present a disguised polyphony, where an explicature of the kind, “I was told that…” is necessary for determining what is said actually. These “implicit


F. Macagno and A. Capone

indirect reports” are not part of what is said explicitly, but they are conditions thereof. This paper thus hints at a problematic boundary between what is said and what is meant, and the relationship between the former and the knowledge or assumptions necessary for determining the specific semantic representation of an utterance. The last section of this volume presents some new trends in the developments of pragmatics. One of the new frontiers of philosophical approaches to pragmatics is constituted by disciplines such as argumentation theory and discourse analysis, which are increasingly taking into account pragmatic topics and applying pragmatic methods and theories. These domains and these new explorations provide new perspectives and new challenges to the traditional approaches. Argumentation theory is essentially pragmatic, as arguments are defined based on their pragmatic nature (solve a difference between interlocutors, see Walton 1990), or the complex speech act that they express (van Eemeren and Grootendorst 1984). However, without approaching problems such as relevance, presupposition, and implicatures, argumentation can only describe and analyze limited aspects of discourse. Facing these limitations, scholars in argumentation theory have moved to explore pragmatics and theoretically interact with it. Similarly, the analysis of discourse inescapably faces the reconstruction and assessment of implicit messages. Pragmatics can offer a method of analysis, or a theoretical background for developing new analytical proposals. In this framework, the theoretical developments aiming at discovering and proposing new dialogues between related fields of study are crucially important. Walton’s paper focuses on a crucial topic for the two domains of pragmatics and argumentation theory, relevance. In argumentation, relevance has been always analyzed and described in terms of probative relevance. Probative relevance is the capacity of an argument to support another to make it stronger or to attack the other by casting doubt on it or even to discredit it as worthless. However, in “Assessing Dialectical Relevance Using Argument Distance” Walton shows how this concept is vague and at the same time problematic, unless defined more precisely considering other conditions, such as inferential distance – defined as the number (quantity) and acceptability (quality) of the argumentative inferences needed for connecting a premise to a conclusion. The modeling of inferential systems with a formal model of argument, Carneades Argumentation System, allows determining when an argument is relevant and when it is poorly relevant or irrelevant. Pragmatics becomes in “The Communicative Functions of Metaphors Between Explanation and Persuasion” the framework that allows using argumentation theory for analyzing medical discourse, and more specifically the uses of metaphors  in medical consultations. Argumentation (persuasion) and explanation are considered in the literature the two prototypical uses of metaphors. However, the definition of explanation and its distinction from persuasion is problematic, both at a theoretical and at an empirical level. Considering the most accepted definition of explanation as a transfer of understanding, they observe that in medical discourse, metaphorical explanations are very frequently use to persuade the interlocutor. But this is not their only function, as through many examples the authors show how other dialogical

Introduction: Inquiries in Philosophical Pragmatics – Theoretical Developments


goals can be pursued through them. Metaphorical utterances thus provide a theoretical challenge, that they address by drawing on the notion of dialogue move developed in argumentation theory, and distinguishing between the cognitive level of understanding (and the transfer thereof) from the pragmatic/dialogical ones. The pragmatic dimension of argumentation is explored in “Stereotypes Favour Implicatures and Implicatures Smuggle Stereotypes: The Case of Propaganda” from a cognitive perspective, focusing on the role of stereotypes and the tacit communication thereof in persuasive messages. As Lombardi Vallauri points out, the literature has clearly underscored the crucial effects that the implicit mechanisms of meaning conveyance such as presuppositions and implicatures have on the addressees’ attention on the content. This cognitive effect of implicitness can be used for persuasive and manipulative purposes. Through an analysis of advertising texts and political propaganda, this last essay shows how implicatures and stereotypes are frequently combined in persuasive messages. This combination is explained as a twofold interaction: a stereotype can be more easily inferred than less accepted (and acceptable) information, while stereotypes can be more easily accepted when they are communicated implicitly. The ideas proposed and defended in these twelve chapters offer an overview not only of the theoretical topics in pragmatics, but also of some new directions and methods of inquiry. Acknowledgments  This work was supported by the Fundação para a Ciência e a Tecnologia (research grant no. PTDC/FER-FIL/28278/2017).

References Ashworth, J. (1973). The doctrine of exponibilia in the fifteenth and sixteenth centuries. Vivarium, 11(2), 137–167. Austin, J. L. (1962). How to do things with words. Oxford, UK: Oxford University Press. Goffman, E. (1981). Forms of Talk. Philadelphia, PA: University of Philadelphia Press. Horn, L. (2011). Only XL: The assertoric asymmetry of exponibles. In E.  Cormany, S.  Ito, & D. Lutz (Eds.), Proceedings of semantics and linguistic theory (SALT) (pp. 198–222). Ithaca, NY: CLC Publications. Huang, Y. (2014). Pragmatics. Oxford, UK: Oxford University Press. Jaszczolt, K. (2018). Pragmatics and philosophy: In search of a paradigm. Intercultural Pragmatics, 15(2), 131–159. Kecskes, I. (2013). Intercultural pragmatics. Oxford, UK: Oxford University Press. Stalnaker, R. (1970). Pragmatics. Synthese, 22(1–2), 272–289. BF00413603. van Eemeren, F., & Grootendorst, R. (1984). Speech acts in argumentative discussions: A theoretical model for the analysis of discussions directed towards solving conflicts of opinion. Dordrecht, Netherlands: Floris Publications. Walton, D. (1990). What is reasoning? What is an argument? Journal of Philosophy, 87, 399–419.

Three Mistakes About Semantic Intentions Michael Devitt

There is much talk of intentions in semantics. Intentions to refer are said to determine reference. The central idea of Gricean “intention-based-semantics” is that speaker meanings are constituted by the speakers’ communicative intentions. Finally, it is common, particularly among Griceans, to believe that there is a constraint on what a speaker can intend to communicate by an utterance. I think that all of this is mistaken.

Mistake (I): Intending to Refer Stephen says, “Grice is great”; or, pointing to Grice, “He is great” or “That man is great”. Who does Stephen refer to with these “singular referring terms”? One answer that is very popular among philosophers is that he refers to “whoever he intends to refer to” (e.g. Bach 2001; Kaplan 1989; Neale 2016; Schiffer 1972, 1981). Another answer that has some popularity among philosophers is that he refers to “the particular person he has in mind” (e.g. Almog 2012; Devitt 1974; Donnellan 1966, 1968). Both answers are initially appealing because they are in line with what the folk might say in answer to our question. I think that having-in-mind is indeed a helpful starting point for a good answer, but no more than that. Intending-­ to-­refer is not even that. I shall begin by arguing for this and then make three more objections to the intending-to-refer hypothesis. I conclude that it should have no place at all in a theory of language.

M. Devitt (*) City University of New York, New York, NY, USA e-mail: [email protected] © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 F. Macagno, A. Capone (eds.), Inquiries in Philosophical Pragmatics, Perspectives in Pragmatics, Philosophy & Psychology 27,



M. Devitt

Objection 1: Implausible Starting Point1 What is wrong with intending-to-refer as a starting point? Think of what is required for someone doing A to intend to V to x; for example, what is required for someone leaving her apartment to intend to bicycle to the Met. This intention is a thought, a “propositional attitude”, about bicycling and so contains the concept bicycling. Similarly, for a speaker uttering a singular referring term in a sentence to literally intend to refer to x with the term. She must have a thought containing the concept reference. So, according to the intending-to-refer hypothesis, she can’t refer without thinking about reference! There is no reason to believe this, no reason to believe that any expression of a thought about something must be accompanied by a further thought about reference. That is far too intellectualized a picture of referring and is psychologically implausible. Referring is a cognitive skill, mere know-how; or so I have argued (Devitt 1981a, 1996, 2006, 2011). One could refer without even having a concept of reference (Devitt 1981a, p. 97). This is not to deny that referring is intentional, in some sense: it’s an action like walking, kissing, etc. (And we would make no theoretical progress by saying that the speaker refers to x in virtue of intentionally referring to x: referring is intentional.) And it is not to deny that any normal adult speaker with a minimal education could probably tell you after the event what she was “talking about”. That’s a very easy bit of semantic knowledge. Yet even that easy bit is surely beyond the capacity of many organisms that nonetheless refer: for example, humans at the age of three2; bees, prairie dogs, vervet monkeys, and other nonhuman species that have what cognitive ethologists call “referential languages”. In the face of this, one could of course restrict the intending-to-refer hypothesis to adult humans, but that seems ad hoc. In any case, to repeat, it is not plausible that such metalinguistic intentions are constitutive of anyone’s referring. Having-in-mind does not have this problem as a starting point. For, having x in mind simply requires that the part of the thought expressed that causes the use of the term refers to x. That is the first objection to including intending-to-refer in a theory of language. There are three more objections, two of which apply also to having-in-mind.

Objection 2: Incomplete Even if intending-to-refer was a plausible starting point it would immediately raise the question: In virtue of what did the speaker intend to refer to x in using the singular term rather than to y or z? Without an answer to this question, the intending-to-­ refer hypothesis could not advance our theory of reference significantly: the  Objections 1 and 2 draw on my (Devitt 2015, pp. 110–11).  Developmental evidence suggests that the capacity to have metalinguistic thoughts comes later, in middle childhood; e.g. (Hakes 1980). 1 2

Three Mistakes About Semantic Intentions


explanatory problem has simply been moved a short distance from the reference of the utterance to the reference of the intention. Now, of course, the having-in-mind hypothesis raises an analogous question and so is similarly incomplete: In virtue of what did the speaker have x in mind? In virtue of what does the part of the thought that causes the use of the term refer to x. What reality is this somewhat vague bit of folk talk getting at? I have argued that, for the sort of having-in-mind and singular reference exemplified in Stephen’s utterances, the reality is causal (Devitt 1974, 1981b, 1981a).3 A part of Stephen’s thought determines that he had x in mind (in this sense) in virtue of standing in a certain sort of causal relation to x, a relation involving the perceptual grounding of someone’s thought in x and, perhaps, reference borrowings (Kripke’s wonderful insight).4 The details of this causal relation for demonstratives, pronouns, and names will differ a bit, of course.5 (Note that It follows from this that the reference of the term is determined by a mental state of the speaker. So, contrary to the tradition arising from formal semantics – see, e.g., David Lewis (1987) – the “indices” of the context external to the speaker’s mind play a reference-determining role only to the extent that relations to that context partly constitute that mental state.)

Objection 3: Redundant Of course, the intending-to-refer hypothesis could be completed by an answer to the in-virtue-of-what question even if it seldom is. Thus the hypothesis could adopt a causal answer along the lines I have just sketched: applied to our example, Stephen intends to refer to Grice in using ‘Grice’, ‘he’, or ‘that man’ in virtue of that use being immediately caused by an intention that stands in the specified causal relation to x. Suppose that the intending-to-refer hypothesis is indeed completed by such an

3  I was clear from the start that the vague talk of “having in mind” is but a “stepping stone” to a causal theory of reference, not something that features in the theory (Devitt 1974, p. 202). Still, Andrea Bianchi’s criticisms (Bianchi 2020) have made it clear to me (Devitt 2020) that I should have been more careful in my handling of this stepping stone. I should not have claimed, early, to offer an “analysis” (Devitt 1974, p. 202) of “having in mind” and, late, “an explanation – better, an explication – of this somewhat vague folk talk” in causal terms (Devitt 2015, p. 111). The folk talk arguably covers more than the causal relation we want; see section “Objection 4: Misleading” below on how it can mislead. So my causal explanation was, in reality, an explanation of a restricted, “technical”, notion of having-in-mind. 4  Joseph Almog has recently proposed a similar causal explanation of having in mind (Almog 2012, pp. 177, 180–82), attributing it to Donnellan without evidence; for discussion, see (Devitt 2015, p. 111). As Julie Wulfemeyer has aptly remarked recently: “The grounding cognitive relation went largely unexplained by Donnellan” (Wulfemeyer 2017, p. 2). 5  And the causal relation determining a proper name’s semantic reference differs from, but is built upon, that determining its speaker reference (Devitt 2015, p. 126, 2020, sec. 5).


M. Devitt

answer. We can see then that the hypothesis is redundant, an unhelpful excursion. For, intentions are thoughts (propositional attitudes) and the explanation of in virtue of what Stephen’s intention refers to Grice is an explanation of in virtue of what any of his thoughts refer to Grice. So it is an explanation of in virtue of what his original thought about Grice’s greatness, expressed using ‘Grice’, ‘he’, or ‘that man’, referred to Grice. And that would explain why Stephen referred to Grice in using one of those singular referring terms. We would have solved our original problem without any need to appeal to metalinguistic intentions.

Objection 4: Misleading I have proposed a causal theory for singular referring terms. The tradition proposed description theories for some or all of them. Take a normal use of a term ‘E’ where the speaker refers to x on either theory: so the use is causally connected to x in the way specified by the causal theory and x is uniquely described by the associated reference-determining description, ‘the F’, specified by the description theory. Then, if we are prepared to attribute a referential intention to the speaker at all – setting aside the worry in Objection 1 – we should all agree that the speaker using ‘E’ intends to refer to the F. But now suppose that the speaker not only has the identifying belief that she would express, “E is the F”, but also several others that she would express, “E is the G”, “E is the H”, and so on. In a quite ordinary sense, we should also say that the speaker, in using ‘E’ intends to refer to the G, to the H, etc. What would those intentions show about who she referred to? Nothing at all, on either theory. First, on a causal theory, whatever descriptions she associates with ‘E’ are quite beside the point of what she refers to. Indeed, the resulting intentions may well mislead us. Thus, suppose that the speaker’s belief that E is the G is false; perhaps ‘G’ uniquely describes not x but y. Then her intention to refer to the G would mislead us into thinking that she referred to y not x. Second, on a description theory, only her intention to refer to the F is a reliable guide to her reference for only it is reference-determining. So, her intention to refer to the G would be as misleading as it would be on the causal theory. The having-in-mind hypothesis can be just as misleading: the speaker in the above scenario might be said to have the F, the G, the H, in mind – “having in mind” is vague enough for that – with the potentiality to mislead us about E’s reference. The possibility that talk of intentions to refer might mislead in this way is not an idle one. In their seminal work in “experimental semantics”, Edouard Machery, Ron Mallon, Shaun Nichols, and Stephen Stich (Machery et al. 2004; “MMNS”) conducted a Gödel experiment in which they asked subjects who John, a character in the vignette, was “talking about”. This question was criticized as ambiguous. MMNS wanted the participant’s intuition about who the name conventionally refers to, its semantic reference. But the participant might take the prompt as asking about

Three Mistakes About Semantic Intentions


who the character means by the name, its speaker reference. So the experiment may not have yielded what was wanted.6 Here is one such criticism: Examples…suggest that a question of the form, “Who is S talking about in using n?” does not have a univocal meaning….In particular, although the communicative intention of the speaker (John) is not made explicit, the vignette may be taken to suggest that he intends to talk about the man who discovered that arithmetic is incomplete since the vignette insists that this is the only piece of information associated with “Godel.” It is thus unclear whether those participants who answer that John is talking about the man who discovered that arithmetic is incomplete are making a genuine descriptivist judgment about the semantic reference of “Godel” or rather making a judgment about speaker’s reference. Consequently, it may be that almost everybody has causal–historical intuitions about semantic reference, but that some participants…report their intuitions about speaker’s reference. (Machery et al. 2015, p. 67)

The vignette may indeed suggest that John “intends to talk about the man who discovered that arithmetic is incomplete” but, in light of our discussion, we can see that this does not provide cause for concern about the alleged ambiguity in MMNS’s question. For, that intention gives no reason for the subjects to take the speaker reference to be different from the semantic reference. On the causal theory, that intention is irrelevant to both the semantic and speaker reference. On the description theory in question – one that has the reference determined by an associated description along the line of ‘the man who discovered the incompleteness of arithmetic’ – the intention is relevant to the speaker reference only because it is relevant to the semantic reference. In sum, if taken literally, the intending-to-refer hypothesis is: (1) implausible; (2) incomplete; (3) redundant once completed; (4) misleading. The having-in-mind hypothesis does not share faults (1) and (3).

Mistake (II): Intending to Communicate What explains the speaker meaning of a sentential utterance? A central idea of Gricean “intention-based semantics” is that this meaning is constituted by the speaker’s intention to communicate a certain content to an audience. As Stephen Schiffer puts it, “Meaning entails audience-directed intentions, and one cannot mean something without intending to be understood” (Schiffer 1992, p.  515). Stephen Neale expands the idea as follows: “In doing x, S, meant that p iff (roughly) for some audience, A, S did x intending A to think that p via A’s recognition that S intended A to think that p” (Neale 2016, p. 281). No doubt speakers typically produce utterances to communicate messages to an audience. And those utterances are intentional acts and might involve intentions of some sort. Furthermore, I’m inclined toward the view that representational systems

 I think that the criticism is mistaken (Devitt and Porot 2018, pp. 1579–80 n. 17).



M. Devitt

like human languages owe their very existence, whether they are innate or conventional, to their role of communicating messages to audiences (2006: 130–31). But it does not follow that the only meaningful use of language is to convey a message to an audience. More importantly, it does not follow that it is essential to its meaningful use that it involve an intention to communicate. And it is not essential, as Noam Chomsky points out: Under innumerable quite normal circumstances—research, casual conversation, and so on—language is used properly, sentences have their strict meaning, people mean what they say or write, but there is no intent to bring the audience (not assumed to exist, or assumed not to exist, in some cases) to have certain beliefs or to undertake certain actions. (Chomsky 1974, p. 62)

Making an utterance is indeed an intentional act but the basic act in question is better described using the popular locution, “expressing a thought”: “there is much to be said for the old-fashioned view that speech expresses thought, and very little to be said against it” (Fodor et al. 1974, p. 375). Expressing a thought can be done in various ways: by speaking, writing, or signing in some other way. And the point of this expressing may not be to communicate a message to an audience. There may be no audience, or the speaker may not think there is. The speaker may be musing (“To be, or not to be”), making notes, trying out a line for a poem, and so on. Even where the speaker is addressing an audience, she may not, for a variety of possible reasons, care if the message gets through. She may even be happier if it doesn’t! Thus, it is common – particularly, I think, among the well-educated English – for people who are bored by ordinary small talk to regard conversation as a challenge to be cleverly and wittily indirect. Such a person certainly does not intend his hearer to grasp his message. Perhaps he will hope that the hearer is smart enough to do so, but he may not: he may anticipate the pleasure of watching the hearer fail. In sum, communicating a thought to an audience is just one form of expressing that thought: it is expressing the thought with the intention of being understood by the audience. If this is right, a speaker means that p in virtue of expressing the thought that p. And it is not necessary that the speaker intends to communicate that thought to some audience. Now some philosophers will surely be tempted to make moves to save the view that communicative intentions are essential: “when making notes one is communicating with one’s future self”; “when musing, one is communicating to a possible audience”; “when having fun at one’s audience’s expense, one is communicating… – tricky, I’ll have to get back to you on that one”. But, first, is any of this psychologically plausible? That is to say, is it plausible that speakers making utterances that are at least apparently non-communicative must have some such thought about some audience or other? Is it really plausible that it is impossible to mean something by an utterance without thinking about an audience? Second, and more important, what is the point of saving the view? Why do we need to go beyond expressing a thought to the more demanding attempting to communicate that thought to an audience in order to explain speaker meaning? Expressing a thought will do the job. There is no theoretical motivation to require an intention to communicate the thought.

Three Mistakes About Semantic Intentions


The idea that speaker meaning is constituted by communicative intentions is the fundamental mistake of intention-based semantics. But that mistake is compounded by the accounts given of such communicative intentions. Here is one of Grice’s later ones: “U meant something by x” is true iff U uttered x intending thereby: 1. 2. 3. 4.

That A should produce response r That A should, at least partly on the basis of x, think that U intended (1) That A should think that U intended (2) That A’s production of r should be based (at least in part) on A’s thought that U intended (1) (that is, on A’s fulfilment of [2]) 5 . That A should think U intended (4). (Grice 1989, pp. 96–97) Even this definition proved insufficiently complex to rule out all the suggested counter-examples (Schiffer 1972). What are we to make of this? I have just indicated the psychological implausibility of the claim that meaning something by an utterance must be accompanied by a thought about an audience A. What about the claim that it must be accompanied by this thought?! It is hardly plausible that this baroque structure of intentions is psychologically real in any speaker. Grice was sensitive to this concern from the start, disclaiming “any intention of peopling all our talking life with armies of complicated psychological occurrences” (Grice 1989, p. 222). But then if there are no such armies we have no theory: we are left with no explanation of speaker meaning. I know of no satisfactory solution to this problem. I think the program is misguided (Devitt and Sterelny 1999, pp. 149–50). I have used the ordinary locution, “expressing a thought” to describe the basic act which constitutes speaker meaning.7 I don’t say that the locution is perfect for the task. We want something that describes the intentional act common to speaking, writing, signaling, emailing, tweeting, and so on, something that abstracts from the differences between these behaviors. Certainly these behaviors are covered by “expressing a thought” but very likely some other ones are too. Thus, we might ordinarily say that a person’s face expresses the belief that her company is boring even if its so doing is unintentional. On the other hand, one might argue that when humans greet (“Hi”), cheer (“Bravo”), abuse (“Get lost”), and the like, they are expressing some mental state but often not a thought. So, English seems not to have an ordinary locution that perfectly describes the action we want to pick out for our theoretical purposes. And that is hardly surprising: the folk are not bent on explaining speaker meaning. Still, “expressing a thought” seems closest to what we want. So I will stick with it, perhaps narrowed to exclude the unintentional and widened to include the likes of greetings. A further complication should be mentioned. Humans are not the only animals that make meaningful “utterances” in a language; consider, for example, bees and prairie dogs. What constitutes the “speaker meaning” of a particular bee dance? Not

7  This discussion of “expressing a thought” has benefited from the criticisms of Elmar Geir Unnsteinnson, Daniel Harris, and Gary Ostertag.


M. Devitt

presumably the content of the thought expressed but rather the content of some “lesser” representational state in the bee that causes the dance. So to cover nonhumans perhaps we should take speaker meaning to be something like “the content of the inner representational state causing that use of language”. But I set this complication aside. So what is it to “express a thought”, in this perhaps slightly technical sense? I have nothing more to say about this. And I don’t think that philosophers should expect to have anything more to say at this time. So, expressing a thought is partly unexplained in the theory, and to some extent a “primitive”. This is not to say that expressing a thought is inexplicable. Psycholinguists studying language production are in effect trying to throw light on it (see, e.g., Vigliocco and Vinson 2006). The point is just that this is now beyond philosophers. In light of this, in virtue of what does a sentential utterance have its speaker meaning? In virtue of the causal story of its production. Thus the speaker meaning of a word in that utterance is determined by the concept it expresses on that occasion. The “speaker syntax” of the utterance’s sentence is determined by the structure of the underlying thought that the utterance expresses and by the way in which the utterance was produced from that underlying thought.

Mistake (III): Constraints on Intentions Griceans standardly believe that there is a constraint on what a speaker can intend to communicate by an utterance, a belief that reflects the belief that there is a constraint on intentions in general. The literature contains many versions of this constraint, some of them astonishingly strong. Here is a sample. Talking of intentions in general, Grice claimed in one place that 1. A condition on ‘X intends to do A’ is that X “is sure that he will in fact do A” (Grice 1971, p. 266).8 Talking of intentions to refer, Stephen Schiffer claimed that 2. “The speaker cannot intend to refer to a particular female [by ‘she’] unless he expects his hearer to recognize to which female he is referring” (Schiffer 2005, p. 1141). Reinaldo Elugardo and Robert Stainton make the weaker, hence more plausible, claim that the speaker’s expectations must be about what the hearer can do rather than what she will do: 3. “The intentions that a speaker can have are importantly constrained by her reasonable expectations about what the hearer can figure out” (Elugardo and Stainton 2004, p. 445).9

 See also Harman (1976); Velleman (1989).  See also Brand (1984).

8 9

Three Mistakes About Semantic Intentions


And Grice, in another place, urges a similarly weaker constraint: talking about intentions to mean something by a hand wave HW, Grice proposes that 4. A speaker “must (logically) be in a position, when uttering HW, to suppose that there is at least some chance that these intentions will be realized” (Grice 1989, p. 125). These are all “positive” claims, requiring the intender to have a certain belief (or “be in a position” to have one). Elsewhere Grice’s constraint is a much weaker “negative” one, requiring the intender not to have a certain belief: 5. “One cannot in general intend that some result should be achieved, if one knows that there is no likelihood that it will be achieved” (Grice 1989, p. 101; emphasis added). In an early paper Stephen Neale urges a slightly weaker negative claim: 6. The formation of genuine communicative intentions by U is constrained by U’s expectations: U cannot be said to utter X M-intending A to φ if U thinks that there is very little or no hope that U’s production of X will result in A φ-ing”. (Neale 1992, p. 552; emphasis added)10 Later Neale urges a constraint that is much the same as Grice’s negative one: 7. “I cannot intend what I believe to be impossible (Neale 2004, p. 77). We can capture well enough the range of possible constraints as follows, with two positive ones followed by two negative ones, the stronger first in each case: X cannot intend to A unless: P1: X believes that she will A. [cf. (1), (2)] P2: X believes that she can A, has some chance of A-ing. [cf. (3), (4)] N1: X lacks the belief that she will not A. N2: X lacks the belief that she cannot A, has no chance of A-ing. [(5), (6), (7)] Before assessing these constraints, we should think about the force of “cannot intend”. Perhaps it yields a constitutive constraint: if the constraint is not met, it is not metaphysically possible for any mental state of X, whatever else it might be, to be an intention to A. That seems very strong. So perhaps “cannot intend” yields a weaker normative constraint: if the constraint is not met, it is not metaphysically possible for X to rationally intend to A (though X might irrationally intend to). I shall focus on the normative constraints. Clearly, anything that counts against a normative constraint will count against the corresponding constitutive constraint. Consider the positive constraints, P1 and P2. Neither is plausible. First, and simply, X might have no beliefs at all about the likely outcome of trying to A. Perhaps she has no relevant evidence; or she has some evidence but is unable or unwilling to assess its significance. Yet she might still think, rationally, that it’s worth trying to A anyway; “I don’t know if I can but I might as well give it a go”, she says to herself. Second, as Keith Donnellan (1968, p. 212, n. 10) and Stephen Schiffer (1972, p. 69) bring out, if X’s life is at stake she might try anything however bleak she thinks the


 See also Mele (1992).


M. Devitt

prospects of success. Suppose, for example, that X is shipwrecked and her only hope is to swim to shore. Although X is too realistic to believe that she will reach shore (P1), or even that she can reach shore (P2), she may still form the rational intention to reach shore. For similar reasons, the negative constraint, N1, is not plausible. Even if X believes that she won’t reach shore however hard she tries, if she is in dire enough circumstances it may not be just possible for her to try to do so, it may be rational for her to try: “What have I got to lose?”, she says to herself. But what about the last and weakest constraint, N2? Could X really intend to swim to shore even if she believes that she has no chance of doing so, that her doing so is impossible?11 Initially it may well seem not. Still, on further thought, even this constraint seems dubious. Suppose X has an appropriate amount of epistemic humility and is well aware of her own fallibility. So she realizes that her belief about the impossibility of her reaching shore may be wrong: Then, given what is at stake, she may still intend to swim to shore in the hope that she is wrong. And so she should. So, I’m inclined to think that intentions are not subject to any of the constraints contemplated by Griceans, whether construed as constitutive or normative. I suspect that we should settle for a weak psychological, not constitutive or normative, constraint on intentions along the following lines: the less a person believes that she will succeed in A-ing if she tries then, other things being equal, the less likely it is that she will intend to A. Finally, why is it so common among Griceans to insist on some constraint on what a person can intend? The answer seems to be: the Humpty Dumpty worry. If speaker meaning is simply a matter of what the speaker intends to communicate, then we need some constraints on that intention to avoid the unacceptable consequence that the speaker might mean absolutely anything by an utterance. I think that the above psychological, not constitutive or normative, constraint on intentions is sufficient for the purpose. Thus it is unlikely that Humpty Dumpty would intend to communicate to Alice that that is a nice knock-down argument by saying “There’s glory for you” because he should believe that he is very unlikely to convey that message to Alice by that expression, given the expression’s conventional meanings in English.

References Almog, J. (2012). Referential uses and the foundations of direct reference. In J.  Almog & P. Leonardi (Eds.), Having in mind: The philosophy of Keith Donnellan (pp. 176–184). Oxford, UK: Oxford University Press. Bach, K. (2001). You don’t say? Synthese, 128(1–2), 15–44. Bianchi, A. (2020). Reference and causal chains. In A. Bianchi (Ed.), Language and reality from a naturalistic perspective: Themes from Michael Devitt (pp. 121–136). Cham, Switzerland: Springer.  Neale recently concluded a subtle discussion of suggested constraints by endorsing this one (Neale 2016, pp. 278–80).


Three Mistakes About Semantic Intentions


Brand, M. (1984). Intending and acting: Toward a naturalized action theory. Cambridge, MA: MIT Press. Chomsky, N. (1974). Reflections on language. New York, NY: Pantheon Books. Devitt, M. (1974). Singular Terms. The Journal of Philosophy, 71(7), 183–205. Devitt, M. (1981a). Designation. New York, NY: Columbia University Press. Devitt, M. (1981b). Donnellan’s distinction. In P. French, T. Uehling, & H. Wettstein (Eds.), Midwest studies in philosophy, volume VI: The foundations of analytic philosophy (pp.  511–524). Minneapolis, MN: University of Minnesota Press. Devitt, M. (1996). Coming to our senses: A naturalistic program for semantic localism. Cambridge, UK: Cambridge University Press. Devitt, M. (2006). Ignorance of language. Oxford, UK: Oxford University Press. Devitt, M. (2011). Methodology and the nature of knowing how. The Journal of Philosophy, 108(4), 205–218. Devitt, M. (2015). Should proper names still seem so problematic. In A. Bianchi (Ed.), On reference (pp. 108–143). Oxford, UK: Oxford University Press. Devitt, M. (2020). Stirring the possum: Responses to the Bianchi papers. In A.  Bianchi (Ed.), Language and reality from a naturalistic perspective: Themes from Michael Devitt (pp. 371–455). Cham, Switzerland: Springer. Devitt, M., & Porot, N. (2018). The reference of proper names: Testing usage and intuitions. Cognitive Science, 42(5), 1552–1585. Devitt, M., & Sterelny, K. (1999). Language and reality: An introduction to the philosophy of language. Cambridge, MA: Mit Press. Donnellan, K. (1966). Reference and definite descriptions. The Philosophical Review, 75(3), 281–304. Donnellan, K. (1968). Putting humpty dumpty together again. The Philosophical Review, 77(2), 203–215. Elugardo, R., & Stainton, R. J. (2004). Shorthand, syntactic ellipsis, and the pragmatic determinants of what is said. Mind & Language, 19(4), 442–471. Fodor, J., Bever, T., & Garrett, M. (1974). The psychology of language: An introduction to psycholinguistics and generative grammar. New York, NY: McGraw-Hill. Grice, P. (1971). Intention and uncertainty. Proceedings of the British Academy, 57, 263–279. Grice, P. (1989). Studies in the way of words. Cambridge, MA: Harvard University Press. Hakes, D. (1980). The development of metalinguistic abilities in children. Berlin, Germany: Springer. Harman, G. (1976). “Practical reasoning.” Edited by Joseph Raz. The Review of Metaphysics, 29(3), 431–463. Kaplan, D. (1989). Afterthoughts. In J.  Almog, J.  Perry, & H.  Wettstein (Eds.), Themes from kaplan (pp. 565–614). Oxford, UK: Oxford University Press. Lewis, D. K. (1987). Philosophical papers: Volume I. Oxford, UK: Oxford University Press. Machery, E., Mallon, R., Nichols, S., & Stich, S.  P. (2004). Semantics, cross-cultural style. Cognition, 92(3), 1–12. Machery, E., Sytsma, J., & Deutsch, M. (2015). Speaker’s reference and cross-cultural semantics. In A. Bianchi (Ed.), On reference (pp. 62–76). Oxford, UK: Oxford University Press. Mele, A. (1992). Springs of action: Understanding intentional behavior. Oxford, UK: Oxford University Press. Neale, S. (1992). Paul Grice and the philosophy of language. Linguistics and Philosophy, 15(5), 509–559. Neale, S. (2004). This, that, and the other. In A. Bezuidenhout & M. Reimer (Eds.), Descriptions and beyond (pp. 68–182). Oxford, UK: Oxford University Press. Neale, S. (2016). Silent reference. In G. Ostertag (Ed.), Meanings and other things: Themes from the work of Stephen Schiffer (pp. 229–342). Oxford, UK: Oxford University Press. Schiffer, S. (1972). Meaning. Oxford, UK: Oxford University Press. Schiffer, S. (1981). Indexicals and the theory of reference. Synthese, 49(1), 43–100.


M. Devitt

Schiffer, S. (1992). Belief ascription. The Journal of Philosophy, 89(10), 499–521. Schiffer, S. (2005). Russell’s theory of definite descriptions. Mind, 114(456), 1135–1183. Velleman, D. (1989). Practical reflection. Princeton, NJ: Princeton University Press. Vigliocco, G., & Vinson, D. P. (2006). Speech production. In Encyclopedia of cognitive science (pp. 182–189). Chichester: Wiley. Wulfemeyer, J. (2017). Bound cognition. Journal of Philosophical Research, 42, 1–26. https://doi. org/10.5840/jpr2017531106.

Common-Knowledge-Based Pragmatics Richard Warner

In The Art of Statistics, David Spiegelhalter writes, “Personally, I rather like acting as if all that occurs around us is the result of some random pick from the all the possible things that could happen” (Spiegelhalter 2019, p. 93). When you read that, do you reason to the conclusion that Spiegelhalter believes (or intends you to believe) that he rather likes acting as if all that occurs around us is the result of some random pick from the all the possible things that could happen? Paul Grice and many who follow him answer “Yes” (Grice 1989; 2001). Pragmatics makes a similar assumption. “Pragmatics involves perception augmented by some species of “ampliative” inference—induction, inference to the best explanation, Bayesian reasoning, or perhaps some special application of general principles special to communication, as conceived by Grice … —but in any case a sort of reasoning” (Korta and Perry 2015). The reasoning is sophisticated. For Grice, for example, an audience’s response to a speaker’s utterance involves reasoning from the premise that the speaker expects the audience to recognize that the speaker intends to produce a certain response in the audience. Pragmatics also typically posits sophisticated reasoning. The problem is that it is implausible to attribute any but very minimal reasoning to speakers and audiences, far less than pragmatics regularly attributes. I offer an approach that supports pragmatic inquires while significantly reducing the amount of reasoning attributed. The approach appeals to common knowledge. Common knowledge is the recursive belief state in which Professor of Law, Chicago-Kent College of Law. This article is a companion piece to “Meaning, Reasoning, and Common Knowledge,” (Warner 2019b). I am gratefully indebted to Stephen Schiffer for his comments and encouragement and to Steven Wagner for many insightful conversations over the years. R. Warner (*) Chicago-Kent College of Law, Chicago, IL, USA e-mail: [email protected] © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 F. Macagno, A. Capone (eds.), Inquiries in Philosophical Pragmatics, Perspectives in Pragmatics, Philosophy & Psychology 27,



R. Warner

people are know s­ omething, know they know it, know they know they know it, ad  infinitum. I incorporate common knowledge in a model of communication which entails Grice’s characterization of conversational implicature and thereby gives, to that extent, a general account of such implicatures.

A Coordination Problem Model of Communication I model a range of communicative activities as coordination problems. The explanation of a coordination problem appeals to two concepts. The first is a group-action goal. A group-action goal is a goal whose realization requires coordinated group action. For example, if Victor and Victoria both prefer to meet for dinner, neither can realize that goal on their own. The second concept is that of a dominant preference. One dominantly prefers something at a particular time if and only if at that time one prefers it more than one prefers other options that conflict with its realization. A coordination problem for a group is a situation in which there a group-action goal which each member of the group dominantly prefers to realize (Chwe 2013, p. 3). The problem is to achieve the required coordination. I use the notion of M-intending to define a special type of coordination problem. For my purposes, the following simple definition suffices (using ‘utter’ and its cognates to cover both linguistic and non-linguistic items). For all propositions p,1 a speaker S means that p by an utterance U for an audience A if and only if by uttering U, S M-intends A to believe that p. A speaker S M-intends that p for an audience A by uttering U if and only if S utters U intending (1) that A believe that p; (2) that A recognize S’s intention (1), and (3) that this recognition be part of A’s reason for believing p. A speaker S means that p by producing X if and only if S M-intends that p by producing X (Grice 1957, p. 337). These definitions are only appropriate for cases of asserting that p. Later work by Grice and Stephen Schiffer refines the analysis to meet a variety of counterexamples, and it extends the analysis to the full range of speech acts. To illustrate coordination problems in communication situations, suppose Alice and Bob are talking after the concert at which their friend Victoria performed a clarinet solo. When Bob sees Victoria approaching them, he wants to know her mood. He knows Alice spoke with her immediately after the concert ended, so, in the instant before Victoria joins them, he asks, “How did she feel about her 1  The notion of a proposition is faces difficulties. In addition to Quinean doubts about relying on synonymy relations for individuation, there are the problems Stephen Schiffer groups under vagueness (Schiffer 2017). We could eliminate talk of propositions if we had a theory that explained how the brain combines semantically interpreted lexical representations into ever more complex structures in ways that underlie thought and language. The ‘p’ in ‘S means p’ could then be interpreted as ranging over those structures. Lacking such a theory, we have to make do with the promissory note drawn on such a theory—the concept of a proposition. The concept is a stopgap measure. As Schiffer notes, “What could be the point of trading in facts about meaning for facts about the content of beliefs if one ends up with nothing to say about the latter?” (Schiffer 1987, p. 2).

Common-Knowledge-Based Pragmatics


­performance?” Where P is a set of propositions that are possible answers to Bob’s question, both Alice and Bob dominantly prefer that Alice M-intend that p for one of the propositions, and that Bob believe it in response.2 The dominant preferences mean that Alice and Bob face a coordination problem. Call such problems communication coordination problems. A speaker and an audience are in a communication coordination problem whenever, for some proposition p, they both dominantly prefer the group-action goal of S′ M-intending that p and that A’s believing it in response. How do speakers and audiences solve communication coordination problems? The Gricean answer is that they reason their way to the solution.3 To illustrate, consider the following coordination problem. As she and Victor drive in their car with their children, Victoria says “Let’s get the kids a snack,” M-intending thereby that they should get the children a snack. Let P be a set of propositions that Victor could relevantly M-intend in response. The propositions in P spread across a spectrum from complete agreement to complete disagreement with varieties of partial agreement in between. Victoria and Victor both dominantly prefer the group-action goal of Victor’s M-intending a proposition in P and Victoria’s believing it in response. On a Gricean view, Victoria and Victor solve the coordination problem by reasoning in complementary ways. Suppose, for example, Victor utters ‘Okay, but I veto I-C-E C-R-E-A-M’ thereby M-intending that he agrees but vetoes ice cream. To solve the coordination problem, Victoria must in response believe that Victor agrees but vetoes ice cream. A Gricean explanation of how she arrives at that belief is that she reasons in something like the following way. “Victor uttered ‘Okay,’ one of the standard meanings of which is that the speaker agrees with some contextually specified proposition. The best explanation of Victor making such an utterance is that he M-intends that he agrees. Victor’s utterance includes ‘but I veto I-C-E C-R-­ E-A-M.’ That utterance type standardly means the speaker vetoes ice cream, and the best explanation of Victor’s uttering it is that he M-intends, as limitation on his agreement, that he vetoes ice cream. The best explanation of his spelling out the word ‘ice cream’ is that he M-intends that the children should not realize we are talking about ice cream.” Turning to Victor, how does he know his utterance will lead Victoria to respond by reasoning her way to the relevant belief? Prior to his utterance, Victor reasons his way to the conclusion that Victoria will reason in more or less the foregoing way.

2  Typically, it will not be a single proposition p that is involved, but a class of propositions. For some relevant considerations, see (Warner 2019a). 3  The ‘Gricean’ label is convenient and not inaccurate, but my highly simplified discussion does not do justice to the subtlety and sophistication of Paul Grice’s work. Were I presenting these views to Grice, I would describe them as “Something someone might interpret him as having thought at some time.”


R. Warner

Two Problems with the Gricean Answer The first problem is that it is implausible that people reason in the required ways.4 It is implausible to see Victoria and Victor as consciously reasoning in the ways outlined above. At best, their reasoning is unconscious. The problem with that suggestion is indeterminacy. I offered one version of how Victoria and Victor might reason, but with a little ingenuity, one can produce a variety of options. Which one of the possibilities is Victoria’s reasoning? None of them. She could have reasoned in a variety of ways, and there is (exceptional cases aside) no way to identity one of them as the way she did in fact reason. And there is no need to do so. A sufficient explanation of Victoria’s response is that she is disposed to reason in something like various possible explicit reasoning sequences. The disposition explains her response. In general, explaining speaker-­ meaning interactions only requires attributing relevant dispositions to speakers and audiences. For speakers, the disposition is to infer that the audience will believe a certain proposition if the speaker utters a certain sentence. For audiences, the disposition is to infer from the fact that a speaker uttered a certain sentence that the audience should believe a certain proposition. M-coordination is not a matter of complementary reasoning, but of complementary dispositions. For the second objection, grant for the sake of argument that speakers and audiences engage in relevant complementary reasoning about M-intentions. That still is not enough to reliably solve coordination problems. The difficulty is that parties may not be aware of how each of them has reasoned. Suppose, for example, that Victoria reasons in the way outlined earlier. To solve the coordination problem, Victor needs to realize that she did so. He may not. Suppose, for example, the children are from Victoria’s first marriage. She and her children are just getting to know Victor, who is assuming an increasingly important role in Victoria’s life. Victor is under the mistaken impression that the only snack the children like is ice cream, so he thinks that to veto ice cream is in effect not to agree to getting the children a snack. He mistakenly takes all this to be clear to Victoria, and, in an ill-advised attempt at cleverness, what he actually M-intends by uttering ‘Okay, but I veto I-C-E C-R-E-A-M’ is that he does not agree that they should get the children a snack. When Victor utters ‘Okay, but I veto I-C-E C-R-E-A-M,’ Victoria however reasons as to the conclusion that Victor agrees except for ice cream, and, intending to confirm their agreement, she utters “Sounds good,” which Victor interprets as her M-intending that she agrees with him on no snack. When Victoria stops to buy cookies, she is puzzled by Victor’s surprise, and Victor is puzzled by her puzzlement. The discoordination results from their failure to solve their earlier coordination problem. There is no proposition in in P meeting the condition that Victor’ M-intends it, and Victoria’s believes it in response.5  I discuss this point at length in (Warner 2019b).  Additional examples of discoordination include Stephen Schiffer’s counterexamples in (Schiffer 1972) to Grice’s definition of speaker meaning. Those examples have the added feature that it is 4 5

Common-Knowledge-Based Pragmatics


In general, solving coordination problems typically requires that the parties know what beliefs and intentions the other parties have. Deciding whether to attend a political protest is a well-known example. Imagine you are one of a number of protesters all of whom dominantly prefer to realize the group-action goal of enough of them attending the protest (Easley and Kleinberg 2010, p. 514): You are aware of a public demonstration against the government that is planned for tomorrow. If an enormous number of people show up, then the government will be seriously weakened, and everyone in society—including the demonstrators—will benefit. But if only a few hundred show up, then the demonstrators will simply all be arrested (or worse), and it would have been better had everyone stayed home.

You will go to the protest if you know enough others will go. But the same is true of them. Each will go if he or she knows enough others go. So you will go if you know that enough others know that enough others will go. What is true of you is again true of the others. Each of them will go if each knows that enough others know that enough others will go. So, you will go if you know that enough others know that enough others know that enough others will go. But the same is true of them, … and so on ad infinitum. The question arises for any coordination problem. Parties solve coordination problems if they coordinate their actions to realize a goal. How does a party know that the others will coordinate? Game theory typically solves this problem by assuming the parties’ preferences are common knowledge among the parties. I propose a similar solution. As explained below, when people interact as speaker and audience, common knowledge arises from their perceptions of each other as fulfilling the roles of speaker and audience. The common knowledge approach requires attributing reasoning to speakers and audiences, but it keeps those attributions to a plausible minimum. Thus the common knowledge approach is an alternative to attributing extensive reasoning to speakers and audiences. To avoid misunderstanding, I should emphasize that we I not saying common knowledge is necessary for solving coordination problems. What common knowledge does is greatly facilitate their solution. When it is available, using relevant common knowledge is in fact that way people typically solve coordination problems. To support these claims, we first need to see how common knowledge arises.

Common Knowledge Generators I begin with an example. The game theorist Michal Chwe notes that, during a game in 1996 (Chwe 2013, p. 41): implausible to see the speaker as meaning any appropriate proposition even though the speaker fulfills the definition. Grice discusses the examples in “Meaning Revisited” in (Grice 1989), where he expresses skepticism about the need to appeal to common knowledge (mutual knowledge* in the terminology of Meaning). Thinking about Grice’s observations lead to my appeal to common knowledge generators.


R. Warner baseball fans at Cleveland’s Jacobs Field [looked] up to see an airplane pulling a banner advertising anonymous HIV testing. Obviously the irony here is the airing of such a sensitive issue as AIDS publicly and even festively on a bright sunny day at the ballpark … [The underlying purpose is that] I would be more likely to get an HIV test if I knew that doing so was not unusual, but I wouldn’t find this out through everyday conversation; at the ballpark, looking up at the plane, however, it is obvious to all that everyone is seeing the same thing.

The display of the banner makes it common knowledge among the people in the stadium that anonymous testing is available. The reason is that the display of the banner meets two conditions: (1) the display of the sign results in everyone who is paying attention knowing that anonymous testing is available; and, (2) the display ensures that everyone who is paying attention knows that everyone who is paying attention knows that anonymous testing is available. The display is an example of a common knowledge generator. A common knowledge generator is a process meeting these two conditions: (1) the process results in all members of a group G knowing X, and (2) the process ensures that all members of G know that all members of G undergo the process and as a result come to know X.  The Appendix shows how derive an infinite series of knowledge statements from (1) and (2). The essential point, however, is that, given (1) and (2), perceiving the sign is sufficient for people to realize they have common knowledge without their having to generate any of the potentially infinite series of knowledge claims. One just needs to realize that one could generate any desired level of knowledge. The existence and role of common knowledge generators receives some empirical confirmation in Thomas et al. (2014, p. 657). They provide evidence for the claim that “common knowledge has its own representation, it need not contain multiple levels of embedded knowledge; it could consist of a single mental symbol which means, ‘We have common knowledge.’” It is worth noting that there is nothing unusual about having a finite representation of an infinite sequence. Suppose 10 year-old Johnny says, “If you start with zero and count one number at a time, you can go on forever.” Johnny knows a finite set of facts that entail (alternatively, he has mental symbols that mean) that the operation of adding the number one to each number starting with zero can in principle be repeated infinitely many times. I will take it for granted that people have the capacity to recognize a wide range of common knowledge generators and form a “mental symbol which means, ‘We have common knowledge.’” The claim is an empirical one—but also a plausible one. What else would you expect from reflective beings for whom thinking about thinking is routine? In the next section, I explain how common knowledge generators arise in communication coordination problems. The explanation entails Grice’s general characterization of conversational implicatures. More precisely, it entails that speakers and audiences involved in communication coordination problems typically commonly know that a relevant instance of the conditions holds. Grice appeals to the general notion of implicature to characterize conservational implicatures. He does not offer a definition of implicature, but the following definition is adequate here. First, define saying that p: suppressing qualifications, one says that p in uttering a sentence S if and only if S means p and one utters it in order to M-intend that p (Grice 1989,

Common-Knowledge-Based Pragmatics


p.  24). Then one can define one central case of implicature this way (again ­suppressing qualifications): One implicates that q in uttering S if and only if one says that p in uttering S and in uttering S also M-intends that q, where p and q are distinct (see Grice 1989, pp. 24–26). Grice also appeals to the Cooperative Principle in characterizing conversational implicatures. It is the principle that, other things being equal, you should “make your conversational contribution such is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged.”6 Referring to the Cooperative Principle, Grice characterizes conversational implicatures as follows (Grice 1989, p. 26): [One] who, by (in, when) saying (or making as if to say) that p has implicated that q, may be said to have conversationally implicated that q provided that (1) he is to be presumed to be observing the conversational maxims, or at least the Cooperative Principle; (2) the supposition that he is aware that, or thinks that, q is required in order to make his saying or making as if to say p (or doing so in those terms) consistent with this presumption; and (3) the speaker thinks (and would expect the hearer to think the speaker thinks) that it is within the competence of the hearer to work out, or grasp intuitively, that the supposition mentioned in (2) is required.

The following account of common knowledge generators in communication coordination problems entails (1) – (3).

 enerating Common Knowledge in Communication G Coordination Problems Speakers and audiences typically have common knowledge that solves communication coordination problems. The common knowledge generators arise from two sources: the social roles of speaker and audience, and the context in which the speaker makes his or her utterance.

Social Roles The relevant social roles are speaker and audience of a natural language involved in a communication coordination problem. To understand their function as common knowledge generators some preliminary comments about social roles generally are in order. Social roles are socially recognized patterns of thought and action associated with standards of permitted, expected, or required behavior (Parsons 1991, Chap. 6). As the sociologists Peter Berger and Thomas Luckmann note “[i]n the common stock of knowledge there are standards of role performance that are accessible to all members of a society, or at least to those who are potential performers of the roles in question” (Berger and Luckmann 1966, p. 73).  Grice, 26.



R. Warner

For my purposes here, I take it for granted that proper performance of the roles of speaker and audience requires knowledge of standard meanings of an appropriately wide range the words and sentences of the relevant natural language. I also take it for granted that—at least for speakers and audiences in communication coordination problems—proper performance requires observing the Cooperative Principle. This may seem wrong. After all, rationality alone—without any appeal to standards associated with a social role—requires conformity to a the Cooperative Principle in communication coordination problems.7 What additional explanatory power does one gain by claiming that conformity to the Cooperative Principle is also required for proper performance? The answer is that it makes it common knowledge for a speaker and an audience that both conform to the Cooperative Principle. As Berger and Luckmann note (Berger and Luckmann 1966, p. 73): not only are the standards of role X generally known, but it is known that these standards are known. Consequently every putative actor of role X can be held responsible for abiding by the standards, which can be taught as part of the institutional tradition and used to verify the credentials of all performers and, by the same token, serve as controls.

To put the point in our terms, social roles are common knowledge generators. You acquire social roles through instruction and acculturation. Those processes typically meet the conditions for a common knowledge generator: (1) the processes result people knowing standards of proper performance, and (2) the processes ensure that people know that people who undergo the processes know the standards.8 Consider Victor and Victoria. They present themselves to each other in the roles of speaker who is competent English language user and audience who is competent English language user. The presentations serve as generators that make it common knowledge that Victor has uttered a sentence that has, among its standard meanings, “Okay, but I veto ice cream,” and that they are both observing the Cooperative Principle. Now how do Victor’s specific M-intentions become common knowledge? Through the context as a common knowledge generator.

Context I take it for granted that having the capacity to reliably identify contextual features that generate relevant common knowledge is part of standards of proper role performance for speakers and audiences. When Victor utters ‘Okay, but I veto I-C-E C-R-­

7  Given a dominant preference to realize the group-action goal of the speaker M-intending that p and the audience in response believing that p, it is rational, other things being equal, to “make your conversational contribution such is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged.” 8  For a detailed explanation, see (Sloan and Warner 2020).

Common-Knowledge-Based Pragmatics


E-A-M,’ the context—we may assume—ensures that (1) Victoria and Victor know that the best explanation of Victor’s uttering is that he M-intends that he (as the contextually specific reference of the indexical ‘I’) agrees with getting the kids something, and (2) they know (1). Assume the context also ensures that it is common knowledge that Victor vetoes ice cream but does not want the children to know he mentioned it. To see how, note that Victor’s spelling ‘ice cream’ appears to be complicate communication for no apparent reason. As such, it is an apparent violation of the Cooperative Principle. However, assume that the context ensures that (1) given that Victor is observing the Cooperative Principle, Victoria and Victor know that the best explanation of Victor’s spelling out ‘ice cream’ that he M-intends that he vetoes ice cream and wants the children not to know that; and, (2) Victor and Victoria know (1). Such cases fulfill the second condition of Grice’s characterization of conversational implicatures: “the supposition that [the speaker] is aware that, or thinks that, q is required in order to make his saying or making as if to say p (or doing so in those terms) consistent with [observing the Cooperative Principle].” Now what about the third condition? That condition requires that “the speaker thinks (and would expect the hearer to think the speaker thinks) that it is within the competence of the hearer to work out, or grasp intuitively, that the supposition mentioned in [the second condition] is required.” This follows from the second condition’s being common knowledge.

Common Knowledge-Based Pragmatics A common knowledge-based approach to pragmatics would answer at least the following question: What informative generalizations are possible about the role of contexts in generating common knowledge relevant to conversational implicatures? The question is important. As Thomas et al. note (Thomas et al. 2014, p. 672): In recent decades, psychologists have recognized that cooperation is one of the hallmarks of the human species, and that its game-theoretic demands have shaped our emotions, our morality, our social relationships, and our language. Much has been learned about these domains of psychology from a focus on the problem of altruistic cooperation and the mechanisms of reciprocity. We hope that comparable insights are waiting to be discovered by psychologists as they investigate the problem of mutualistic cooperation, and the mechanisms of common knowledge are—as we might say—put out there.

Appendix Assume the banner is flying over the stadium. Assume (1) everyone paying sufficient attention sees and thereby comes to know the banner is flying over the stadium, and (2) everyone knows (1). To see how (1) and (2) give rise to an infinite


R. Warner

sequence of knowledge levels, suppose Alice and Bob are sitting together in the stadium, and they are looking up at the sign. Then: First level: Second level: Third level:

Alice knows1 that Bob sees the sign. Bob knows1 that Alice sees the sign. Alice knows2 Bob knows1 that Alice sees the sign. Bob knows2 Alice knows1 that Bob sees the sign. Alice knows3 Bob knows2 Alice knows1 Bob sees the sign. Bob knows3 Alice knows2 Bob knows1 Alice sees the sign.

How do Alice and Bob take this infinite series of steps? Start with the first level. Alice knows that Bob sees the sign by reasoning this way: “I see Bob looking up toward the sign. He has normal perceptual abilities, so I can conclude that he sees the sign.” Bob reasons the same way about Alice. This gives us the first level: Alice knows that Bob see the sign, and Bob knows that they see the sign. They get to the second level by reasoning about each other’s reasoning at the first level. Alice reasons: “At the first level, I started from the fact I saw Bob’s looking at the sign, and I reached the conclusion that Bob sees the sign. Bob sees looking at the sign, and he has normal perceptual and reasoning capacities, so he will have reasoned just as I did to his conclusion that I see the sign.” For Alice to realize that fact about Bob is for her to reach the second-level conclusion: “I know2 Bob knows1 I see the sign.” Bob will reason in the same way to his secondlevel conclusion that he knows2 Alice knows1 he sees the sign. Thus: Alice knows2 Bob knows1 Alice sees the sign, and Bob knows2 Alice knows1 Bob sees the sign. Alice and Bob get to the rest of the levels the way they get from the first level to the second: by reasoning about their reasoning at the level below. For any level n, Alice reasons about Bob’s n−1 level reasoning to reach the conclusion that Alice knowsn Bob knowsn−1 … that Alice sees the sign, and Bob reasons in the same way to reach the conclusion that that Bob knowsn Alice knowsn−1 …. that Bob sees the sign. There is an obvious problem of course. No one reasons in this way. It is extremely unlikely that Alice and Bob reason in a way that even approximates the reasoning we have attributed to her. The answer is that Alice and Bob do not need to derive the infinite sequence. They commonly know that they see the sign as long as each knows (1) that the other sees the sign and thereby comes to know the banner is flying over the stadium, and (2) each knows that the other knows (1). Then, to start with Alice, for any number n, she has the capacity to generate the sequence Alice knowsn Bob knowsn−1 … that Alice sees the sign, and Bob has the capacity to generate the sequence Bob knowsn Alice knowsn−1 … that Bob sees the sign. It was convenient to characterize that capacity by representing Alice and Bob as reaching successive knowledge levels through explicit reasoning. But it is the capacity that matters. It means that Alice and Bob are capable of ruling out any possibility of doubt or deception with regard to their seeing the sign at any level of knowledge.

Common-Knowledge-Based Pragmatics


References Berger, P., & Luckmann, T. (1966). The social construction of reality: A treatise in the sociology of knowledge. New York: Anchor Books. Chwe, M. S.-Y. (2013). Rational ritual: Culture, coordination, and common knowledge. Princeton: Princeton University Press. Easley, D., & Kleinberg, J. (2010). Networks, crowds, and markets reasoning about a highly connected world. New York: Cambridge University Press. Grice, P. (1957). Meaning. The Philosophical Review, 66(3), 377–388. 2182440. Grice, P. (1989). Studies in the way of words. Cambridge: Harvard University Press. Grice, P. (2001). In R.  Warner (Ed.), Aspects of reason. Oxford, UK: Oxford University Press. Korta, K., & Perry, J. (2015). Pragmatics. In The stanford encyclopedia of philosophy. Metaphysics Research Lab, Standford (CA): Stanford University. archives/win2015/entries/pragmatics/. Parsons, T. (1991). The social system. London: Routledge. Schiffer, S. (1972). Meaning. Oxford, UK: Oxford University Press. Schiffer, S. (1987). Remnants of meaning. Cambridge, MA: MIT Press. Schiffer, S. (2017). Gricean semantics and vague speaker-meaning. Croatian Journal of Philosophy, 17(3 (51)), 293–318. Sloan, R., & Warner, R. (2020). The privacy fix: How to preserve privacy in the onslaught of surveillance. Cambridge, UK: Cambridge University Press. Spiegelhalter, D. (2019). The art of statistics: How to learn from data. New York: Basic Books. Thomas, K.  A., DeScioli, P., Haque, O.  S., & Pinker, S. (2014). The psychology of coordination and common knowledge. Journal of Personality and Social Psychology, 107(4), 657–676. Warner, R. (2019a). Indirect reports in the interpretation of contracts and statutes: A Gricean theory of coordination and common knowledge. In A.  Capone, M.  Carapezza, & F.  L. Piparo (Eds.), Further advances in pragmatics and philosophy: Part 2 – Theories and applications (pp. 25–40). Cham: Springer. Warner, R. (2019b). Meaning, reasoning, and common knowledge. Intercultural Pragmatics, 16(3), 289–303.

The Primacy of Semantics and How to Understand It Nenad Miscevic

Introduction In the realm of meaning, what territory belongs to pragmatics, and what to semantics? The vehicle of content is primarily semantic, and is run by its ‘semantic engine’ so to speak, the tradition claims. I shall call the followers of the tradition “semanticists”, since for them the encoded semantic content is sufficient or almost sufficient for determining truth conditions. However, the dissenting voices have been present for almost a century. The content is to a large extent determined by pragmatic factors, and these go way beyond the traditional list of time, place, speaker and the like. Let us remind ourselves of already famous from the literature.1Consider the utterance of “Everyone went to Paris”,2

In order to know its truth conditions one must know who counts as “everyone”. The pragmaticist suggests that it is a matter of wide pragmatic considerations. Therefore, such considerations enter the crucial jobs of determining and recognizing truth-­ conditions, originally and traditionally reserved for semantics. The other famous examples are, of course, weather reports like “It is raining” (Perry and Blackburn 1986). Which area is relevant for the truth-value of this utterance? For me it is usually the town where I live, Rijeka, but it can be also Budapest

 Many of them are listed and brilliantly discussed by Kent Bach in his pioneering (Bach 1994).  Recanati’s famous example, whose ancestor appeared in (1989, p.  314). See also Recanati (2004, p. 10). 1


N. Miscevic (*) Central European University, Vienna, Austria e-mail: [email protected] © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 F. Macagno, A. Capone (eds.), Inquiries in Philosophical Pragmatics, Perspectives in Pragmatics, Philosophy & Psychology 27,



N. Miscevic

where I have been teaching a trimester per year, for almost two decades. Again, the pragmaticist suggests that it is a matter of pragmatic considerations, which again determine the truth-conditions of the sentence. Only a lot of pragmatics can assure a definite truth value! And there is no firm guidance from encoded, semantic meaning. I shall call authors like him (Bezuidenhout 2004; Carston 2002; Sperber and Wilson 1995), “pragmaticists”; Neale calls them “heavy handed pragmatists”, and there are other nicknames in the literature. It is all a matter of pragmatic considerations; our sentences taken in isolation have no truth value! The debate between semanticists and pragmaticists strongly marks the present-­ day philosophy of language (see, for example the Springer series “Perspectives in Pragmatics, Philosophy & Psychology” which nicely illustrates the state of the game. In the debate, my own sympathies are on the side of semanticists, and I would like to propose a defense, built upon analogies with semantics of anaphora (and its wider cognates). Here is then the preview. The next section prepares the ground by briefly presenting the semantics/pragmatics “territorial wars”, and points to a few typical weaknesses on each side. Section “The Guidance Proposal” is dedicated to my own proposal, that I call “the guidance view”, according to which the semantic component guides and drives truth-conditional interpretation; it is the “semantic engine” doing the main job in the process of interpretation.

The Territorial Wars: Semantics Versus Pragmatics The Pragmaticist Proposal The analytic philosophy of language has been marked by the semantics/pragmatics contrast since the mid-twentieth century, when the semantics-centered Fregean tradition had to face Wittgenstein’s insistence on language use as the determining factor. In the last two or three decades the contrast has provoked a kind of territorial wars, semanticists claiming the central territory of linguistic phenomena, in particular of “meaning” for themselves, and pragmaticists coming with the opposite claim. Let me start with the pragmaticists, and then turn to recent semanticist answer(s); all this with apologies for brevity. To stay for a moment with two examples from the Introduction, the sentences like ““Everyone went to Paris”, as well as sentences making up weather reports, do not have their truth-conditions fixed by their lexical meaning and the composition rules used to put them together. Recanati, one of the pioneers of pragmaticism, has a lengthy discussion of such utterances in his characteristically titled Truth-conditional pragmatics (2010, pp. 77–125). Here, in order to get a proposition with clear truth-conditions, an enrichment is needed, and it comes from pragmatic circumstances: who are the persons whose whereabouts are the topic of conversation in the Paris example, or location of the speaker, or the place

The Primacy of Semantics and How to Understand It


she has in mind, in the case of weather reports. Herman Cappelen and John Hawthorne nicely summarize the view (which they otherwise reject): Philosophers such as Sperber and Wilson, Carston, Bach, and Soames (…) think that so-­ called propositional skeletons can be the semantic values of sentences (relative to contexts). These philosophers simply hold that semantics, in some cases, generates semantic values for sentences that fail to reach the level of propositionality’. Such subpropositional semantic values are neither true nor false. (Cappelen and Hawthorne 2009, p. 24)

Examples abound. “The table is covered with books.” normally refers to particular table, either the one that speaker has in mind, or one that is prominent in the conversational situation; none of this is present in the “propositional skeleton” proposed by the sentence, The verbs like “cut” and “open” can be used for a wide variety of actions (“cut the grass and “cut the cake” point to very different kinds of action, as Searle noted in his (Searle 1980, p. 22) see the discussion in Recanati 2010, p. 169). “Light” in “light sound” carries has quite a different sense than in “light luggage”, as Recanati notes in (Recanati 2010, p.  32n). Next, there are non-literal meanings, like the meaning of “swallow” in “The machine swallowed my credit card. Further, and famously, one might use “I haven’t eaten.”

to mean one haven’t eaten breakfast today. And, one might use “You’re not going to die.”

to suggest to one’s child in a certain context that the child will not die from a minor cut. Recanati proposes the following as a general pragmaticist framework for dealing with such cases (Recanati 2010, p. 42). We should “distinguish two types of contextual processes, possibly affecting truth-conditions.” On the one hand, there is ‘saturation,’ through which indexicals and free variables are “assigned a contextual value”. On the other, there is “modulation”, through which the meanng of an expresson is mapped to “distinct meaning”. The clearest case are metaphorical and metonymical processes, like the one involved in interpreting “swallow” when it comes to credit cards. But there are other processes of enrichment, for example the making of the meaning of “cut” more precise: not just ‘linear separation’, but the slicing mode of separation typical, for instance, for cases of bread and cakes. Similar moves have been made by relevantists. Let me start with a pertinent quote, of two passages of Relevance in which Wilson and Sperber offer a sketch of a framework, and an interesting example. Here is the framework. It begins with a linguistic description determined by the grammar, which is considered to be quite universally valid. It then passes to semantics: Second, this linguistic description yields a range of semantic representations, one for every sense of the sentence uttered. Each semantic representation is a schema, which must be completed and integrated into an assumption about the speaker’s informative intention, and can be as complex as the speaker cares to make it. Moreover, each schematic sense is generally quite different from all the others, and can be completed in quite different ways. (Sperber and Wilson 1995, p. 175)


N. Miscevic

It looks as if these semantic representations, associated with the lexicon, are relatively stable. Wilson and Sperber suggest that “the encoded concept helps to activate contextual implications that make the utterance relevant as expected” (Wilson and Sperber 2012, p. 110). In their work on metaphor Wilson and Sperber offer a fine example of how people interpret a metaphor they meet for the first time. In the example, Peter and Marry discuss whom to invite to Billy’s birthday party. Mary says: (1) “Archie is a magician. Let us invite him.”

What if Peter has only one encoded meaning for magician, person with supernatural powers. For that matter, some people may have only a single encoded sense for “magician”: someone with supernatural powers who performs magic. They would still have no difficulty arriving at an appropriate interpretation of (7) by extending the category of “real magicians to include make-believe ones. For other people, the metaphorical sense may have become lexicalized, so that “magician” has the additional encoded sense someone who achieves extraordinary things. Suppose, such a person hears the utterance of (2) “My chiropractor is a magician”

How is she going to react? They would obviously have no trouble arriving at an appropriate interpretation of (29). Mary did not intend her utterance to be understood literally in (7) and metaphorically in (29); her communicative intentions – like those of all speakers – are about content and propositional attitude (Wilson and Sperber 2012, p. 114). In brief, the pragmaticists claim that the semantic meaning is minimal, and insufficient for reconstructing the import of typical ordinary utterances; most importantly, it is insufficient for reconstructing their truth-conditions.

The Semanticists Strike Back: A Radical Proposal The semanticists have been looking for an alternative, that would guarantee the sufficiency of the encoded meaning for determination of truth conditions. The minimalists (Lepore, Cappelen) come close to establishing sufficiency, but at a high price. For them, it is the strongly literal meaning. If Ann says “I haven’t had breakfast.” The truth-condition of what she said is that she never had breakfast. With such truth-conditions, the sentence is normally false; not a very attractive option. Michael Devitt in his recent work (Devitt 2013a, b, c, d; 2018) has a different, but also radical proposal: the totality of literary meaning is really semantic. Most of the properties discussed in the debate, and claimed by the pragmaticists to be pragmatic properties, are in fact semantic, his answer goes. In particular, the seemingly pragmatic determiners of the truth conditions of our utterances are all semantic.

The Primacy of Semantics and How to Understand It


To mention two crucial examples, “Everyone went to Paris” is implicitly restricted to everyone in a certain group the speaker has in mind (Devitt 2018, Chapter 8, p.  6, also in Devitt 2013a, p.  94). Weather reports like “It is raining” implicitly refer to a location that the speaker has in mind (Devitt 2018, Chapter 8, p. 6). And this restriction is semantic, not pragmatic. In this sense, he is a kind of pan-semanticist. His line has a critical and a positive component. The critical component urges a strict distinction between metaphysical issues (like “what constitutes the content of this utterance?) and the epistemological ones (like, “how does the average hearer reconstruct the content). What about variations in meaning, cases like “swallow the credit card”? Devitt agrees, that speakers’ conceptions linked with a given expression vary enormously. ”Rabbit” can mean the animal, the meat (food) or the skin, “cabbage” is a plant, and a kind of food. “The crown” can refer to a monarch. Each of these different contents influences the truth conditions of utterances. But then, we have different meaning and different semantics content in each case. “Crown” in the sense of “monarch” is just a dead figure of speech, and this is part of the semantics of “crown”, not of the conversational context: “The ATM swallowed my credit card” (Recanati 2004, p. 26)

may once have been an example although now it is surely a dead metaphor (Devitt 2018, sec. 8). I agree; I have been arguing that pejoratives used by usual targets, but in friendly conversation (like “We, bitches” do not stand for a new meaning of “bitch” (Perhat and Miscevic 2016). “The ATM swallowed my credit card” (Recanati 2004, p. 26)

may once have been an example although now it is surely a dead metaphor. The main job in his proposal is done by appeal to conventions, and to the thesis that the relevant conventions are all semantic, rather than pragmatic: The rules of human languages are largely conventional…. I think that the conventions associated with a linguistic form—a sound, an inscription, etc.—in a community typically come from the regular use in the community of that form to convey certain parts of messages. That regular use of it in utterances with a certain speaker meaning, leads, somehow or other, to that form having that meaning conventionally in the language of that community. (Devitt 2013d, p. 97)

“Linguistic expression typically gets its meaning from a convention. But conventions do not constitute the meanings of a language.” Devitt claims (Devitt 2018, Chapter 4, p. 1). The contents-meanings that are the object of the current discussion are in fact part of the convention-generated meaning of the expressions under discussion. Now, conventions are very specific The convention says that what you have in mind determines the meaning “Everyone went to Paris” is implicitly restricted to everyone in a certain group the speaker has in mind. (Devitt 2018, Chapter 8, p. 6, also in Devitt 2013a, p. 94) Weather reports like “It is raining” implicitly refer to a location that the speaker has in mind (Devitt 2018, Chapter 8, p. 6)


N. Miscevic

So, what speaker has in mind is not part of the pragmatic context, but part of semantics, thanks to presumably (or allegedly) semantic convention, giving it semantic status.3Devitt notes the parallel with indexicals and the usual and unproblematic manners of saturation: “The first way that the tradition takes account of context relativity is by noting that the references of indexical elements are determined in context...Perhaps many pragmatist phenomena can be accommodated in this indexical way. I think that many can: there are more indexical elements in language than the familiar list. I have already argued for one in Chapter 7: there is a convention of using a description in a way that demands saturation in the context; its reference is fixed, as we might ordinarily say, by “what the speaker has in mind”; or, as I would say theoretically, by an underlying causal-perceptual link to an object. Here, briefly, are some other plausible conventions of saturation in context:” (Devitt 2018, Chapter 8, p. 6)

It is really the character of the indexical that does the semantic job; with Devitt we have a ‘character-mania’ tied to intentions. At this point it is good to note that this is a rare position in the debate. Zoltan Szabo, listing the generally accepted criteria of what makes a content semantic give a place of pride to: Intention-independence. Typically, some but not all of what the speaker conveys is independent of the speaker’s specific intentions to talk about this or that. (Szabó 2005, p. 6)

With Devitt, it is the opposite. Most problems are solved by combining convention and having in mind. For instance, “A restriction on the domain of quantifiers. Thus, the domain of “Everyone went to Paris” is implicitly restricted to everyone in a certain group the speaker has in mind.” Next, it is “Weather reports like “It is raining” implicitly refer to a location that the speaker has in mind. (Devitt 2018, Chapter 8, p. 6) In brief, Devitt proposes an extreme pan-semanticist answer: their semantic content is complete, since it relies on the convention pointing to what speaker “has in mind” (Devitt 2013d, p. 89). The semantic vehicle is complete, pragmatics offers just lining and paint. Devitt things they are sufficient. Let me briefly pass to some problems and questions. Pan-semanticist conventionalism offers a picture of semantic content that makes it look like a Swiss cheese, full of holes; one wonders if there is any word in the language or phrase in discourse worth mentioning that does not contain a slot, regulated by convention-plus-having­in mind. The first problem with such conventionalism-intentionalism is that in many cases there are no definite conventions regulating how to fill the slot, rather, there is some

3  Let me quote a footnote offered by Devitt at this point: Fn 3 Some prefer to say that the reference is determined by what the speaker “intends to refer to”. This can be just a harmless difference but it may not be. Having x in mind in using the term simply requires that the part of the thought that causes that use refers to x. In contrast, for a speaker literally to intend to refer to x, given that intentions are propositional attitudes, seems to require that she entertain a proposition containing the concept of reference. So she can’t refer without thinking about reference! This would be far too intellectualized a picture of referring. Uttering and referring are intentional actions, of course, but it seems better to avoid talking of intentions when describing them (Devitt 2013d, p. 89).

The Primacy of Semantics and How to Understand It


determination, or quasi-determination, a kind of guidance that makes one reading more acceptable than the other. It often comes from the context of conversation. A related problem is the multiplicity of dimension of variation that has been noted in Wittgensteinian tradition.4 Let us take a look at both kinds of problems. So, consider first the case of mistake or slip of tongue. Suppose the conversation is about the pope Frances, who (actually) does not want to make our Croatian collaborationist archbishop from the times of the Second World War into a saint. My friend says: “He is quite intransigent.” My thoughts have wondered, I actually think intensely about the former pope, Benedict. And I say: “He loves the Croatian archbishop.” What did I say? To whom did I refer? In conversation, it seems that I said that pope Francis is quite intransigent. But this is not what I had in mind. If my friend said: “Are you crazy, Frances does not love the archbishop!”, I would not claim that I said that Benedict does, but that I stupidly “meant” Benedict, talking about Frances. It is the conversation, not convention about having in mind that hold the upper hand. Devitt and Hermogenes are on the wrong track. Call it “the Mistake Problem”. A variant of it are malapropisms. To quote from the web, “cylinders for ‘syllables,’ anecdote for ‘antidote,’ facilities for ‘faculties’”. bathtub-effect-language-1689163 Take “it is raining” and assume that the area is given by speaker’s intention, what s/he has in mind. Miss Malaprop says “My relative urgently needs an anecdote”, having in mind “antidote”. Unfortunately for her and for her relative, what she says, the semantic content, is that the relative needs an anecdote, period. Another, less damaging but still relevant variant is the Mental Reservation Problem. Ivo and Pero are talking on the phone about weather, and Pero asks about the weather “Ivo, should I come to Rijeka, or is it raining?” Ivo is in Rijeka, but he imagines he is in London; he knows it is raining there, concentrates and says “It is raining.” Is he lying to get rid of Pero, or did he just “say what he had in mind”, truly? Well, Ivo had London in mind, so his sentence was, by Devitt’s standard, true since its conventional meaning, by the same standard, is that it is raining in London. But this is not how we take the sentence. Ivo’s mental reservation is just a trick, a way of lying. His sentence says that it is raining in Rijeka- it is Pero’s question that determines the place, not what Ivo has in mind. There is no having-in-mind convention that Devitt hopes for. So much about the first problem. Let us now pass to variation. What about intensity? In rainy areas like north England, a few drops of rain don’t count as verifying “It is raining here”, but in an area hit by draught, occasional drops of rain are precious: “It is raining, thanks Allah!” a desert dweller might thankfully say. Are there several dimensions for convention? Here is an example concerning time. Suppose I say: “I haven’t eaten breakfast.”

 Charles Travis is the most creative active author in this tradition (see, for instance Travis 2006).



N. Miscevic

in the discussion where my wife wants to know if I have eaten breakfast yesterday morning before measuring my blood sugar level. In fact, I am lying, I haven’t eaten breakfast the day of conversation, but I did the day before. What convention is being violated by my statement? It seem that it is more the context of conversation than any standing convention that determines the time in question. This brings us to a more general point: questions obviously determine place, time, relevant group of people and the like. But there is no need for explicit questions, it can be the context of conversation, like in blood sugar case. Now, what determines other dimensions, like intensity (of the rain or whatever)? There is a lack of standing conventions of the kind the semanticist needs. So, suppose we want to add various contextual factors (who talks to whom, preceding questions, etc.) as determinants of primary semantic material (meaning), and other contextual factors for determining less prominent dimensions we want to import into the meaning. And we claim that each such factor produces a new meaning. None of them are noted in the dictionary. Is not the semantic Ockham’s beard growing beyond proportion, and no razor is available? Even if we accept Devitt’s strictures against what he calls the Third Methodological Flaw of Linguistic Pragmatism, the embrace of “Modified Occam’s Razor” (Devitt 2018, Chapter 6) we might agree that one must stop somewhere; not every variation along every dimension produces a new meaning or a standing sense, a concept. It seems that the pan-semanticist proposal, combining convention and having-in-­ mind would not work. We, semanticists, have to look elsewhere, even if Devitt’s radical stance stays as a source of inspiration. Remember Kent Bach’s car metaphor for the contrast between different semanticist viewpoints: if we compare the content of an utterance with say, a car, the least ambitious pro-semantic theoretician will say that the semantic part is a mere skeleton; it has solidity but it does not propel the whole thing (see Bach 1994, p. 126). Devitt’s answer goes to the opposite extreme: pan-semanticism and conventionalism. In terms of the car comparison, semantic gives you the whole thing, only unpainted: the encoded semantic content is almost sufficient for truth conditions. Now, to continue with comparing and pass to my own view, I agree with his semanticist inspiration, but I am afraid he has gone too far. I would propose a more modest for semantic content, namely guidance. In terms of the car, it is like the powertrain of a modern automobile, comprising engine (with exhaust system), transmission, drive shaft, suspension and the wheels.

The Primacy of Semantics and How to Understand It


The Guidance Proposal The Isolation Strategy and Why It Is Misleading Let me start from an interesting fact. All main pragmaticists we have been mentioning (and some semanticists) consider the target sentence-utterance whose semantic content is being discussed by them in (relative) isolation, and then claim that they don’t give us their truth-conditions. Recanati and Bach are good examples. Devitt concentrates on the mind of the speaker, and neglects the conversational context. Call this approach “isolation strategy”. Normally, sentences like (1) I’ve had breakfast. (2) You are not going to die. (3) It’s raining. (4) The table is covered with books. (5) Everybody went to Paris. and the rest would most often be said in situations of ongoing communication where a question has been asked, or a pointing has been made, and the like. Consider (1) I’ve had breakfast.

When are such sentences normally produced? Often as an answer to the question “Did you have breakfast today?”, or “You look hungry, did you have breakfast?” If we incorporate the question, we end up having the TC-content: Recanati recognizes this (Recanati 2010, p. 124), and then changes the topic. “You look hungry, did you have breakfast?” is a usual question, and everyday knowledge about periodicity of having breakfast indicates that what is meant is “today”: we again have the TC-content. Or consider (2) You are not going to die.

In what situation(s) are such things said? For instance, as reaction to “Mom, it hurts so much!”

And here “it” specifies the antecedent, “hurts” specifies the topic, and the truth-­ conditions are that the child will not die from the hurt. Alternatively, the child is intensely looking at its finger, and trying to direct parent’s attention. The direction of gaze indicates the topic. Very rarely the child will just cry, with no indication of cause, no attempt to direct the attention of the parent to the finger. Finally, and most ironically, the weather reports. Consider Recanati’s diagnosis: In the case of “rain” we find that the contextual specification of the place of rain results from free enrichment, (…), because it is not mandatory. In the weatherman example, no place of ran 15 specified (…) In all cases, a process of free enrichment may take place in virtue of which the speaker tacitly refers to a particular place as the place of the described event, but this process is entirely pragmatic and is therefore irrelevant to the semantics of the event predicate. (Recanati 2010, p. 95)

But actual weather reports on TV give you a map with rather precise contours!!! But with ordinary statement of


N. Miscevic (3) It’s raining.

there is a rich area of possibilities. It could be a comment out of blue. Then, and only then is it indeterminate the way our authors see it. Normally, it could be and often is an answer to a question: “What is the weather like in Budapest?”,

and then we have an anaphora: “It is rainy…”. Or, “Take a look. Is it snowing?” “…is it sunny?”

It is quasi-anaphoric and suggests the area immediately surrounding “you”. Finally, the indefinite reference: (4) The table is covered with books.

Said in what situation? Not in isolation, specially not in a complete one. A normal context would involve questions like “I see a table in the corner. Can I put my things on it?”

The same holds for other popular examples like (5) Everybody went to Paris. (question: Where are the colleagues?), or: I am not ready. In brief, it is hard to imagine that the utterances unprovoked by any recognizable antecedent are the rule. But if they are not, the discussion has been wrong-headed. Isolationism misrepresents the normal context of our target sentences. I would like to argue that once we give up on isolationism, it is easy to notice linguistic connections of the target utterances with other sentences in the context. Even Recanati mentions some of the problematic sentences (“I have eaten”, in the context of speaker declining dinner, taken from Wilson and Sperber 2012, p. 207). But in these cases the link seems to be linguistic and thus arguably semantic, rather than pragmatic. Again, once we give up on isolationism, we might get more interested in situations in which our target sentences are responses to something just said; the semantics of such responses is very different from the semantics of isolated utterances (sentences and often sub-sentences). They get saturated from the preceding material. I suggest that we take this link to what has been said as the most relevant link in many cases of saturation. If we do this, the phenomenon we encounter first is anaphora.

Anaphora and Guided Saturation The clearest case of semantic determination from the context of conversation is the case of anaphora. Let me use material from Stainton (Stainton 2005, 2006); he is more of a pragmaticist, so I cannot be accused of using biased evidence from linguistics. Here is how he introduces the case of elliptic speech exemplifying anaphora.

The Primacy of Semantics and How to Understand It


Imagine Steve being asked the question ‘What language does Mary write in?’ and he says: “In Latin”. Steve obviously believes that Mary writes in Latin, and in addition, he thinks that having said “In Latin” he has suggested that Mary writes in Latin. The phenomenon is known as anaphora, and speakers normally have no problem with it. If one asked Steve what he informed his interlocutor about, he would produce the judgment “that Mary writes in Latin.” (Stainton 2006, p. 33)

Here we have the example of anaphora, where the work is done by the syntax, and the resulting intuition that the meaning (in the example “Mary is writing in Latin” is semantic. In looser situations, where there is no pronounced antecedent, the listener and the judge are guided by the canon of the strict case. Let me use Stainton’s example (from Stainton 2006, p. 34). Rob points to a boat and says “Pretty fast”. Stainton contrast this with the stricter situation, in which there is a clear syntactic antecedent (He presumably has in mind cases like the following short conversation: “How do you find the speed of the boat? Pretty fast”) But note that the looser situation is analogous to the stricter situation. What did Rob say? That the boat is pretty fast, intuition replies. And it is almost correct. It seems that the listener’s and judge’s intuitions proceed by analogy. Let me than uses anaphor as the model for partial determination or guidance: it assumes that anaphor is semantic, shows that anaphor guides the hearer in determining the truth-conditional content, argues that most problematic cases are anaphor-­ like (and the rest can be dealt with), Let me call my proposal “guidance view”. The main steps are easy to grasp: First, assume that anaphora is semantic, Second, show that anaphora determines the truth-conditional content guides the hearer in recognizing the determination and the content, Third, argue that most problematic cases are anaphora-like (and the rest can be dealt with) and that in quasi-anaphoric cases the hearer proceeds in an analogical fashion. If this holds, it follows Fourth, that guidance view is very close to being the right one.

By guidance I here mean objective guidance, or quasi-determination, not mere epistemic help. I admit that the construction of content is literally and stricto sensu pragmatic, but it is strongly determined-guided by semantic elements. Let me put the third, crucial point in context and defend it briefly. I am taking anaphora as a model. I also assume the following distribution of situations. First, complete out-of-the-blue utterances are extremely rare in normal situations, and when they appear, they are typically uninterpretable. Second, normal anaphora is quite frequent: elliptic sentences very, very often appear in reaction to the preceding discourse, which completely determines the slot-filling. In between these two extremes, we find quasi-anaphoric situations, in which there is a verbal antecedent, but it does not clearly determine the slot-filling in a linguistically-semantically unproblematic way. Finally, there are situations in which verbal antecedents are replaced by other events in interaction: the common direction of the gaze, pointing or almost pointing gestures, and the like. Let us illustrate the variety with the key examples from the debate. (1) I’ve had breakfast.


N. Miscevic

Suppose, the question was: “Did you have breakfast today?”, and the answer was: “I’ve had breakfast.” This is a clear case of anaphora: the deleted item TODAY is present in the syntactic structure. What about “You look hungry, did you have breakfast?”

Here we have a quasi-anaphora: everyday knowledge about periodicity of having breakfast. The same with You want some food? “I’ve had breakfast.”

It is a quasi-anaphora. Sentence needs saturation, and anaphora-like component (question activating the periodicity knowledge) is guiding the saturation. (2) You are not going to die.

Said in what situation? Perhaps, as reaction to “Mom, it hurts so much!” Mom says that the child is not going to die from it, the hurt or the cause of it. It is very close to anaphora, maybe it is even one. If there is nothing said by the child, Mom might take direction of child’s sight as the indication; again quasi-anaphora, not so distant from a genuine one. (3) It’s raining.

Said in what situation? If as an answer to a direct question. (say, “What is the weather like in Budapest?”) it is anaphora. What if it’s the answer to the following: “Take a look. Is it snowing?” …is it sunny? It’s quasi-anaphora referring to the area immediately surrounding you, accessible for your looking. (4) The table is covered with books.

Said in what situation? “I see a table in the corner. Can I put my things on it?” anaphora: the deleted item is present in syntactic structure (5) Everybody went to Paris Said in what situation? “Where are the colleagues?”5

Consider now two kinds of problems we raised for pan-semanticist conventionalism. First the case of mistake or slip of tongue. First, the pope(s) example. If the conversation is about the pope Frances, but I actually think intensely about the former pope, Benedict, and I say: “He loves the Croatian archbishop.” it is the conversation, not convention about having in mind that offers an antecedent. Frances is chosen by quasi-anaphora as the referent of what I said. Second, malapropisms. Miss Malaprop says “My relative urgently needs an anecdote”, having in mind “antidote”. Here the literal meaning wins, we don’t even need special guidance. Third, the Mental Reservation Problem. In our example, Ivo had London in mind, but it did not help. It is Pero’s question that determines the place, in a direct anaphoric way.

5  These are clearly cases of procedural knowledge. Now, Bezuidenhout (2004) argues that procedural knowledge is pragmatic; I disagree but I have to leave it to another occasion.

The Primacy of Semantics and How to Understand It


Similarly with variation. What about intensity? A few drops of rain fall somewhere in north England. In the local conversation, they don’t count as verifying “It is raining here”, period. Similarly, with our example concerning time. Remember, I am supposed to have said “I haven’t eaten breakfast.” in the discussion where my wife wanted to know if I have eaten breakfast yesterday morning before measuring my blood sugar level. It is her question that counts, in pretty much anaphoric way. My answer, if properly filled in syntactico-semantic fashion is “I haven’t eaten breakfast yesterday morning before measuring my blood sugar level.” And it comes out false, as commonsensically assumed. Now, one might object to the guidance proposal from two directions. Someone like Michael might ask: isn’t this just a version of conventionalism? And the answer is: No, it’s less intentionalistic than what you see as conventionalism, it is more linguistic, focusing upon linguistic context of conversation. A pragmaticist, like Stainton, might attack it from the opposite direction: the “guidance” is just a pragmatic phenomenon that has little or nothing to do with semantics. But consider the analogy with indexicals. Their content was first seen as pragmatic, but soon, already with Montague, theoreticians recognized the strong and systematic determination of content, and started counting saturation for main indexicals (I, now) as being semantic or almost. I propose we do the same with quasi-anaphora: the proximity to the pure syntactico-semantic determination (i.e. proximity to anaphora stricto sensu) suggests an analogous semantic treatment. After having discussed the metaphysics of meaning, let me quickly turn to epistemology, Let me use my proposal about the nature of objective guidance, the powertrain work, to the topic of epistemic guidance in producing intuitions about what is said. The problem is that many authors agree that people’s intuitions about what is said (what is semantic, to put in in the jargon) and what is meant are unreliable: people tend to see most of the stuff as semantic. Here is Kent Bach: I conclude that intuitive and related cognitive considerations do not undermine a minimalist conception of what is said. As Jerry Fodor says, “No doubt, intuitions deserve respect, … [but] informants, oneself included, can be quite awful at saying what it is that drives their intuitions. … It is always up for grabs what an intuition is an intuition of” (1998, p. 86). In the case of seemingly semantic intuitions, they are largely irrelevant to determining what is said. They are influenced by semantically irrelevant information, they tend to be insensitive to relevant distinctions, and they are likely to be biased in favor of understandings corresponding to things that people are relatively likely to communicate. Or so it seems to me, at least intuitively. (Bach 2002, p. 32)

Let me give a fresh example, tied to the semantic-pragmatic distinction, having to do with intuitions that concern what speaker said vs. what she implied by her given utterance. The testing of speaker’s intuitions is in principle simple. Take as the example the utterance of I have three children

Its literal meaning, the semantics, is the proposition that the speaker has at least three children, i.e., no less than three but possibly more. But when you ask native naïve “judges” what was said, they claim that it was that the proposition that speaker


N. Miscevic

has exactly three children, which is by most theoreticians taken to be implied, not literally said.6 Bach draws a very bleak conclusion about such intuitions: People’s spontaneous judgments or “intuitions” provide data for semantics, but it is an open question to what extent they reveal semantic facts and should therefore be explained rather than explained away. Since, as I am suggesting, they are often responsive to non-semantic information, to what is implicit in what is said but not part of it, they should be treated cautiously. (Bach 2002, p. 24)

He warns that they should not be given the respect accorded to them by authors like Recanati, who invites us to preserve the naïve intuitions in our theorizing, or like the two authors just quoted in the footnote (Gibbs and Moise 1997, p. 51) who conclude that pragmatics influences “people’s understanding of what speakers say” (Gibbs and Moise 1997). Devitt joins Bach in his criticisms. Here is my contrary proposal: the naïve judge is close to truth. Start with the example: the utterance of “I have three children” guides the hearer in the direction of ‘exactly three’. The ‘exactly three’ reading is (i) Well established in linguistic practice, with both speakers and hearers; this is what we normally say in order to suggest ‘exactly three’, except in non-­ ordinary circumstances (ii) In conformity with the immediate, lexical, meaning of “three” (iii) Least demanding in terms of processing. The lexical meaning suggests the ‘exactly three’ reading, and this guides hearer’s 2nd person and judge’s 3rd person unconscious reasoning. The resulting intuition is to a large extent correct. Ironically, Devitt would also say that the ‘exactly three’ reading gives the correct meaning; I find it completely enigmatic why he refuses to enlist these correct intuitions in his own favor. He is probably being intellectually honest and consistent; since he has been criticizing intuitions on other topics, he refuses to rely on them when they support his own view. What explains the approximate correctness of naïve intuitions? My hunch is that the epistemology here follows metaphysics: if semantic elements determine, or almost determine the total semantic content, then the knowledge about them and determination guides hearer’s intuitions, without forcing them. To return to earlier examples, if a child cries “Mom, it hurts so much!” and the Mom replies “You are not going to die” the observer is led to connect “die” with “…from the hurt”, in a quasi-anaphoric way. This being led then reflects in his intuition: Mom said that the child will not die from the hurt. The ascription of saying this to the Mum is the result of the guidance: the intuition is being guided by the conversational antecedent, which determines the potential cause of dying. Epistemic guidance parallels 6  The participants were explicitly instructed to choose the paraphrase that best reflected what they thought represented the speaker’s “said meaning”, apart from whatever they were shown earlier about what various linguists and philosophers have often claimed about “said” meaning (Gibbs and Moise 1997, p. 64).

The Primacy of Semantics and How to Understand It


quasi-linguistic determination. Consider the parallel with indexicals. “Who spilt the coffee?” “I did.” The observer translates it into the third person on the level of what is said: “Nenad said that he did it.” Determination is paralleled by epistemic guidance; obviously a general and quite rational phenomena on the guidance account.

Conclusion The semanticists are right. The vehicle of content is primarily semantics-connected, and is run by its ‘semantic engine’ so to speak. But what about problematic examples, like I’ve had breakfast.(2) You are not going to die. (3) It’s raining.(4) The table is covered with books. and (5) Everybody went to Paris.? Here I have sketched an intermediate semanticist position that I find preferable: semantic content provides the guidance to truth-conditions (or, on the objective side, a semi-determination of them), the rest is pragmatics. Less intentionalistic than other approaches, since not any “having in mind” counts but rather linguistically-­ conversationally (co)-determined) components. We have called this “the guidance view”; I have used anaphora as the model for guidance. In terms of vehicle metaphor, semantic content provides and directs the main force, it is the powertrain of the vehicle. The rest is pragmatics. The strategy to be followed by guidance-theorist is summarized in the following sketch of an argument (I hope to develop it further in the future): 1 . Anaphora is a semantic phenomenon. 2. Anaphora determines the truth-conditional content and guides the hearer in recognizing this determining. 3. The most problematic cases analyzed by pragmaticists are anaphora-like, only with more guidance and less strict determination. 4. The rest can be dealt with, therefore 5. The guidance view is the right one. To return to our car metaphor, semantics is not mere skeleton: semantic content provides the guidance and the main force. It is a powertrain as opposed to mere lining. The semantic body of the car and the motor dictate questions concerning filling of empty slots, offers linguistic directives and semantic directives from close concepts, helping to produce canonical answers. So we have guided saturation as the main semantic or almost-semantic process in the conversation. The proposal uses anaphora as the model for guidance: it assumes that anaphora is semantic, shows that anaphora guides the hearer in determining the truth-­ conditional content, argues that most problematic cases are anaphora–like (and the rest can be dealt with), and it follows that guidance view is very close to being the right one


N. Miscevic

Finally, on the epistemic side, the “the guidance view” suggests that naïve intuitions about what is said are guided by the same cognitive-linguistic mechanisms that actually provide the semantic content.

References Bach, K. (1994). Conversational Impliciture. Mind & Language, 9(2), 124–162. Bach, K. (2002). Seemingly semantic intuitions. In J. Campbell, M. O’Rourke, & D. Shier (Eds.), Meaning and truth (pp. 21–33). New York, NY: Seven Bridges Press. Bezuidenhout, A. (2004). Procedural meaning and the semantics/pragmatics interface. In C. Bianchi (Ed.), The semantics/pragmatics distinction (pp. 101–131). Stanford, CA: CSLI publications. Cappelen, H., & Hawthorne, J. (2009). Relativism and monadic truth. Oxford, UK: University Press. Carston, R. (2002). Thoughts and utterances: The pragmatics of explicit communication. Oxford, UK: Blackwell Publishing Ltd. Devitt, M. (2013a). Good and bad Bach. Croatian Journal of Philosophy, 13(38), 169–200. Devitt, M. (2013b). Linguistic intuitions are not ‘the voice of competence’. In M.  Haug (Ed.), Philosophical methodology: The armchair or the laboratory? (pp.  268–293). London, UK: Routledge. Devitt, M. (2013c). Three methodological flaws of linguistic pragmatism. In C.  Penco & F.  Domaneschi (Eds.), What have you said? Pragmatics and the debate on the content of thought (pp. 285–300). Stanford, CA: CSLI Publications. Devitt, M. (2013d). What makes a property ‘semantic’? In A. Capone, F. L. Piparo, & M. Carapezza (Eds.), Perspectives on pragmatics and philosophy (pp. 87–112). Cham, Switzerland: Springer. Devitt, M. (2018). Overlooking conventions the trouble with linguistic pragmatism. Gibbs, R., & Moise, J. (1997). Pragmatics in understanding what is said. Cognition, 62(1), 51–74. Perhat, J., & Miscevic, N. (Eds.). (2016). The word that bears a sword. Zagreb: Kruzak. Perry, J., & Blackburn, S. (1986). Thought without representation. Proceedings of the Aristotelian Society, Supplementary Volumes, 60, 137–166. Recanati, F. (1989). The pragmatics of what is said. Mind & Language, 4(4), 295–329. https://doi. org/10.1111/j.1468-0017.1989.tb00258.x. Recanati, F. (2004). Literal meaning. Cambridge, UK: Cambridge University Press. https://doi. org/10.1017/CBO9780511615382. Recanati, F. (2010). Truth-conditional pragmatics. Oxford: Oxford University Press. Searle, J. (1980). The background of meaning. In J. R. Searle, F. Kiefer, & M. Bierwisch (Eds.), Speech act theory and pragmatics (pp. 221–232). Dordrecht, Netherlands: Springer. Sperber, D., & Wilson, D. (1995). Relevance: Communication and cognition. Oxford, UK: Blackwell Publishing Ltd. Stainton, R. (2005). In defense of non-sentential assertion. In Z. Szabo (Ed.), Semantics versus pragmatics (pp. 383–457). Oxford, UK: Oxford University Press. Stainton, R. (Ed.). (2006). Words and thoughts: Subsentences, ellipsis, and the philosophy of language. Oxford, UK: Oxford University Press. Szabó, Z. (Ed.). (2005). Semantics versus pragmatics. Oxford, UK: Oxford University Press. Travis, C. (2006). Thought’s footing: A theme in Wittgenstein’s philosophical investigations. Oxford, UK: Oxford University Press. Wilson, D., & Sperber, D. (2012). Meaning and relevance. Cambridge, UK: Cambridge University Press.

‘Few’, ‘A Few’, ‘Only’: Negative Quantifier Noun Phases and Negative Polarity Items – The Horn-Atlas Debate 1991–2018 Jay David Atlas To Jacob Mey

Historical Introduction In the late 1960s Larry Horn published a youthful and eventually classic paper “A Presuppositional Analysis of only and even” (Horn 1969). At Princeton in 1970 I read the paper with great admiration and interest, but it was only 20 years later that I was free to think more seriously about Focal Particles in English. A University of Groningen, The Netherlands, graduate student in Linguistics Peter I.  Blok had become so interested in Topic Noun Phrases (in Atlas 1988) as a topic for his PhD dissertation (Blok 1993) that one of his teachers Dr Jaap Hoepelman invited me to give a paper at a conference on Topic and Focus at the University of Stuttgart, Germany, in the summer of 1989. That was when I met the University of Groningen semanticists Jaap Hoepelman, Frans Zwarts and Sjaak de Mey and their student Blok. On the last afternoon of the five-day conference, Hans Kamp and I gave the last two lectures. In the Question period, the clever graduate student Peter Blok offered me one of the great moments for a professor: He asked a question, saying that he had read all of Professor Sir Peter Strawson’s published work on the topic that I was discussing, but he had never encountered the Strawson view that I had

J. D. Atlas (*) Pomona College, Claremont, California, CA, USA e-mail: [email protected] © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 F. Macagno, A. Capone (eds.), Inquiries in Philosophical Pragmatics, Perspectives in Pragmatics, Philosophy & Psychology 27,



J. D. Atlas

discussed in my paper: How did I know that this was Strawson’s view? I paused for effect before I answered: “Because I asked him.” When shortly thereafter I sent Larry Horn a copy of the Stuttgart lecture, he emailed me his reply, which in its entirety read “You waited twenty years?”1

“Only Proper Name” and “Few N” Sentences: Entailments, Implicata, and Monotonicity In those early days it was clear that sentences like (1)  a. No one ever doubted that Socrates was wise. would license the occurrence of a Negative Polarity Item in English like ever, which was ungrammatical in non-negative sentences like *Plato ever doubted that Socrates was wise. But why should a sentence with only Socrates, e.g. (1b), (1)  b. Only Socrates ever doubted that Socrates was wise. do so as well? The grammatical occurrence of the English Negative Polarity Item ever in the sentence was to be explained by the semantical properties of a negative element of the meaning of only Socrates (See Ladusaw (1980), Linebarger (1981), Zwarts (1995, 1996)). Geach (1962, 1980) had made the logical negativity of Only Socrates was wise seem plausible by suggesting that logical economy and theoretical smoothness would be served by analyzing (2): (2) Only Socrates was wise. by (3): (3) No one other than Socrates was wise.

1  The Stuttgart lecture appeared as Atlas (1991). There ensued a debate in the semantics and pragmatics literature between Professor Larry Horn and me that has lasted some years. For highlights see Atlas (1996) and Horn (2018). University of Groningen Professor of Dutch Linguistics Jack Hoeksema once reduced a lunch-table of Groningen linguists and graduate students to raucous laughter when he ironically remarked, “There are two necessary conditions on receiving the PhD in Linguistics in Groningen: (1) The student must be able to read an Atlas paper. (2) The student must disagree with all of Atlas’s intuitive linguistic judgments.” So in the presence in Lisbon of so many excellent linguists and philosophers, and notably Professor Larry Horn, I tried first to offer some theoretical arguments and then in the Hoeksema spirit some linguistic data.

‘Few’, ‘A Few’, ‘Only’…


Zwarts’s (1996) De Morgan Taxonomy of Negative Noun Phrases By Frans Zwarts’s (1996) De Morgan taxonomy of negative Noun Phrases (NP), No one other than Socrates meets the first condition for being a Zwarts’s anti-additive quantifier, viz. the first (Ia) of the following De Morgan entailments, analogous to ¬ (A ∨ B) -||- (¬A ∧ ¬B).2 I. Anti-Additive Quantifiers (e.g. No students) a. Downwards Entailing (Downward Monotonicity) NP (VP1 or VP2) ||- NP VP1 AND NP VP2 b. Closure under Finite Disjunctions NP VP1 AND NP VP2 ||- NP (VP1 or VP2) For example, (4) of (Ia): (4) No other than Socrates smokes or drinks ||- No one other than Socrates smokes AND no one other than Socrates drinks. The other two Zwarts anti-multiplicative De Morgan entailments are analogous to ¬ (A ∧ B) -||- (¬A ∨ ¬B), the second of which is Downward Entailing. II. Anti-Multiplicative Quantifiers (e.g. Not all students) a. NP (VP1 and VP2) ||- NP VP1 OR NP VP2 b. Downwards Entailing (Downard Monotonicity) NP VP1 OR NP VP2 ||- NP ( VP1 and VP2 ) The NP No one other than Socrates satisfies the second of the two previous conditions, also a Downwards Entailing condition (IIb), e.g. (5): (5) No one other than Socrates smokes OR no one other than Socrates drinks ||- No one other than Socrates smokes and drinks. Example (4) of (Ia) and Example (5) of (IIb) illustrate that Monotonic Decreasing quantifiers are closed under subsets: necessarily they hold true of the subsets of the extensions (of predicates) of which they hold true. That is, the following should be an entailment: (6) a. NP VP1 ||- NP (VP1 and VP2)   b. No one other than Socrates smokes ||- No one other than Socrates smokes and drinks.

2  ‘||-’ means ‘semantically entails’. ‘-||-’ means ‘logically equivalent’. ‘||-≫’ means ‘conversationally implicates’.


J. D. Atlas

By contrast, Monotonic Increasing quantifiers are closed under supersets: necessarily they hold true of the supersets of the extensions (of predicates) of which they hold true. (7) Upwards Entailing (Monotonic Increasing) Quantifiers (e.g. Socrates, Some students) a. NP (VP1 and VP2) ||- NP VP1 AND NP VP2 b. NP VP1 OR NP VP2 ||- NP (VP1 or VP2 ) For example, a. Socrates (smokes and drinks) ||- Socrates smokes AND Socrates drinks; b. Socrates smokes OR Socrates drinks ||- Socrates (smokes or drinks). The medieval Scholastic logicians noticed that Only Socrates is wise could be paraphrased by, or at least was logically equivalent to, a conjunction of two obvious entailments Socrates is wise and no one other than Socrates is wise. This would be a conjunction of two sentences one of which contains the Upwards Entailing quantifier Socrates and the other of which contains the Downwards Entailing quantifier No one other than Socrates. The sentence Socrates and no one other than Socrates is wise is thus effectively a Non-Monotonic Quantifier sentence. What the Scholastic logicians took for granted was that Socrates and no one other than Socrates is wise was synonymous with Only Socrates is wise. By contrast, Peter Geach’s view was as follows:

“Only Proper Name”: The Standard View (Geach 1962, Pre-2002 Horn) 1. Only Socrates is wise -||- No one other than Socrates is wise This is an explicit way to use the negative quantifier No one other than Socrates to “exclude” those who are not Socrates from those who are wise. 2. Only Socrates is wise ||/- Someone is wise (since No one other than Socrates … does not entail Someone … in Classical Logic) 3. Only Socrates is wise ||-≫ Socrates is wise (i.e. the former implicates, but does not entail, the latter) 4. No interesting semantic relationship is noted between Only Socrates is wise and Socrates is wise (because (3)). 5. Only Socrates is Downwards Monotonic (because (1)) 6. In the Standard theory of Negative Polarity Licensing (Ladusaw 1980; Linebarger 1981; Zwarts 1995), (5) implies that Only Socrates licenses NPIs, e.g. any, ever, in sentences like Only Socrates ever liked Xantippe. By contrast:

‘Few’, ‘A Few’, ‘Only’…


“Only Proper Name”: The Stuttgart View (Atlas 1991, 1996) 1. Only Socrates is wise -||- Exactly one individual, and no one other than Socrates, is wise. This is a uniqueness statement, with an additional subordinate clause employing a Downwards Monotonic negative quantifier allegedly to “exclude” those who are not Socrates from being wise. It is logically equivalent to the Medieval analysis, but its logical syntax is different. 2. Only Socrates is wise ||- Someone is wise (unlike (2) in section ““Only Proper Name”: The Standard View (Geach 1962, Pre-2002 Horn)” above) 3. Only Socrates is wise ||- Socrates is wise (unlike (3) section ““Only Proper Name”: The Standard View (Geach 1962, Pre-2002 Horn)” above) 4. There is a semantic entailment relation between Only Socrates is wise and Socrates is wise (unlike (4) in section ““Only Proper Name”: The Standard View (Geach 1962, Pre-2002 Horn)” above). 5. Only Socrates is neither Upwards nor Downwards Monotonic (unlike (5) in section ““Only Proper Name”: The Standard View (Geach 1962, Pre-2002 Horn)” above). 6. On the Standard Theory of NPI licensing, there is no reason to expect that Only Socrates licenses NPIs. But it does license NPIs! It was obvious to me that a new theory of NPI licensing was needed. In 2002 Horn accepted the Atlas semantics of the Stuttgart View, and then Horn (2002) attempted to construct a pragmatic theory of NPI licensing to meet my observations in section ““Only Proper Name”: The Stuttgart View (Atlas 1991, 1996)”, (6), just above.

The Post-Horn (1996) Pragmatic View 1. Only Socrates is wise asserts No one other than Socrates is wise. 2. Only Socrates is wise ||/- assertorically Someone is wise 3. Only Socrates is wise ||/- assertorically Socrates is wise but: Only Socrates is wise ||- background Socrates is wise, which is “assertorically inert.” Backgrounded, assertorically inert entailments are the new substitutes for conversational implicata: Only Socrates is wise ||-≫ Socrates is wise 4. No interesting assertoric relationship is noted between Only Socrates is wise and Socrates is wise. 5. Only Socrates is pragmatically (assertorically) Downwards Monotonic (because (1)). 6. On Horn’s pragmatic theory of NPI licensing, Only Socrates licenses NPIs (because (5)).


J. D. Atlas

For a long time “Only Proper Name” quantifiers were thought to be exceptional cases, until I began to look more carefully at other supposedly Downwards Monotonic quantifiers, one of which Zwarts had even used as a classic example of a Downwards Monotonic quantifier, viz. Few N. I too had thought of it as Downwards Monotonic, till I examined the data more carefully. The Standard View of Few N runs as follows:

 he Standard View of Few N: Chomsky (1972), Horn (1989, T pp. 244–250), Zwarts (1996) 1. Few N are F -||- Not many N are F; and more strongly, Few N are F is synonymous with Not many N are F. Few men are philosophers -||- Not many men are philosophers Few men are philosophers is synonymous with Not many men are philosophers. 2. Few N are F ||/- Some N are F Few men are philosophers ||/- Some men are philosophers 3. A Horn Scale and its usual implicata: Some N … ||-≫ Not many N …, would also predict from the corresponding Atlas and Levinson (1981) Negative Scale the converse implicatum Not many N … ||-≫ Some N (viz. not none).3 In that case, on the basis of the synonymy claim in (1), we would have: Few N … ||-≫ Some N …. Few men are philosophers ||-≫ Some men are philosophers. But on the grounds of the symmetry of the Horn and Atlas-Levinson implicata of Some and Not all, or Some and Not many, we would expect by virtue of (1) the implicatum Some N … ||-≫ Few N. E.g. Some men are philosophers ||-≫ Few men are philosophers. But that is an incorrect Generalized Conversational Implicature. So either the well-grounded symmetry of these Horn and Atlas-Levinson Scalar implicata is undermined, or Few N is not synonymous with Not many N. The latter conclusion seems the more plausible. 4. The obvious semantic relationship between the related lexical items Few N and A few N in English is unexplained by the pair Not many N and A few N. 5. Few N is Downwards Monotonic (because (1)) 6. On the Standard theory of NPI licensing, Few N licenses NPIs (because (5)) By contrast:

 See S. Levinson (1983, pp. 123, 134) and J. D. Atlas and S. C. Levinson (1981, p. 39, n. 9).


‘Few’, ‘A Few’, ‘Only’…


The Atlas View of Few N4 1. a. Few N … ||- Not many N … b. Not many N … ||/- Few N… ; c. Not many N … ||-≫scalar A few N … Few men are philosophers ||- Not many men are philosophers Not many men are philosophers ||/- Few men are philosophers Not many men are philosophers ||-≫scalar A few men are philosophers 2. Few N … ||- Some N … Few men are philosophers ||- Some men are philosophers 3. Few N … ||/-≫ Some N …, though Not many N … ||-≫scalar Some N … . Few men are philosophers ||/-≫ Some men are philosophers though Not many men are philosophers ||-≫scalar Some men are philosophers “Few and in fact none …” is not a cancelation of an implicatum of ‘Few’; it is a literal contradiction and so a correction of asserting ‘Few’. 4. There is in fact an interesting semantic relationship in English between Few N and A few N. a. Few N … ||- A few N … Few men are philosophers ||- A few men are philosophers b. A few N … ||/- Few N … A few men are philosophers ||/- Few men are philosophers c. A few N … ||-≫ scalar Not many N … A few men are philosophers ||-≫ scalar Not many men are philosophers d. Few N … -||- Only a few N…, and more strongly, Few N are F is synonymous with Only a few N are F. Few men are philosophers -||- Only a few men are philosophers Few men are philosophers is synonymous in English with Only a few men are philosophers.5 5. Few N is neither Upwards nor Downwards Monotonic; it is Non-Monotonic, like Only Socrates.

 For antecedents of the current view, see Atlas (2013, 2015, 2017, 2018).  Professor Larry Horn is quite doubtful of the correctness of this synonymy claim, so caveat emptor.

4 5


J. D. Atlas

6. Few N licenses NPIs like any, ever,… with the exception of Bolinger’s (1972) minimizers in English, e.g. give a tinker’s dam, give a red cent.6 The licensing is not predicted by the Standard theory of NPI licensing, since by (5) Few N is Non-Monotonic.

Assertion and the Division of Semantic Labor Suppose that Horn were to accept my analysis of Few N sentences. To do so he would have to abandon the use of Not many N in his analysis. The licensing of some types of NPIs is then predictable from a Horn Pragmatic theory if Horn is prepared to argue that what is asserted in stating Few boys are drunk is None but a few boys are drunk and that A few boys are drunk is not assertorically entailed but is entailed only in the background. But what he could not then predict is the inability of Few boys are drunk to permit the felicitous occurrence of Bolinger’s minimizers, as I have noted of Few N sentences in Atlas (2018). But that is a minor blemish in an otherwise impressive Pragmatic Analysis by Horn of the grammatical occurrence of NPIs. What is now at stake in Horn’s Pragmatic Theory of NPI Licensing is the division of semantic labor between the assertoric and the backgrounded components of his analysis, analogous to the division between the entailed and the implicated components in his earlier theory. What I now want to address is this Dualism of the Assertoric and the Backgrounded. To do that, I would have to put words into Horn’s mouth that he has not uttered and possibly would never utter, but Peter Geach, obligingly, has uttered them.

 eter Geach’s Account of “Only Proper Name” Sentences: P Excluding Force In 1962 Peter Geach (1962, 1980) had introduced a subtlety into his account of Only Socrates is wise. Geach’s analysis of Only Socrates is wise ignored the positive, affirmative component of the sentence-meaning, viz. Socrates is wise. Geach (1980, p. 208) says of the Medieval analysis Socrates is wise and no one other than Socrates is wise: It is formally much more convenient to treat the exclusive proposition as having precisely the exclusive force of its supposed second conjunct and not to read as implying < Fa>. ...The exclusive force that distinguishes this class of propositions belongs entirely to the second conjunct, e.g. to ‘no one other than Socrates is wise’ of the exponible ‘Socrates is wise & no one other than Socrates is wise’ of ‘Only Socrates is wise.’ 6  The point about miminizers is made in Atlas (2018), an earlier and longer version on which this essay was based.

‘Few’, ‘A Few’, ‘Only’…


I have never doubted Geach’s view that his suggestion is formally much more convenient in attributing the excluding Fregean “force” of the utterance of Only Socrates is wise to the excluding force of the second conjunct No one other than Socrates is wise of the medieval paraphrase if the second conjunct were asserted alone. It is much more formally clean to take the excluding force of the assertion Only Socrates is wise to come packaged in the exponible’s second, Downwards Entailing conjunct rather than being smeared throughout the original ‘only’ sentence ‘Only Socrates is wise’. This would be especially tempting if one thought, incorrectly as many did and do, that the Fregean assertoric force of an asserted conjunction │─ (A & B) could be represented by the concatenation (represented by ‘^’) of two Fregean assertions “│─ A,” “│─ B,” by “and” thus: “│─ A ^ & ^ │─ B,” roughly taking “I assert that A and I assert that B” to be the same as “I assert that (A and B).”7 One would find it formally much pleasanter to think of the Excluding Force of the utterance “Only Socrates is wise” as packed into one nice propositional package after contemplating the force-less Fregean bivalent truth-function “the Horizontal”, e.g. ─ (Socrates is wise), like this: “─ (Socrates is wise) ^ and ^ │─ (No one other than Socrates is wise).”8 Roughly, assuming ‘Socrates is wise’ is true, this amounts to the ungrammatical *“ ‘Socrates is wise’ being true, and I assert that no one other than Socrates is wise.” It is particularly odd that Peter Geach should have continued to take this view in the 1980 edition of Reference and Generality, as in his 1963 Howison Lecture “Assertion” in the University of California, Berkeley, he points out that Frege explicitly rejected the idea that in asserting a conditional

the speaker is asserting either the antecedent P or the consequent Q. But Geach (1965, 1972) notably fails to consider the analogous question for the assertion of a conjunctive sentence whose truth-value is determined by the truth-function represented by ‘&’. By contrast with his Fregean view about conditionals he does not bother to notice that for the conjunction


does not require, or in my view even permit, that the speaker make two assertions: asserting

and asserting . He does criticize P.F. Strawson for the doctrine in Individuals (1959) that predicating consists in asserting a predicate of something. Geach (1972, p. 266) says, “Accepting the Frege point, we know that no term of any proposition gives the proposition assertoric force; for the same term might occur without any change of sense in an unasserted occurrence of the proposition.” If this is true for an n-place predicate, n = 1, 2, …, it is just as true for an n = 0 predicate, viz. for a closed sentence. So just how for Geach is the truth-function “Conjunction” of an unasserted Noun clause and an asserted sentential clause, < that P & │─ Q,> going to give the whole sentence

assertoric force? It cannot.

 See Stalnaker (1974, p. 11) for the error, and the criticism in Atlas (2011, p. 50, n. 19).  The Frege assertion marker ‘│─’ is a sequence of two Frege symbols, the truth-value function symbol ‘─’ for a function that takes an expression and assigns it the truth-value T if the expression is a sentence and is true, and assigns any other expression the truth-value F, and the judgment symbol ‘│’ for making a judgment of truth. 7 8


J. D. Atlas

So Geach must admit that though the exponible ‘Socrates is wise and no one other than Socrates is wise’ means ‘Only Socrates is wise,’ the assertion of ‘Only Socrates is wise’ cannot be the same as the assertion of ‘Socrates is wise and no one other than Socrates is wise,’ especially if one thought the assertion Socrates is wise and no one other than Socrates is wise required the assertion of the two clauses concatenated with ‘and.’9 So on Geach’s account whenever a speaker thinks that he or she has asserted the propositional content of ‘Only Socrates is wise,’ he or she has not done so; one has asserted only ‘No one other than Socrates is wise’. In fact in the history of English, on Geach’s view, no one has ever asserted the propositional content of ‘Only Socrates is wise.’ Not in those words, anyway! It may be formally more convenient to attach the excluding force of asserting ‘Only Socrates is wise’ only to the second conjunct ‘No one other than Socrates is wise,’ but if one does, utterances of the literal meaning of ‘Only Socrates is wise’ have no assertoric force at all. That is clearly a reductio ad absurdum of Geach’s view.

What Are Geach’s Options? One way out for Geach is to claim that ‘Only Socrates is wise,’ if it is to be assertible with excluding force, just means, i.e. is just synonymous with, the second conjunct ‘No one other than Socrates is wise.’ It does not mean, is not synonymous with, is not even logically equivalent to, the conjunction ‘Socrates is wise and no one other than Socrates is wise.’ And so, in that view, the sentence ‘Only Socrates is wise’ cannot as a matter of meaning entail that Socrates is wise. For reasons that have been explored in detail in more than a decade of linguistic discussion, especially by Horn (2018) and myself, ‘Only Socrates is wise’ does analytically entail that Socrates is wise. So something else must be said. One option is to deny Geach’s Frege point: to the contrary, the assertion of any constituent of a sentence can make the whole sentence an assertion, bringing along the unasserted constituents and transforming them into asserted constituents. If so, everything is asserted. One has to drop the formulation of the debate in terms of the asserted/ non-asserted constituents. That means dropping the formulation of the debate in terms of what is asserted, e.g. No one other than Socrates is wise, versus what is only entailed, “inertly” entailed, presupposed, implicated, presumed, or not-­ at-­issue, e.g. Socrates is wise. Since 1969 Larry Horn has explored these theoretical options in great depth and with great perspicacity. We are not about to give up the distinction between the asserted and the presupposed (however that notion is glossed). Another option is to abandon Geach’s notion of a constituent carrying an excluding or any other kind of force, and to abandon Geach’s spurious distinction between a predicate and a sentence, viz. hating predicates that are said to be assertoric but

 Many philosophers have taken that view. I do not. See Atlas (2011).


‘Few’, ‘A Few’, ‘Only’…


loving sentential clauses that are said to be assertoric. If an utterance of ‘Only Socrates is wise’ has excluding force, What is its excluding force? and Why does an assertion of ‘Only Socrates is wise’ have it?

The Atlas Dis-Solution of the ‘Only’ Problem Unlike Larry Horn, and unlike Peter Geach and the rest of the Medieval monks, I do not think that assertions of ‘Only Socrates is wise’ have excluding force. In my view the considerations just discussed show that the category of “excluding” expressions is mistakenly applied to ‘Only Socrates’ sentences. If one took the original semantic intuition of exclusion seriously, the underlying reason for it must be that excluding something is best constituted by negating something, or complementing something, in the sense of distinguishing a set from its complement set. What if the whole semantic issue of exclusion was a mistake from the very start? I suggest that describing ‘Only Socrates’ as an “excluding” expression and hence in some fashion a “negating” expression was a mistake. In fact, I believe that the same mistake is made about . Moreover, Bertrand Russell’s Theory Definite Descriptions for sentences like The king of France is bald is really a theory for The only king of France is bald.10

 The first version of the essay on which this lecture is based was presented in the Semantics Workshop, Department of Linguistics, New York University, New York, New York, 13 October 2013. I am indebted to Stephanie Harves, Chris Barker, Larry Horn, and the participants in the Workshop. A second version was delivered in a Colloquium in the Department of Dutch Linguistics, University of Groningen, The Netherlands, 9 November 2015. I am indebted to Professor Jack Hoeksema, Rector Magnificus Emeritus Dr. Frans Zwarts of the University of Groningen, and to Dr. S. C. Levinson, Co-Director of the Max Planck Institute for Psycholinguistics, Nijmegen, The Netherlands. A new version was delivered as one of two lectures that I gave at “Radical Pragmatics 40 Years On: Workshop with Professor Jay Atlas” sponsored by Newnham College, Cambridge, and The Department of Theoretical and Applied Linguistics, Cambridge University, 17 March 2017, Cambridge, England, and organized by Kasia Jaszczolt, Professor of Linguistics and Philosophy of Language. I am indebted to Professor Kasia Jaszczolt, Dr. Roberto Soleo, Dr. M.M. Leezenberg, Professor Ruth Kempson, Dr. David Good, and the other participants in the Workshop. My thanks to all for their kind hospitality and for their invaluable comments and engaging conversation. This version was adapted for an oral presentation at Pragmasophia II from the longer version Atlas (2018). I am indebted to Professor Larry Horn for critical comments on Atlas (2018) and for his remarkable contributions to linguistic theory, as well as his friendship, personal and professional; to Professors Kasia Jaszczolt, Paul Saka, Daniel Hoek, Yan Huang, Douglas Walton, Fabrizio Macagno, Alessandro Capone, Kenneth Taylor, Robin Jeshion, Michael Nelson, Sarah Blackwell, Julia Kolkmann, and Jacob Mey for their conversation and comments at Pragmasophia II: Second International Conference on Pragmatics and Philosophy, University of Lisbon, Portugal, 18–21 September, 2018. This essay is dedicated to Professor Jacob Mey.



J. D. Atlas

References Atlas, J.  D. (1988). What are negative existence statements about? Linguistics and Philosophy, 11(4), 373–394. Atlas, J. D. (1991). Topic/comment, presupposition, logical form and focus stress implicatures: The case of focal particles ‘only’ and ‘also’. Journal of Semantics, 8(1–2), 127–147. https:// Atlas, J. D. (1996). ‘Only’ noun phrases, pseudo-negative generalized quantifiers, negative polarity items, and monotonicity. Journal of Semantics, 13(4), 265–332. jos/13.4.265. Atlas, J. D. (2011). Expressing regret and avowing belief: Sadock’s expositive adverbials, Moore’s paradox, and performative and quasi-performative verbs. In E. Yuasa, T. Bagchi, & K. Beals (Eds.), Pragmatics and autolexical grammar: In honor of Jerry Sadock (Linguistics today) (Vol. 176, pp. 35–58). Amsterdam, Netherlands: John Benjamins Publishing. Atlas, J.  D. (2013). ‘Few’, ‘A Few’, and ‘Only’ noun phrases, non-monotonic quantifiers, and negative polarity items. Lecture. Semantics Workshop, Department of Linguistics, New York University, 18 October 2013. Atlas, J. D. (2015). ‘Few’, ‘A Few’, and ‘Only’ noun phrases, non-monotonic quantifiers, and negative polarity items. Colloquium. Department of Dutch Linguistics, University of Groningen, The Netherlands, 9 November 2015. Atlas, J. D. (2017). ‘Few’, ‘A Few’, and ‘Only’ at the semantics/pragmatics boundary. Lecture. Radical Pragmatics Forty Years On: Workshop with Professor Jay Atlas. Department of Theoretical and Applied Linguistics, University of Cambridge, UK and Newnham College, Cambridge, UK, 17 March 2017. Atlas, J. D. (2018). ‘Few’, ‘Hardly Any’, and ‘Only’ at the semantics/pragmatics boundary: negative quantifier noun phrases and negative polarity items – conjectures and refutations in the Horn-Atlas Debate (1991–2018). MS. V. 3.0. 6 September 2018. Department of Philosophy, Trinity University, San Antonio, Texas, USA. September 2018. Atlas, J. D., & Levinson, S. C. (1981). It-clefts, informativeness, and logical form: An introduction to radically radical pragmatics (revised standard version). In P. Cole (Ed.), Radical pragmatics (pp. 1–61). New York, NY: Academic Press. Blok, P. I. (1993). The interpretation of focus: An epistemic approach to pragmatics. Groningen dissertations in linguistics 6. Groningen: Rijksuniversiteit te Groningen. Bolinger, D. (1972). Degree words. The Hague, Netherlands: Mouton. Chomsky, N. (1972). Deep structure, surface structure, and semantic intepretation. In Studies on semantics in generative grammar (pp. 62–119). The Hague, Netherlands: Mouton. Geach, P.  T. (1962). Reference and generality. An examination of some medieval and modern theories. Ithaca, NY: Cornell University Press. Geach, P.  T. (1965). Assertion. The Philosophical Review, 74(4), 449–465. https://doi. org/10.2307/2183123. Geach, P. T. (1972). Assertion. In P. T. Geach (Ed.), Logic matters (pp. 254–269). Berkeley, CA: University of California Press. Geach, P.  T. (1980). Reference and generality. An examination of some medieval and modern theories (3rd ed.). Ithaca, NY: Cornell University Press. Horn, L. (1969). A presuppositional analysis of ‘only’ and ‘even’. In Proceedings of the 5th annual meeting of the Chicago linguistic society(pp. 97–108). Chicago, IL: Chicago Linguistic Society. Horn, L. (1989). A natural history of negation. Chicago, IL: University of Chicago Press. Horn, L. (1996). Exclusive company: ‘Only’ and the dynamics of vertical inference. Journal of Semantics, 13, 1–40., Horn, L. (2002). Assertoric inertia and NPI licensing. In Proceedings of the 38th Annual Meeting of the Chicago Linguistic Society Part 2 (pp. 55–82). Chicago, IL: Chicago Linguistic Society.

‘Few’, ‘A Few’, ‘Only’…


Horn, L. (2018). An (Abridged) atlas of negation: polar landscape in an era of climate change. In K. Turner & L. Horn (Eds.), Pragmatics, truth, and underspecification: towards an atlas of meaning (Current research in the semantics pragmatics interface) (Vol. 34, pp. 3–53). Leiden, Netherlands/Boston, MA: Brill Publishing Company. Ladusaw, W. (1980). Polar sensitivity as inherent scope relations. New  York, NY: Garland Publishing. Levinson, S. C. (1983). Pragmatics. Cambridge, UK: Cambridge University Press. Linebarger, M. C. (1981). The grammar of negative polarity. Bloomington, IN: Indiana University Linguistics Club. Stalnaker, R. (1974). Pragmatic Presuppositions. In M. Munitz & P. Unger (Eds.), Semantics and philosophy (pp. 197–213). New York, NY: New York University Press. Strawson, P. F. (1959). Individuals. London, UK: Methuen. Zwarts, F. (1995). Nonveridical contexts. Linguistic Analysis, 25(3–4), 286–312. Zwarts, F. (1996). A hierarchy of negative expressions. In H. Wansing (Ed.), Negation: A notion in focus (Perspectives in analytical philosophy 7) (pp. 169–194). Berlin, Germany/New York, NY: Mouton de Gruyter.

Speech-Act-Theoretic Explanations of Problems of Pure Indexicals Etsuko Oishi

Problematic Cases of the Personal Pronouns I and You Personal pronouns are indexicals along with demonstrative pronouns such as this and that, adverbs such as here, now and tomorrow, and adjectives such as actual and present (Kaplan 1989, p.  489). Referents of these indexicals change from one speech situation to another. Among these indexicals, Kaplan identifies two types, true demonstratives (he, she, that, etc.) and pure indexicals (I, here, now, etc.). A referent of true demonstratives is fixed by the speaker’s demonstrations or intentions, whereas a referent of pure indexicals does not depend on such demonstrations or intentions. For example, the character of I determines the content of each of its tokens to be the speaker producing it, where a contextual factor is sufficient to determine the content. Kaplan’s description applies to straightforward cases, but there are some puzzling uses of I. The script of the American film Sommersby1 provides an example. Sommersby is based on the French film The Return of Martin Guerre, which in turn is based on the true story of Martin Guerre. Sommersby is set in the Reconstruction period following the U.S. Civil War. In the film Jack is presumed dead after he fails to return from the war. His wife, Laurel, is set to marry her neighbour, but he mysteriously returns after six years. Jack returns with a change of heart. He is now kind and loving to Laurel and their young son. After a while, two

 It is a 1993 American romantic period drama film directed by Jon Amiel.


E. Oishi (*) Department of Liberal Arts, Faculty of Science, Tokyo University of Science, 1-3 Kagurazaka, Shinjuku-ku, Tokyo, Japan e-mail: [email protected] © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 F. Macagno, A. Capone (eds.), Inquiries in Philosophical Pragmatics, Perspectives in Pragmatics, Philosophy & Psychology 27,



E. Oishi

U.S. Marshals arrest Jack on the charge of murder, which carries the death penalty. Laurel’s attempts to save her husband focus on the question of his identity: whether or not this ‘Jack’ is who he claims to be. Laurel and Jack’s lawyer agree to argue that her husband is an impostor. This would save him from hanging for murder, but he would still be imprisoned for fraud and military desertion. One witness testifies that he is Horace Townsend, an English teacher and con artist from Virginia. Jack fires the lawyer and sets about re-establishing himself as the real Sommersby. When Laurel is called as a witness and asked how she knows that Jack is not the real Sommersby, she answers ‘… because I never loved him the way I love you!’ Then the following exchange begins: (1) Jack: So tell me, Laurel, from your heart … … Am I your husband? Laurel: (a long pause) Yes, you are. Jack: Thank you. The question is what is the content of the utterance ‘Am I your husband?’, to which Laurel answers ‘Yes, you are.’ If Jack’s utterance is the question asking whether or not Jack Sommersby is Laurel’s husband, Laurel would not say ‘Yes, you are.’ If it is the question asking whether or not Horace Townsend is Laurel’s husband, and Laurel affirms that he is, Jack would not say ‘Thank you.’ There must be the content that is asked by Jack and answered affirmatively by Laurel, which Jack appreciates. It is not obtained by replacing ‘I’ and ‘you’ with either ‘Jack Sommersby’ or ‘Horace Townsend.’ This is a unique problem because it is not the issue of context-specificity: the referent of an indexical changes from one speech situation (i.e. context) to another. The referent of ‘I’/‘you’ in these utterances does not change from, say, Jack Sommersby to Horace Townsend, or to any third person. This is not the case of metonymies, such as ‘The ham sandwich wants more coffee’, or other figurative speeches, either. There is no figurativeness involved when ‘I’ in ‘Am I your husband?’ refers to the speaker, and ‘your husband’ refers to the husband of the hearer. There is something unique about the use of these personal pronouns. There are other problematic cases of I. Sidelle (1991) refers to one case as the Answering Machine Paradox. The character of each of I, here and now is a function from a contextual parameter to the referent of the expression, and the tokens ‘I’, ‘here’ and ‘now’ refer, respectively, to the utterer, the location of utterance and the time of utterance. The sentence ‘I am not here now’ is uttered truly if and only if the utterer is not at the location of utterance at the time of utterance, and, because of this, the sentence ‘I am not here now’ may never be uttered truly; an utterer is always at the location of utterance at the time of utterance. This conclusion is at odds with the intuition that there are true instances of the sentence ‘I am not here now’ when it is uttered on a telephone answering machine—hence the Answering Machine Paradox—and written on a post-it note and stuck on an office door when its inhabitant is not in residence.

Speech-Act-Theoretic Explanations of Problems of Pure Indexicals


Corazza et al. (2002) give another example. Joe is not in his office one  day and Ben notices that a number of students keep approaching his door and knocking. They then stand around and look bemused for a while before leaving. Taking pity on these poor students wasting their time, Ben decides to attach his ‘I am not here today’ note to Joe’s door. The trick works; the students, instead of knocking and waiting, take one look at the note and then leave. What does ‘I’ refer to? Corazza et al. claim, at the moment a student looks at the note, it would be strange to deny that it refers to Joe. Sidelle proposes to solve the paradox by proposing that an utterance can be deferred: an utterance takes place when the message is heard by the listener (the decoding time) not when it is recorded (the encoding time). If the individual who recorded the message (the referent of ‘I’) is not at the place where the machine is located (the referent of ‘here’) at the time the call is made (the referent of ‘now’), then that utterance of ‘I’m not here now’ can express a truth. Corazza et al. propose that, for any use of the personal indexical, the contextual parameter of the agent is conventionally given. When Ben sticks his own post-it note saying ‘I am not here today’ to Joe’s office door, Ben is exploiting the convention that ‘I’ on a notice on someone’s office door refers to the office’s usual incumbent. These proposals, that is, deferring an utterance and relying on social conventions, however, do not help to explain our example: interpreting the time of the utterance as any later times does not solve the problem, and there are not any social conventions that specify the referents of ‘I’ in ‘Am I your husband?’ and ‘you’ in ‘Yes, you are’. Furthermore, the problem with Kaplan’s character-token model of indexicals extends to the second-person pronoun you. A conversation in the British family film Paddington2 illustrates this point. A young bear, Paddington, reaches London’s Paddington Station, where he meets the Browns, who take him home temporarily. The following conversation occurs just after they arrive at the Browns’ house: (2) Mrs. Brown: You see, when a young person comes to this country, I’m afraid they don’t just move in with the first people they meet. Paddington: No? Mrs. Brown: You need a proper guardian. Paddington: What’s that? Mrs. Brown: A grown-up who takes you into their home and looks after you. Paddington: Like you? Mrs. Brown: Erm … Yes, well, I suppose so. Paddington: But not you?

2  It is a 2014 live-action animated comedy film written and directed by Paul King. The film is based on the stories of the character Paddington Bear created by Michael Bond.


Mr. Brown:

E. Oishi

No. We don’t do that.

According to Kaplan’s model, the character you determines the content of each of its tokens to be the hearer, so the referent of ‘you’ in ‘Like you?’ and the referent of ‘you’ in ‘But not you?’ are one and the same person, Mrs. Brown. If so, the conversation does not make sense: Mrs. Brown is and is not a guardian. It seems that Kaplan’s model does not give us the whole story of the use of the personal pronouns I and you: something much more complex is going on when the speaker refers to her/himself by the pronoun I and refers to the hearer by the personal pronoun you. The following section, section “The Analyses Given by Goffman (1981) and Austin ([1962]1975)”, discusses the speaker’s and the hearer’s discursive roles, in which Goffman’s (1981) distinction among an animator, an author and a principal, as well as Austin’s ([1962]1975) felicity conditions, is utilized. On the basis of the proposed idea in section “The Analyses Given by Goffman (1981) and Austin ([1962]1975)”, the problematic uses of the personal pronouns I and you are re-interpreted in section “Analysing the Problematic Cases”. This is followed by concluding remarks.

 he Analyses Given by Goffman (1981) and Austin T ([1962]1975) Goffman (1981, p. 144) differentiates three functions of the speaker: an animator, an author and a principal. An animator is ‘a body engaged in acoustic activity’, an author is ‘someone who has selected the sentiments that are being expressed and the words in which they are encoded’, and a principal is ‘someone whose position is established by the words that are spoken, someone whose beliefs have been told, and someone who is committed to what the words say’. The distinction might not be recognized when the speaker is an animator, an author and a principal at the same time. However, it becomes evident when these functions are performed by different people. When a spokesperson reads the Prime Minister’s announcement, the spokesperson is an animator, the writer of the announcement is an author, and the Prime Minister is a principal. It is then reasonable to assume that, when only one person is involved in making an utterance, s/he is performing all of these functions. In a face-to-face communication, the animator function can be easily recognized: the speaker engages in acoustic activity. S/he might do this, for example, enthusiastically, happily, or reluctantly, with an American accent, without pause and so on. The distinction between an author and a principal is more subtle. An author selects what sentiments to be expressed by what words so that her/his position as that of a principal is to be established, her/his beliefs as those of the principal are to be expressed, and s/he, as the principal, is to be committed to what the words say. Austin ([1962]1975, pp. 14–15) also analyses complexity of the speaker as the performer of a speech act. In his felicity conditions in the following, Austin

Speech-Act-Theoretic Explanations of Problems of Pure Indexicals


distinguishes a certain person in certain circumstances who utters certain words to produce a certain conventional effect (in [A.1]) from a particular person in a given case (in [A.2])3: (A.1) There must exist an accepted conventional procedure having a certain conventional effect, that procedure to include the uttering of certain words by certain persons in certain circumstances, and further, (A.2) the particular persons and circumstances in a given case must be appropriate for the invocation of the particular procedure invoked. (B.1) The procedure must be executed by all participants both correctly and (B.2) completely. (Γ.1) Where, as often, the procedure is designed for use by persons having certain thoughts or feelings, or for the inauguration of certain consequential conduct on the part of any participant, then a person participating in and so invoking the procedure must in fact have those thoughts or feelings, and the participants must intend so to conduct themselves, and further (Γ.2) must actually so conduct themselves subsequently. Doesn’t Austin provide the felicity conditions to characterize only performatives? They are first introduced to specify the conditions for performatives to produce their conventional effects. Austin, however, later abolishes the performative-constative distinction, and classifies illocutionary acts in general according to the types of effect that they bring about: verdictives, exercitives, commissives, behabitives and expositives (Austin [1962]1975, pp.  148–164). The felicity conditions are, therefore, reasonably assumed to be general conditions under which illocutionary acts are happily performed and their conventional effects are brought about.4 For a terminological convenience, a certain person in certain circumstances who utters certain words to produce a certain conventional effect (in [A.1]) is referred to as the addresser of an illocutionary act, while a particular person in a given case (in [A.2]) is simply referred to as a speaker.5 A speaker in a given case, who is appropriate for the invocation of the procedure of bringing about a particular illocutionary effect, invokes the procedure as the addresser of the illocutionary act. As the felicity conditions [Γ.1] and [Γ.2] illustrate, the speaker expresses the thoughts or feelings as those of the addresser of the act, and intends to inaugurate, and subsequently inaugurates, consequential conduct as the addresser of the act. Goffman’s distinction between an author and a principal seems to correspond to Austin’s distinction between a speaker and an addresser, while Goffman’s animator  Searle (1969, pp.  62–71) reanalyzes Austin’s felicity conditions and categorizes them as the preparatory rule and the sincerity rule, which are, along with the propositional content rule and the essential rule, interpreted as rules for the use of the illocutionary force indicating devices. In Searle (1979), the essential rule is replaced by illocutionary points. Searle (1969, 1979) does not discuss complexities concerning the speaker. 4  See Sbisà (2002, 2018) for analyses of Austin’s felicity conditions. 5  Sbisà’s (1984) distinction between actors and actants is parallel to the speaker-addresser distinction. 3


E. Oishi

corresponds to a performer of a locutionary act (Austin [1962]1975, p. 94). Goffman models the speaker’s functions of an author and a principal to describe the speaker’s expressive function, while Austin describes how conventional effects are brought about by a speaker’s utterance. Goffman and Austin both postulate entities which appear/function in communication/discourse, principals and addressers, who are distinguished from entities in the world, authors and speakers. There are parallel complexities involving the hearer. When an illocutionary act is performed, a particular person to whom a speaker directs the utterance is a hearer, who is an appropriate addressee at whom the invocation of the procedure of bringing about a particular illocutionary effect is targeted. The hearer as the addressee of the illocutionary act secures the production of the illocutionary effect by her/his uptake (Austin [1962]1975, pp. 116–117),6 and understands the thoughts and feelings and the intention to inaugurate consequential conduct that the speaker expresses as the addresser. A person who listens to the speaker’s locutionary act can be referred to as the listener of the locutionary act. The following table illustrates correspondences among (i) the distinctive functions of the speaker described by Goffman, (ii) the distinctive aspects of the performer of a speech act described by Austin and (iii) the distinctive aspects of the receiver of a speech act (Table 1): The study concerning the distinction between a principal/addresser and an author/speaker, and interplays between them, is largely neglected, with some exceptions of discourse-anchored studies such as Bull and Fetzer (2006), in which the distinction among an animator, an author and a principal is utilized. Characteristics of a principal/addresser are described as those of an author/speaker, or the other way round. Traces which indicate the distinction are treated as special uses of linguistic expressions. For example, the function of discourse marker well is described as indicators of (i) divergence (Jucker 1993; Schiffrin 1987), (ii) acceptance (Blakemore 2002; Bolinger 1989; Carlson 1984) and (iii) ‘gestural’ interjection (Schourup 2001). Aijmer (2013) provides a corpus-based analysis and describes the functions of well as (i) coherence, (ii) involvement and (iii) politeness. Norrick (2001) describes the staging function of well in oral narratives. Oishi (2020a), on the other hand, claims that by using well, a speaker indicates a gap between the speaker as the addresser/ addressee of the preceding illocutionary act and the speaker as the addresser of the present illocutionary act, while connecting the addresser (principal) with the speaker (author) and describing interplays between them.

Table 1  Correspondences between Goffman’s terminology and Austin’s Goffman (1981) An animator An author A principal

Austin ([1962]1975) The performer of a locutionary act A speaker An addresser

The listener of a locutionary act A hearer An addressee

 Sbisà (1984) describes the importance of the hearer’s uptake in Austin’s speech act theory.


Speech-Act-Theoretic Explanations of Problems of Pure Indexicals


Some studies of de se attitudes grasp interplays between a principal/addresser and an author/speaker. De se attitudes are self-knowledge, knowledge about oneself ‘when one thinks of oneself in the first-person way’ (Ninan 2010, p. 551), and have attracted attention from philosophers and linguists (Boër and Lycan 1980; Castañeda 1966; Chierchia 1989; Higginbotham 2009; Lewis 1979; Perry 1977, 1979; Stalnaker 1981), and have recently reemerged as an issue (Cappelen and Dever 2013; Feit and Capone 2013; García-Carpintero and Torr 2016; Huang and Jaszczolt 2018).7 To have de se attitudes, the speaker has to have the external perspective which grasps her/him as ‘her/himself’, which Perry (1979) analyzes by using the concept of limited accessibility.8 De se attitudes can be described by making use of Goffman’s distinction: an author has the external view of the principal who sees her/him as ‘her/himself’ and describes ‘her/himself’, by which her/his position is established as that of the principal by the words spoken, her/his beliefs are told as those of the principal, and s/he is committed to what the words say as the principal. They can be also illustrated by adopting Austin’s distinction: a speaker has an external view as that of the addresser of an illocutionary act, in which the speaker sees her/him (from the addresser’s perspective) as ‘her/himself’ (from the speaker’s perspective), and brings about the illocutionary effect as the addresser. The speaker expresses the thoughts or feelings about her/himself as those of the addresser of the act, and intends to inaugurate, and subsequently inaugurates, consequential conduct as the addresser of the act.9 In the connection with the arguments on de se attitudes, Recanati (2016, p. 142) tries to characterize indexical thoughts, which he describes as ‘thoughts whose (absolute) truth-conditions are not fixed once and for all but depends on the context’ (the emphasis is Recanati’s.) Indexical thoughts seem to be thoughts of a principal/ addresser, which are distinct from those of an author/speaker. Indexical thoughts are the thoughts expressed by words used, as Goffman specifies, and, therefore, without being connected with the sentiments that an author expresses by them, they are not specific enough10: they depend on sentiments an author expresses by using the words. Indexical thoughts can be described as the thoughts of the addresser who 7  The recent debate was triggered by Cappelen and Dever (2013, pp. 3, 50), who assert ‘… there is no such thing as essential indexicality, irreducibly de se attitudes, or self-locating attitudes’, and different actions, such as Perry’s ceasing to follow the trail and rearranging the torn sack, and the professor’s moving to the room in which the department meeting is being held (Perry 1979), do not guarantee distinct beliefs. According to them ‘a given agent has an “action inventory”’. Torre (Torre 2016, p. 4) criticizes them in saying “Not only do their available actions differ, but their beliefs about what actions are available to each of them also differ’ (the emphasis is Torre’s.) Unless Perry believes that he himself is Perry, he does not realize the action of ceasing to follow the trail is an available action. 8  According to Perry (2000, p. 38), ‘[…] there is a class of propositions that can only be expressed in special circumstances. In particular, only I could express the proposition I expressed when I said “I am making a mess.” Others can see, perhaps by analogy with their own case, that there is a proposition that I express but, it is in a sense inaccessible to them.’ 9  Oishi (2020b) utilizes the distinction between a speaker and an addresser to explain de se attitudes. 10  In other words, they are not truth-valuable (Recanati 2004).


E. Oishi

performs a particular illocutionary act. These thoughts are specified by the accepted conventional procedure for performing the illocutionary act as the thoughts of the addresser, and, therefore, specific contents are given only when a particular speaker in a given case performs it.

Analysing the Problematic Cases The preceding section illustrates entities in discourse/communication, a principal/ addresser and an addressee, who are distinct from real entities in the world, an author/speaker and a hearer, respectively. If these distinctive entities are referred to by the personal pronoun I and you, communication in which tokens of I and you are used should be more complicated than Kaplan’s model suggests. The problematic cases discussed in section “Problematic Cases Of The Personal Pronouns I and You” are such cases. Sidelle’s (1991) Answering Machine Paradox and the similar problem posed by Corazza et al. (2002) are caused by analyzing the referents of tokens of I only as authors/speakers, and, therefore, by allowing principals/addressers to be referents of them, much more straightforward descriptions are possible.11 According to Goffman’s (1981) model, ‘I’ refers both to the author and to the principal, and, in Sidelle’s (1991) case, the principal ‘I’ in ‘I am not here now’ refers to the author ‘I’, and commits to what is said by the words that the author chose to express her/his sentiments: the author is away from the place of the utterance (where the telephone answering machine is located) at the time of utterance (when a hearer makes a call). Similarly, in the case of Corazza et al. (2002), the principal ‘I’ in ‘I am not here today’ refers to the author ‘I’ whose sentiments (presumed by the animator Ben) are expressed by the words used, and the principal commits to what the words say: the author is away from the place of utterance (Joe’s office) in the day of the utterance (when the note is put). According to Austin’s ([1962]1975) model, ‘I’ refers both to a speaker in a given case and to the addresser who performs an illocutionary act and brings about a conventional effect. By the recorded utterance ‘I am not here now’, the speaker ‘I’ as the addresser ‘I’ performs the illocutionary act of informing a hearer (an appropriate addressee) of her/his (the speaker’s) whereabouts: the speaker is away from the place where the telephone answering machine is located when the hearer makes the call. As the illocutionary effect, the hearer as the addressee is informed of the speaker’s whereabouts. In the case of Corazza et al., the addresser ‘I’, whose locutionary act is performed by Ben for the speaker Joe, performs the illocutionary act of informing a hearer (an appropriate addressee) of his (Joe’s) whereabouts: the

 Oishi (2017) discusses how to solve problems of indexicals, which is developed by the present paper.


Speech-Act-Theoretic Explanations of Problems of Pure Indexicals


speaker (Joe) is away from his office at the time when the hearer visits there. As the illocutionary effect, the hearer is informed of the speaker’s (Joe’s) whereabouts. Another problematic case (Predelli 1998) can be dealt with by allowing the referent ‘you’ to be the addressee of an illocutionary act as well as a hearer. Joe, before leaving home at 8.00 AM, writes a note to his partner: ‘As you can see I am not at home now. Please meet me in six hours in my office’ (the emphasis is Predelli’s.) Joe, expecting his partner to return at 5.00 PM, intends for her to meet him at 11.00 PM. Joe’s partner, however, arrives home late at 7.00 PM. Being aware that she was expected to be home at 5.00 PM, she will meet him at 11.00 PM rather than at 1.00 AM in the next morning, which is 6 hours after she actually reads the note. ‘You’ in ‘As you can see I am not at home now. Please meet me in six hours in my office’ refers to both the hearer, Joe’s partner, and the addressee of the illocutionary acts of informing and requesting. As the addressee of the acts, the hearer is informed that the speaker (Joe) is away from the place where the note is placed (their home) at the time the illocutionary acts are expected to be performed to her (at 5.00 PM), and requested to meet the speaker in his office in 6 hours (at 11.00 PM). The hearer interprets the time referred to by ‘now’ as the time she was expected to be the addressee of the illocutionary acts, rather than as the time the acts are actually performed to her, and, therefore, she will meet Joe at 11.00 PM in his office. To analyze our examples (‘Am I your husband?’/‘Yes, you are’ and ‘Like you?’/‘But not you?’), in addition to acknowledging a principal/addresser and an addressee as the referents of tokens of I and you, respectively, the illocutionary act of identifying should be compared and contrasted with that of stating. Austin ([1962]1975, pp. 161–162) categorizes both types of illocutionary act as expositives: the acts of expounding a view, conducting an argument, and clarifying a usage or a reference. Oishi and Fetzer (2016) regard them as an illocutionary act type whose effect is brought about in discourse; when these expositive illocutionary acts are performed, a view is expounded, an argument is conducted, and a usage or a reference is clarified as their illocutionary effects, and the direction in which the discourse is developed is affected. The following is Austin’s list of expositives: 1 . affirm, deny, state, describe, class, identify 2. remark, mention,?interpose12 3. inform, apprise, tell, answer, rejoin 3a. ask 4 . testify, report, swear, conjecture, ?doubt, ?know, ?believe 5. accept, concede, withdraw, agree, demur to, object to, adhere to, recognize, repudiate 5a. correct, revise 6 . postulate, deduce, argue, neglect, ?emphasize 7. begin by, turn to, conclude by 12

 The question marks in the list are Austin’s.


E. Oishi

7a. interpret, distinguish, analyse, define 7b. illustrate, explain, formulate 7c. mean, refer, call, understand, regard as In his paper ‘How to Talk: Some Simple Ways’ (Austin 1979) Austin shows how the act of identifying is differentiated from that of stating.13 Austin illustrates the simplified world S0, which consists of individual items with only one definite type. The language in S0 permits the utterance of sentences of one form S: ‘I is a T.’ I(tem)-words are correlated by conventions of reference with items that they refer to, and T(ype)-words are correlated by conventions of sense with item-types (and thereby item-types are given names and T-words are given senses). Four distinctive acts are performed by utterances of the simple sentence form S, ‘I is a T’: c-identifying (placing), b-identifying (casting), stating and instancing. All these acts basically match the type of the item referred to by an I-word with the sense of a T-word. How are these distinctions brought about? The first difference is the direction of fit: fitting names to given items, which is the case of c-identifying and stating, or fitting items to given names, which is the case of b-identifying and instancing. The other difference is the onus of match: the type of the item is taken for granted and the question is whether the sense of the T-word is such as really to match it, which is the case of c-identifying and instancing, or the sense of the T-word is taken for granted, and the question is whether the type of the item is really such as to match it, which is the case of stating and b-identifying.14 Austin provides the following diagram (1979, p. 142) to illustrate these distinctions (Fig. 1): Austin (1979, p. 143) adds the following explanation: c-identifying (placing) ‘N fits I’

instancing ‘N fits I’

stating ‘I fits N’

b-identifying (casting) ‘I fits N’

Fig. 1  Four types of speech act performed by ‘I is a T’

 On the basis of this paper, Oishi (2003) analyzes how four types of speech act can be performed in uttering ‘X is a Y,’ and how they express four distinguishable meanings. 14  Oishi (2003) shows that the referential use of definite descriptions occurs in stating while the attributive use occurs in c-identifying. 13

Speech-Act-Theoretic Explanations of Problems of Pure Indexicals


The horizontal line indicates that the direction of fit is the same. […] In the verbalizations given in the diagram, that which is being fitted to is shown by italics, in contrast to that which is being with, which is not in italics. The vertical line indicates that the onus of match is the same. […] In the verbalizations given in the diagram, that, name or item, on the sense or type of which lies the onus of match is put as the subject (Austin 1979, p. 143). (the emphasis is Austin’s)

The difference between c-identifying and stating is relevant to explaining the example of ‘Am I your husband?’ While their direction of fit is the same, that is, names (T-words) are matched to the given items (referred to by I-words), there is a difference in the onus of match. In the case of c-identifying, an item-type is taken for granted, and the question is whether or not the sense of a name (T-word) really matches the item-type. In the case of stating, on the other hand, the sense of a name (T-word) is taken for granted, and the question is whether or not the type of the item really matches the sense. Let us show the ‘Am I your husband?’ example again: (1) Jack:

So tell me, Laurel, from your heart …

… Am I your husband? Laurel: (a long pause) Yes, you are. Jack: Thank you. The uniqueness of Jack’s utterance lies in its illocutionary act type, requesting to c-identify, and the referent of ‘I’. The speaker (Jack) refers to himself as the addresser, and requests Laurel as the addressee to c-identify his type by asking whether or not his type matches the sense of ‘your husband’. In this act, the type of the addresser referred to by ‘I’, in particular, the nature of his relationship to the addressee, is matched to the sense of ‘your husband’, the addressee’s husband. In answering ‘Yes, you are,’ the speaker (Laurel) c-identifies the addressee as her husband, and accepts that they are husband and wife. The act of c-identifying contrasts with that of stating, and the utterance ‘Am I your husband?’ can be interpreted as the act of requesting to state the truth. In this act, the sense of ‘your husband’ is taken for granted as Jack Sommersby, who committed the crime of murder, and an item, the speaker, is referred to by ‘I.’ The question is whether or not the type of the speaker really matches this sense. Laurel’s response, ‘Yes, you are’, is, according to this interpretation, the act of stating/confirming that the type of the speaker is her husband Jack Sommersby. These are the illocutionary act types that the jury and the judge in the film Sommersby interpret these utterances as, and, as the result, Jack is found guilty of the murder and hanged.15  These utterances are interpreted by Jack and Laurel as a sequence of requesting to c-identify and c-identifying, while the jury and the judge at the court interpret them as a sequence of requesting to state the truth and stating it. The film viewers interpret them in the same way as Jack and Laurel. Different illocutionary acts can occur in different frames of interaction (Fetzer 2006).



E. Oishi

An addressee as the referent of ‘you’ (in ‘Like you’) appears in the act of b-­ identifying and a hearer as the referent of ‘you’ (in ‘Not you?’) appears in the act of instancing in the film Paddington. B-identifying and instancing have the same direction of fit: items (referred to by I-words) are matched to given names (T-words). In the case of b-­identifying, the sense of a name (T-word) is taken for granted, and the question is whether or not an item-type really matches the sense. In the case of instancing, on the other hand, the type of an item is taken for granted, and the question is whether or not the sense of the name (T-word) really matches the type. Let us show the conversation in Paddington again: (2) Mrs. Brown: You see, when a young person comes to this country, I’m afraid they don’t just move in with the first people they meet. Paddington: No? Mrs. Brown: You need a proper guardian. Paddington: What’s that? Mrs. Brown: A grown-up who takes you into their home and looks after you. Paddington: Like you? Mrs. Brown: Erm … Yes, well, I suppose so. Paddington: But not you? Mr. Brown: No. We don’t do that. In uttering ‘Like you?’ Paddington performs the act of b-identifying: the sense of the name ‘guardian’ is taken for granted (‘a grown-up who takes you into their home and looks after you,’) and the question is whether or not the item-type referred to ‘you’, the type of the addressee, matches the sense. In the following utterance, Mrs. Brown affirms the (b-)identification: the type of her matches the sense of the name ‘guardian.’ Paddington then performs the act of instancing: the type of the item, the hearer Mrs. Brown, is taken for granted, and the question is whether or not the sense of the name ‘guardian’ matches the type. Mr. Brown denies it: the type of Mrs. Brown is such that the sense of the name ‘guardian’ does not match it.

Concluding Remarks The present paper has shown that the speaker and the hearer should be analyzed not only as real entities in the world but also entities in the discourse as a principal/ addresser and an addressee, respectively (Austin [1962]1975; Goffman 1981). The distinction between entities in discourse (a principal/addresser and an addressee) and entities in the world (an author/speaker and a hearer) is necessary to describe communication in which either a speaker or a hearer is absent in the speech situation where a speech act is performed. Furthermore, the distinction should be

Speech-Act-Theoretic Explanations of Problems of Pure Indexicals


beneficial for describing mediated communication, in which frames of interaction are layered. The distinction provides a useful tool to describe communicative interactions. An author selects what sentiments s/he expresses by what words, and, by means of the selected words, the author’s position as the position of a principal is established, and her/his beliefs as those of a principal are told, and s/he as a principal commits herself to what the selected words say. In speech-act-theoretic terms, a speaker in a given case, who is appropriate for the invocation of the procedure of bringing about a particular illocutionary effect, invokes the procedure as the addresser of the illocutionary act, and expresses thoughts or feelings as those of the addresser, and intends to inaugurate, and subsequently inaugurates, consequential conduct as the addresser. A hearer in the given case understands the locutionary act (i.e., the words used), and, as the addressee of the illocutionary act, understands not only the illocutionary effect but also the thoughts or feelings expressed by the addresser and the addresser’s intention to inaugurate consequential conduct. The discourse is developed in a particular direction by the illocutionary effects and the thoughts/feelings and intention expressed by the speaker as the addresser of the act and understood by the hearer as the addressee of the act. The hearer, that is, the speaker in the next turn performs an illocutionary act in the developed discourse to brings about further effects while expressing her/his thoughts/feelings and the intention for consequential conduct. The addresser of the expositive illocutionary act type in particular brings about direct effects in discourse, which influence the direction in which the discourse is developed. This model illustrates how a speaker (author) and a hearer communicate with each other by means of being an addresser (principal) and an addressee of an illocutionary act, respectively, and bring about illocutionary effects jointly. Communication can be complicated in different stages because of idiosyncratic characteristics of a speaker, a hearer and a speech situation, but types of complication can be specified as a gap between a speaker and an addresser, or between a hearer and an addressee, or between the addresser of the preceding illocutionary act and the addresser of the present illocutionary act.

References Aijmer, K. (2013). Understanding pragmatic markers: A variational pragmatic approach. Edinburgh, UK: Edinburgh University Press. Austin, J. L. (1962). How to do things with words. Oxford, UK: Oxford University Press. Austin, J. L. (1979). How to talk: Some simple ways. In J. L. Austin, J. O. Urmson, & G. J. Warnock (Eds.), Philosophical papers (pp. 134–153). Oxford, UK: Oxford University Press. Blakemore, D. (2002). Relevance and linguistic meaning: The semantics and pragmatics of discourse markers. Cambridge, UK: Cambridge University Press. Boër, S., & Lycan, W. (1980). Who, me? The Philosophical Review, 89(3), 427–466. https://doi. org/10.2307/2184397.


E. Oishi

Bolinger, D. (1989). Intonation and its uses: Melody in grammar and discourse. Stanford, CA: Stanford University Press. Bull, P., & Fetzer, A. (2006). Who are we and who are you? The strategic use of forms of address in political interviews. Text & Talk-An Interdisciplinary Journal of Language, Discourse Communication Studies, 26(1), 3–37. Cappelen, H., & Dever, J. (2013). The inessential indexical: On the philosophical insignificance of perspectives and the first person. Oxford, UK: Oxford University Press. Carlson, L. (1984). “Well” in dialogue games: A discourse analysis of the interjection “well” in idealized conversation. Amsterdam, Netherlands/Philadelphia, PA: John Benjamins Publishing. Castañeda, H.-N. (1966). He’: A study in the logic of self-consciousness. Ratio, 8, 130–157. Chierchia, G. (1989). Anaphora and attitudes de se. In R. Bartsch, J. van Benthem, & P. v. E. Boas (Eds.), Semantics and contextual expression (pp. 1–31). Dordrecht, Netherlands: Foris. Corazza, E., Fish, W., & Gorvett, J. (2002). Who is I? Philosophical Studies, 107(1), 1–21. https:// Feit, N., & Capone, A. (Eds.). (2013). Attitudes de se: Linguistics, epistemology, metaphysics. Stanford, CA: CSLI Publications. Fetzer, A. (2006). “Minister, we will see how the public judges you.” Media references in political interviews. Journal of Pragmatics, 38(2), 180–195. pragma.2005.06.017. García-Carpintero, M., & Torr, S. (Eds.). (2016). About oneself: De se thought and communication. Oxford, UK: Oxford University Press. Goffman, E. (1981). Forms of talk. Philadelphia, PA: University of Philadelphia Press. Higginbotham, J. (2009). Tense, aspect, and indexicality. Oxford, UK: Oxford University Press. Huang, M., & Jaszczolt, K. (Eds.). (2018). Expressing the self: Cultural diversity and cognitive universals. Oxford, UK: Oxford University Press. Jucker, A.  H. (1993). The discourse marker well: A relevance-theoretical account. Journal of Pragmatics, 19(5), 435–452. Kaplan, D. (1989). Demonstratives: An essay on the semantics, logic, metaphysics, and epistemology of demonstratives. In J. Almog, J. Perry, & H. Wettstein (Eds.), Themes from Kaplan (pp. 481–563). Oxford, UK: Oxford University Press. Lewis, D. (1979). Attitudes de dicto and de se. The Philosophical Review, 88(4), 513–543. https:// Ninan, D. (2010). De se attitudes: Ascription and communication. Philosophy Compass, 5(7), 551–567. Norrick, N. R. (2001). Discourse markers in oral narrative. Journal of Pragmatics, 33(6), 849–878. Oishi, E. (2003). Semantic meaning and four types of speech act. In P. Kühnlein, H. Rieser, & H. Zeevat (Eds.), Perspectives on dialogues in the new millennium (pp. 135–147). Amsterdam, Netherlands: John Benjamins Publishing. Oishi, E. (2017). Discursive Functions of Japanese Personal Pronouns. Вестник Российского Университета Дружбы Народов. Серия: Лингвистика (Russian Journal of Linguistics), 21(2), 305–319. Oishi, E. (2020a). Discourse markers as indicators of connectedness between expositive illocutionary acts. Pragmatics and Society, 11(1), 1–23. Oishi, E. (2020b). De se attitudes in discourse. Intercultural Pragmatics, 17(3), 365–387. https:// Oishi, E., & Fetzer, A. (2016). Expositives in discourse. Journal of Pragmatics, 96(April), 49–59. Perry, J. (1977). Frege on demonstratives. The Philosophical Review, 86(4), 474–497. https://doi. org/10.2307/2184564. Perry, J. (1979). The problem of the essential indexical. Noûs, 3–21. 2214792.

Speech-Act-Theoretic Explanations of Problems of Pure Indexicals


Perry, J. (2000). The problem of the essential indexical and other essays. Stanford, CA: CSLI Publications. Predelli, S. (1998). I am not here now. Analysis, 58(2), 107–115. analys/58.2.107. Recanati, F. (2004). Literal meaning. Cambridge, UK: Cambridge University Press. Recanati, F. (2016). Indexical thought. In M. García-Carpintero & S. Torr (Eds.), About oneself: De se thought and communication (pp. 141–178). Oxford, UK: Oxford University Press. Sbisà, M. (1984). On illocutionary types. Journal of Pragmatics, 8(1), 93–112. https://doi. org/10.1016/0378-2166(84)90066-3. Sbisà, M. (2002). Speech acts in context. Language & Communication, 22(4), 421–436. https:// Sbisà, M. (2018). Varieties of speech act norms. In M.  Witek & I.  Witczak-Plisiecka (Eds.), Normativity and variety of speech action (pp. 23–50). Leiden, Netherlands/Boston, MA: Brill Publishing Company. Schiffrin, D. (1987). Discourse markers. Cambridge, UK: Cambridge University Press. Schourup, L. (2001). Rethinking well. Journal of Pragmatics, 33(7), 1025–1060. https://doi. org/10.1016/S0378-2166(00)00053-9. Searle, J. (1969). Speech acts: An essay in the philosophy of language. Cambridge, UK: Cambridge University Press. Searle, J. (1979). Expression and meaning: Studies in the theory of speech acts. Cambridge, UK: Cambridge University Press. Sidelle, A. (1991). The answering machine paradox. Canadian Journal of Philosophy, 21(4), 525–539. Stalnaker, R. (1981). Indexical belief. Synthese, 49(1), 129–151. BF01063917. Torre, S. (2016). De se thought and communication. In M. García-Carpintero & S. Torr (Eds.), About oneself: De se thought and communication (pp. 1–21). Oxford, UK: Oxford University Press.

Prospective Reference Paul Saka

After introducing the notion of prospective reference, and identifying it as a species of speaker reference (Sect. 1), I argue that it is ubiquitous in everyday language, and that confusing it with semantic reference has momentous consequences for commerce, the law, and discourse about abortion (Sect. 2). In particular, the fact that fetuses are unexceptionally called babies does not mean that they already are babies, a point tested by experiment (Sect. 3). Analysis of the data confirms that speakers of English, regardless of their professed stances on fetal personhood, do not normally regard “children”, “human being”, or “person” as extending to fetuses (Sect. 4).

1. Theories of Prospective Reference Prospective reference is best introduced by example. Suppose that I want to bake a cake, that I have prepared the batter, that I have poured it into a pan, and the oven has reached the right temperature. Then I may say: (1) My cake is ready to go in the oven. This use of language seems perfectly natural, and it would seem to be true. At the same time, however, it is mighty peculiar, and it seems to be semantically defective. It is peculiar because cakes, by definition, are baked goods. Since the batter is unbaked (at the time of utterance), it does not qualify as cake, leaving “my cake” without reference. It thus seems that (1) is either false or without truth-value.

P. Saka (*) University of Texas Rio Grande, Edinburg, TX, USA e-mail: [email protected] © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 F. Macagno, A. Capone (eds.), Inquiries in Philosophical Pragmatics, Perspectives in Pragmatics, Philosophy & Psychology 27,



P. Saka

My example is theoretically interesting because it seems to demand an intensional reading without bearing markers of modality or propositional attitude – the traditional cloaks of opacity. It is intensional inasmuch as coreferential substitution seemingly fails: even though cake is arguably identical to a kind of baked batter, it may be that (1) is true while (2) is false: (2) My baked batter is ready to go in the oven. Statement (1) is furthermore intensional in that normal existential inference fails: whereas “a book is on the table” would seem to entail the existence of at least one book and at least one table, (1) does not entail, at the time of utterance, the existence of at least one cake. Statement (1) invites a number of competing analyses. According to what I shall call the narrow view, the definition of “cake” is indeed restricted to baked goods and (1) is semantically defective. According to the broad view, the definition of “cake” is sufficiently capacious to encompass both baked goods and batter, and (1) is perfectly true. According to the ontic indeterminacy view, “cake” has a single underspecified meaning that accommodates the narrow and broad interpretations, with batter lying in a penumbra that gets counted as cake in one context and non-­ cake in another. According to the epistemic vagueness view, there is a fact of the matter but it is inscrutable. According to the simple, semantic polysemy view, the word “cake” possesses multiple senses, whether lexical or derived: a broad use that makes (1) true and a narrow use that makes (1) false. In this case of vertical polysemy, one sense is superordinate to the other; “cake” is comparable to “cow”, which denotes either a bovine or a female bovine, with all senses on a semantic par: “that’s a cow,” said of a bull, is completely true (in one sense) and completely false (in another). According to the idealization view, one or more distinguished default senses dominate inferior senses (cf. Lakoff 1987). The word “cake” can denote batter, according to this view, but such denotation is not as “good”, satisfying, preferred, valid, or robust as its reference to baked stuff. Some of the foregoing options may or may not coalesce, depending on how they are elaborated. Indeterminacy and idealizations may prove to be interdependent, for instance, but it’s also possible that the two call for separate logics (supervaluationist versus fuzzy). Another aside: the idealization view might be developed by means of asymmetric conceptual roles. The assumption that batter triggers cake, but cake does not trigger batter (instead triggering an image-schema of something solid and hence inconsistent with batter), may take us some way toward explaining why (1) seems both true and untrue. This theoretic approach, however, requires a great deal of working out. For example, batter per se (a substance) never qualifies as a cake (an object); rather, in (1) it is a quantity of batter, individuated by a cake tin, that counts as a cake. Although I believe that meaning is far less determinate than it is usually imagined to be, and I believe that ideal models offer a promising approach for psychologically grounding linguistic competence, in the case of prospective reference I shall favor the narrow view. According to the narrow view, (1) is not literally or technically true, but it loosely conveys a truth because speakers can use “cake” to

Prospective Reference


refer to batter, given the perceived similarity or causal relation between the two. Under this speech-act theoretic approach, prospective reference may be defined in terms of speaker intentionality: (df)

Speaker S uses expression “x” to make prospective reference to object o iff S uses “x” to refer to o merely because S imagines o becoming x.

Notice that even if the batter never actually becomes a cake – for instance, if I trip on the way to the oven, and spill the batter all over the floor – then (1) would still seem to be correct. Indeed, even when I expect or know that the batter will never become a cake I may still felicitously call it a cake. For instance, if I suddenly discover that my oven is not working, I may decide to dump my batter so as to focus on repairing or replacing the oven. In that case, calling upon my prior expectations, I may correctly say on my way to the garbage disposal: (3) My cake is ready for the oven, but the oven isn’t ready for my cake! For speaker reference, imagination is all that’s needed. So long as I think of object o becoming an x, I may mentally label o as an x, and thence I may actually speak of o as an x. (For more on imaginary reference, subsuming prospective reference, retrospective reference, partial reference, and broken reference, see Nunberg 1979; Saka, manuscript.) The narrow view is accompanied by a pragmatics view. While (1) is semantically defective, there is a manner of speaking in which it is felicitous. This approach can be developed in two ways. First, it could be said that (1) is figuratively true. Since the cake batter in my pan is similar to a genuine cake in composition, shape, and size, (1) may be regarded as metaphorically true; and since the cake batter is furthermore related to a cake in that the one causally grounds the other, (1) may be regarded as metonymically true. Second, it could be said that (1) is referentially true rather than attributively true, in Donnellan’s terms (1966). Donnellan observes that (4) admits two uses: (4) Smith’s murderer is insane. The so-called attributive use of (4) means “Smith’s murderer, whoever that may be, is insane”; the so-called referential use means “That person I am drawing your attention to, by speaking of Smith’s murderer, is insane.” By the same token, I can draw your attention to the batter in my pan by saying “my cake”, despite our both knowing that cake it is not. Donnellan seems to classify the referential use as a semantic phenomenon, for he regards it as a refutation of Bertrand Russell’s theory of descriptions. It is more plausible, however, to regard it as an unequivocally pragmatic phenomenon, as do Kripke (1977), Searle (1979), Soames (2009), Bezuidenhout (2013), and Leonardi (2019). This puts referential uses in the same orbit as figurative speech, and it invites us to consider the relation between the two. Could referential uses be a special case of figurative speech? Surely not. When you use “Smith’s murderer” to refer to some scapegoat who’s been framed, you aren’t using metaphor, except in the most meaningless way. (The scapegoat and the


P. Saka

true murderer may resemble each other in both being human, and both inhabiting the same district, and in any number of other ways, but if this were enough for metaphorical grounding then you could felicitously refer to countless individuals, even the prosecutor, as “Smith’s murderer”.) Nor are you using sarcasm: you may refer to the scapegoat as “Smith’s murderer” without echoing anyone else’s use of the expression and without knowing or caring whether the alleged murderer actually murdered. Nor are you using metonymy, hyperbole, or any other recognized figure of speech. What’s really going on, in referential uses, is that the speaker believes that the scapegoat is Smith’s murderer, or believes that the hearer or larger society so believes, or adopts the pretense or somehow imagines the scapegoat as a murderer. The parenthetical remark in the preceding paragraph, by the way, refutes the standard account of figurative speech. If metonymy and metaphor are to be defined in terms of relatedness and resemblance, as dictionaries and textbooks often say, then any two terms whatsoever are equally good metonyms and metaphors for anything, since relatedness and resemblance are available in infinite abundance (the bottom quark and Taylor Swift are alike in that neither one is the number 42, yet neither works as a figurative reference to the other). What matters is not relatedness per se, but relatedness in the mind: the minds of the speaker and hearer for language production and comprehension, respectively. For this reason, I would classify metonymy and metaphor as special cases of “referential” use. To summarize the discussion so far: I have given an example of prospective reference and indicated its significance for the literature on opacity, I have given a theory-laden definition of it and identified alternative rival analyses, I have suggested that figurative and “referential” uses are not independent, and I have argued that the former is better understood as an instance of the latter rather than the other way around. In the next section I elaborate on how prospective reference is used in the real world, and what’s at stake.

2. The Practice of Prospective Reference Prospective reference is ubiquitous in everyday speech, it raises interesting issues in the worlds of commerce and law, and it plays an important role in the abortion debate.

2.1. Commerce and Law If you buy a toaster then by definition it should toast, or at least be capable of toasting. Yet the appliance in itself is incapable of toasting; it is only the complex of appliance plus electricity (not to mention slices of bread or whatnot) that can produce toast, and it is only because speakers imagine the appliance becoming something other than it now is – something removed from the box and plugged into an outlet – that they call it a toaster. If you think absurd my allegation that so-called

Prospective Reference


toasters in themselves are strictly speaking not toasters, consider that many consumer products confess “batteries not included” or “requires assembly”. Such qualifications are necessary to prevent fraud; a putative flashlight without a battery or working power source does not emit light and therefore is not really a flashlight, and a “coffee table” consisting of a box of boards and loose screws is not really a table. Consumer product labels are routinely untrue. Untruth does not always spell deceptiveness. Still, merchant deception is more common than one might think. Online product reviews teem with postings from disgruntled consumers who expected more from their purchases than they got. Once, an acquaintance of mine paid for a hotel room whose advertised amenities included internet use. Upon arrival she discovered that she could not access a single website: the hotel provided connectivity, but she was supposed to have lugged her own computer. To her, this was like advertising that the room comes with television, telephone, and heating simply because it comes with radio waves, phone jack, and electricity. For some artifacts, working functionality is required to satisfy the concepts that go with them. Even if only one in a million consumers is fooled, consumers who naively rely on the strict truth of an ad should be favored over merchants who exaggerate, for naivete should not be a crime; making deceptive falsehoods should be. That, however, is not the principle according to which courts normally adjudicate. Instead, courts habitually invoke what’s most “reasonable” and what’s not, notions that mock due process because they are so subjective. In effect, calling an argument unreasonable or absurd, or citing “mischievous” effects, to use legal jargon, is often just a way of saying “I cannot or will not tell you where I think you’ve gone wrong, but I don’t like your conclusion.” When the person saying this wields authority over others, it is a cover for the expression of naked power; more generally, it is an exercise in dogmatic question-begging. If actual fallacies in an argument were identifiable, judges would specify them instead of hiding behind the hollow term “unreasonable”. So-called legal pragmatists openly acknowledge ruling so as to grease the wheels of commerce (e.g. Posner 1981), a policy that systematically favors businesses (which want to close sales) at the expense of customers (who want refunds). What they overlook is that contracts are regarded by most people as binding because they are supposedly based on mutual consent. Mutual consent is absent, however, when there is no actual meeting of the minds. If a contract is ambiguous between readings m1 and m2, and if one party to the contract accepts m1 while the other accepts m2, there is no substantive agreement. In such cases courts should void the contract rather than seek the most “reasonable” interpretation. Alternatively, in cases of ambiguous contracts courts should systematically favor the party that did not author the contract. When multinational corporations craft legal agreements that customers are expected to sign without benefit of attorney, and the contracts prove to be ambiguous or self-contradictory, fault primarily lies with the author – the party that created the confusing and misleading document, the party with superior resources for doing the job, and the only party that might be suspected of attempting deliberate trickery. Yet another alternative to current practice, which sometimes operates


P. Saka

according to winner-take-all, would be for judges more often to “split the difference” in cases involving ambiguous contracts. When what’s “reasonable” conflicts with the language of law or contract, an argument can be given for following a strict kind of “plain meaning rule” of judicial interpretation: (a) The use of figurative language always expands ambiguity, exponentially building upon the set of meanings possible in the literal base. (b) Therefore minimizing ambiguity requires the use of literal language. (c) In order to be fair, laws and contracts must minimize ambiguity. (The fact that some ambiguity is always unavoidable does not mean that we should be indifferent to reducing it.) (d) Therefore, in order to be fair, laws and contracts must use literal language. (e) Therefore, if we are to presume that legislators and parties to contracts are fair then we are to presume that their language is used and meant literally. The argument rests upon the nature of language, but also presupposes that laws aim at fairness. (For those who deny the existence of literal language, the argument may be recast in terms of “language whose interpretation requires the least possible amount of non-demonstrative inference”.) To summarize: I have argued that the practice of prospective reference is deeply embedded and highly respected in our culture, as indicated by the fact that merchants and governments often uphold the non-literal terms of contracts, appealing to the doctrine that correct interpretation is determined by what’s “reasonable” (which all too often means what’s profitable). This doctrine of interpretation is but a cultural choice, however, as indicated by the fact that alternative judicial norms are possible.

2.2. Abortion Discourse When a woman refers to her fetus as a “baby”, do her words denote an actual human baby (and therefore a child and a human being), in which case the fetus has a defeasible right to life? Or does the woman refer in a manner of speaking, in a prospective sort of way, which carries no moral implications beyond the proposition that the fetus has yet to become a human being? Since the semantic question is pregnant with potential moral significance, partisans in the abortion debate frequently disagree over the correct answer. In particular they disagree over the following theses: Fb The (human) fetus is a baby. Fc The fetus is a child. Fh The fetus is a human being. Fp The fetus is a person.

Prospective Reference


To clarify, I use the term “human being” idiomatically, to mean human organism, and not as a composition of “human” plus “being” or “entity”. The idiomaticity of the term is confirmed by the fact that it receives its own entry in dictionaries. Besides, under the compositional reading every cell in my body, and strand of hair, would count as a human being. Clearly, that is not the intended interpretation when it is said that killing human beings is wrong. Just as “cake” invited a number of competing analyses, so too do the terms above. According to the polysemy view, the word “child” possesses multiple senses, one of which makes Fc true and one of which makes Fc false.1 According to the broad view, the definition of “child” is sufficiently capacious to encompass prepubescents stretching backward not only to birth but to conception, and Fc is simply true. According to the narrow view, the definition of “child” is restricted to postnatal beings and thesis Fc is simply false. According to the indeterminacy view, “child” has a single underspecified meaning that accommodates both narrow and broad readings, with penumbral fetuses getting counted as children in some contexts and non-children in others. According to the epistemic vagueness view, whether a fetus is a child is something that no one incontestably knows. According to the idealization view the word “child” may extend to fetuses, but such denotation is not as “good”, satisfying, preferred, valid, or robust as its reference to postnatal beings. Fetuses may be children, just not to the same degree, or as ideally, as infants and older children are; or they may even be so far from the ideal that their category membership is indeterminate, either ontologically or at least epistemically. I shall present data in favor of the narrow analysis: fetuses are not literally children, but may loosely be called such in an act of prospective reference. If this be the case, it would be an unremarkable instance of a common device in natural language. “Baby” can be elliptical for “unborn baby”, and “unborn baby” would be a non-­ intersective phrase akin to “retired colonel”, “celebrity manqué”, “counterfeit money”, and “toy gun”. An unborn baby is not a baby, and colloquially calling it one does not make it one. (Even those who possess mutual knowledge that a banknote is counterfeit might still felicitously refer to it as “money” simpliciter. A counterfeiter may telephone her partner: “the police are on their way – you’ve got 15 minutes to pack up all the money and get out of there.”)2

1  “Child” is surely polysemous, with senses physiological (prepubescent), legal (minor), genealogical (offspring), and others. The question here is whether its multiple senses include one that extends to fetuses and one that does not. 2  Thus, if I am right then non-intersective phrases are meant to be interpreted according to speaker reference and do not call for special semantics, contrary e.g. to Partee (2010).


P. Saka

3. Experimental Data To investigate prospective reference I conducted some experiments. The hypotheses, designs, and interpretations are my own; for assistance in administering surveys and compiling results, I gratefully credit my students Oscar Quintanilla and Daphne Zuniga. Due to resource limitations, our work focuses exclusively on Fc, Fh, Fp, and: F The (human) fetus is a child, human being, or person. Setting aside thesis Fb should hardly matter to the abortion debate, where humanity and personhood are what’s fundamentally at issue rather than babyhood per se. My experimental work presupposes the following quantificational principle: QP

If the number of j’s in a given situation is independent of the number of k’s then j’s do not qualify as k’s and k’s do not qualify as j’s.

Since the number of beagles in a room is independent of the number of frogs, beagles can’t be frogs. In contrast, the number of beagles in a room does impact the number of dogs because beagles do count as dogs. This is so elementary that we can say, for all rational subjects S: QP’ If S believes the number of j’s in a given situation is independent of the number of k’s  then S believes that j’s do not qualify as k’s and k’s do not qualify as j’s. In particular, if S believes that the total number of children in a situation is independent of the number of fetuses then S believes that fetuses do not qualify as children. Truism QP’ is helpful because it licenses an experimental paradigm that instills a high level of seriousness in subjects. Subjects may say that famous persons adorn our currency, but what they really mean is that images of famous persons are on currency. To get at their true beliefs we can ask them to count the number of famous persons in a room, in which case they will ignore celebrities on banknotes. Approaching their task with a quantificational frame of mind makes subjects more objective, careful, literal, rigorous, and precise. Using the same paradigm, we can ask subjects to count the children or persons in a given situation. Subjects who think persons include fetuses will want to number fetuses among the total, and when the number of fetuses is unknowable then counting will be impossible. In my first study, I asked my students “How many persons are in the room?” Opponents of thesis Fp should have counted the roughly 30 persons who were visibly present, and cited an exact number. Proponents of Fp should have reported that they did not know or could not tell, given that any number of the female students might have been pregnant. Neither side performed as it should have. Instead, the subjects ignored the announcement that they could stand up and move around; and from their seated positions, they almost universally failed to see all of their classmates, with nearly every student giving a distorted number. However, give numbers they did; not one suggested that the true answer was impossible to know due to possible pregnancies or fetuses.

Prospective Reference


For my subsequent studies I decided to use crisp images projected on a large screen at the front of the room. So as not to conflate reality and representation, we never asked “How many persons are on the screen?” So as not to imply that answers could be garnered by visual means, we never asked “How many persons are represented on the screen?” Instead we asked object-level questions such as “How many people are in the hot tub?” and “How many persons are in the tent?” We included questions that were obviously impossible to answer, e.g. “How many dogs are in the car?” (with exterior image of car), to test whether subjects admitted ignorance where information was insufficient. A disheartening number of our subjects guessed, and they persisted in guessing despite emphatic instructions not to. Finally we achieved compliance, excepting a few outliers, by training subjects with a few opening slides, as detailed below. Even so, during the pilot phase certain test slides garnered anomalous responses and had to be modified or removed. As one example, we asked “how many pacemakers are in the bus stop shelter?”, accompanied by an image of six people in a bus stop shelter that also features two prominent route maps. Some subjects answered “two”, which suggests to us ignorance of what pacemakers are, combined with inexplicable false confidence. As another example, we asked “how many people are on the couch?”, accompanied by an image of five people sitting on a couch, including one sitting in the lap of another. Many subjects answered “four”, yet to our way of thinking the correct answer is “five”, with the lap-sitter obviously contributing to the weight pressing down on the couch. The on relation does not require contiguity. When I’m riding in a bus I’m on the road, even though it is the tires that are touching the asphalt, not me. When a stack of books is on a table, the top book is on the table, even though mediated by the book below it. If the on relation required direct contact the correct answer to the pilot question would be “zero”, as every single person in the image is wearing clothes; the only thing they are physically touching is their clothes. At any rate, we changed the question to “how many people are on the couch (directly or in directly)?”, and we verified that subjects then counted as expected. The general point is that subjects are prone to giving bizarre responses, but we’ve averted most of the whackiness by putting considerable effort into developing test material. In the end, our test material consisted of two instruments, a political questionnaire and a countability test. The questionnaire served to prime subjects with the idea of fetal personhood and it revealed subjects’ attitudes thereto. The countability test asked subjects whether it is possible for them to tell how many children (human beings, persons) are in situations where the number of fetuses is unknowable, and either to specify an exact number if possible or to explain why it’s not possible. Subjects who claim that it is possible, and specify a number, demonstrate that they are employing a narrow conception of what a child is. Subjects who claim that it is not possible to determine the true number, and offer an explanation bearing any conceivable relation to fetuses or pregnancy, demonstrate that they are employing a broad conception of what a child is. Subject responses that deny countability, while giving an irrelevant reason, were excluded from analysis as uninterpretable (e.g., “it’s not possible to tell how many human beings are on the jet ski because the dog is not wearing glasses”).


P. Saka

The political questionnaire has two versions. The simple version primes just once by asking subjects to agree or disagree with various controversial arguments, one of which is “A fetus is a child / a human being / a person, and destroying it is homicide. Therefore abortion should be restricted.” The full-page 10-minute intensive form is devoted entirely to the abortion issue, with 18 items including: The fetus is a child; abortion is actually a form of infanticide. The fetus is a human being. Killing it is murder. The fetus is already a person, and killing it is murder. The intensive questionnaire served to prime subjects at greater length and to separate out attitudes toward Fc, Fh, and Fp as distinguishable theses. Subjects who agree with a given argument accept its premise, and those who express uncertainty express openness to the possibility that the premise is true; together, these two groups, the assenters and the undecided, count as sympathizers of F/Fc/Fh/Fp. Sympathizers form a single group in that they should pattern together on the countability test: sympathizers should all judge it impossible to say how many children, human beings, or persons there are when the presence of a fetus is unknowable, i.e. in situations with at least one woman of child-bearing age. The countability test begins with a statement of purpose: “In this experiment we are trying to measure your awareness of what you do know and what you don’t know. If you can determine the exact answer, please give it; if you cannot, DO NOT GUESS.” This is followed by four training slides that encourage subjects to recognize and acknowledge instances of ignorance. For example, one slide pictures a sliced cake accompanied by the question, “How many cherries are in the cake? Is it possible for you to tell the answer?” In the training, we told subjects that it is impossible to tell the right answer because the slide does not show the entire inside of the cake. After the training slides come the query slides, which call for subject response. Roughly two-thirds of the query slides are fillers, half of them designed to give subjects clear examples where counting is definitely possible and half where counting is definitely not possible. The latter further served as comprehension-and-­ compliance check tests to filter out subjects who did not acknowledge ignorance where they should have. For example, subjects who claimed they could count how many pairs of underwear a fully dressed man has on, when the true answer could at best be only surmised from convention, were excluded from our analysis. Finally, six slides were interspersed to probe how subjects conceptualize children, human beings, or persons. For example, the question “How many persons are inside the bus stop shelter? Is it possible for you to tell the answer?” is accompanied by a clear image of a bus stop shelter containing two females of child-bearing age and four males. The countability test has two versions. While the basic test depicts no visible pregnancies, the salient test depicts two obvious pregnancies. (Originally I planned to run a neutral condition with no priming at all. Actually running the neutral condition seems unnecessary, however, because results from our first primed condition

Prospective Reference


Table 1  Condition 1: single verbal prime Term of interest in questions “Children” “Children” “Human beings” “Humans” “People” “Persons” Total

Uninterpretable responses 3 6 2 3 1 3 18

“Broad” responses 0 0 0 1 0 0 1

“Narrow” responses 143 140 144 142 145 143 857

“Broad” responses 2 0 3 2 0 0 7

“Narrow” responses 119 116 116 118 120 121 710

Table 2  Condition 2: intensive verbal priming Term of interest in questions “Children” “Children” “Human beings” “Humans” “People” “Persons” Total

Uninterpretable responses 1 6 3 2 2 1 15

proved to be so stark.) Altogether, 445 volunteer subjects participated, all students in lower-division liberal arts classes at my institution. Of these, 423 (94%) passed the series of topic-neutral comprehension-and-compliance check tests; the failures were excluded from the following analysis. Of the analyzed subjects, proponents and opponents of F/Fc/Fh/Fp on the political questionnaire were evenly split, 42% to 43%, with 15% being uncommitted. Condition 1 used the simple political questionnaire, followed by the basic countability test (Table 1). In response to six critical test items, 146 subjects generated 6 × 146 = 876 data points, of which 2% were uninterpretable (unanswered items and the like). Of the 858 interpretable responses, 857 (99.9%) claimed that it was possible to count the number of children, human beings, and persons in situations that contained unknown numbers of fetuses. One subject did mention the relevance of pregnancy on one question, but did not consistently follow this thought through on any of the other items. Condition 2 used the intensive political questionnaire and the basic countability test (Table 2). From 122 subjects we collected 732 data points, of which 2% were uninterpretable. Of the 717 interpretable responses, 710 (99%) claimed that it was possible to count the number of children, human beings, and persons in situations that contained unknown numbers of fetuses. Three subjects did mention the relevance of pregnancy, but not one did so consistently across all key items. Condition 3 used the salient countability test followed by the simple political questionnaire (Table 3). From 135 subjects we collected 816 data points, of which


P. Saka

Table 3  Condition 3: visual priming (images of very pregnant women) Term of interest in questions “Children” “Children” “Human beings” “People” “People” “Persons” Total

Uninterpretable responses 8 5 1 2 19 4 39

“Broad” responses 0 0 0 1 0 0 1

“Narrow” responses 127 130 134 133 116 131 776

5% were uninterpretable. Of the 777 interpretable responses, 776 (99.9%) claimed that it was possible to count the number of children, human beings, and persons in situations that contained unknown numbers of fetuses. One subject did mention pregnancy as relevant to one item, but did not regard it as relevant to any of the other items. Even with priming by images of pregnancy and thoughts of fetal personhood, in some cases considerable priming, the overwhelming majority of subjects claimed that they could tell how many children, human beings, and persons were in situations where the number of fetuses was unknown. Five subjects deviated from this pattern (two proponents of F/Fc/Fh/Fp, two opponents, and one uncommitted), but were inconsistent in doing so. Even they, more often than not, gave “narrow” responses on the countability tasks.

4. Significance of the Data So far I’ve assumed that meanings are associated with concepts, and that concepts are mental. Thus, judgments and beliefs are relevant to semantics. This assumption continues through Sect. 4.1, and its rejection is explored in Sect. 4.2.

4.1. Traditional Lexical Semantics My experiments can be used to test the competing analyses of prospective reference (Sect. 2.2), which yield partially distinct predictions. According to the narrow view, which denies thesis F, all subjects should count children without regarding fetuses. According to the broad view, which asserts F, all subjects should try to include fetuses in the count of children, and should be uncertain of the total whenever the number of fetuses is unknown. According to the semantic polysemy view, “child” possesses different lexical senses, all equally valid semantically, and on any given occasion its interpretation depends on psychological factors such as frequency, recency, and relevance. According to the epistemic vagueness view, either ignorance

Prospective Reference


of category boundaries should lead subjects to admit their inability to count, yielding the same results as the broad view, or it should lead to widespread guessing, yielding the same results as the polysemy view. To extract predictions from the indeterminacy view, consider an example of scalar vagueness such as “How many large circles are there?” In context (5), where the middle circle is liable to count as large, the answer would be “four”, in context (6) the same-sized middle circle counts as small and the answer is “two” and in (7) we might expect answers to divide between “one” and “three”, with perhaps some “I don’t knows”, depending partly on the speaker’s personality.

o o o


o o

ooo o oo o


o o o





By the same token, if the boundaries of “child” are unfixed to the point that fetuses are neither definitely inside nor definitely outside, we should expect applications of “child” to vary from context to context, with psycho-contextual factors again playing a decisive role, as it does for polysemy. Predictions of the idealization view depend on how the approach is developed. If a fetus is to child as an ostrich is to bird, it is an atypical yet fully fledged member of the given category, and the account collapses with that of the broad view. If idealizations yield graded category membership, where fetus is distant enough from the center of child to render categorical judgments iffy or variable, then the boundary between child and non-child would presumably shift according to context, in which case subjects would respond as they would under the polysemy, indeterminacy, and epistemic-vagueness views. In particular, we would expect speakers to judge borderline examples according to their vested interests, and for abortion critics to prefer the broad extensions of “child”. The experimental results flatly refute the broad view. Furthermore, they render the polysemy, indeterminacy, epistemic-vagueness, and idealization views hard to swallow. In condition 1 the political questionnaire explicitly floats the proposition that children include fetuses, so if an inclusive sense of the word “child” truly existed, or an inclusive use would be truthful, it would surely be primed. In condition 2 the political questionnaire dwells at length upon fetuses, amplifying a recency effect with a frequency effect. In condition 3, visible pregnancies are salient, thus raising their relevance. With primes in either the verbal and visual channels, we should predict that a non-trivial number of respondents would categorize fetuses as children. The narrow view comes closest to matching the evidence, by far. The match is not 100%, but falls well within the range afforded by the “noise” in any social-­ science experiment. What’s more, the exceptions as do occur can plausibly be accounted for. Perhaps, not to mention ideological bias, speakers carelessly mistake prospective uses of “baby” and “child” for literal uses. Evidently this is possible, for


P. Saka

over the years many of my students have insisted that play money is money, literally and actually, that toy guns are guns, and that former presidents are presidents (“Why is Clinton still called Mr. President if he isn’t?”). To decide which side is right, an important methodological principle to keep in mind is the second-order criterion of preference: If theories θ1 and θ2 equally explain a given phenomenon, but θ1 is better at explaining why some theorists mistakenly subscribe to θ2 than θ2 is at explaining the reverse, then θ1 is preferable to θ2 (a clarification of Saka 2007: 15). Now proponents and opponents of F can both explain the others’ positions (there exists motivated reasoning, respectively, to accord moral status to the fetus and to deny it). These considerations, so far, cancel each other out. However, opponents of F can add that “unborn baby”, in reference to fetuses, fools careless thinkers into regarding fetuses as actual babies. In short, it is easier to explain how the assertion of thesis F may illusorily appear to be persuasive to proponents of F than it is to explain how the denial of F may illusorily appear to be persuasive to opponents of F. This constitutes evidence against F.

4.2. Semantic Externalism According to semantic externalism, the denotation-conditions of a term are fixed by facts outside the linguistic community’s ken (Kripke 1972; Putnam 1975). Just as everyone once failed to recognize that the earth is actually a planet, i.e. they failed to recognize that the denotation of “planet” includes the earth, perhaps today’s language-­users err when they fail to categorize fetuses as children. The experimental data reported above is accordingly irrelevant. Against this objection I have three arguments. First, externalism holds that denotational boundaries hew to the edges of natural kinds, which are best apprehended by expert specialists. One prediction is that the more that speakers know about biology, the more likely they will accurately categorize instances of “human being”. I don’t have formal data, but my sense is that physicians and physiologists use “human being” just the same as everyone else. Another prediction is that the more that speakers know about morality, or about moral philosophy, the more likely they will accurately categorize instances of “person”. The problem here is that the two sides of the abortion issue both claim to possess greater moral sensitivity and understanding than the other  – except that this isn’t really a problem, as even those who oppose abortion end up in my experiment regarding fetuses as non-persons. In short, invoking a division of linguistic labor and appealing to experts changes none of my earlier conclusions. Second, in reply to the charge that today’s speakers may be mistaken, and that some future uncontested experts will limn the true boundaries of personhood, I clarify that my denial of F, (8), pragmatically amounts to the assertion of (9): (8) A fetus is not a child, human being, or person.

Prospective Reference


(9) To the best of current knowledge, i.e. according to prevailing consensus as indicated by the experimental data, a fetus is not a child, human being, or person. Even if (8) can be challenged by a desperate appeal to ignorance, such disingenuous skepticism does not touch (9). Third, I challenge the legitimacy of semantic externalism. According to externalism, reference is initially grounded in some canonical sample, and then extended by the relation of type-identity. Assuming that some referent, call it c, was “baptized” with the term “water” (perhaps even by means of baptism with water!), “water” then came to mean anything of the same kind as c. The problem is that c instantiates many kinds all at once. To qualify as “the same” as c, must a substance share the same elements as c (hydrogen and oxygen), the same molecular structure (bonded H2O), the same isotopes (the same ratio of deuterium), the same phase state (liquid), the same general size and kinetic state (puddle versus pond or stream), or the same practical functionalities from an anthropocentric point of view, e.g. regarding impurities, temperature, pressure, physical accessibility, price...? (For naturalists, anthropic kinds are natural kinds too!) The intrinsic properties of a substance do not and cannot tell us which properties human speakers bestow their attention on when they choose their verbal conventions; that is a mental act or habit, an internal matter. By the same token, if we were to ask whether a fetus is the same kind of thing as an uncontroversial canonical human being, an adult, the question would be meaningless until “same” is qualified. Fetus and adult are type-identical relative to the type or class of living things (both are alive), they may be type-identical relative to blood-type, they are type-distinct relative to complete genome type, and they are type-distinct in that one has hopes and dreams while the other does not. The denotation of “human being” does not get fixed by attachment to one canonical sample, or to any finite set of samples, without overarching intentions or beliefs. This is not to say that intentions and beliefs are omnipotent. Speakers can and do make mistakes. Mistakes are possible, however, only when speakers misapprehend the facts. Galileo’s pious opponents denied that the earth is a planet because they denied that the earth moves, not because they failed to realize that “planet” denotes a class of objects that wander through the heavens. Likewise, if fetuses are children it must be for some trait that speakers mentally attach to “child”, and proponents of F should be able to specify what trait it is that they have in mind. In summary, the countability experiments overwhelmingly refute the broad view: fetuses do not unequivocally qualify as children, as human beings, or as persons. The experiments also challenge the polysemy, indeterminacy, epistemic-vagueness, and idealization views. The fact that speakers so uniformly claim they can count children, human beings, and persons without regard to fetuses tells us that fetuses are not those things. Although entire communities can and do make systematic errors, there is no evident reason to suspect that one is being made in this case. At least, the doctrine of semantic externalism gives us no such reason. Calling a fetus a baby is but an act of prospective reference.


P. Saka

Acknowledgements  For helping to make this work possible, I gratefully thank Oscar Quintanilla, Daphne Zuniga, and the organizers and audience of the International Conference on Pragmatics and Philosophy (Lisbon, 2018).

References Bezuidenhout, A. (2013). The (in)significance of the referential-Attributive distinction. In A. Capone, F. L. Piparo, & M. Carapezza (Eds.), Perspectives on pragmatics and philosophy (pp. 351–366). Cham, Switzerland: Springer. Donnellan, K. (1966). Reference and definite descriptions. The Philosophical Review, 75(3), 281–304. Kripke, S. (1972). Naming and necessity. Cambridge, MA: Harvard University Press. Kripke, S. (1977). Speaker’s reference and semantic reference. Midwest Studies in Philosophy, 2, 255–276. Lakoff, G. (1987). Women, fire, and dangerous things. Chicago, IL: University of Chicago Press. Leonardi, P. (2019). Descriptions in use. In A. Capone, M. Carapezza, & F. L. Piparo (Eds.), Further advances in pragmatics and philosophy: Part 2. Theories and applications (pp.  137–153). Cham, Switzerland: Springer. Nunberg, G. (1979). The non-uniqueness of semantic solutions: Polysemy. Linguistics and Philosophy, 3(2), 143–184. Partee, B.  H. (2010). Privative adjectives: Subsective plus coercion. In R.  Bauerle, U.  Reyle, & T.  E. Zimmermann (Eds.), Presuppositions and discourse: Essays offered to Hans Kamp (pp. 273–285). Leiden, Netherlands: Brill Publishing Company. Posner, R. (1981). The economics of justice. Cambridge, MA: Harvard University Press. Putnam, H. (1975). The meaning of ‘meaning’. In H. Putnam (Ed.), Philosophical papers (Vol. 2, pp. 215–271). Cambridge, UK: Cambridge University Press. Saka, P. (2007). How to think about meaning. Dordrecht: Springer. Searle, J. (1979). Referential and attributive. In Expression and meaning: Studies in the theory of speech acts (pp. 137–161). Cambridge, UK: Cambridge University Press. Soames, S. (2009). Donnellan’s referential/attributive distinction. In S. Soames (Ed.), Philosophical studies (Vol. I, pp. 149–168). Princeton, NJ: Princeton University Press.

Categorization, Memory and Linguistic Uses: What Happens in the Case of Polysemy Grazia Basile

Introduction Polysemy has been considered a semantic universal since the beginning of the 1960s (see Ullmann 1963) and is such a pervasive phenomenon in historical-natural languages that virtually every word is polysemic and may be so in numerous ways.1 This phenomenon mainly concerns the more common words whose meaning varies depending on the context rather than the specific terms whose meaning is less subject to variation. In this paper we will describe the phenomenon of polysemy both in its semiotic specificity – linked to the limitlessness of the noetic field of historical-natural languages and to the relevance of the community of speakers – and taking into account the suggestions of contextualist approaches, for which the meaning of linguistic utterances varies systematically depending on the linguistic and extralinguistic context.2

 Polysemy is an ordinary language and ordinary life phenomenon (cf. Nerlich and Clarke 2001).  Using Sperber and Wilson’s (1998, 197) words, polysemy is “the outcome of a pragmatic process whereby intended senses are inferred on the basis of encoded concepts and contextual information”. 1 2

G. Basile (*) Dipsum – University of Salerno, Fisciano, SA, Italy e-mail: [email protected] © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 F. Macagno, A. Capone (eds.), Inquiries in Philosophical Pragmatics, Perspectives in Pragmatics, Philosophy & Psychology 27,



G. Basile

Polysemy as a Semiotic Necessity Polysemy is a phenomenon that is central to the description of how languages function and reveals how closely intertwined semantics and pragmatics are in historical-­ natural languages.3 It should be understood as the coexistence of two or more correlated – through, in Wittgensteinian terms, family resemblances – senses from the same linguistic form (usually a word, but also a morpheme or a grammatical construction)4 and thus, despite the theoretical limitlessness of the lexicon, human beings show a “trend in saving” with regard to the number of lexemes in a given language. Polysemy is an indispensable resource of the economy of languages, which André Martinet (1960) defined as the unstable balance between the ever-­ changing needs of communication and natural human inertia: two essential forces contributing to the optimization of the linguistic system. If each word or each term were connected to a single meaning, we would need an infinite number of words or terms in order to speak, and comprehension would inevitably be compromised; however with a few thousand words we can talk about anything and everything. This is possible because languages reveal two fundamental needs in action: maximum individuation and minimal effort. If the need for maximum individuation acted alone, we would witness the proliferation of millions and millions of different words, each one specific to every different situation or category of entities. However, such situations are impossible because, at the same time, languages are subject to the balancing action of the principle of minimal effort (enunciated by Martinet 1960), according to which human beings consume energy only to the extent necessary in order to achieve the intended goals. In short, in an ideal language (cf. Lyons 1977, 140) every thing, event, fact or even every class or cultural category would have a unique and univocal designation; but, by virtue of the already mentioned principle of economy and since human experiences are characterized by considerable richness and complexity, we cannot have a one-to-one correspondence between signs and referents. In this context polysemy constitutes a naming process that is regular, economic, natural and typical of historical-­natural languages. 3  “Polysemy is a phenomenon that exposes the multiple relations and connections between syntax, semantics and pragmatics, and between language, cognition and social interaction. It can only be studied if we try not to isolate one from another” (Nerlich and Clarke 2003, 16). 4  This kind of approach accounts for the intuition that the different functions of a given item are semantically related, and contemporary allows for the fact that new senses may be created, while others may disappear. The not easy issue, as the literature shows us, is the distinction between monosemy and polysemy, which also involves the issue of how the senses of a polysemic word are represented in our mental lexicon. Cf. the famous debate between Jackendoff (1992) and Fodor (1998) concerning the English verb keep. Ray Jackendoff argues that keep must be polysemous, as it has different meanings in constructions such as keep the money, keep the car in the garage, and keep the crowd happy. Jerry Fodor, for his part, argues in favour of a monosemy account of keep in which it means “keep” in all cases, so that the apparent differences in meaning are simply an artefact of the different contexts in which the verb appears (cf. Falkum and Vicente 2015, 1).

Categorization, Memory and Linguistic Uses


Polysemy is therefore a semiotic necessity and a consequence of both linguistic economy and the limitations of our memory, responding perfectly to the functional needs of speakers, who are thus able to talk about everything and understand each other by using just a few thousand words. It is not, therefore, an irrationality of languages but the fruit of a rationalization, of a tendency to save (cf. De Mauro 19753, 213). Polysemy is also a consequence of the vagueness and indeterminacy of meanings: vagueness takes the form of a semiotic condition, making it possible to extend the boundaries of meaning of each morpheme and to embrace new and unpredictable senses without changing code but only changing the code locally in some points as a function of new impulses to signification to which users are sensitive (cf. De Mauro 19953, 102) and thus in relation to their different habitudines of use and understanding of signs (cf. De Mauro 19953, 61). It is thanks to the vagueness and indeterminacy of meanings that new meanings of the same word can arise, in close relation both to the possibility of enlarging and changing the boundaries of the noetic field of a code, as well as to the the spurs to signification that start from users.5 The vagueness of meanings and the essential “link”, so to speak, with users (who, in Saussurian terms (cf. de Saussure 1959), are internal to the functioning of a language) are the semiotic conditions that make polysemy possible; to these must be added the necessary connection to the context of use in order to define the possible meanings of a word. Context is not something that is added subsequently and from the outside, possibly changing the meaning of any word or phrase but – according to a top-down process (starting from the cognitive context of the speakers who interact in a communicative situation, from their encyclopedic knowledge, from their tacitly shared knowledge and from the recognition of their communicative intentions) – becomes a necessary component of their meaning. Languages can therefore be said to be creative codes characterized by flexibility and semantic dilatability and capable of guaranteeing their adaptability to heterogeneous changes and situations.6 A useful concept in this regard is adaptedness, which is typical of the epistemology of biology and states that the history of life is characterized by a trend towards a growing adaptive complexity resulting in the appearance of more complex organisms than the previous ones (cf. Pievani 2005, 207).

5  Users, both speakers and listeners, are the, so to speak, “epistemic guarantees” of the possibility of creating new senses of the same word. On this see Carston (2019, 161): “much polysemy originates in the process of communication: speakers use existing words to express new senses and coin new words, and hearers are able to pragmatically infer the ad hoc senses and words intended”. 6  Cf. Gilles Fauconnier and Mark Turner, according to which polysemy “is not just an accident of history or of synchrony, but rather an essential manifestation of the flexibility, adaptability, and richness in meaning potential that lie at the very heart of what a language is and what it is for. It is also a symptom (rather than a primitive component) of the way in which various cognitive operations allow for creativity at many levels” (Fauconnier and Turner 2003, 80).


G. Basile

Perspectives on the Study of Polysemy The problem of polysemy has elicited a number of different responses. It has been discussed in historical terms in an attempt to identify a common etymon for polysemic senses and, more recently, it has been described in the framework of structural semantics where polysemy exists if there is at least one common semantic component among the various meanings (without reference to contexts). Some scholars have given an explanation based on semantic-logical derivation principles, identifying a basic meaning as well as other meanings derived by extension, narrowing, etc., while other scholars have preferred an explanation based on syntactic-distributive criteria, and so on. Numerous contributions have been made in the field of cognitive semantics. In summary, in the context of cognitive linguistics, talking about polysemy implies: (i) viewing meaning/sense as categorizazion, (ii) recognizing the importance of context for meaning/senses and that linguistic and encyclopedic knowledge are hard to keep separate, and (iii) incorporating prototype theory into linguistics (Gries 2015, 472).

In recent years, moreover, research in cognitive semantics has shown that the meanings of polysemic words are motivated in part by our metaphorical and metonymic structuring of experience and can be explained in terms of some basic conceptual and cultural metaphors (cf., among others, Norrick 1981; Gibbs 1994, 40–41).7 Words do not accumulate meanings randomly but follow certain paths or patterns that are natural to human cognition and whose structure determines our acquisition and change of experience, knowledge and language. The correspondences between the properties of polysemy and the structures of human cognition mean that polysemy can provide crucial information on basic cognitive processes (cf. Deane 1988, 327; Cuyckens and Zawada 1997). The study of polysemy can therefore be considered – in our opinion – as an “exceptional observatory” in order to understand how human cognition works.8 In this regard Cecil H. Brown and Stanley R. Witkowski say: Several studies have identified polysemous equations that occur widely in languages […]. While documentation of uniform tendencies in polysemous naming is of interest itself […].

7  Of particular interest to linguists has been the interrelation of the various senses of a polysemous lexeme, that gives raise to chains of senses, in which sense a is related to sense b by virtue of some shared attribute(s), sense b is related to sense c, which is related to sense d and so on. As pointed out by Cuyckens et al.: “One of the major issues in cognitive lexical semantics over the past two decades has been the analysis of polysemous lexical items in terms of a family-resemblance network of multiple, interrelated senses of usage types. [...] The links between the different senses in a lexical network are manifold (conceptual/semantic overlap, metaphor, metonymy, image-schema transformation) and are supposed to represent the cognitive principles behind the processes of meaning extension” (Cuyckens et al. 1997, 35). 8  This assumption not only reveals that linguistic analysis can shed light on aspects of the mind, but also underlies “that language cannot be insulated from general cognition and that linguistic analysis should therefore be informed by cognitive principles” (Cuyckens and Zawada 1997, 36).

Categorization, Memory and Linguistic Uses


Such frameworks can help delineate shared processes underlying human cognition (Brown and Witkowski 1983, 72),

and furthermore: Polysemy is ubiquitous in language and its investigation has considerable potential for illuminating human cognition. In addition, the regular patterns of lexical change […] indicate that the lexicon is amenable to systematic investigation as are other components of language. Most importantly, the study of these regular lexical patterns can contribute significantly to knowledge of the processes and capacities which underlie human language and culture [italics added] (Brown and Witkowski 1983, 84).

As examples of regular patterns of lexical change, we can see some patterns of historically and culturally determined polysemy,9 such as concrete-abstract (for example head “the upper part of the human body of an animal, typically separated from the rest of the body by a neck, and containing the brain, mouth, and sense organs” vs “an aptitude for or tolerance of: she had a good head for business; a head for heights”); part of the body-object (for example tooth “each of a set of hard, bony enamel-coated structures in the jaw of most vertebrates, used for biting and chewing” vs “a projecting part on a tool or other instrument, especially one of a series that function or engage together, such as cog on a gearwheel or a point on a saw”); animal-meat butchered or cooked (for example lamb “a young sheep” vs “the flesh of a lamb as food”) etc. based on metonymical processes.10 In the context of pragmatic studies, important contributions start from the attempt of Geoffrey Nunberg to ask the question in a pragmatic rather than a semantic perspective: roughly speaking, homonymy is a phenomenon of meaning, while polysemy is a phenomenon concerning pragmatics. For Nunberg, a pragmatic consideration of polysemy allows us to explain how a name or general term can be used to refer to something in the absence of a linguistic convention for doing so (and as such, it will perforce be an account of those metaphorical word-uses that are not judged normal or acceptable, as well) (Nunberg 1979, 154).

9  Cf. Mahesh Srinivasan and Hugh Rabagliati who have investigated a large number of systematic polysemy patterns in English (e.g. animal for meat, material for artifact, container for content, etc.), and asked whether and how these were manifested cross-linguistically. They explored whether 27 distinct patterns of polysemy found in English are also present in 14 other languages and the results showed low variability regarding the presence of the polysemy patterns investigated across languages. However, with regard to the specific senses of the patterns that were istantiated across languages, they found considerable variability: for instance the pattern material for artifact is cross-linguistically present, but it is carried out in different ways in different languages: the linguistic form for “glass”, for example, refers to a drinking vessel in English, a car window in Spanish and a mirror in Russian. This demonstrates that languages develop different senses (conventions) but that there are commonalities in patterns and senses across languages (cognitive biases) (cf. Srinivasan and Rabagliati 2015; Falkum and Vicente 2015). 10  Metonymy can be considered as a regular source of polysemy “which results when a particular metonymic usage becomes entrenched and conventionalized” (Langacker 2008, 70). See, for example, the case of church that can profile either a building used for religious meetings or a religious organization that meets in such buildings (ibid.).


G. Basile

Moreover – says Nunberg – linguists have postulated that “words have ‘meanings-­ in-­the-language’ because it seemed necessary to do so, in order to explain how novel utterances are understood”, but it is too strong a hypothesis that “would force us to make assumptions about our language over and above what simple understanding would require. We do not have to know what a word names to be able to say what it is being used to refer to” (Nunberg 1979, 177).

Lexical Pragmatics and Attention to Use In recent years (since the 1980s) the studies of lexical pragmatics have spread and multiplied in order to explain how “linguistically-specified (‘literal’) word meanings are modified in use” (Wilson 2003, 273). The field of lexical pragmatics11 – which is our theoretical background – “specifically studies the interaction between an expression’s linguistically-encoded meaning and aspects of the context” (Falkum and Vicente 2015, 10). In particular, according to the theorists of relevance, languages are “faulty” codes since there is almost never a biunique correspondence between signal (expression) and content: ambiguities, polysemy, deixis, referential indeterminacy, lexical and grammatical differences are an insurmountable obstacle to communication, which cannot be entrusted to processes of encoding / decoding, however complex they might be.12 For this reason, a sentence is not interpreted by activating decoding strategies, but is understood as a clue from which it is possible to trace the intentions of the sender. In this perspective, the inferential dimension pervades the entire process of communication and – in particular – it plays a fundamental role in the development and proliferation of polysemy in verbal communication. An approach of this kind is not particularly fruitful in meanings already determined or fixed by use, i.e. systematic,13 conventionalized, present in our mental lexicon recorded in dictionaries, such as the entry to freeze meaning both “(with reference to a liquid) to turn or be turned into ice or another solid as a result of extreme cold: e.g. in the winter the milk froze” and “store (something) at a very low temperature in order to preserve it: e.g. the cake can be frozen”, and “become suddenly motionless or paralysed with fear or shock: e.g. she froze in horror”, and “hold (something) at a fixed level for a period of time: e.g. new spending on defence  It is a field of research that aims at giving a systematic explanation of pragmatic phenomena related to the semantic sub-specification of lexical elements (cf. Blutner 1998). 12  In the relevance-theoretic framework polysemy “is treated as a mainly communicative phenomenon, which arises as a result of lexical concepts underdetermining the situation-specific concepts that are comunicated by them, as part of hearers’ search for optimal relevance in the process of utterance interpretation” (Falkum and Vicente 2015, 11). 13  The first who speaks of regular polysemy is Apresjan (1974), who defines regular (or systematic) the polysemy of a word if there is at least one other word with a common pattern. 11

Categorization, Memory and Linguistic Uses


was to be frozen” etc.; nor is it useful in meanings belonging to technical languages, as in the case of kinesis “in biology: an undirected movement of a cell, organism, or part in response to an external stimulus” vs “in zoology: mobility of the bones of the skull, as in some birds and reptiles”, that do not pose problems of interpretation, as they select different and more predictable contexts of use. On the other hand it may have a strong heuristic value in cases of lexical-­ pragmatic processes such as the widening and narrowing of meaning, metaphorical extension, etc., usually studied separately, which were included by the theorists of pertinence (cf., among others, Carston 2002; Wilson and Sperber 2002) in a single investigative approach aimed at accounting for the fact that “the concept communicated by the use of a word often differs from the concept encoded” (Wilson 2003, 274). These are cases of words used in everyday language and which are subject to a process of lexical modulation in which their meaning is less predictive “in the absence of a linguistic convention for doing so”, as argued by Nunberg (cf. par. 2) and is reviewed and reformulated at local level, forcing the receiver to necessarily proceed through inferences and to review and recalculate the meaning of the whole sentence (cf., for instance, Ervas 2012). In such cases it is crucial the communicative aspect of polysemy, that is, the interaction between linguistically-encoded content and contextual information to understand the speaker-intended lexical meanings. Consider some of these cases: (a) Extension of lexical category: cases in which a word is used in a new lexical category; for example He is pianoing Christmas songs, where the noun piano takes the verbal form pianoing and means “he is playing the piano” (cf. Bushnell and Maratsos 1984; Clark 1982); (b) Approximation: cases of so-called interpretive broadening, in which the interpretation of a word with a precise meaning is extended to indicate a vague use, as in the case of the vague uses of numbers (for example a thousand used with unspecified meaning to refer to a large number or a large quantity: I’ve told you a thousand times!, Today I have a thousand things to do), or of geometric terms (e.g. square to refer to a figure, surface, etc., of a more or less square shape: Give me a small square of chocolate). In this regard, Robyn Carston speaks of broadening when the semantic field of the term widens, passing from a subset of its possible senses to a larger whole,14 as it does for example in the case of crazy in That boy is crazy, in which we do not mean that the boy in question has a particular psychiatric pathology but that he has a strange, bizarre or unwise

 Among the examples reported by Carston cf. There is a rectangle of lawn at the back, to refer not to a rectangle in the proper sense, with four right angles and with the two opposite sides alike, but to a lawn that is roughly a rectangle, or even This steak is raw, to imply that the steak, presumably served in a restaurant, is not really raw but is less cooked than the speaker wants and so on. In these cases “the concepts encoded […] are generally agreed to have clear category borders as a matter of their semantics, and what is going on is that speakers are using them loosely for a particular comunicative purpose” (Carston 2002, 329, see also Carston 1997).



G. Basile

behaviour. The semantic field of crazy is thus modulated and expanded in the more general sense of strange. In such cases, as Carston points out, the lexically encoded concept in the logical form of the utterance is replaced by an ad hoc concept, pragmatically derived from the lexical one, and […] this new non-lexicalized concept is a constituent of the proposition expressed by the speaker of the utterance (Carston 2002, 28).15

(c) Narrowing of meaning: cases in which a word is used to transmit a narrower interpretation than that semantically codified; in this regard we speak of narrowing16: for example, to drink meaning “drink wine or alcohol” in a sentence like John drinks like a fish,17 to sleep meaning “be inactive or dormant” in a sentence like Instead of getting a job you sleep18 etc. (d) Unconventional metaphors: cases such as John in our family is a street light, in which the meaning constitutes a metaphorical use of the word street light (more or less with the meaning of “reference person, who acts as a shining beacon for the others”), inferred from the physical properties of the street light (“a light illuminating a road, typically mounted on a tall post”19) and considering the context of enunciation. (e) Unconventional metonymy: more extemporaneous cases of metonymy (cf. Falkum, Recasens, and Clark 2017), for example in the case where a waiter in a café says to the person at the cash desk, The ham sandwich asked for the bill, to refer, obviously, not to a ham sandwich but to the customer who ordered a ham sandwich, wants to pay and has asked for the bill.20 In the cases of broadening and narrowing (but also in the others) the interpretive and active process of both the issuer and the receiver21 comes into play, since they are “outcomes of a single interpretive process which fine-tunes the interpretation of

 Carston recently returned to this subject saying that “some ad hoc senses become frequently used and may eventually be fully conventionalised, hence ‘semantic’ in that they can be retrieved directly from the lexicon […] and so are potentially input to pragmatic processes of meaning modulation” (cf. Carston 2019, 52). 16  Cfr. Carston (Carston 2002, 333): “Cases of concept narrowing contribute to the propositional form of the utterance; that is, enrichment is taken to be one of those pragmatic processes, along with reference assignment and disambiguation, that are involved in arriving at an explicitly communicated assumption, an explicature”. 17  Cf. the Italian bere, the German trinken, the French boire and the Spanish beber which behave similarly to English. 18  Cf. the Italian dormire, the German schlafen, the French dormir which behave similarly to English. 19  On this subject cf. Falkum (2015, 84); cf. also Deamer (2013) and Özçalişkan (2005). 20  In this regard Nunberg (1995) talks about transfers of meaning. Cf. also Jacob Mey, according to whom “my utterance, in its final analysis, receives its meaning not only from what I put into it, but to an equally high degree from what the other gets out of it” (Mey 1993, 236). 21  In particular, as receivers, “we aim at specifying meaning, at selecting, out of the many possible intepretations, one interpretation that coheres with our background default assumptions and other contextual considerations [italics in text]” (Kittay 1989, 80). 15

Categorization, Memory and Linguistic Uses


almost every word [italics in text]” (Wilson and Carston 2007, 231). These are meaningful lexical paths also from a pragmatic-cognitive point of view; in fact, polysemy is seen as “variation in the construal of a word on different occasions of use” (Croft and Cruse 2004, 109), and the meaning, linked to construal operations similar to those that govern conceptualization, is seen as something that is dynamic. A cognitive-pragmatic perspective  – in accordance with what Ludwig Wittgenstein argues in Philosophical Investigations (“One cannot guess how a word functions. One has to look at its use and learn from that [italics in text]” -Wittgenstein 1958, 340) – seems the most fruitful to account for the dynamic between lexical meaning and its context of use. Semantic analysis alone cannot fully recover utterance meaning and pragmatic enrichment is needed to complete this process (cf. Jaszczolt 2010, 458–62), bypassing the strictly semantic dimension and showing how fuzzy the boundaries between semantics and pragmatics are.22 In short, “what a word means often depends on the context in which it is used, and people pick up much of their vocabulary knowledge from context” and “contextual variation in meaning is pervasive in natural language, and much of this variation is irregular and/or language specific [italics added]” (Nagy 1995, 1).

Some Proposals for a “Dynamic” Interpretation of Polysemy We can say that polysemy is a sort of litmus test that shows us that a logical or semiological science without a subject is impossible (cf. De Mauro 19953, 64). And in this context polysemy is structured as a principle of cognitive organization23 that acts on the lexicon in synergy – especially in cases that we can define as context-­ driven (as in the examples seen above John in our family is a street light and The ham sandwich asked for the bill) – with the variability of the context of enunciation. Let us take another example: the adjective light in contexts such as you’re light as a feather (“of little weight; not heavy”), her eyes are light blue (“pale”), passenger traffic was light (“relative low in density, amount, or intensity”), stick to a light diet (“low in fat, cholesterol, sugar, or other rich ingredients”), a light Italian wine (“not strongly alcoholic or heavy on the stomach”) etc. The question that arises is whether each of these meanings constitutes an autonomous meaning of light or whether they should be understood under a single and generic meaning as “that produces a less intense physical-perceptive effect”.

 Cf., among others, Jaszczolt (2010). Katarzyna M. Jaszczolt, in particular, speaks of a Default Semantics, according to which a representation of utterance meaning is created as a merger of the output of a variety of linguistic and non-linguistic sources (cf. Jaszczolt 2005). 23  Cf. George Lakoff who sees polysemy as a consequence of our cognitive organisation: “Polysemy arises from the fact that there are systematic relationships between different cognitive models and between elements of the same model. The same word is often used for elements that stand in such cognitive relation to one another” (Lakoff 1987, 13). 22


G. Basile

In cases like these, it is as if we are faced with some sort of pre-meaning (Croft and Cruse 2004, 110), a complex semantic potential (cf. Violi 2003, 331), which is upstream from the various interpretations that are fully specified once the context of application is given. As hearers, “we aim at specifying meaning, at selecting, out of the many possible interpretations, one interpretation that coheres with our background default assumptions and other contextual considerations [italics in text]” (Kittay 1989, 80). Usually, the speaker selects one of the senses of a polysemic word and the hearer automatically and without problems disambiguates polysemic words in and thanks to the context (cf. Nerlich and Clarke 2001, 4). The question that arises at this point is whether or not these cases (which Ullmann appropriately defines shifts in application – see Ullmann 1962, 159–61) are discernible from actual polysemy: it is indeed difficult to find a criterion by which the two phenomena can be clearly distinguished, and none of those proposed is decisive.24 Therefore, most scholars – including Ullmann (1957) and later Allan (1986, 153), Cruse (1986, 71), Tuggy (1993), Geeraerts (1993), Brisard et al. (2001) etc. – believe that this distinction, like the one between polysemy and homonymy, should be conceived not in a dichotomous but in a “fluid” way, i.e. in terms of a continuum (see also Lehrer 1990, 208; Geeraerts 1993) in which polysemy (i.e. the existence of several correlated but distinct meanings) represents the intermediate area between the extremes of homonymy (in which there are several unrelated meanings) and semantic modulation (in which there are more extensive senses related to a single meaning). We propose the following scheme: _______________________________________________________________________  homonymy 1

page “one or both sides intense of paper in a book, magazine etc.” 2 page “a boy or young man employed in a hotel or club to run errands, open doors, etc.”

polysemy arrest “the action of seizing someone and taking him into custody” and “a stoppage or sudden cessation of motion”

extensions of a monosemic meaning light “that produces a less physical-perceptive effect”: you’re light as a feather, her eyes are light blue, passenger traffic was light etc.

The representation in terms of a continuum makes it possible – in our opinion – to account for the fact that there is a gradation between: (i) cases of full polysemy, in which the meaning of a lexeme is articulated in more stable and lexicalized meanings (such as arrest in the meanings of “the action of seizing someone and taking him into custody” and “a stoppage or sudden cessation of motion”, or cases of technical-specialist meanings such as moment “a very brief period of time; an exact point in time” vs “(in Physics) a turning effect produced by a force acting at a distance on an object”); (ii) cases where a meaning has extensive uses that are so widespread and established as to be recorded as such in lexicographical resources (for example   A critical examination of the criteria for distinguishing polysemy and vagueness is in Geeraerts (1993).


Categorization, Memory and Linguistic Uses


cloud in the meaning of “an indistinct or billowing mass, especially of smoke or dust”); (iii) cases in which a meaning can assume particular nuances and even entirely casual and contingent meanings in specific contexts, as happens for example when we use cloud in reference to anything that recalls the shape or look of a cloud (for example a cloud of lace, a cloud of white feathers, the almond was a cloud of flowers and perfumes, a cloud of colors etc.), even if we must be aware that we may encounter borderline cases often difficult to classify as, say, polysemy vs. context-conditioned variation, or, alternatively, as polysemy vs. homonymy (cf. Taylor 2003, 35). This type of representation can be said to account for the mobile boundaries of words and the fact that “there are always occasions when one wants to say something new, interesting, subjective, which has never been said before” (Nerlich and Clarke 2001, 6). We advance the hypothesis that there is a sort of lexical adjustment between what is given semantically, syntactically, and what we infer from the surrounding pragmatic context of discourse.25 It is a process which is by its nature dynamic and for the most part approximate within a so-called polysemic continuum, so that we go from the more autonomous and stabilized senses in the langue (and thus encoded in dictionaries, as we have seen in such cases as to freeze, kinesis or arrest) to the more extemporaneous ones totally dependent on the context (context-driven),26 i.e. linked to the particular context of enunciation (as in the case of cloud).27 In short, there is an oscillation from highly lexicalized cases to completely pragmatic ones (cf., in this perspective, Nunberg 1992, 393). In any case there is a process of modulation – as we have seen in the cases of broadening and narrowing  – linked to an objective of economy of expression in which the speaker, instead of facing the laborious task of completely codifying the lexical meaning that he has in mind – a thing that he could do if he wanted to – is often led to choose a cheaper form of expression out of a sort of pragmatic relaxing (cf. Lasersohn 1999, 522).

 Since, generally speaking, “the sense of an expression is generated through the interaction of that expression’s meaning with the context, both linguistic and extralinguistic” (Recanati 1997, 107). 26  The problem – as Richard C. Anderson and Andrew Ortony argue – “is not with the macro-distinctions embodied in homonyms but with micro-distinctions, like ball in The quarterback threw the ball, Arnold Palmer lost his ball, The guard dribbled the ball and so on” (Anderson and Ortony 1975, 177). 27  As Jarno Raukko has pointed out, it is important “to see polysemy as patterns of flexibility in (lexical) meaning in much the same way as it is accepted that situational (utterance or discourse level) meaning is not fixed, inexact and negotiable” (Raukko 2003, 161). 25


G. Basile

Conclusions As we have seen, the overwhelming majority of polysemic words in the lexicon of historical-natural languages can shed some light on our categorization of objects, events, properties, etc. of reality, as well as on the way the enérgeia of these words activates and organizes our memory and guides us in linguistic uses and mutual understanding. In doing so the context (both linguistic and extralinguistic) plays an essential role, to the extent that we can say that historical-natural language is ineliminably context-sensitive. Using a slogan proposed by Charles Travis we can say that “content is inseparable from point” (Travis 2006, 33). Words have no meaning per se (as by virtue of a virtus significativa inherent to them – cf. De Mauro 19753, 10), but “what is communicated in our words lies, inseparably, in what we would expect of them. How our words represent things is a matter of, and not detachable from, their (recognizable) import for our lives” (Travis 2006, 33). And our lives are made up of continuous interactive exchanges where the speaker relies on implicit cooperation and understanding with the listener, on the basis of a common background of shared knowledge, such that the listener is led “to pragmatically infer her intended lexical meaning” (Falkum 2015, 94). In this epistemological context we can therefore understand polysemy as a fundamentally communicative phenomenon, which arises (mainly) as a result of a merging between encoded lexical concepts (which underdetermine the speaker’s intended concepts) and the necessary and constant recourse to our pragmatic inferential capacity, which is closely related to our capacity to attribute mental states (specifically, communicative intentions) to others. Among the speakers there is a continuous negotiation of multiple meanings so that, according to Brigitte Nerlich and David D. Clarke, we can conclude that the three general functions of polysemy are to inject the language you use with subjectivity […]; to jointly reappropriate the language we use as a shared, inter-subjective system for the expression of meaning, and (as a hidden result of these processes); to remotivate the language we use so that we can go on using it for ever new comunicative purposes [italics in text] (Nerlich and Clarke 2001, 14).

References Allan, K. (1986). Linguistic meaning. London, UK: Routledge & Kegan Paul. Anderson, R., & Ortony, A. (1975). On putting apples into bottles  — A problem of polysemy. Cognitive Psychology, 7(2), 167–180. Apresjan, Ju. D. (1974). Regular polysemy. Linguistics, 12(142). ling.1974.12.142.5. Blutner, R. (1998). Lexical pragmatics. Journal of Semantics, 15(2), 115–162.

Categorization, Memory and Linguistic Uses


Brisard, F., Van Rillaer, G., & Sandra, D. (2001). Processing polysemous, homonymous, and vague adjectives. In H.  Cuyckens & B.  Zawada (Eds.), Polysemy in cognitive linguistics (pp. 261–284). Amsterdam, Netherlands/Philadelphia, PA: John Benjamins. Brown, C. H., & Witkowski, S. R. (1983). Polysemy, lexical change and cultural importance. Man, 18(1), 17–89. Bushnell, E., & Maratsos, M. (1984). Spooning and basketing: Children’s dealing with accidental gaps in the lexicon. Child Development, 55(3), 893–902. https://doi. org/10.1111/j.1467-8624.1984.tb03826.x. Carston, R. (1997). Enrichment and loosening: Complementary processes in deriving the proposition expressed? In Pragmatik. Implikaturen und sprechakte (Vol. 8, pp. 103–127). Wiesbaden, Germany: Springer. Carston, R. (2002). Thoughts and utterances: The pragmatics of explicit communication. Oxford, UK: Blackwell Publishing Ltd. Carston, R. (2019). Ad hoc concepts, polysemy and the lexicon. In K. Scott, B. Clark, & R. Carston (Eds.), Relevance, pragmatics and interpretation (pp. 150–162). Cambridge, UK: Cambridge University Press. Clark, E. V. (1982). The young word maker: A case study of innovation in the child’s lexicon. In W. Eric & G. Lila (Eds.), Language acquisition: The state of the art (pp. 390–425). Cambridge, UK: Cambridge University Press. Croft, W., & Cruse, A. (2004). Cognitive linguistics. Cambridge, UK: Cambridge University Press. Cruse, D. A. (1986). Lexical semantics. New York: Cambridge University Press. Cuyckens, H., & Zawada, B. (Eds.). (1997). Polysemy in cognitive linguistics. Amsterdam, Netherlands: John Benjamins. Cuyckens, H. et al. (1997). Towards an Empirical Lexical Semantics. In: B. Smieja & M. Tasch (eds.), Human Contact through Language and Linguistics (pp. 35–54). Wiesbaden, Germany: Peter Lang. Deamer, F. (2013). An investigation into the processes and mechanisms underlying the comprehension of metaphor and hyperbole. London, UK: University College London (University of London). Deane, P. (1988). Polysemy and cognition. Lingua, 75(4), 325–361. https://doi. org/10.1016/0024-3841(88)90009-5. De Mauro, T. (19753). Introduzione alla semantica. Roma-Bari, Italy: Laterza.  De Mauro, T. (19953). Minisemantica dei linguaggi non verbali e delle lingue. Roma-Bari, Italy: Laterza. Ervas, F. (2012). Polisemia e ambiguità lessicale nella traduzione. In S. Ferreri (Ed.), Lessico e lessicologia. Proceedings of the XLIV international congress of studies of the Italian Linguistics Society (SLI) (Viterbo, 27–29 Sept. 2010) (pp. 295–306). Roma, Italy: Bulzoni. Falkum, I. (2015). The how and why of polysemy: A pragmatic account. Lingua, 157, 83–99. Falkum, I., & Vicente, A. (2015). Polysemy: Current perspectives and approaches. Lingua, 157, 1–16. Falkum, I., Recasens, M., & Clark, E. (2017). ‘The moustache sits down first’: On the acquisition of metonymy. Journal of Child Language, 44(1), 87–119. S0305000915000720. Fauconnier, G., & Turner, M. (2003). Polysemy and conceptual blending. In B. Nerlich, Z. Todd, V. Herman, & D. Clarke (Eds.), Polysemy: Flexible patterns of meaning in mind and language (pp. 79–94). Berlin, Germany/New York, NY: Mouton de Gruyter. Fodor, J. (1998). Concepts: Where cognitive science went wrong. Oxford: Oxford University Press. Geeraerts, D. (1993). Vagueness’s puzzles, polysemy’s vagaries. Cognitive Linguistics, 4(3), 223–272. Gibbs, R. (1994). The poetics of mind: Figurative thought, language and understanding. Cambridge, UK: Cambridge University Press.


G. Basile

Gries, S. (2015). Polysemy. In E. Dąbrowska & D. Divjak (Eds.), Handbook of cognitive linguistics (pp. 472–490). Berlin, Germany/Boston, MA: Mouton de Gruyter. Jackendoff, R. (1992). What is a concept, that a person may grasp it? In R.  Jackendoff (Ed.), Languages of the mind: Essays on mental representation (pp.  21–52). Cambridge, MA: MIT Press. Jaszczolt, K. (2005). Default semantics: Foundations of a compositional theory of acts of communication. Oxford, UK: Oxford University Press. Jaszczolt, K. (2010). Semantics-pragmatics interface. In L.  Cummings (Ed.), The pragmatics encyclopedia (pp. 458–462). London, UK: Routledge. Kittay, E. F. (1989). Metaphor: Its cognitive force and linguistic structure. Oxford, UK: Oxford University Press. Lakoff, G. (1987). Women, fire, and dangerous things. Chicago, IL: University of Chicago Press. Langacker, R. (2008). Cognitive grammar: A basic introduction. Oxford, UK: Oxford University Press. Lasersohn, P. (1999). Pragmatic halos. Language, 75, 522–551. Lehrer, A. (1990). Polysemy, conventionality, and the structure of the lexicon. Cognitive Linguistics, 1(2), 207–246. Lyons, J. (1977). Semantics (Vol. 1). Cambridge, UK: Cambridge University Press. Martinet, A. (1960). Eléments de linguistique générale. Paris, France: Librairie Armand Colin. Mey, J. (1993). Pragmatics: An introduction. Oxford, UK: Blackwell Publishers. Nagy, W.  E. (1995). On the role of context in first-and second-language vocabulary learning. Technical report no. 627. Champaign, IL: Center for the Study of Reading. Nerlich, B., & Clarke, D. (2001). Ambiguities we live by: Towards a pragmatics of polysemy. Journal of Pragmatics, 33(1), 1–20. Nerlich, B., & Clarke, D. (2003). Polysemy and flexibility: Introduction and overview. In B. Nerlich, D. Clarke, Z. Todd, & V. Herman (Eds.), Polysemy: Flexible patterns of meaning in mind and language (pp. 3–30). Berlin, Germany/New York, NY: Mouton de Gruyter. Norrick, N. (1981). Semiotic principles in semantic theory. Amsterdam: John Benjamins. Nunberg, G. (1979). The non-uniqueness of semantic solutions: Polysemy. Linguistics and Philosophy, 3(2), 143–184. Nunberg, G. (1992). Systematic polysemy in lexicology and lexicography. In H.  Tommola, K. Varantola, T. Salmi-Tolonen, & J. Schopp (Eds.), Proceedings of the 5th EURALEX international congress (pp. 386–396). Tampere, Finland: Tampereen YIiopisto. Nunberg, G. (1995). Transfers of meaning. Journal of Semantics, 12(2), 109–132. https://doi. org/10.3115/981574.981599. Özçalişkan, Ş. (2005). On learning to draw the distinction between physical and metaphorical motion: Is metaphor an early emerging cognitive and linguistic capacity? Journal of Child Language, 32(2), 291–318. Pievani, T. (2005). Introduzione alla filosofia della biologia. Roma-Bari, Italy: Laterza. Raukko, J. (2003). Polysemy as flexible meaning: Experiments with English get and Finnish pitää. In B. Nerlich, Z. Todd, V. Herman, & D. Clarke (Eds.), Polysemy: Flexible patterns of meaning in mind and language (pp. 161–193). Berlin, Germany: Mouton de Gruyter. Recanati, F. (1997). La Polysémie Contre Le Fixisme. Langue Française, 107–123. de Saussure, F. (1959). In W.  Baskin (Ed.), Course in general linguistics. New  York, NY: Philosophical Library. Sperber, D., & Wilson, D. (1998). The mapping between the mental and the public lexicon. In P.  Carruthers & J.  Boucher (Eds.), Language and thought: Interdisciplinary themes (pp. 184–200). Cambridge, UK: Cambridge University Press. Srinivasan, M., & Rabagliati, H. (2015). How concepts and conventions structure the lexicon: Cross-linguistic evidence from polysemy. Lingua, 157, 124–152. lingua.2014.12.004. Taylor, J. (2003). Cognitive models of polysemy. In B. Nerlich, D. Clarke, Z. Todd, & V. Herman (Eds.), Polysemy: Flexible patterns of meaning in mind and language (pp.  31–47). Berlin, Germany/New York, NY: Mouton de Gruyter.

Categorization, Memory and Linguistic Uses


Travis, C. (2006). Thought’s footing: A theme in Wittgenstein’s philosophical investigations. Oxford, UK: Oxford University Press. Tuggy, D. (1993). Ambiguity, polysemy, and vagueness. Cognitive Linguistics, 4(3), 273–290. Ullmann, S. (1957). The principles of semantics. New York, NY: Philosophical Library. Ullmann, S. (1962). Semantics: An introduction to the science of meanings. Oxford, UK: Basil Blackwell. Ullmann, S. (1963). Semantic universals. In J.  Greenberg (Ed.), Universals of language (pp. 172–208). Cambridge, MA: MIT Press. Violi, P. (2003). Significati lessicali e pratiche comunicative. Una prospettiva semiotica. Italian Journal of Linguistics, 15, 321–342. Wilson, D. (2003). Relevance and lexical pragmatics. Italian Journal of Linguistics, 15, 273–292. Wilson, D., & Carston, R. (2007). A unitary approach to lexical pragmatics: Relevance, inference and ad hoc concepts. In N. Burton-Roberts (Ed.), Pragmatics (pp. 230–259). Basingstoke, UK: Palgrave. Wilson, D., & Sperber, D. (2002). Truthfulness and relevance. Mind, 111(443), 583–632. https:// Wittgenstein, L. (1958). Philosophical investigations. Oxford, UK: Basil Blackwell.

Interpretation of Copredicative Sentences: A Rich Underspecification Account of Polysemy Marina Ortega-Andrés

Introduction Polysemy is a very common phenomenon in language. Actually, most of the open-­ class words we normally use are polysemous: they are associated with several related senses. A consequence of the breadth of the term is that linguists and philosophers have distinguished between many different kinds of polysemy. Consider the following examples: (1) (a) The school caught fire. (b) The school starts at 8 am. (2) (a) The book is too thick (b) The book is very interesting. (3) (a) Sally is a chameleon. (b) The common chameleon is also found on the peninsular area. The word school means different things in (1a) and (1b): while in (1a) it refers the building, in (1b) it refers to the time-schedule. The word book in (2a) refers the physical object, while in (2b) it refers to the informational content. In (3), the word chameleon also means two different things: in (3a), it predicates a particular M. Ortega-Andrés (*) University of the Basque Country, Leioa, BI, Spain e-mail: [email protected] © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 F. Macagno, A. Capone (eds.), Inquiries in Philosophical Pragmatics, Perspectives in Pragmatics, Philosophy & Psychology 27,



M. Ortega-Andrés

property of Sally -that she is very good at changing her personality and adapting depending on the context-, while in (3b) the word chameleon refers to an animal. All these uses of the words are conventional1 senses of regular2 polysemous words. The words book and school are inherent polysemous, that is: they typically allow copredication; while the word chameleon is not inherent polysemous, it does not typically copredicate. Copredication is the phenomena in which the same polysemous nominal expression comes along with simultaneous predications for two (or more) different meanings or senses of the word in a sentence. Here there are some examples of copredicative sentences: (4) The books are thick and interesting. (5) The school caught fire and was celebrating 4th of July when the fire started. In (4), the word books refers to the physical object, but also the content or information that it transmits. In (5), the word school refers to the building but also the people inside the building. Some empirical investigations show that there is a difference between how inherent polysemy, other kinds of conventional polysemy and homonymy3 are represented in the lexicon (see Frisson and Pickering 2001; Frisson 2009; Klepousniotou 1  The distinction between conventional and non conventional polysemy is similar to the distinction between semantic polysemy and pragmatic polysemy (see Gibbs and Gerrig 1989; Falkum 2011). It is based on the idea that conventional senses are those that have been lexicalized or encoded in the mental lexicon after a process of conventionalization (Carston 2015). Once they have been lexicalized, they only have to be selected from a list of words when the interpreter encounters the word. On the contrary, when the hearer encounters a new sense of a word, it has to be pragmatically generated (Carston 2013; Recanati 2004). Following this idea, the label “conventional polysemous word” is not essential to the words themselves: most open-class words are conventionally polysemous because they have many different related and conventional senses. 2  According to Apresjan (1974) the polysemy of a word a with the senses Ai and Aj is regular if there exists at least one word b with the related senses Bi and Bj, being semantically distinguished in exactly the same way as Ai and Aj. For example:

( a) The manifesto was signed by the University (b) I have a meeting with Laura at the university. (c) The school caught fire. (d) The school celebrates the end of the year party tomorrow. In (a) the word University has the meaning “people that work or represent an University- institution”; while in (c), school means the group of people that work/represent the school-institution. In (b), university means the building of the university and in (d) school also means the building of the school. Those are different regular polysemous words that have two related senses that are related in the same way. 3  The distinction between polysemy and homonymy has been widely discussed by linguists, philosophers and psychologists. Traditionally, the main difference is that senses or meanings of homonymous words are not related (apart from historically), while senses of polysemous words are related (see Falkum 2011). Most of dictionaries differentiate polysemy and homonymy by introducing them in different entries. For example: Bank1: (a) slope of land that borders water. Bank2: (a) an institution; (b) a building.

Interpretation of Copredicative Sentences…


and Baum 2007; Klepousniotou et al. 2008; Pickering and Frisson 2001; Pickering et al. 2006; Pylkkänen et al. 2006). It seems that related senses of inherent polysemous words like book, school, etc. are more closely related than other senses, and that this relation affects the way they are accessed. This fact has been used for claiming that related senses of a polysemous word share a representation, while meanings of homonymous words are stored in different representations. However, there is no agreement about this conclusion. The reason is that while some studies argue in favor of Separate Sense Storage (Katz 1972; Klein and Murphy 2002; Foraker and Murphy 2012) –which claims that different senses are different representations in the lexicon-, others have presented some evidence for the idea there is a “single representation” for different senses of a polysemous word in the mental lexicon (Beretta et al. 2005; Frazier and Rayner 1990; Frisson and Pickering 2001; Pickering and Frisson 2001; Pylkkänen et al. 2006). One Representation Models propose that different senses of some polysemous words are lexicalized in the same representation. Most of them claim that for interpreting the specific sense of the word, an underspecific meaning has to be accessed first. According to these theories, the difference between homonomy and polysemy is that senses of polysemous words share the same underspecific meaning. What is this underspecific meaning? Some researchers have maintained that it is a “core meaning” (Rabagliati and Snedeker 2013; Ruhl 1989), that is, a very abstract and general meaning that is shared by all the senses of a word and that must be specified in context. One example is the one that Spalek (2012) proposes for romper (break in Spanish), according to her, the meaning of the word romper is so underspecific that can vary in context from “disintegration” to “ceasing”. Recanati (2004, 2012) has also maintained an underspecification hypothesis. As an example, he proposes that the standing meaning of the word cut may be “effect a linear separation affecting the integrity of (some object) by means of an edged instrument” (Recanati 2012, 185). Alternative underspecification accounts have claimed that the underspecific meaning is not a definition nor a core meaning, but something not conceptual, like a pointer or a schematic meaning (Carston 2013, 2015, 2016) or an instruction for accessing and assembling concepts (Pietroski 2005). These theories postulate that the general representation is very thin. Contrary to thin semantic theories, Ortega-Andrés and Vicente (2019) propose an informational structure that contains all potential senses of the word and that needs to be accessed and activated before selecting the specific sense (I will call this The words bank1 and bank2 are homonymous: they have the same word-form (phonetics and spelling), but different meanings that are unrelated. In contrast, the word bank2 is polysemous, because the two senses of the word (a and b) are somehow related: the building is the place where the events and activities that are involved the institution take place. However, the distinction seems not to be so strict. It is always possible for someone to intuitively see a relation between two senses while some other people do not see that relation. The classification between more related senses and less related senses does not seem to be so categorical. What is clear is that a word form can have a huge range of different meanings and some of them are going to be “semantically closer” than others.


M. Ortega-Andrés

hypothesis the Activation Package model). The Activation Package model explains the interpretation of the polysemous words by an activation process: first, the information in the structure is activated; second, the specific relevant sense that is used is selected. Therefore, it could be considered that the whole structure is the underspecific meaning that has to be accessed and the aspect of the structure that is selected is the specific sense of the word. Also from a rich semantic perspective, Pustejovsky (1995) proposes a Generative Lexicon (GL), which is a system where every lexical item has a semantic structure of four levels: eventual structure, qualia structure, argumental structure and lexical inheritance. In this view, polysemy is normally generated by internal mechanisms, the role of context in communication is minimal. The explanation about how related senses of a polysemous word are generated tries to be purely internal to the linguistic system. Given the huge variety of hypotheses, it is not that easy to conclude which of them actually explains the empirical results about the representation of polysemous words. Most of the empirical work about the representation of polysemy discuss the core meaning approaches and Separate Sense Storage approaches. In this chapter, I argue that the evidence of the experiments suggest that senses of polysemous words may be represented on the same representation when they are very closely related, but that there are some cases of polysemy that seem to be more similar to homonomy. More precisely, in this chapter I argue that the Activation Package model explains the empirical results about inherent polysemous words and gives a plausible answer to the question of how copredicative sentences are interpreted. I show that the proposal fits with the conclusions of empirical studies about representation and gives a good account to the puzzle comparing to other approaches. The structure of the chapter is the following: In Sect. 2, I discuss different theories of underspecification of word meaning. I classify these theories depending on (i) the kind of information encoded in the general meaning of the word and on (ii) how specific senses of inherent polysemous words are accessed. I differentiate between three kinds of underspecification theories: core meaning approaches, thin semantic theories and rich semantic models. In Sect. 3, I analyze a set of experiments whose results have been used as arguments in the debate between the Separate Sense Storage model and underspecification approaches of word meaning. In Sect. 4, I use the notion of “activation packages” from the Activation Package model (Ortega-Andrés and Vicente 2019; Vicente 2019) to explain the empirical results about the interpretation of inherent polysemous words and for copredicative sentences.

Classification of Underspecification Models The Separate Sense Storage proposes that different senses of a word are stored in separated entries or representations in the lexicon (Katz 1972). The hypothesis has been criticized because it does not give an account of more creative polysemous

Interpretation of Copredicative Sentences…


words (see Pustejovsky 1995). Moreover, almost every word is polysemous, which means that open class words have several related senses. So, the hypothesis requires that the speaker has a huge capacity for storing all senses and it makes very difficult for them to distinguish between those senses that are already stored as part of the meaning of a word and those that are created in context (see Falkum 2015). Contrary to Sense Separate Storage, some Underspecification theories of word meaning propose that when hearers encounter a polysemous expression, they do not opt for a particular sense, but rather access an underspecific representation, which is specified only if the context requires it (see Frisson 2005). The nature of the underspecific meaning is still not clear. It may be a “core meaning” that contains the “overlapping” features between all the senses (Klepostinoutou et al. 2008); it could be something schematic (Frisson and frazier (2005); a very abstract meaning that summarizes all senses of the word (Recanati 2004); a pointer that gives access to the conceptual system (Falkum 2011); or it may be the case that the underspecific meaning is a rich structure and the process of “specificating” the meaning consist in selecting the relevant information (Ortega-Andrés and Vicente 2019) or generate the specific sense from the stored information by internal mechanisms (Pustejovsky 1995). So, depending on the kind of information that is contained in the “underspecific meaning” of the polysemous word, we can differentiate three kinds of theories: (a) Core meaning theories: these theories claim that the underspecific meaning of the word contains lexical information that is shared by all different senses of the polysemous word and that has to be accessed before the interpretation of the specific sense. The core meaning may contain more or less information (and be closer to rich semantic approaches or to thin semantic approaches). (b) Thin semantic theories of word meaning consider that the underspecific meaning of the word contains minimal information. In many cases, it has no semantic information at all. (c) Rich semantic theories: the general underspecific meaning is very rich. The senses are parts or aspects of the general meaning of the word and the specification process consists on selecting or generating the sense from this information, depending on the case.

Core-Meaning Approaches The core meaning is a general meaning that is shared by all conventionalized uses of the word (Ruhl 1989). The main difficulty of theories that propose a core meaning is how to decide what semantic information should be part of the core and what information should be considered “external” to the underspecific representation. There are many different ways for defining what the core meaning of a word is. The main reason why core meaning approaches have become so popular is that they seem to explain very successfully the fact that polysemous words that have


M. Ortega-Andrés

many senses are easier to process than meanings of homonymous words. For example, Klepousniotou (2008) describes the core meaning as a memory structure that encompass all semantic features that are common across multiple senses of the polysemous word. The existence of such “overlapping meaning” would explain why those related senses facilitate the interpretation of each other, while meanings of homonymous words compete. Within the core-meaning approach, we can postulate different mechanisms for accessing the specific sense of the word. Pragmatic theories propose that there is a pragmatic mechanism that allows the interpreter to access the specific sense. For instance, Blutner (1998, 2004) proposes that every lexical item has an underspecific representation that contains restrictions to concepts that are interpreted by pragmatic mechanisms. According to this proposal, the specific senses of words that refer to institutions (building, process, etc.) are not part of the lexical representation, they are aspects (or realizations) provided by abduction rules. The mechanisms for interpreting the specific senses are based on contextual and encyclopedic knowledge that is not stored in the lexicon. Consider the following sentence: (6) The school has a flat roof. The underspecific representation of nouns that refer to institutions contains the minimal information that specifies that there is a school that has different realizations. In the case of the word school, The sense “building” is understood as a realization of the school. The building-interpretation is accessed by an abductive inferential process from the predicational ambient of (6). More internally linguistic theories propose that the interpretation of the specific sense depends on internal linguistic mechanisms. For example, Spalek (2012) argues for a lexicon in which the meaning of polysemous words is a core meaning that is so general that can be shared by all potential uses of the word. She studies different uses of the verb and extracts a general definition. Moreover, she maintains that for determining the meaning of a phrase-verb, it is not necessary to appeal to background and world knowledge. The process consists on determinate an underspecific meaning of the verb that gets specific when it is accompanied by the argument. So, the underspecific meaning of the word romper would be something like ceasing or breaking an event or entity. The main difference between pragmatic theories and internal emantic theories is how the specific sense is interpreted: if the mechanisms that allow the interpreter to access the specific meaning of the polysemous word are external to the lexicon, they are considered lexical-pragmatic theories. If the mechanisms that are used are internal to the lexicon, then they are considered “internal linguistic mechanisms” (see also Falkum 2015). Therefore, lexical semantic theories need to propose a richer account of word meaning: if the mechanism that are involved in the interpretation of the specific sense are internal to the linguistic system, more general information is going to be needed in the lexicon. On the other hand, pragmatic theories tend to minimize the information encoded in the lexical entry of the word, they are “thinner semantic theories”. It is for this reason that the interpretation process has to be external to the internal information.

Interpretation of Copredicative Sentences…


Thin Semantic Approaches A natural reasoning is that the underspecific meaning, which has to be accessed before specifying or accessing the particular sense of the word, is very thin. Thin semantic theories claim that the underspecific representation is so meagre that needs to be specified in context to become into a full-fledged concept (see Falkum and Vicente 2015). According to Falkum and Vicente (2015), thin semantic theories about word meaning are those that claim that lexical, or standing meanings of words are impoverished with respect to their occasional meaning. The idea is that lexical meanings only contain the necessary information for constraining the range of concepts that words can express (Falkum 2011; Carston 2013; Travis 2008). One way to reduce the information of the lexical representation of words is to postulate that the underspecific meaning is so thin that it has no content at all. For example, Pietroski (2008) proposes that meanings are instructions for how to access and assemble concepts. The underspecific meaning of words does not have anything to do with the content of sentences. Understanding an expression is a matter of recognizing that expression as a certain concept–construction–instruction. What characterizes underspecific meaning-instruction of polysemous words like book is that they have to be linked with various concepts. According to Carston (2013, 2015) and Falkum (2011), only “ad hoc concepts” are fully specific. The schematic meaning has to be specified and adjusted in order to accommodate to the specific sense of the word that is used in the occasion of utterance. What characterizes the proposal is that the underspecific meaning is non-­ conceptual while the specific senses are fully conceptual entities. The “content” of a word is a non-conceptual lexical meaning, a concept schema, pointer or indicator, whose function is to constrain the general pragmatic process of accessing or constructing the fully specific meaning. Carston (2016), who has recently shown some skepticism towards the thin semantic hypothesis, has criticized thin semantic theories because in these proposals the schematic meaning apparently has no role to play in interpretation. So, in a conversational context, when a speaker says Could you please give me the book?, the hearer has to interpret, on the basis of the concepts that they has, the context and relevance expectations, that what is referred as book has to be a physical object, because informational content cannot be given in the way they is asking for it. So, they first has to access a pointer and then access their concept BOOK. As Carston (2016) already notices, in these theories the speaker has enough information for selecting BOOK without accessing the schematic meaning. So, it is not clear what the role of that pointer or schematic meaning is. It seems unlikely that during the interpretation process the hearer has to access a non-­ conceptual meaning that does not give any new information to constraint the hearer.


M. Ortega-Andrés

Rich Underspecific Structures Contrary to thin semantic theories, the GL and the Activation Package model propose rich semantic accounts of word meaning. According to these theories, most of the information that is needed for interpreting the specific sense of a polysemous word is internal to the lexical representation of the word. The encoded information associated with the word is rich, so the information is detailed and precise and includes what is mostly considered conceptual or encyclopedic knowledge (see Hogeweg 2012). Therefore, instead of an abstract underspecific meaning that is shared by all conventional senses of the word, these theories propose a semantic structure that contains the general information (including world knowledge) that is needed for interpreting the specific sense of the word. The Generative Lexicon (GL) exemplifies a semantic approach in which most of the information required for interpreting a specific sense is internal to the semantic representation. According to GL, there is a set of generative devices that connect the information given in the qualia structure, which contains the information needed for generating the specific senses. It is for this reason that this theory has also been considered a rule-internal theory of polysemy. So, lexical items are provided with mechanisms for fitting to novel environments and most of cases of polysemy are generated by internal mechanism that combine these pieces of information. For example, the word bake4 has the qualia “changing the state of x”. When the word bake appears with a noun of a natural kind, like potato, the meaning of the phrase baking a potato is “to change the state of the potato”. However, when the word bake appears with cake, whose agentive qualia is “baking”, then bake a cake means “to create the cake” and not to change its state. GL proposes that not all senses of polysemous words are represented in the same way. So, while in some cases of polysemy senses are derived by internal generative rules, -for instance, rule producer cases of polysemy like content/container are expected to be generated by internal generative mechanisms-, idiosincratic cases of inherent polysemous words (for instance: book) have their copredicative senses already encoded in the qualia structure. For example, here is the structure of book: As it is shown in Fig. 1, the lexical item of the word book has two arguments in its argument-structure, which means that the noun could have these two senses in a sentence. The information given in the qualia structure is based on both arguments: it is the compound of the two aspects what is read, what is hold and what is written. The qualia structure in GL contains four specific kinds of information. However, there is no principle reason to limit the information that tells us what a certain entity is to a given, fixed, number of (four) features (see Ortega-Andrés and Vicente 2019). Pustejovsky concedes that some entities are characterized by three features. For

4  According to GL, the meaning of verbs is thin (see also footnote 5), while the meaning of nouns is very rich. Therefore, GL explains phrasal meanings using a compositional mechanisms, between verbs and nominals: the meaning of baking the potato and baking the cake is a result of interaction between the rich meaning of the noun, and the thin meaning of the verb.

Interpretation of Copredicative Sentences…


example, if a concept is an artifact concept, it will have a function (a telic quale); if it is a natural kind concept, it will not. The Activation Package model is aless restricted rich underspecification theory: the information stored in the structure depends on ontological relations between possible denotations of the word. It is not limited to four different kinds of information. In the Activation-Package model, the rich structure contains all the conventional senses of the word, so its information is not going to be shared by all of them. For example, the word school has several related senses: people that work for the institution, the rules of the institution, the building, the temporal organization, etc. These senses of the word school have many different (even incompatible features). The general meaning of the polysemous word contains all potential conventionalized senses. When the word is interpreted, the informational structure is accessed and the information that it contains is activated, forming an activation package. The predicational ambient and the context allow the selection of the specific sense that is used in each specific case. The GL and the Activation Package model propose that polysemous words stand for complex structures that contain general knowledge about what the word refers. They introduce inside the “lexical meaning” the general knowledge that thin semantics considers as part of the conceptual repository associated with the word. The aspects that are taken into consideration inside the lexical structure of the word are the aspects or features that characterize a certain kind. So, for instance, in the Activation Package model, the knowledge structure associated with an institution includes information about its telos, its social realization, and its physical realization. Further specific knowledge derives from the particular telos of the institution (see Ortega-Andrés and Vicente 2019; Vicente 2019). However, there are some specific differences between both theories. For example, GL understands the interpretation process as “generative” mechanisms in which many senses of words are generated by mechanisms that combine and transform particular pieces of information of its structure with the information contained in other lexical structures. On the other hand, the Activation Package model explains the interpretation of the inherent polysemous word in terms of activation and sense-selection. In next section, I discuss some empirical studies that study the plausibility of core-meaning for explaining the interpretation of polysemy. Some of the results are difficult to explain from any One Representation Model, even thin and rich semantic theories. However, if we consider that different kinds of polysemous words are stored differently, then the evidence can be explained from a rich semantic perspective.

Fig. 1  Structure of “book”


M. Ortega-Andrés

Analysis of Empirical Results In this section, I discuss different kinds of experiments about the processing of polysemy. The results of some of them have been considered to be in favor of Separate representations, while others have been considered in favor of underspecification accounts of polysemy. Klein and Murphy (2002) run five experiments and conclude that their results are incompatible with a core meaning model. They used phrases that include a polysemous word and a modifier. For example, paper (polysemous word) was paired with the modifiers wrapping, shredded, liberal, and daily. Experiment 1 had two phases: first, participants had to study the phrases and, secondly, they had to decide whether the capitalized word in each phrase had appeared in the first part of the study. For example, they had to decide whether the following words appeared in the first part or not: liberal PAPER, wrapping PAPER, daily PAPER, etc. In the second experiment, the task was to make a sense/nonsense judgment on phrases that were similar to those used in the first memory experiment. For example, subjects first saw daily paper or wrapping paper. After that, they saw liberal paper and had to decide whether or not the expression was consistent with the previous one. So, the phrase daily paper was thought to be consistent with liberal paper because both phrases refer to a published paper. On the contrary, the word paper in wrapping paper means a different kind of paper (the one that is used for cover presents and gifts), so they were expected to be consider inconsistent. Experiment 3 was identical to Experiment 2, but adding homonymous words. In Experiment 4 they only used the modifiers from Experiment 3 (both polysemous and homonymous items). They compared modifiers taken from consistent pairs (like wrapping and shredded) with modifiers taken from inconsistent pairs (like liberal and shredded). Participants had to decide whether they were words or non words. The studies show that using a word in a sense facilitated comprehension of the same sense in a second sentence; however, using the word in a different sense, there was inhibition in comprehension. In addition, the effects of sense consistency in polysemy were similar to the effects of meaning consistency in homonyms. So, Klein and Murphy (2002) interpret that their results are difficult to make compatible with a “core meaning” approach. The idea is that if there was a core meaning, there would not be any “consistency effect” in the selection of the word in the second phase, because they would have to access the core meaning in any case. The results are not only difficult to explain by the core meaning approaches, its evidence also contradicts any underspecificacion theory of representation. If the participants have to access an underspecific meaning before selecting the specific sense, then the consistency effect should not play any role in the interpretation. So, all the theories I have presented in the previous section predict that there is no inhibitory effect between senses of the same representation. Nevertheless, it is important to have in mind that the experiment has been criticized by Klepousniotou (2002; Klepousniotou et  al. 2008) because it does not

Interpretation of Copredicative Sentences…


distinguish between different kinds of polysemous words. According to them, polysemous words should be classified depending on how closely related their senses are. They replicate the study considering this factor and suggests that the higher overlapping polysemous words are (see Table 1), the more likely to be part of the same representation they are. Klepousniotou et al. (2008) used a similar method as Klein and Murphy (2002): participants had to judge whether ambiguous words embedded in word pairs (e.g., tasty chicken) made sense as a function of a cooperating, conflicting, or neutral context. However, they distinguished between three kinds of polysemous words: (i) highly overlapping meanings; (ii) moderate overlapping polysemy and (iii) low-overlap polysemy. Examples are shown in Table 1. The result of the experiment showed that high-overlapping polysemous words differ from moderate- and low-overlapping ambiguous words in comprehension. It seems that senses that high-overlap facilitate the interpretation of the others, while in the case of moderate and low overlapping senses, there is some inhibition effect. Given the huge diversity of results in previous experiments about the interpretation of polysemous words, Foraker and Murphy (2012) also run four different reading task experiments. In Experiment 1, the participants had to read sentence pairs. The first sentence provided a context for the polysemous word, which ended the sentence with the polysemous word. The second sentence (target) is an unfinished sentence that disambiguates the polysemous word of the first sentence. The target sentence started with a NP that was closely associated with the dominant (more frequent sense) or subordinate sense (less frequent sense). An example of the materials appears in Table 2. Participants had to read each sentence in a screen and press the space bar once they have fully understood it. For the first sentence, the shorter neutral context sentences were read quicker than the others. For the target sentences, they found the following results: (i) following the dominant context, the dominant sense was read faster than the subordinate sense; (ii) following the subordinate context, the subordinate sense was read faster than the dominant; (iii) following the neutral context, the dominant sense was significantly read faster. The effects of consistent vs. inconsistent biasing contexts confirm that reading time of the target sentence was influenced by readers’ interpretation of the earlier context sentence. Moreover, a frequency effect was found in the interpretation of the second predicate when the contextual sentence and the target sentences were inconsistent: sense frequency affected the easy recovering of the sense in inconsistent contexts. When there is inconsistency between the contextual sentence and the disambiguating sentence, the dominant sense was easier to recover. Table 1   Types of polysemous words Overlapping

Ambiguous word Dominant sense Modifier 1 Modifier 2 Low overlapping Appeal Sex Universal Moderate overlapping Atmosphere Tense Informational High overlapping Article History Well-written

Subordinate sense Modifier 1 Modifier 2 Legal Pening Upper Polluted Submitted Popular


M. Ortega-Andrés

Table 2  Materials for experiment 1 Context sentence Context dominant Context dominant Context subordinate Context subordinate Context neutral Context neutral

Sense completion Context sentence Dominant sense The fashion designers discussed the cotton Subordinate The fashion designers discussed sense the cotton Dominant sense The farm owners discussed the cotton Subordinate The farm owners discussed the sense cotton Dominant sense They discussed the cotton Subordinate sense

They discussed the cotton

Target sentence The fabric ripped a second time The crop failed a second time The fabric ripped a second time The crop failed a second time The fabric ripped a second time The crop failed a second time

In Experiments 2 and 3, they incorporated a context, a polysemous word, and a disambiguating region in a single sentence. This allows them to test whether interpreting a polysemous word within one sentence -see sentences in (7)- is the same as processing across sentences. Experiment 2 investigated one-sentence materials using self-paced reading, while Experiment 3 used eye-tracking: (7) (a) The fashion designers discussed the cotton after the fabric ripped a second time; (b) The fashion designers discussed the cotton after the crop failed a second time; (c) The farm owners discussed the cotton after the fabric ripped a second time; (d) The farm owners discussed the cotton after the crop failed a second time; (e) They discussed the cotton after the fabric ripped a second time; (f) They discussed the cotton after the crop failed a second time. The results were similar to Experiment 1. The biasing context was effective and consistent conditions were read faster than the inconsistent ones. So, accessing the dominant sense of a polysemous word, prior dominant context, provided no processing advantage over neutral context. That means, when subjects read the predicate the fabric was not what they had been hoping for the reading time was equal for a dominant context (the fashion designers discussed the cotton) and for a neutral context (they discussed the cotton). Foraker and Murphy concluded that the results are compatible with the idea that readers represent the different senses of polysemous words in separated representations rather than a core meaning, because the dominance effect that they reflect. The reason is that following the idea that interpreters first access an underspecific meaning, then in the neutral context the selection/generation of the specific sense would have to be in the second predicate. Does this mean that senses of polysemous words are not stored together? The results of some experiments go in that line, however, it is still possible that not all

Interpretation of Copredicative Sentences…


polysemous words are equally encoded or represented. Remember from the experiment that Klepostinoutou et al. (2008) run that not all polysemous words show the same results. So, as Foraker and Murphy (2012) already admits, it is possible that some forms of polysemy are represented iseparately nd that more related senses could be represented together. Why is it the case that some senses of polysemous words inhibit each other? According to thin semantic approaches (remember the previous section), the lack of dominance effect is explained because these senses share a pointer that facilitate the interpretation of other senses. So, thin semantic theories need to postulate that these senses that show inhibitory effect are stored in different representations, so they do not share the same thin underspecific meaning. On the other hand, rich semantic theories also postulate that not all senses of polysemous words involve the same interpretation process. So, for instance, GL proposes that some senses of polysemous words have to be generated by internal mechanisms while others (like the two senses of the word book) are directly accessed in the qualia structure. From the perspective of the Activation Package model, senses that show inhibitory effect do not form activation packages. It is for this reason that they compete with each other. So, even when the senses of the word cotton are ontologically related, their ontological relation is not strong enough for forming activation packages. According to this theory, the main difference between inherent polysemous words and other polysemous words is that the senses of words that copredicate form stable activation packages (see Ortega-Andrés and Vicente 2019). There are some other experiments that suggest that senses of inherent polysemous words are stored differently from other cases of polysemy. Frisson (2005) run two experiments: a sensicality task and an eye tracker study. The sensicality task was based on Klein and Murphy (2001) study. The experiments that Frisson (2005) run tested frequency effect in a specific kind of word. He used inherent polysemous words of the kind book. The experiment consisted of a list of 24 pairs adjective + polysemous nouns presented in pairs with the prime constructions and the target constructions expressing a coherent interpretation. Adjectives selected one sense of the word: either abstract (scary book) either concrete (bound book). Participants had to chose whether they made sense or not by pressing buttons. The first experiment did not show any significant difference between switching in one direction or the other, which means that dominance has no effect in processing the words. However, it showed that there is a cost in switching in either direction. Frisson interprets that these results are more consistent with the underspecification approach than with the Separate Sense Selection theory. The reason is that if these senses were represented in the same way as meaning of homonymous words, there would be some frequency effect. Frisson also run an eye tracker experiment using copredicative sentences of the kind book. In this experiment, eye movements were recorded when participants read sentences in which a polyseme was disambiguated to a specific sense, following a neutral context, a sense was repeated or a sense was switched. Here are some examples of the sentences Frisson used:


M. Ortega-Andrés

(8) (a) Mary told me the book was scary and that she valued it a lot. (b) Mary told me that the book was bound and she valued it a lot. (9) (a) Mary told me that the science-fiction book was scary and that she valued it a lot. (b) Mary told me that the gift-wrapped book was bound and that she valued it a lot. (10) (a) Mary told me that the bound book was scary and that she valued it a lot. (b) Mary told me that the scary book was bound and that she valued it a lot. The sentences in (8) exemplify the neutral condition, where there is no switch nor repetition of senses; sentences in (9) exemplify sense repetition: (9a) repeats the informational content sense of the word book (the dominant or more frequent sense), and (9b) the physical object sense (the subordinate or less frequent). In the sentences in (10) there is sense switching (copredication): from physical object to informational content in (10a), and from informational content to physical object in (10b). The main discovery of the second experiment (the eye tracker) is that it shows that there is some processing cost in the selection of the second sense of some co-­ predicative sentences involving book-type polysemes. Moreover, switching from the subordinate (less frequent sense) to the dominant sense (more frequent) resulted in longer reading times than switching from the dominant sense to the subordinate sense. Again, the results of the experiment have been interpreted by Frisson as been in favor of One Representation theories and against the Separate Sense storage. Actually, the underspecification theories seem more appropriate for explaining the results we have for inherent polysemous words. From the perspective of the Activation Package model, senses that are in explanatory dependency ontological relations are more likely to form co-activation patterns (see Ortega-Andrés and Vicente 2019). So, one plausible explanation is that the senses of the word book form such activation patterns while other cases do not. Moreover, there are some experiments on neural activity that suggest that different neurological areas are activated when inherent polysemous words are interpreted. Tao (2015) run an experiment investigating the neural representation with a fMRI experiment comparing inherent polysemous words of the kind book with other words that refer to abstract objects and concrete objects but that are not polysemous (for example: book vs dream and book vs chair). Participants of the experiment had to read on a screen a list of phrases in Italian and they had to judge whether the phrases were meaningful. For instance: consult the book, have the idea. For the concrete sense, he found that there was activation in the ventral ATL when the inherent polysemous words were interpreted but not when other words were interpreted. This result is interpreted as that the meaning representation of inherent

Interpretation of Copredicative Sentences…


polysemous words, when in context, instantiates to a more specific representation, which makes sense because the vATL seems to play a critical role in representing this complex conceptual knowledge (see Tao 2015, but also Patterson and Ralph 2016). Tao (2015) repeated the experiment using words of the kind lunch (event+food) in Italian. The task of the experiment was similar to the first one, but the results were even more impressive. As in the first experiment, he found that the neural distinction between the concrete and abstract interpretations of the inherent polysemous words differed from the concrete-abstract distinction observed in the other words. Moreover, inherent polysemous words of the kind lunch showed an important effect in the right ATL, comparing to book. This is surprising because the right hemisphere is not normally associated with linguistic task. However, it is true that in previous experiments about polysemy (Pylkkänen et  al. 2006) it has been shown that the interpretation of some phrases with polysemous words -for instance: liberal paperthere is M350 longer latency in the right hemisphere than in the left. Moreover, it has been previously suggested that this area (right ATL) is associated with social thinking (Snowden et al. 2004; Zahn et al. 2009). It seems to me that these results are compatible with rich semantic accounts. [So] Thus, we have seen in the previous section that according to rich semantic theories there is general knowledge associated in the lexical representation of the word that needs to be accessed for interpreting the correct sense of the polysemous word (remember previous section). These events (lunches, dinners, etc) are normally social events, therefore, from a rich semantic perspective, it makes sense that interpreting the meaning of these words require some “social thinking”. However, more investigation would be needed for being conclusive about this hypothesis. Finally, Tao run a third experiment using inherent polysemous words of the kind church (building+organization). In this case, the experiment consisted in a MEG, whose results showed that the MEG gamma-band frequency power could distinguish the neural correlates of the concrete and abstract interpretations and a notably the divergence occurred 400  ms and later post-stimulus, which suggest that the selection of the specific meaning of the word occurs at the later integration stage in the interpretation of the word. According to the Activation Package model, the interpretation of the inherent polysemous word requires that the whole lexical structure get activated, and that the correct sense of the word is finally selected (see Ortega-­ Andrés and Vicente 2019). So, it could be the case that the selection of the correct sense of the word occurred at the final stage because of the activation process that is required to happen during the interpretation of the word. In the following section I show how the Activation Package model explains the representation of inherent polysemous words and interpretation of copredicative sentences.


M. Ortega-Andrés

 epresentation, Interpretation and Copredication R in the Activation Package Model As we have seen in the previous section, empirical evidence is not conclusive about how we interpret polysemous words. Nevertheless, it seems to be a general assumption that senses of inherent polysemous words are linked in a way that they do not inhibit each other, at least not as much as the meaning of homonymous words. The Activation Package model proposes a rich semantic account in which specific senses of inherent polysemous words are parts or aspects of a rich semantic-­ conceptual structure. The structure contains general knowledge about what the word means. For example, the general knowledge about schools includes information about the kind of entity a school is, its physical and temporal realization or implementation, the kind of people that take part in it, and its organizational structure, etc. Such stored information is intended to capture the prototypical knowledge that we have about schools and the different aspects of the informational structure relate to each other. The following structure represents the kind of information that could be encoded in the lexical entry of school (Ortega-Andrés and Vicente 2019): Different senses of the word school are aspects in the informational structure (Fig. 2) that have to be selected in each case, forming the specific sense of the word. Interpreting the meaning of a polysemous word in a sentence is a process of activation and selection, that ends when the specific sense of the word is selected. There is a double activation process: when the interpreter encounters the word school in a sentence, the whole structure (Fig. 2) is activated, which means that the information represented in Fig. 2 is accessible for being selected. The predicational ambient and the extralinguistic context selects the most relevant aspect of the structure. Consider the following sentence: (9) The school caught fire. In (9), the predicate on fire describes a property that has to be predicated of a physical object. The predicates give the information that allows the interpreter to select from the structure the aspect “building”. When the sentence (9) is interpreted, the correct sense of the word school is selected when the predicate caught fire selects the aspect builsing from the structure. However, before that, all aspects in the structure of the word have been previously activated. The aspects or pieces of knowledge are linked forming activation patterns that depend on how their relations are conceptualized. What characterizes the Activation Package structure is that the aspects given in the structure activate each other depending on how they are internally linked. Different realizations of the word school are normally in the following relations: I. The participants of the institution (students and teachers) are those that normally participate in the events and activities associated with the institution. II. The social organization is formed by some rules and representative roles that regulate the events that the participant should commit.

Interpretation of Copredicative Sentences…


Fig. 2  Structure of “school”

III. The temporal organization sets a timetable for those particular events in which the participants of the institution normally participate. IV. The physical realization is the place where the events associated with the institution occur. The participants of those events and activities are occupants of the inside of the building when they participate in those activities. Considering relations (I)–(IV), the hypothesis answers the question about why some senses of inherent polysemous words allow copredication and others do not. Ortega- Andrés and Vicente propose that copredication is more likely to occur between those aspects form activation patterns. It seems that their proposal could go forward and give a more specific answer by analyzing the kind of ontological dependency relations that they assume when they describe the realization relations between aspects. For example, when the aspect “institution” is selected, it activates its realizations. The reason is that the institution depends somehow on its realizations for being such institution. For example, for being an institution, the school needs a social realization, which means that the aspect “institution” activates the aspect “organization”, forming an activation package. See the following sketch of an activation pattern: Figure 3 does not intend to show all the activation patterns that may occur in the interpretation of the word school. I have included only some of the patterns that follow from the ontological relations given in (I)–(IV). It does not mean that there could not be more. Actually, I have excluded many aspects from the informational structure given in Fig. 2. The reason is that it is difficult to predict how the activation


M. Ortega-Andrés

pattern would work in every case. For example, the dependency relation between the aspect “institution” and the “social organization” is more determinant that in the case of “organization” and “building”. There is no direct ontological dependency between the organization and the building: buildings do not require social organizations for their existence and social organizations do not necessarily require a building. However, considering (II) and (IV) there is an agential relationship such that the participants of the organization that are committed with the organization rules are normally located in the building when they are organized (for a similar approach about ontological agential relations between aspects of institutions, see Arapinis and Vieu 2015). This agential relationship may generate an activation pattern, however, it would be expected that the activation was not as strong/common as in the case of “institution-organization”. The aspect “social organization”should activate the rules because there is no possible organization without rules. The aspects “institution” and “social institution” also activate the aspect “participants” because there is no possible institution without the agents that participate or work in it. Generally, the participants are the occupants of the building, which also explains why the aspect “organization”may also activate the aspect “participants//occupants” and they activate the aspect “building”. So, it seems that following Figs. 1 and 2, we can predict how the activation pattern should work in the interpretation of some copredicative sentences. Consider again the following sentence: (5) The school caught fire and was celebrating 4th of July when the fire started. When the hearer encounters the NP the school, the informational structure (see Fig. 2) is activated. The activation of the structure means that all the information it contains is accessible for the interpretation of the word. The predicate caught fire selects the aspect “building”. The selection of the specific sense does not deactivate all the other aspects. As it is shown in Fig. 2, there is an ontological relation between the participants and the building: participants of the institution (students and teachers) are very commonly the occupants of the building that is the physical realization of the institution. The ontological explanatory/dependency relation generates an activation package between these aspects. Therefore, they form an activation package (see Fig. 3) with the senses “participants”, “occupants” and “building”. The selection of one aspect means the strong activation of the other, which facilitates the selection of the second one. In this case, after the sense “building” activates the aspects “occupants” and “participants”, forming one activation package, the predicate selects the aspect “participants”. So, the sentence is interpreted as follows:

Fig. 3  Activation patterns in the interpretation of “school”

Interpretation of Copredicative Sentences…


(5′) The school-building caught fire and the school-participants was celebrating 4th of July when the fire started. Now, how does this theory explain the results given by Frisson and frazier (2005)? Remember that the experiment showed that readers spent longer times when the switch is from the less frequent (physical object) to the more frequent sense (content) sense than vice versa. The lexical representation of the word book has the formal quale “informational content” and the telos “reading”. The volume that some how contains the informational content is the physical realization of the informational content. So, it makes sense that the sense “content” facilitates the interpretation of the sense “physical realization” because the formal quale depends on its physical realization for being read. However, once the physical aspect is selected it is more difficult to recuperate the sense “content”. One consequence of the model is that it predicts that aspects that are more closely related by ontological dependency relations are more likely to generate felicitous copredicative sentences than others. Still, it is important to have in mind that the proposal still has many open questions to answer.5 For example, it may be the case that some other factors influencing the possibility of copredication: for instance, Murphy (2019) postulates that the order in which the senses of the word appear affect, depending on whether the first sense is more abstract or more concrete. In any case, considering the empirical evidence we have, it seems that the interpretation of many inherent polysemous words is a matter of activation and selection of senses. According to the Activation-Package model, senses of inherent polysemous words form activation packages, so they are easier to interpret. Meanings of homonymous words do not form activation packages because they are in different structures. It is for this reason that they show inhibition effects that polysemous words do not show and they do not copredicate. These are only tentative explanations, however, they give a plausible answer to the question of how senses of inherent polysemous words are represented.

5  Moreover, there are some problematic cases. The word newspaper has been thought has having two structure: the “paper” structure and the “institution” structure. If that was the case, they should not copredicate, however, there are some particular cases in which they copredicate:

( a) ? The newspaper fired its editor and fell off the table (b) ? That newspaper is owned by a trust and is covered with coffee. (c) John used to work for the newspaper that you are reading. (d) The newspaper has been attacked by the opposition and publicly burned by demonstrators. There are some cases in which the senses “institution” and “content” also copredicare. Sentences (a) and (b) are usually taken to be bad, while (c) and (d) are regarded as better or even good. According to some researchers (Arapinis and Vieu 2015; Ortega-Andrés and Vicente 2019) there are two structures involved. So, there are two inherent polysemous words: newspaper as institution and newspaper as text or printed paper. Following the GL, Arapinis and Vieu (2015) propose that the aspect of newspaper as institution is accessible via the agentive qualia of newspaper as copy, which specifies how the copy was created, and, conversely, the copy aspect is accessible from the telic quale of the newspaper as institution, which specifies what the institution is for.


M. Ortega-Andrés

Conclusions The present chapter gives a classification of underspecification accounts of word meaning and discuss them. Moreover, I have analyzed some empirical results in the debate between underspecification accounts and the Sense Selection Storage approach. The experiments show that senses of inherent polysemous words are interpreted differently from other polysemous words and from homonymy. The results are not conclusive about the kind of approach that we must adopt. However, there are theoretical reasons for adopting a rich semantic underspecification account. I have argued that the Activation Package model proposes a rich semantic underspecification account that is able to explain these results by appealing to activation patterns. Acknowledgements  Versions or parts of this paper were presented at the Seminars at University of the Basque Country’s SemLin and at the UCL. I am very grateful to audiences at those events, in particular to Tim Pritchard and Robyn Carston. My gratitude goes also to my colleagues Aitor Lidardi, Elena Castroviejo, Javier Ormazabal, Laura Vela-Plo Myriam Uribe-Etxeberria and, especially, to my supervisor Agustín Vicente who provided insightful comments. Funding for this paper was provided by a FPI grant from the Spanish Ministry of Economy (BES-2015-071378) and the projects: PGC2018-093464-B-100 and IT1396-19

References Apresjan, Ju. D. (1974). Regular polysemy. Linguistics, 12(142). ling.1974.12.142.5. Arapinis, A., & Vieu, L. (2015). A plea for complex categories in ontologies. Applied Ontology, 10(3–4), 285–296. Beretta, A., Fiorentino, R., & Poeppel, D. (2005). The effects of homonymy and polysemy on lexical access: An MEG study. Cognitive Brain Research, 24(1), 57–65. cogbrainres.2004.12.006. Blutner, R. (1998). Lexical pragmatics. Journal of Semantics, 15(2), 115–162. Blutner, R. (2004). Pragmatics and the lexicon. In L. Horn & G. Ward (Eds.), The handbook of pragmatics (pp. 488–514). Oxford, UK: Blackwell Publishing Ltd.. Carston, R. (2013). Word meaning, what is said and explicature. In C. Penco & F. Domaneschi (Eds.), What is said and what is not (pp. 175–204). Stanford, CA: CSLI Publications. Carston, R. (2015). Contextual adjustment of meaning. In N. Riemer (Ed.), The Routledge handbook of semantics (pp. 195–210). New York, NY: Routledge. Carston, R. (2016). The heterogeneity of procedural meaning. Lingua, 175–176(May), 154–166. Falkum, I. L. (2011). The semantics and pragmatics of polysemy: A relevance- theoretic account. London, UK: University College London. Falkum, I. L. (2015). The how and why of polysemy: A pragmatic account. Lingua, 157, 83–99. Falkum, I. L., & Vicente, A. (2015). Polysemy: Current perspectives and approaches. Lingua, 157, 1–16.

Interpretation of Copredicative Sentences…


Foraker, S., & Murphy, G.  L. (2012). Polysemy in sentence comprehension: Effects of meaning dominance. Journal of Memory and Language, 67(4), 407–425. jml.2012.07.010. Frazier, L., & Rayner, K. (1990). Taking on semantic commitments: Processing multiple meanings vs. multiple senses. Journal of Memory and Language, 29(2), 181–200. https://doi. org/10.1016/0749-596X(90)90071-7. Frisson, S. (2009). Semantic underspecification in language processing. Language and Linguistics Compass, 3(1), 111–127. Frisson, S. (2005). About bound and scary books: The processing of book polysemies. Lingua, 157, 17–35. Frisson, S., & Frazier, L. (2005). Carving up word meaning: Portioning and grinding. Journal of Memory and Language, 53(2), 277–291. Frisson, S., & Pickering, M. J. (2001). Obtaining a figurative interpretation of a word: Support for underspecification. Metaphor and Symbol, 16(3–4), 149–171. 8.2001.9678893. Gibbs, R., & Gerrig, R. (1989). How context makes metaphor comprehension seem ‘special’. Metaphor and Symbol, 4(3), 145–158. Hogeweg, L. (2012). Rich lexical represe ntations and conflicting features. International Review of Pragmatics, 4(2), 209–231. Katz, J. (1972). Semantic theory. New York, NY: Harper and Row. Klein, Devorah E, & Murphy, Gregory L. (2001). The representation of polysemous words. Journal of Memory and Language, 45(2), 259–282. Klein, D.  E., & Murphy, G.  L. (2002). Paper has been my ruin: Conceptual relations of polysemous senses. Journal of Memory and Language, 47(4), 548–570. S0749-596X(02)00020-7. Klepousniotou, E. (2002). The processing of lexical ambiguity: Homonymy and polysemy in the mental lexicon. Brain and Language, 81(1–3), 205–223. brln.2001.2518. Klepousniotou, E., & Baum, S.  R. (2007). Disambiguating the ambiguity advantage effect in word recognition: An advantage for polysemous but not homonymous words. Journal of Neurolinguistics, 20(1), 1–24. Klepousniotou, E., Titone, D., & Romero, C. (2008). Making sense of word senses: The comprehension of polysemy depends on sense overlap. Journal of Experimental Psychology: Learning, Memory, and Cognition, 34(6), 1534–1543. Murphy, E. (2019). Acceptability properties of abstract senses in copredication. In M. Bolognesi & G. Steen (Eds.), Perspectives on Abstract Concepts: Cognition, language and communication, (pp. 145–165). Amsterdam, Netherlands: John Benjamins. Ortega-Andrés, M., & Vicente, A. (2019). Polysemy and co-predication. Glossa: A Journal of General Linguistics, 84(3). Patterson, K., & Lambon Ralph, M. A. (2016). The hub-and-spoke hypothesis of semantic memory. In G. Hickok & S. Small (Ed.,), Neurobiology of language (pp. 765–775). New York, NY: Academic Press. Pickering, M., & Frisson, S. (2001). Processing ambiguous verbs: Evidence from eye movements. Journal of Experimental Psychology: Learning, Memory, and Cognition, 27(2), 556–573. Pickering, M., McElree, B., Frisson, S., Chen, L., & Traxler, M. (2006). Underspecification and aspectual coercion. Discourse Processes, 42(2), 131–155. s15326950dp4202_3. Pietroski, P. (2005). Meaning before truth. In G. Preyer & G. Peter (Eds.), Contextualism in philosophy: Knowledge, meaning, and truth (pp. 255–302). Oxford, UK: Oxford University Press. Pietroski, P. (2008). Minimalist meaning, internalist interpretation. Biolinguistics, 2(4), 317–341.


M. Ortega-Andrés

Pustejovsky, J. (1995). The generative lexicon. Cambridge, MA: MIT Press. Pylkkänen, L., Llinás, R., & Murphy, G.  L. (2006). The representation of polysemy: MEG evidence. Journal of Cognitive Neuroscience, 18(1), 97–109. https://doi. org/10.1162/089892906775250003. Rabagliati, H., & Snedeker, J. (2013). The truth about chickens and bats: Ambiguity avoidance distinguishes types of polysemy. Psychological Science, 24(7), 1354–1360. https://doi. org/10.1177/0956797612472205. Recanati, F. (2004). Literal meaning. Cambridge, UK: Cambridge University Press. Recanati, F. (2012). Compositionality, flexibility, and context dependence. In E.  Machery, M. Werning, & W. Hinzen (Eds.), The Oxford handbook of compositionality (pp. 175–191). Oxford, UK: Oxford University Press. Ruhl, C. (1989). On monosemy: A study in linguistic semantics. New York, NY: State University of New York Press. Snowden, J. S., Thompson, J. C., & Neary, D. (2004). Knowledge of famous faces and names in semantic dementia. Brain, 127(4), 860–872. Spalek, A. (2012). Putting order into literal and figurative uses of verbs: “Romper” as a case study. Borealis–An International Journal of Hispanic Linguistics, 1(2), 140–167. https://doi. org/10.7557/ Tao Y. (2015). The neural representation of polysemy: The case of dot-objects. Doctoral dissertation, University of Trento. Travis, C. (2008). Occasion-sensitivity: Selected essays. Oxford, UK: Oxford University Press. Vicente A. (2019). Chomskyan arguments against truth-conditional semantics based on variability and co-predication. Erkenntnis, 84(3),1–22. Zahn, R., Moll, J., Paiva, M., Garrido, G., Krueger, F., Huey, E. D., & Grafman, J. (2009). The neural basis of human social values: Evidence from functional MRI. Cerebral Cortex, 19(2), 276–283.

First Person Implicit Indirect Reports in Disguise Alessandro Capone

Introduction In this paper, I deal with implicit indirect reports. First of all, I discuss implicit indirect reports involving the first person. Then, I prove that in some cases second person reports are implicit indirect reports involving a de se attribution. Next, I draw analogies with implicit indirect reports involving the third person. I establish some similarities at the level of free enrichment through which the explicature is obtained and I propose that the explicature is syntactically active, given that it sanctions anaphora. An implicit indirect report is a report which does not explicitly display features of indirect reports (e.g. the verb ‘say’ or the presence of a reported speaker), but implies an evidential base requiring the structure of an indirect report. Most importantly, in this paper I demonstrate that such structural elements are active from a syntactic point of view in that they allow anaphora under certain conditions. Although it is the speaker’s meaning that matters in these cases, insofar as it intrudes into the explicature and it requires a certain (compulsory) logical form, the elements of the logical form implied at the level of the explicature are syntactically active. Furthermore, they sometimes require syntactic slots such as the experiencer and, furthermore, and somewhat I first considered the issue of implicit indirect reports when reading a paper by Elizabeth Holt that was submitted for my collection on indirect reports and pragmatics (Holt 2016). There was little discussion about implicit indirect reports because the focus was on indirect reports in general, but that was sufficient to allow me to give some consideration to this topic in Capone (2016) (implicit embeddings), and now in the present paper. A. Capone (*) University of Messina, Messina, Italy e-mail: [email protected] © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 F. Macagno, A. Capone (eds.), Inquiries in Philosophical Pragmatics, Perspectives in Pragmatics, Philosophy & Psychology 27,



A. Capone

surprisingly, in the case of second person reports what is being implied is a structure hosting a de se implicit attribution which allows an internal perspective. Such implicit indirect reports with de se ramifications are to be considered as logophoric structures that present the perspective of a particular person, and in general the experience is linked to a time which is posterior to the event being narrated in the indirect report. The issues addressed by this paper go beyond the topic being explicitly discussed, as issues are raised on the nature of the explicature involved in semantic/ pragmatic analysis and the possibility that elements of the explicature are syntactically active. Such considerations can also be extended to other types of explicatures and have to be pursued in due course. I also find it rather surprising that from a speaker’s intentions we can go on to reconstruct the explicature and its syntactic configuration, an aspect which is not normally discussed in the literature. That the explicature should have a compulsory syntactic configuration as a result of the speaker’s intentions is somewhat novel in the literature. For example, when I discussed belief reports and their opacity (Capone 2008), I found it useful to explain the explicature by pointing to a syntactic configuration, which would solve lots of problems arising from the introduction of modes of presentation. I proposed to analyse an utterance such as “John believes that Mary is clever” as consisting of an articulated sentence and an appositive sentence conjoined to it. The appositive sentence merely expresses a concatenation of modes of presentation, while the simple sentence only represents referential interpretation of an articulated sentence. Anaphoric links between elements of the appositive sentence and nodes of the simple sentence would allow us to reconstruct pragmatic opacity. This resolves the thorny problems of the logical form of belief reports, because the pragmatics of the utterance is helped by the reconstructed syntactic analysis of the actual sentence uttered and the sentence which is in the air, an unarticulated sentence rather than an unarticulated constituent. Concerning indirect reports in disguise, we can say that a syntactic analysis is coupled with a certain semantic-pragmatic interpretation and helps guide the interpretation. That syntactic analysis is conducive to meaning, but the syntax is not at the level of the articulated sentence but is part of a pragmatically reconstructed sentence. The fact that unarticulated pragmatically reconstructed constituents should have some syntax is not surprising, though some may be prepared to deny this. If an unarticulated constituent is a sentence, it must have some syntax, but this does not come from the explicit logical form but from the mind of the reader who reconstructs the interpretation. We assume that more examples than are under discussion in this paper can be subjected to a similar analysis, as most cases in which one cannot directly know another person’s mind but gets to know it through what a speaker has said, potentially constitute cases that can be analysed as implicit indirect reports.

First Person Implicit Indirect Reports in Disguise


Even innocent remarks such as, John has a pain in his stomach, can be analyzed as implicit indirect reports, given that the question arises as to how we know what is happening in John’s mind or body if he has never told us what has happened to him. Since we cannot know what he feels like telepathically, it must be reasonable to assume that we know this by some other means and, presumably, through what John has said to us about his corporeal sensations. A speaker’s intentions prevail and a deeper logical form has to be reconstructed. The Wittgensteinian idea that at least a number of utterances must have a logical form that is different from what is superficially testified by the utterance is, therefore, vindicated. I received a stimulating comment by one of the reviewers of this paper, which runs like this: The idea is very stimulating and interesting, but perhaps the author extends a little too much the possibility of considering statements, for example in the first person, as “implicit indirect reports.” In fact, if we consider the author’s example “John has a pain in the stomach” it can be considered as an indirect report because, as maintained by the author, “John could have told me he had a stomachache.” And therefore the expression is correctly understood as an implicit indirect report. Although this could be true in many cases, I do not think that I can state that this kind of sentence is always an implicit indirect report. For example, I can say that “John has pain…” because I saw John contorting like when people have stomach ache, or because I know he has ingested a poison. Further, how do we know that a person feels depressed? Canonical evidence is provided by the person’s utterance which expresses feelings and state of depression. Thus, ‘You feel depressed’ is pragmatically equivalent to ‘You say you feel depressed’, which is an indirect report. Thus, utterances such as (5) can be considered indirect reports in disguise. This is true but only if we consider the sentence “You are depressed” based on the previous utterance “I feel depressed.” But what if just you look depressed?

I have great respect for this position. However, it actually helps build up the case that the inferences I am talking about are pragmatic, since they are defeasible in certain contexts, or because they are sensitive to contextual information, which plays some role in promoting them or in demoting them. I suspect, nevertheless, that the reviewers have somehow interpreted me as saying or proposing that these are default inferences, in which case they would be right that at least in certain cases they would not arise. However, I suspect that these are genuinely contextual inferential phenomena, where the context plays a role not only in cancelling an inference but also in promoting it. Needless to say, I agree with the reviewer that there are contexts in which we look at someone and know that she is depressed, without waiting to hear her utterance. But in some cases (e.g. telephone conversations), linguistic information is essential and in reconstructing the inferential layers of the response we see that some pragmatics is required to make a report an indirect report in disguise. All in all, I agree with the reviewer that here there may be different cases to note and discuss.


A. Capone

The Scope of Pragmatics Semantics deals with aspects of meaning that are independent of context.1 These are relatively stable and provide a structure upon which further meanings can accrue to the utterance.2 Normally, semantics is fueled by words and the syntactic glue that combines them. Scholars in pragmatics believe that semantics is underdetermined, which means that even if we know the semantics of a certain expression, we do not know enough to know what the world is like (exactly). We more or less know what the world must be like, but this may not be enough for the purpose of knowledge. Capone (2013) has expressed the view that there may be a certain degree of exaggeration in this under-determinacy claim. At least certain words have stable meanings. If we know, for example, that there is a rose in Capone’s house, we know what kind of flower we can expect there to be in that house. However, I admit that many examples are unlike this simple one and may advance the view that there are pragmatic increments to utterance interpretation which are fuelled through explicatures – that is, inferential enrichments are often of the free type (see cases of belief reports, knowing how attributions, indirect reports, attributive/referential attributions, de se attributions, etc.). I also agree that enrichments processes may go beyond reference assignment and ambiguity reduction. My view is that rationality can provide suitable expansions to sentential meanings.3 When one takes into account what the speaker can rationally mean, we can reconstruct what s/he says. Very often such expansions serve to enrich the lexicon or the syntax. In other words, they are powerful ways of maximizing the linguistic resources of a language. Pragmatics is like a set of tools that can amplify the power of the language user. But, of course, this is no more than a metaphor, albeit a useful one, particularly for those languages such as pidgins which are impoverished and are stripped of semantic and grammatical resources. Thus, pragmatic enrichments range from reference fixing (saturation processes) and ambiguity resolution to free enrichments of the expansive type, (often) aimed at resolving logical problems such as blatant falsehood, contradiction and absurdity. These increments are the pragmatic components of explicatures. There has been a debate within the literature about whether these increments are ­cancellable

1  Although, in some cases, semantics appears to be like an instruction which takes context as input, in a particular way, and gives a specific truth-evaluable content as output. 2  Semantics provides a platform on which further meanings can be constructed. Most importantly, the unenriched logical form is capable of working as a premise in an act of reasoning which is conducive to fuller interpretation, provided that basic operations such as reference fixing and disambiguation have occurred. 3  Synthetically, rationality is what leads to processes of reflective and unreflective inference in which a number of premises, including the literal meaning and the disseminated contextual clues, are put together (combined) in an argument that leads to a full interpreted proposition. The Gricean maxims or suitable equivalents (in expanded or compressed form) also work as premises in the argument. Since in an argument, we normally need something that leads from a set of premises to a conclusion, that is, a warrant, we can assume that in pragmatic inference the role of the warrant is played by the necessity of having speakers’ intentions that obey canons of informativity.

First Person Implicit Indirect Reports in Disguise


or not. Pragmatists (notably Carston 2002) believe that pragmatic inferences, explicatures included, ought to be cancellable, with cancellability being the hallmark of pragmatic inference. Other scholars, for example, Capone (2003, 2006, 2009) proposed that the pragmatic components of explicatures ought not to be cancellable due to strong intentionality and to the fact that they are expected to resolve logical problems, which would remain if the explicature was cancelled (Jaszczolt 2006 writes about entrenched meanings, while she is not particularly explicit about the cancellability of explicatures). Thus, explicature cancellation would result in anomalies at the discourse level, in the cases of explicatures dealt with by Capone. (Of course, if the notion of explicature is extended, as proposed by relevance theorists, to cases of conversational implicatures (e.g. the quantifier ‘some’), then Capone’s considerations are not immediately applicable). Capone’s considerations make sense if a strong notion of explicature is considered and if attention is confined to cases like those originally discussed in Carston’s work. In this paper, I am going to discuss the issue of implicit indirect reports. Thus, I need to provide some useful background that will enable the reader to make progress in the understanding of what is to follow. Indirect reports are usually micro-­ narrations relating utterances (an utterance is surely an event from a Davidsonian perspective) without doing so verbatim, that is, by reporting the exact words proffered by the original speaker.4 They are usually summaries of stories, in which small details can be omitted, although the reporter is not allowed to offer a perspective that is totally removed from the perspective of the original speaker. Indirect reports are implicitly logophoric, in that they must be aimed at reproducing the perspective of the original speaker – at least they should not alter it too much and they should not present a self that is drastically different from the self of the original speaker. For example, it is not usually licit to replace some words with epithets or slurring expressions, even if the denotation is the same, because doing so would amount to representing the original speaker as someone who would use epithets or slurs, which is not (or need not be) the case. The perspective should not be altered – thus, indirect reports could be considered implicitly logophoric. Indirect reports are usually employed when the hearer is removed in space and time from the event of the utterance being reported. The context that facilitates the understanding of the pronominals contained in the indirect report is usually (with some exceptions) the context of the hearer, since the hearer has to have access to referents and the best way to have such access is to situate pronominals in a context which is accessible to the hearer. The original context of the utterance, being removed in time and space from the hearer, cannot be useful for the purpose of constructing reference. Furthermore, presuppositions need to be satisfied and they have to be satisfied in a context that is accessible to the hearer, not in one that is only accessible to the reporter or to the original speaker. For the time being, these considerations may suffice to allow the hearer to proceed with reading the presentation of the topics being discussed in this 4  Some believe that verbatim direct reports are a fiction, given that due to memory limitations we always make changes in the reported utterance. However, it should be taken for granted that, at least in the written medium, verbatim quotation does makes sense.


A. Capone

paper. Finally, indirect reports can either be explicit or implicit. If they are implicit, one usually has to reconstruct who the reporter is and when the report was proffered. Usually, one has to first of all resort to some contextual clues to understand that an implicit indirect report is needed and, secondly, in order to reconstruct the structure of the indirect report. Implicit indirect reports are more common than imagined, even though one can find very little in the literature by which to construct a theory of implicit indirect reports. Some clues can be found in Holt (2016) and in Capone (2016), who devotes a chapter to simple sentences, substitution failure and implicit indirect reports. The present paper is also a small step in the direction of the theory of implicit indirect reports.

First Person Implicit Indirect Reports Suppose A says: (1) I am depressed.

This may overlap, or otherwise be very different in content from (2) (2) I feel depressed.

If it does not overlap with (2), it can be said to be uttered on the basis of a warrant provided by some other speaker’s utterance, such as ‘You look depressed/You are depressed’. Something very similar can happen in the second person. A can say: (3) You are depressed.

This may overlap with either (4) or (5): ( 4) You look depressed. (5) You feel depressed.

If it overlaps with (5), (3) (You are depressed) is an implicit indirect report, something said on the basis of the warrant given by an utterance of (5) by the person whose state of mind is being described by (3). How do we know that a person feels depressed? Canonical evidence is provided by the person’s utterance which expresses her feelings and state of depression. Thus, ‘You feel depressed’ is pragmatically equivalent to ‘You say you feel depressed’, which is an indirect report. Thus, utterances such as (5) can be considered to be indirect reports in disguise. An utterance is an indirect report in disguise when accepting it amounts to accepting an utterance on which it can be based, and without which there would be no evidence for the utterance in question. An indirect report normally has an implicit base, which is the explicit direct report on which it is based. Without such an implicit base, the utterance makes no sense, it has no evidential support, and amounts to an admission that the speaker is saying something without an evidential base. The considerations above can be taken to introduce the issue of implicit indirect reports, but also the issue of the specularity relationship between the first and the second person (singular). In a dialogue, the addressee can act as a mirror and can be

First Person Implicit Indirect Reports in Disguise


a source of self-knowledge for the first person subject. Anything that is first person can be referred to as second person, and anything that is second person can be referred to as first person. In this kind of game, anyone who uses the third person (e.g. ‘He is depressed’), can be seen as speaking and acting from outside the game. Normally, a subject who uses the first person does not refer to himself by name; and a subject who uses the second person to refer to the addressee does not use a name, but prefers to use a pronominal. (The reason why we use pronominals for ourselves, rather than names, is possibly linguistic economy, given that a name needs a different kind of contextualization and can potentially pick up many referents, while the pronominal ‘I’ can, at most, pick up one referent in the context of the utterance, relative to the utterance). Someone who is outside the game can use either a pronominal or a name to refer to the participants in the I-you language game. A dialogical game involves the use of pronominals by the actors in order to refer to the actors who are playing the game. Anyone outside the dialogic game can use a name for the actors in the game. It is as if the use of pronominals vs. names marks an imaginary boundary between the I/you and the they, i.e. those who are inside the game and those who are outside. Now, consider the following utterance: (6) I never wanted to jump off the swing.

The speaker was three years old at the time of the event. He has pictures of that event, but he does not remember the event through memories from the inside (he does not have memories of himself being on the swing and not wanting to jump off when his sister wanted to join in the game).5 As Higginbotham (2003) states, he has external memories that he was on the swing, because his mother and father told him about that event (alternatively he has some pictures as evidence of that event), he remembers the event from the outside, but he does not remember it from the inside. According to Higginbotham (2003) and Capone (2010), there is a difference between (7) and (8). ( 7) I remember that I was on the swing (8) I remember being on the swing.

The control construction [I remember PRO being on the swing] is even more first-personal than the one with the first-person construction, as it involves a genuine self-based perspective which allows remembering from the inside. If the memory is linked with joy (or, alternatively, with pain), one should remember the joy or the pain if one has remembered it from the inside. But if one remembers the event because someone has told one about it, and that person forgot to describe the associated sensations (experienced from the outside), one would not be able to remember them. 5  Interesting interpretative issues emerge. Is (6) equivalent to ‘I remember I never wanted to jump off the swing’? The other interpretative possibility is: ‘I remember never wanting to jump off the swing’, which is even more first-personal than the former interpretation. As Higginbotham (2003) would state, the latter interpretation is more logophoric in that it involves the internal perspective of the speaker qua the experience and subject of the act of remembering.


A. Capone

Returning to (6), given that the speaker was three years old (as attested by the pictures that he still owns), he could not (normally) remember the event from the inside and, thus, the recipient of the utterance will reason that the event is a memory based on someone else’s recollections. Thus, it is not a direct report, but must be an implicit indirect report. This means that (6) must be embedded in something like (9), (9) [I was told (by my mother) that] I never wanted to jump off the swing.

The sentential fragment flanked by the square brackets has been reconstructed. It is not part of the logical form of the sentence, but it is part of the explicature through which the utterance is understood. We may well, at this point, ask ourselves how active these components of the explicature can be. We may want to know if something like the following, which allows explicit anaphora, is licit. (10) [I was told (by my mother) that] I never wanted to jump off the swing; thus, at one point, I started asking her whether my sister was unhappy about that.6

As far as we are aware, a structure like (10) involving an anaphoric link between ‘her’ and ‘my mother’ does not seem to be licit. The use of ‘her’ must be sanctioned either by an anaphoric site (‘my mother’) or by a referent salient in context. If neither of the two cases materializes, the utterance is uninterpretable and, thus, fails to be licit. However, it is possible that the explicature should be minimal and, thus, we should have (11) rather than (10): (11) [I was told (by someone) that] I never wanted to jump off the swing; thus, at one point, I started asking her whether my sister was unhappy about that.

Utterance (11) does not appear to be much better than (10), but this may be due to the fact that ‘her’ has specific features while ‘someone’ has non-specific features. In fact, (12) appears to be much better than (11) because the null pronominal 0 in the object position with respect to ‘ask’ has unspecific features. (12) [I was told (by someone) that] I never wanted to jump off the swing; thus, at one point, I started asking 0 whether my sister was unhappy about that.

It is true that we have now reached a dilemma. How can we know if 0 has independent reference or whether it is anaphorically linked with ‘someone’? They may happen to have the same reference by pure chance. Yet it cannot be as a result of pure chance, as demonstrated by the following more specific interpretative act: (13) [I was told (by someone (who knew the event in that he participated in it) that] I never wanted to jump off the swing; thus, at one point, I started asking 0 (who knew the event in that he participated in it) whether my sister was unhappy about that.

6  Vividness should be taken to be an inferential syntactic structure which is associated with past tense reports (especially second person reports) requiring a logophoric interpretation. The structure is hosted through the syntactic device of a relative clause construction in an event which is implicit in the semantic/syntactic analysis of the main verb expressing the substance of the report, while we can say that the implicit event of the semantic analysis (a Davidsonian event structure) potentially hosts the syntactic construction that expresses vividness. A vividness implicit structure must be triggered by a number of contextual clues that say something about the participation of the speaker in the reported event.

First Person Implicit Indirect Reports in Disguise


The conclusion we have reached is fairly important, provided no flaws can be found in the argument, because it demonstrates that the explicated part of the explicature, that is the implicit component of the indirect report, is syntactically active at the level of anaphoric connections. Of course, the price we had to pay in order to reach this conclusion was to resort to minimality. We saw that, by choosing a more explicit (and specific) explicature, we could not obtain the same result. Clearly, for (13) to make sense, ‘someone’ and 0 must be restricted in the same way, because someone who was not present at the event (someone selected at random in an arbitrary way) could not necessarily be able to reply to the question being asked. This does not mean that, in deriving the explicature, we cannot further enrich the minimal explicature through further expansion. We can indeed, but we have to be careful to maintain the minimal explicature in the expanded construction so that anaphoric phenomena can still be licensed.

Second Person Indirect Reports and Implicit ‘De se’ Attributions Now consider a variation on (6) (I never wanted to jump off the swing): (14) You never wanted to jump off the swing.

The person who utters (14) can be my father (or mother) who has vivid memories of the events in question (sitting on the swing, not wanting to jump off the swing, etc.). However, if my sister says (14) (or something like (14)), given that she (my sister) was also too young to remember, it is likely that spoken by my sister, (14) is a sort of echo of what my father or my mother used to say in their narrations of the past (I remember this story recurred several times in my life). Thus, spoken by my sister, the utterance should be seen as having an implicit base, which is an indirect report like: (15) I was told (several times) that you never wanted to jump off the swing

or (16) Dad used to tell us that you never wanted to jump off the swing.

Our readers may now become impatient and begin to ask why we are making such a fuss about (14), which is a report, and possibly an indirect report in disguise. The fuss we are making about (14) is justified, because if (14) is merely based on an indirect report (and thus is an indirect report in disguise) its vividness is taken away. Vividness is like the direct participation in an event and is the recording of sensations linked to that event and experienced in that event. Thus, if (14) were not an indirect report, it would have a vividness that (14) qua an indirect report lacks: the sensations associated with the event and experienced from the inside. My sister, seeing that I never wanted to jump off the swing, may have experienced sensations of envy. Now, while the report may be not explicit about these experiences of envy, unless it is based on an indirect report, it should have a structure that reserves a slot for the experience at the syntactic level. Thus, our claim is that the implicit indirect


A. Capone

report (14) would be different, syntactically, from the report (14) which is not based on an implicit indirect report. Thus, (14) which is interpreted as being something different from an indirect report, would have the following structure: (17) You never wanted to jump off the swing in event e experienced by myself.

Adopting logical notation: There is an e, e < utterance U (You never wanted to jump off the swing) and you never wanted to jump off the swing in e and e has at least two participants (you, I) and e was experienced from the inside by myself.

If (14) is interpreted as being an implicit indirect report, it should have a different structure along the lines of (18). (18) There is an event e, and e < U (e is in the past with respect to U) and e has at least two participants (You, the indirect reporter) and e is the event of your not wanting to jump off the swing (at any of times t……tn of interval T) and U was reported to me by the indirect reporter in event e’, such that e < e’ < U.

Now, although the event is narrated from an external point of view being based on someone’s else’s narration, it is not to be excluded that there may be further structure to this utterance, given that the speaker, in reporting what was reported to her, may be an experiencer with respect to the event of the indirect report through which she knew the reconstructed event. In fact, in listening to the report, she may experience envy. Now, if I am correct, there should be a syntactic slot for the experiencer attached to the indirect report, along the following lines: There is an event e, and e < U (e is in the past with respect to U) and e has at least two participants (You, the indirect reporter) and e is the event of your not wanting to jump off the swing (at any of times t……tn of interval T) and U was reported to me by the indirect reporter in event e’, such that e < e’ < U and I was the experiencer with respect to the indirect report of U.

Could there be a syntactic motivation for this analysis? Perhaps data like the following will provide a syntactic motivation: (19) You never wanted to jump off the swing and I resented that.

In (19), the pronominal ‘that’ is surely anaphoric, but anaphoric to what? The utterance could be ambiguous in two ways: (20) You never wanted to jump off the swing and I resented that (the fact that you never wanted to jump off the swing).

This is a double implicit indirect report, as the feeling of resentment may have been reconstructed through the memory provided by the indirect reporter, who did not restrict herself to telling a story of an external event but also gave expression to feelings which were expressed at those times by one of the participants. Thus, (20) should be pragmatically equivalent to (21) You never wanted to jump off the swing and I resented that (the fact that you never wanted to jump off the swing), as my mother told me.

First Person Implicit Indirect Reports in Disguise


Thus, in (21) there are two indirect reports and the events reported may belong more or less to the same period (or they may be part of a sequence in the same period in the past). A different interpretation of (19) would be the following: (22) You never wanted to jump off the swing (mum used to tell us) and I resented that.

Could ‘that’ be now anaphorically linked to ‘Mum used to tell us?’. This is not very plausible, although we are not sure whether it is possible or impossible. But (22) appears to be different from (21), due to the temporal specification of the event of resenting. While in (21) the event of resenting is located in the past with respect to the indirect report (my mother told me), in (22) it seems to be in the future with respect to the event of mum’s telling of the story. Now it ought to be clear that (22) has further structure: (23) You never wanted to jump off the swing (mum used to tell us and I was the experiencer in the event of her telling us and I experienced x during and after her telling us) and I resented that.

It is clear at this point that the resentment concerns the experience x and not the real event, because the speaker cannot remember the event of resenting S’s not wanting to jump off the swing, from the inside. Instead, the speaker can experience the event of resentment from inside by hearing the indirect report, and only after hearing the indirect report. What is being resented is not the fact but the experience of the fact experienced through the indirect report. When we look at all this, we clearly see that, after all, (14) is not only an indirect report in disguise and not only a case of a double indirect report, but it is also a case of a ‘de se’ indirect report (in disguise). Reference to participation or experience of an event from the inside takes us back to Higginbotham’s (2003) view of de se reports. De se reports with PRO, according to Higginbotham, involved direct participation and experience of an event from the inside. Now, if this is the case, we can propose a minor, but not negligible or insignificant, modification of (23): (24) You never wanted to jump off the swing (mum used to tell us) and I remember experiencing x, I was the experiencer in the event of her telling us and I experienced x during/after her telling us) and I resented that.

Now, it seems to me that (24) has the structure of a de se report and, thus, satisfies certain desiderata of de se reports with respect to internal participation and from experiencing a memory from the inside.7 We have seen that by reasoning on the utterance and on what the evidential base for it ought to be, we have reconstructed an explicature which has a relatively complex logical form. Could this explicature be cancelled? Now, while Capone believes that explicatures are not cancellable for certain motivations, such as with issues of strong intentionality and logical well-formedness, we have specific reasons here to maintain that the explicature is not cancellable. In fact, cancelling the explicature 7  It is most interesting that these considerations conform with Jaszczolt’s (2018) view that “both I and you can be regarded as pertaining to modes of reference de se or self- ascription” (her argument is that when one attributes a quality (or predicate) to the second person, the attribution must be recovered by the addressee by using the first person).


A. Capone

means proceeding backwards in the enrichment process. But we should remember why we needed the enrichment process: we reconstructed an evidential base and a reasonable way to do so was to reconstruct the discourse that preceded the utterance (14). Unless an evidential base is reconstructed, the utterance remains up in the air. How does the speaker know that X never wanted to jump off the swing? The event being referred to belongs to a time in the past during which the speaker was too young to remember what happened. Thus, she has to rely on her mother’s memories and the narrations available to her memory. Reconstructing the evidential base is not easy. An objection we may receive is that a hearer may not be interested in the evidential base for the utterance. When a speaker speaks, we trust him and accept what he says, unless there are reasons for not doing so. Why should we care about his evidence? We may believe that the issue of evidence is negligible, particularly if the events in question do not involve a substantial modification of our habits, attitudes or actions. Why should we care how one knows a fact? That he indeed narrates a fact is sufficient warrant for believing that the fact is to be considered. But this is like admitting that, if someone were to be challenged by the question “How did you come to know this?”, he would not be able to reply. But this is not what is involved in the praxis of asserting things, narrating stories or facts. The evidential base may be implicit, but we should always be able to make it explicit, should it be required.

Operation Cases Now, consider the following utterance. (25) When I was operated on, the surgeons had to use a special tube so that I could breathe.

By the standards established by the previous examples, this can also be considered as an implicit indirect report. As everyone knows, when one is being operated on, one is unconscious and, thus, cannot be aware of the techniques or instruments being used. The speaker was himself told by his doctors (after the operation was over) that they had to use a special tube so that he could breathe, because a normal tube was too thick. All he remembers is the preparations for the operation and the moment he was given an injection of narcotic – he slept and did not see any of the events happening to him. Given that this is a stereotypical situation, the recipient of (25) is able to assess that the speaker does not know this fact first-hand, but that he has second-hand knowledge and, thus, his report is based on another person’s report (the surgeon’s). Therefore, the recipient is able to reconstruct the utterance (26): (26) (I was told by one of the surgeons that) When I was operated on, the surgeons had to use a special tube so that I could breathe.

Now, suppose that the recipient replies: (27) This happened to me too.

An opponent of the explicature view might respond that such a reply will be anaphoric to only one portion of the utterance – that is, the explicit part, but cannot

First Person Implicit Indirect Reports in Disguise


be anaphoric to the implicit part. In other words, ‘This’ is anaphoric to the fact ‘that the surgeons used a special tube so that I could breathe’, except for the reference of ‘I’, which should be switched to the current speaker and the reference of ‘the surgeons’, which has to refer to the surgeons in a different context (the context of the recipient’s memories). The form of anaphora being used here is quite sloppy. It is not the specific fact that is being taken up anaphorically, but the general structure of the fact, while the references of certain NPs are to be anchored in a new context. However, the opponent of the explicature view may maintain that even this kind of sloppy anaphora cannot be applied to the implicit component of the explicature. Thus, ‘This’ cannot refer to the whole utterance ‘I was told by the surgeons that when I was operated on, a special tube was used so that I could breathe’. But are we sure that this must be the case? Consider the following case: (28) A: (I was told by one of the surgeons that) when I was operated on, the surgeons had to use a special tube so that I could breathe. B: This happened to me too. My surgeon told me the same story.

(28B) can be interpretatively ambiguous. It can either mean ‘This fact (the surgeons used a special tube) happened to me too’ or ‘This fact (I was told that the surgeons used a special tube) happened to me too’. In the case in which ‘This’ refers to the story by the surgeons, anaphora must be said to be sanctioned by an implicit component of the explicature. B’s utterance simply makes the explicature explicit and disambiguates that section of the utterance which is being referred to anaphorically. We have made enough progress in showing that the explicature is syntactically active in sanctioning anaphora, either through a part of it (a nominal) or through the whole utterance as reconstructed through the explicature.

Extending the Analysis The story we have told so far is similar to that proposed by Capone (2016), a propos of simple sentences and substitution failure. The Fregean apparatus has so far been applied to belief reports and verbs that introduce intensional contexts. We would normally expect replacements of NPs to be licit in simple sentences (salva veritate), but not in intensional contexts, such as ‘Mario believes that Elizabeth is in Rome’. Given that Mario does not believe that Elizabeth is Queen Elizabeth in disguise, even if we know this identity we cannot freely replace ‘Elizabeth’ with ‘the Queen of England’, because Mario would never give his assent to that substitution, given what he knows and believes. However, Saul (2007) discusses various cases in which even in (apparently) simple sentences Leibniz’s Law does not work and substitutions are not licit, as they result in sentences having different truth conditions. One such case is: (29) Clark Kent went into the phone booth and Superman came out.


A. Capone

If we replaced ‘Superman’ with ‘Clark Kent’ we would obtain, according to Saul, a statement having different truth-conditions. How can this be? We do not need to invoke universal opacity, as Saka p.c. does, to explain cases like these. Capone’s solution (in Capone 2016) was to posit that an intensional context is created through an explicature, as in the following case: (30) (The story says that/we are told that) Clark Kent went into the phone booth and Superman came out.

In other words, (29) is an implicit indirect report. Capone (2016) invoked the machinery of explicatures and, in particular, free enrichments to explain this example (and similar ones). Given the implicit indirect report in which there is a verb of saying, we now have an intensional context, and this suffices to explain why substitutions of coextensive NPs cannot occur (salva veritate). Now, given that this case, even if it does not involve the first person, reminds us of those cases which were previously discussed concerning the first person, we want to establish whether the explicated (implicit) parts of the explicature in (30) are also syntactically active. This would be a positive step towards demonstrating that a case can indeed be made for the presence of an explicature. Certainly, it only takes a small stretch of the imagination for us to conceive examples like the following: (31) Clark Kent went into the phone booth and Superman came out and a similar story can be said of Spiderman.

The reconstructed explicature in (31) is spelled out in (32) (32) (The story says that/we are told that) Clark Kent went into the phone booth and Superman came out and a similar story can be said of Spiderman.

Now, while we have made some progress in demonstrating that the explicature is syntactically active because it is a site for anaphoric uptake (similar), we should reply to the skeptic’s objection that in this case ‘story’ can be taken in a (more) factual sense. We do not need to presuppose that the story we are talking about is a fictional one, as it may be a story about real world facts. We can accept that there may be an ambiguity of this sort. But the purpose of explicatures is also to settle ambiguities and to select one reading out of possible number. Thus, in one reading, the story is not a factual one, but a fictional one. And this is all that is required to explain that substitution of an NP with a coextensive one (salva veritate), in such an ad hoc intensional context, is not licit.

Conclusion This paper shows us how we should deal with explicatures. It is not sufficient to state that there is an explicature. We would have to extend such a story by demonstrating that the explicature has certain logical properties and is syntactically active. One way of showing that an explicature is syntactically active is to demonstrate that it can sanction anaphora. Possibly, we need to further discuss this important topic,

First Person Implicit Indirect Reports in Disguise


but at least we have shown that we need to create a new chapter on the theory of explicatures. Most importantly, we have noticed that a large gap may exist between what is said and what is intended, and that superficial syntax is a poor guide to the complex logical form of the utterance. When we talk about the logical form of the utterance, we are referring to the explicature and its syntactic configuration. After all, even if explicatures are pragmatic devices (consisting of a sentence and a pragmatic component that can be added to the sentence or can supersede the sentence altogether), they need to have a syntactic structure, and we have proven that speakers’ intentions can help shape the syntactic structure of an explicature.

References Capone, A. (2003). On Grice’s circle. RASK.  International Journal of Language and Communication, 19, 3–32. Capone, A. (2006). On Grice’s circle (a theory-internal problem in linguistic theories of the Gricean type). Journal of Pragmatics, 38(5), 645–669. Capone, A. (2008). Belief reports and pragmatic intrusion (the case of null appositives). Journal of Pragmatics, 40(6), 1019–1040. Capone, A. (2009). Are explicatures cancellable? Toward a theory of the speaker’s intentionality. Intercultural Pragmatics, 6(1), 55–83. Capone, A. (2010). Between Scylla and Charybdis’: The semantics and pragmatics of attitudes ‘de Se’. Intercultural Pragmatics, 7(3), 471–503. Capone, A. (2013). Explicatures are NOT cancellable. In A. Capone, F. L. Piparo, & M. Carapezza (Eds.), Perspectives on linguistic pragmatics, perspectives in pragmatics, philosophy & psychology 2 (pp. 131–151). Cham, Switzerland: Springer. Capone, A. (2016). The pragmatics of indirect reports: Socio-philosophical considerations. Cham, Switzerland: Springer. Carston, R. (2002). Thoughts and utterances: The pragmatics of explicit communication. Oxford, UK: Blackwell Publishing Ltd.. Higginbotham, J. (2003). Remembering, imagining, and the first person. In A.  Barber (Ed.), Epistemology of language (pp. 496–533). Oxford, UK: Oxford University Press. Holt, E. (2016). Indirect reported speech in interaction. In A. Capone, F. Kiefer, & F. L. Piparo (Eds.), Indirect reports and pragmatics (pp. 167–187). Cham, Switzerland: Springer. Jaszczolt, K. (2006). Meaning merger: Pragmatic inference, defaults, and compositionality. Intercultural Pragmatics, 3(2), 195–212. Jaszczolt, K. (2018). Pragmatic indexicals. In M. Huang & K. Jaszczolt (Eds.), Expressing the self: Cultural diversity and cognitive universals. Oxford, UK: Oxford University Press. Saul, J. (2007). simple sentences, substitution, and intuitions. Oxford, UK: Oxford University Press.

Assessing Dialectical Relevance Using Argument Distance Douglas Walton

Introduction Relevance is an important concept for disciplines such as philosophy (Grice 1975), cognitive science (Giora 1985, 1997; Sperber and Wilson 1995), education (Erduran 2008; Macagno 2016; Macagno et al. 2015; Nussbaum and Edwards 2011; Rapanta et al. 2013), logic (Epstein 1979), argumentation theory (Walton 2004), linguistics (Blakemore 2002; Carston 2004), linguistics and artificial intelligence (Hobbs 1979; Lascarides and Asher 1993), discourse analysis (Taboada 2009), and applied argumentation theory (Walton 2003b; Walton and Macagno 2019). Although there was a significant body of literature on relevance logic that has been influential in formal logic (Anderson and Belnap 1975), this development does not appear to have much effect on the pragmatic approaches to relevance developed in the last decades in informal logic and linguistics, where relevance is seen as a concept employed in communicative contexts of argument use. This pragmatic approach is centrally directed to dialogues containing argumentation, benefiting from recent work in computational argumentation studies. Macagno (2018) holds that assessment of relevance depends on the argumentation schemes and implicit premises used to make up a chain of argumentation connecting one move in a dialogue with another. This paper gives additional support to this pragmatic approach by refining it with some argumentation tools from artificial intelligence. The tools are applied to a series of illustrative examples. Macagno (2018, 42) used three factors to assess the relevance of one proposition or argument (including other speech acts as well) to another: (1) the number of

D. Walton (*) University of Windsor, Windsor, ON, Canada e-mail: [email protected] © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 F. Macagno, A. Capone (eds.), Inquiries in Philosophical Pragmatics, Perspectives in Pragmatics, Philosophy & Psychology 27,



D. Walton

inferential steps required to get from the one to the other, (2) the types of argumentation schemes used in making the transitions at each step, and (3) the implicit premises required in the arguments. The key factor underling these factors is what Macagno calls inferential distance: “Inferential distance can be defined as the number (quantity) and acceptability (quality) of the argumentative inferences needed for connecting a premise to a conclusion” (Macagno 2018, 57). In this view, “relevance can be more or less acceptable or effective from an argumentative point of view” (Macagno 2018, 57), depending on “the probative force of the type of argument and the number and acceptability of the needed implicit premises or intermediate conclusions”. This paper will be primarily concerned with the relevance of arguments, as opposed to the relevance of other speech acts such as questions, replies to questions, and so forth. Relevance is a very broad concept used in virtually all fields and commonly used in everyday discourse in such a broad, vague and often confusing way that the task of trying to make sense of it is intimidating. The term has more specific meanings in both logic and law, but also in both these instances it has proved very hard to define precisely with any unanimity regarding what it should be taken to mean generally and how it should be applied as a tool to include or exclude arguments as admissible in particular cases. At any rate, the attempt of this paper is to apply current argumentation theory and some resources recently developed in computational argumentation, to making sense of relevance of one argument to another, or one part of an argument to another part. Three examples are analyzed to show how the argument mapping technology works. The first one is an example of an irrelevant argument from a logic textbook in a section on fallacies of relevance (Sect. 4). The second is from a trial where an objection of irrelevance is made (Sect. 5). The third one is from a parliamentary debate where the Speaker of the House asks a Member of Parliament to show that her argument is relevant and she responds to his request by arguing that it is relevant (Sect. 6).

The Pragmatic Nature of Dialectical Relevance Dialectical relevance of an argument is defined in (Walton 2004) in the following pragmatic manner: an argument is dialectically relevant if and only if it is an appropriate move, or part of a connected sequence of appropriate moves used in a multiagent dialogue exchange where each move is normatively supposed to contribute to the goal of the dialogue. The term ‘dialectical’ simply refers to argumentation in a context of dialogue, such as a persuasion dialogue or one of the other types of dialogue. Many would say that this term can be equated with the term pragmatic. Certainly the two terms are closely related, but they are contested terms, and scholars in different fields might be inclined to dwell on perceived differences between them.

Assessing Dialectical Relevance Using Argument Distance


The definition of dialectical relevance above is explicitly normative in nature, meaning that it is meant to refer to abstract models of dialogue of the kind currently found in the field of computational argumentation where a model consists of a set of goals and rules representing what is supposed to take place in rational argumentation. Such formal models can have various purposes, but what has been most prominent in recent argumentation studies is to apply the model to short spans of text from natural language discourse to determine, based on this evidence, whether, how and why the argument found in the text conforms to the norms of the model, conflicts with them, or departs from them. The norms basically specify what is a good or a bad argument or other move in dialogue with regard to how it fits with the norms of the abstract model of communication. A formal model of dialogue is defined as an ordered 3-tuple {O, A, C} where any particular sequence of argumentation is modeled as moving through three stages. O is the opening stage where the issue to be discussed is formulated, A is the argumentation stage, where pro and con arguments are put forward by the two sides, and C is the closing stage where the issue is ideally resolved, or at least where the discussion is terminated. Dialogue rules, often called protocols, define what types of moves are allowed at each stage. The familiar dialogue typology (Walton 2003a, 9) is displayed in Table 1. The collective goal of a persuasion dialogue is to elicit the most powerful pro and con arguments on both sides so that the reasons behind the arguments put forward on both sides are elicited by interactions between the pro side and the con side. The main benefit of this exchange of arguments is to give both participants greater depth of understanding of the subject being debated. In order to achieve this benefit by motivating the participants, persuasion dialogue is highly adversarial. ‘Winning’ means producing an argument that is strong enough to discharge the burden of persuasion set at the opening stage. There are definite rules, goals and procedures that enable this normative model to exclude bad arguments and include ones that are good, according to the standards set in the model. Argumentation has adopted the Gricean view that relevance needs to be evaluated pragmatically in relation to a context of argument use where two or more parties exchange speech acts, moves in a dialogue framework where the dialogue itself has communicative goals. Table 1  Standard dialogue typology Type of dialogue Persuasion Inquiry Discovery Negotiation Information Deliberation Eristic

Initial situation Conflict of opinions Need to have proof Need an explanation Conflict of interests Need information Practical choice Personal conflict

Participant’s goal Persuade other party Verify evidence Find a hypothesis Get what you want Acquire information Fit goals and actions Hit out at pponent

Goal of dialogue Resolve issue Prove hypothesis Support hypothesis Settle issue Exchange information Decide what to do Reveal deep conflict


D. Walton

The current way of understanding relevance in argumentation theory arises from Grice’s maxim of relation. According to this maxim, participants in a conversation share a common communicative purpose, a common goal underlying their verbal interaction (Grice 1975, 45): Our talk exchanges do not normally consist of a succession of disconnected remarks, and would not be rational if they did. They are characteristically, to some degree at least, cooperative efforts; and each participant recognizes in them, to some extent, a common purpose or set of purposes, or at least a mutually accepted direction.

Grice formulated his underlying conversational principle as follows (Grice 1975, 47): I expect a partner’s contribution to be appropriate to the immediate needs at each stage of the transaction. If I am mixing ingredients for a cake, I do not expect to be handed a good book or even an oven cloth (though this might be an appropriate contribution at a later stage).

Macagno (2018), following Grice, advanced a pragmatic approach to relevance that conceives of it as a relationship between pairs of speech acts put forward as moves in a dialogue. In his view, the thread of dialogue connecting a sequence of argumentation provides the context from which evidence can be collected and used as part of the evidence to bear (pro or con) on a claim that one speech act is relevant to another. One aspect of this theory that should be observed at the outset is that the same argument can fit the norms of one type of dialogue while violating the rules of another. One of the most obvious examples is the argument from appeal to force or threat of force, typically frowned on in logic, and even standardly classified as a fallacy in the logic textbooks, the ad baculum fallacy. This type of argument is generally considered to be irrelevant in normative models of persuasion dialogue whereas it is generally regarded as a standard form of argument in negotiation dialogue. For example, in union management negotiations, the one side threatens to go on strike and the other threatens to cut wages or lay off staff. It is not that arguments from threats are always acceptable in negotiation, but in contrast to the situation in persuasion dialogue, they are generally accepted as relevant moves in argumentation. One aspect of this theory is that there can be dialectical shifts as argumentation proceeds in a dialogue. According to the theory of Walton and Krabbe (1995), in some cases the second argument is structurally embedded in the first, meaning that it contributes to the goal of the first dialogue, and therefore the shift does not indicate a normative failure in the sequence of argumentation from the one type of dialogue to the other. On the other hand, some shifts are an indication that a fallacy of relevance has been committed, because the argumentation in the second type of dialogue does not contribute to the goal of the first dialogue.

Assessing Dialectical Relevance Using Argument Distance


Probative and Topical Relevance It is important right at the beginning to clarify a distinction between two kinds of relevance called probative relevance and topical relevance. Two arguments are said to be probatively relevant to each other if one can be used to support the other to make it stronger or to attack the other by casting doubt on it or even to discredit it as worthless. One such argument may be said to be directly relevant to another if it probatively supports or attacks the other in one step. But one argument can also be indirectly relevant to another if the two are connected by a sequence of arguments in more than one inferential step. Probative relevance of this kind can be contrasted with topical relevance. It is well known that in order for an argument to be deductively valid it is not necessary for the premises to be relevant to the conclusion. For example, if the premises of a deductively valid argument are inconsistent with each other, any proposition at all follows deductively from them, even if the premises are about bananas and the conclusion is about world peace. This was the reason why the logicians have worked for so long on devising relevance logics, even though to this day they have not proved applicable in a general way to argumentation in natural language discourse. However there is one branch of relevance logic called relatedness logic (Epstein 1979) that has provided a formal model of topical relevance. Two propositions are said to be topically relevant to each other in one species of relatedness logic if the two share subject matter overlap. For example the proposition bananas are yellow shares a topic with the proposition that over 20,000 bananas were imported from Honduras to Ontario in 2015. These two propositions are topically relevant to each other because they share a common topic, bananas. They are not probatively relevant to each other, because you can’t prove or disprove the one exclusively from the other. Of course, in a special case with a fairly large knowledge base, you may find a sequence of inferences connecting the two so that one could be probatively relevant to the other one given that enough of these connections are drawn from the information in the knowledge base. Both probative and topical relevance need to be seen as relative to the information in the knowledge base of the case at issue. In other words, topical relevance in argumentation can have a pragmatic aspect as well. Another important point of contrast is that probative relevance is transitive, whereas topical relevance is not. Probative relevance is definitely transitive, as shown by the many examples of chaining of probative arguments in this paper below. Transitivity for probative arguments means that if argument 1 is probatively relevant to argument 2, and argument 2 is probatively relevant to argument 3, then it follows that argument 1 is probatively relevant to argument 3. However, transitivity fails for topical relevance, as shown by the following example from (Epstein 1990, 77) the proposition that John loves Mary is topically relevant to the proposition that Mary has two apples and Mary has two apples is topically relevant to the proposition that 2 + 2 = 4. Nevertheless it does not follow that the proposition John loves Mary is relevant to the proposition 2 + 2 = 4.


D. Walton

Even though we can make use of topical relevance occasionally, all three of the examples in this paper will be concerned with probative relevance. To model argumentation we will use the rules, conventions and argument mapping styles of the Carneades Argumentation System (Gordon 2010; Walton and Gordon 2012), a formal and computational argumentation system. In this system an argument is defined as a set of premises and a conclusion. The definition of an argument also includes the notion of an exception, because the system is meant to be useful to model defeasible argumentation in which there are exceptions to rules. In Carneades, an argument can be either pro or con a statement or pro or con another argument. To explain the basic ideas behind relevance as straightforwardly as possible, in this paper we will only be concerned with simple examples that illustrate relevance or irrelevance, and although the examples are defeasible arguments, we will not be concerned with exceptions, or with arguments that are meant to attack and defeat another argument. All of the arguments considered will be pro arguments where the premises are meant to support the conclusion, or where one argument is meant to support another.

The Nuclear Waste Example Relevance is a highly abstract and general notion that has proved to be hard to pin down with precision as a concept that can be used for studying argumentation formally. One good way to get insight into the meaning of relevance as a logical notion is to look at examples given in the logic textbooks to illustrate fallacies of relevance. These are cases where an argument is put forward that is irrelevant but might pass inspection as being relevant by someone who doesn’t have enough time to carefully inspect a clever argument directed against him. Many of the different fallacies long recognized in logic are classified as fallacies of relevance, including the ad hominem argument (personal attack), the ad baculum argument (appeal to force or threats), the ad misericordiam argument (appeal to pity), and so forth. This means that these kinds of arguments typically fail, or at any rate very often fail, because they are failures of relevance. But there are basically two different species of fallacies of relevance that have long been recognized that do not fit into any of these other more specific categories. These are the fallacy of wrong conclusion (ignoratio elenchi) and the red herring fallacy. In this type of fallacious argument, an arguer presents an argument that is otherwise valid and reasonable except that its conclusion is a different proposition from the one that is supposed to be proved. The red herring is a fallacy of distraction where the line of argumentation goes in a wrong direction away from the proper argumentation that would take it towards the collective goal of the dialogue. The next two examples given below will illustrate both of these variants. The following example is a close paraphrase of one that was used some time ago in a leading logic textbook (Hurley 2003, 127). The name of the company has been

Assessing Dialectical Relevance Using Argument Distance


anonymized to XY, and CR indicates some source of consumer information not specified here. CR reports that XY batteries last longer than YZ batteries. But do you realize that corporation XY also makes nuclear weapons, has left tons of nuclear waste and failed to dispose of that waste properly. This action is dangerous to human life in our country, showing that XY is an unethical company. Obviously, the CR report is wrong.

One could see how this argument might be persuasive to someone who does not have enough time in the circumstances to think carefully about the reasoning used in it. On reflection however we can see that the argument arrives at a conclusion other than one that is supposed to be proved. The nature of the error can be revealed by examining the structure of the argumentation visually presented in Fig. 1. The arguments, a1, a2 and a3 are shown in rounded nodes (circles), while the propositions that make up the premises or conclusion of an argument are represented as rectangular nodes. a1 and a2 are pro (+) arguments while a3 (−) is a con argument. Directed to an audience concerned about buying products that might have harmful consequences, the argument from a1 to a3 is a reasonable basis for justifying the conclusion that we should not buy XY batteries. But as you look carefully at the text of the argument as quoted above, you need to ask yourself what the conclusion of the argument is supposed to be. There are two textual indicators. One is that the passage opens with a statement that the CR reports that XY batteries last longer than YZ batteries. The other is the last sentence, stating that obviously CR is wrong. Both these statements strongly suggest that the conclusion of the argument is the proposition that CR is wrong. But it is easy to become confused and assume that the argumentation from a1 and a3 is directed to attacking the proposition that we should buy XY batteries. The problem is that the irrelevant argument a2 may seem relevant because it is confused with the relevant argument a3. The premises of argument a1 seem reasonable, as representing what was perceived to be the situation at the time. Let’s say that in 1994 that XY was a well-­ known large corporation. The premise that the social cost of XY’s irresponsible behavior has been tremendous is supported by an additional argument based on

Fig. 1  The nuclear waste example


D. Walton

the premise that we are left with thousands of tons of nuclear waste. We make no claim here, of course, that any of these premises are true or even plausible, but for the purposes of the example we can appreciate how they might have been acceptable assumptions for the audience to whom the textbook was directed in 1994. So let’s say, for the sake of the example, that argument a1 is reasonable, meaning that it is a valid argument with acceptable premises. Using these assumptions, argument a1 is relevant when its conclusion is taken to be statement that we should not buy XY batteries. But the problem of relevance is that as the text clearly indicates, the real proposition supposed to be proved is the statement that CR is wrong. Hence one can see the appropriateness of calling this error the fallacy of wrong conclusion. In the case of the nuclear waste example it is fairly easy for a reader of the logic textbook to see why each of these arguments is considered to be logically irrelevant. It is a realistic enough example, but it has been made to be easy for purposes of explaining fallacies of relevance to students who will be the main readers of the textbooks. In many other realistic cases however, such as the next two to be considered, the argumentation can be more complex.

The Bank Robbery Example The following dialogue quoted from (Mauet, 2005, 533), repurposed from its use in (Walton and Macagno 2016, sec. 4), can illustrate some dialectical and logical argumentation features of the evidential objection of irrelevance in law. Prosecutor: Mrs. Higgins, were you at the bank on June 1, one week before the robbery? Mrs. Higgins: Yes. Prosecutor: Did you see the defendant that day? Defense lawyer: Your Honor, we object on relevance grounds. May we approach? Judge: Yes [Lawyers come to the bench.] Prosecution, where’s this testimony going? Prosecutor: The witness will testify that she saw the defendant outside the bank that day, that he was looking at the bank, but never came inside. Judge: What does that prove? Prosecutor: It shows that the defendant was already planning the robbery and casing the bank. Defense lawyer: The fact that he was outside the bank hardly proves he was planning to rob it. Judge: The objection is overruled. The witness may answer.

In this case the three participants, the judge, the prosecutor and the defense lawyer, are engaged in a metadialogue about an earlier dialogue that took place in the trial between the prosecutor and the defense lawyer. The issue of the metadialogue is whether the witness testimony evidence which the prosecutor started to introduce into the trial is relevant or not. The defense lawyer objected on grounds of relevance. To consider the objection the judge asked the question, where is this testimony going. This question is a key indicator of possible failure of relevance of an argument. Logically speaking, the problem is one of how you get from the proposition that Mrs. Higgins saw the defendant looking at the bank to some conclusion that can

Assessing Dialectical Relevance Using Argument Distance


be used as evidence to help prove the ultimate conclusion at issue, the proposition that D robbed the bank. The problem is: how do we get from A to B? The prosecutor bridged the gap when he stated that it shows that the defendant was planning the robbery by carrying out the action standardly called casing the bank. In this case, the issue of relevance is based on some projected sequence of logical reasoning that starts from an item of evidence, in this case testimonial evidence, and moves forward through a sequence of logical reasoning to the ultimate probandum, the ultimate thesis to be proved by the prosecution (this case being a criminal trial). Walton and Macagno (2016, sec. 4) used this case to illustrate the profiles of dialogue technique, where a dialogue is mapped onto a pair of graphs that can be used to explain fallacies and other factors important for evaluating arguments, such as the factor of relevance. In this case the type of relevance involved is probative relevance, that is, relevance about proving or disproving something at issue. The trial can be modeled as a persuasion dialogue where the issue is whether or not the defendant committed the crime. The prosecution must prove that the defendant committed the crime using a sequence of evidence-based argumentation that proves this thesis beyond reasonable doubt. So once again we have to ask the question of how we get from A to B. How do we get by a sequence of logical reasoning from the witness testimony to the prosecutor’s thesis that the defendant robbed the bank? Without using the whole profiles technique, it is enough here to simply map the logical features of the dialogue into a standard argument diagram that incorporates argumentation schemes. In the next example, two argumentation schemes are used. This is shown in Fig. 2, with the notation pr, which is displayed in two of the round argument nodes representing the argumentation scheme for practical reasoning. The other round argument node contains the notation wt, which stands for argument from witness testimony. The plus sign indicates that the argument is a pro argument supporting its conclusion. D stands for the defendant and H stands for Mrs. Higgins. Scheme for Argument from Witness Testimony Major premise Minor premise Conclusion

Witness W says that proposition P is true W is in a position to know whether P is true P may be presumed to be true

Here is a simple version of the scheme for argument from practical reasoning (pr in Fig. 2) from (Walton et al. 2008, 94–95). Scheme for Basic Practical Reasoning Major premise Minor premise Conclusion

I (a rational agent) have a goal G Carrying out this action A is a means to realize G Therefore, I ought (practically speaking) to carry out this action A


D. Walton

The way practical reasoning is used generally does not conform to the requirements of the simplest scheme above, however. Instead, it fits into a sub-scheme that has this form: Argument from Means and Action to Plan Major premise Minor premise Conclusion

Carrying out action A is a means for agent ϕ to achieve goal G ϕ carried out action A ϕ was planning to achieve goal G

This variant of the practical reasoning scheme can be called argument from means and action to plan (MAP). Two instances of it are shown in Fig. 2. The dashed borders of the rectangles around three of the propositions indicate that they are implicit premises that can be assumed to hold on the basis of common knowledge and that are required to enable the conclusion to be derived from the premises by logical reasoning. Figure  2 shows by example how it is possible to automatically fill in such premises by using argumentation schemes. What is very nice about this example is that by mapping the dialogue onto an argument diagram, the chain of argumentation needed to establish relevance is clearly displayed. Let’s start towards the right side of the diagram and move toward the ultimate conclusion at the left, the proposition that the defendant was planning to rob the bank. First we have the introduction of the argument from witness testimony to show that D was looking at the bank. The next step of the argument is to use the MAP scheme for practical reasoning, by bringing up the implicit premise that looking at a bank is a means of casing it, and then using these two premises to derive the conclusion that D was casing the bank. From there, by another instance of MAP practical reasoning, the conclusion that the defendant was planning to rob the bank can be derived. We can see that this conclusion does not by itself prove that

Fig. 2  Relevance in the bank robbery example

Assessing Dialectical Relevance Using Argument Distance


the defendant robbed the bank, but it is an element that can be combined with further arguments that lead to this ultimate conclusion for the prosecution. That is enough to show that the witness testimony should be accepted as relevant at this stage of the dialogue between the three parties. What does our analysis of this example tell us about Macagno’s notion of inferential distance? As shown in Fig. 2, there is a sequence of argumentation starting from some evidence that H saw D looking at the bank and that moves the line of argumentation forward to support the conclusion that D robbed the bank. The subarguments in this line of argumentation fit argumentation schemes that bring up implicit premises that have to be added in order for the argumentation to connect up and reach the ultimate conclusion. Already here we can see how schemes can work as tools to prove that these missing premises are relevant. This way of modeling relevance fits with Macagno’s use of the notion of inferential distance to evaluate arguments for and against relevance. We can visualize inferential distance by eliciting the sequence of inferential steps it takes to get from a given argument (or premise) to the ultimate conclusion. Distance, in Macagno’s theory, also depends on whether the intervening premises are “unacceptable or poorly acceptable” and whether the “inferences that proceed from premises or result in intermediate conclusions can be doubtful” (Macagno 2018, 57). This factor applies in this example where the defense lawyer argued that the fact that D was outside the bank “hardly proves he was planning to rob it”. This statement clearly shows a weakness in the argument, but one that was rebutted by the prosecutor. What is so nice about this example is that, as often happens in trials, one side makes an objection that the other side’s argument is irrelevant, and the other side replies by offering an argument to support the claim that his argument is relevant. In other words, not only are we examining the pro and con argumentation in a text, we are also using pro and con argumentation at a metalevel to reasonably judge whether the ground-level arguments in the text fit the requirements needed to be relevant or not. The tools useful at the metalevel include argument maps, argumentation schemes and some notion of distance that can tell us how one argument is related to another by an intervening sequence of inferences joining the one to the other.

The Tuna Fish Example The second example comes from a debate in the Canadian House of Commons where the question set for the debate was to consider specific amendments to a bill called the Family Allowances Act. There was a specific proposal for increasing certain government benefits paid to families through this bill that was to be argued for or against. During the middle of the debate, an MP (Member of Parliament) who


D. Walton

was against the amendments mounted a lengthy argument that attacked the government for a number of faults. This attack alleged that government officials failed to apply health standards to block the sale of cans of substandard tuna fish that should not have passed inspection from being made available to consumers. This issue had been covered by the media at the time and was called the “tuna scandal” by some. The Acting Speaker of the House questioned the relevance of these remarks in the exchange quoted below from (Walton 1989, 205–207). Speaker: Order, please. The Hon. Member knows that we are debating the amendment to the Bill on family allowances. I do not know why we are debating tuna fish. I hope the Hon. Member will get back on track. MP: Mr. Speaker, my reference to tuna relates to the health and welfare of Canadians, which is also being dealt a fatal blow as a result of this particular legislation on family allowances.

The MP was not at all discouraged by the allegation that her remarks about the tuna scandal were irrelevant and she went on to try to find a connection between the tuna scandal and the issue of amending the Family Allowances Act (Walton 1989, 206). MP: With regard to the main question, which is family allowances, we are talking about the people who would be most affected by this cutback. The government says that $22 is not a lot for Canadian families. Well, families and single parents who are struggling to raise small children, often surviving on tuna, are being directly assaulted by the government’s anti-family budget measures [my italics].

The expression “often surviving on tuna” was an attempt by the MP to make a connecting sequence of argumentation between the amendment and the tuna scandal. But any such sequence of argumentation, if there was one, was very weak and questionable. Instead of making a tuna connection, it looked very much like the MP’s argument could be called a red herring. There is a rule of relevance for debates in the Canadian House of Commons, but basically it is left to the discretion of the Speaker of the House to make such interventions on a case by case basis. Basically, Rule of Order for relevance directs the Speaker to exclude any discussion that strays from the question being debated in the House. The Speaker can warn an MP to desist, or if this warning is not acted upon, he can direct the MP to stop speaking. Textual evidence of the Speaker’s recognition of the problem and his efforts to solve it is given when he says, “I do not know why we are debating tuna fish”, and talks about “back on track”. Tracking the sequence of argumentation, and the textual evidence in this case, including the protracted dialogue between the MP and the Speaker of the House is quite an extensive job. Extensive comments on different aspects of the case with regard to relevance have already been given in (Walton 2004) as well as (Walton and Macagno 2016; Macagno 2018). There is topical overlap between the tuna scandal issue and the issue about the housing bill, as argued by the MP.  Let us offer a possible

Assessing Dialectical Relevance Using Argument Distance


interpretation of the line of argumentation to see how it might cross the distance between the explicitly given premises through a hypothetical sequence of argumentation that would connect it up with the MP’s claim that the topic of tuna fish is relevant to the amendments to the government’s Bill being debated. Starting at the right side, there are three premises that form a linked argument a1 supporting the conclusion that the changes to the bill would deprive small children of tuna fish. The weakest of these three premises is the bottom one, stating that often these children are surviving on tuna fish. This proposition seems fairly implausible, and certainly calls out for some evidence to back it up before it should reasonably be accepted in the context of the debate. From that point in the sequence we go to argument a2, which is to implicit premises. The conclusion of argument a2 is then used in the next argument a3 to support the MPs claim that tuna is relevant to the debate on the Bill. The two implicit premises and argument a2 also do not seem very plausible, even though if they were accepted, they would provide a reasonable basis for supporting the conclusion that families are being assaulted by the government’s anti-family budget measures. The attempt to draw up an interpretation of the MP’s argument that would join up its explicit premises to lead to her defense of her claim that tuna is relevant to the debate on the Bill displays at least a possible account of the distance. This account reveals the weaknesses in the argument by bringing out premises that are not very plausible and that cast a burden of proof on it. The speaker has put a burden of proof in place on the MP to show why the tuna fish issue is relevant, and she has responded to that request, but the interpretation of the sequence of argumentation and its distance in Fig. 3 reveal some of the weaknesses in this line of argument. This suggests that the burden of proof could be shifted back to the MPs side to answer further questions about how these questionable premises could be supported by some evidence.

Fig. 3  Tracking the distance in the tuna fish example


D. Walton

Modeling Probative Distance with Carneades Next we turn to two abstract examples that illustrate relevance, and failure of relevance in argumentation. The examples concern arguments made up of inferences from premises to conclusions, linked to other arguments in a format that is called argument chaining in computing. Such examples also concern deductive and probabilistic forms of argument, but in realistic argumentation. In order to apply relevance in everyday conversational reasoning in natural language, non-statistical forms of argument are in the forefront, such as argument from witness testimony, argument from expert opinion, argument from precedent, practical reasoning from goal to action, or from action to goal, and so forth. In Carneades, arguments are combined by recursive definitions of applicability of an argument and acceptability of its conclusion. An argument is said to be applicable if and only if all its premises are given as accepted or if each of them is an acceptable conclusion of another argument, and none of its exceptions is accepted or is an acceptable conclusion of another argument. To use this approach in a way that will be more familiar to readers, an argument will be said to be probatively relevant if all its premises are accepted, and there is a dialectically valid argument leading from the premises to the conclusion, where the dialectical validity of an argument is determined by argumentation schemes, including defeasible argumentation schemes as well as deductively valid argumentation schemes. Many different phenomena of argumentation can be modeled by Carneades, but in this paper we will only be concerned with probative relevance. Basically, a set of premises in an argument are probatively relevant to the conclusion if they can be used to prove the conclusion, or at least to support it well enough to satisfy the standard of proof required. But as shown in the examples, and any given instance where relevance is an issue, the context needs to be taken into account because arguments are connected to other arguments forming a sequence of argumentation leading to an ultimate conclusion that is supposed to be proved in that context. So relevance is all about the relationship over the sequence of a particular argument, or premise of an argument, in relation to the proposition that is supposed to play the role of the ultimate conclusion in that case. As shown by the three examples we considered, it is important in studying relevance to work with defeasible arguments that give probative weight towards supporting a conclusion leading to tentative acceptance of the conclusion that may have to be withdrawn later if new evidence comes in. As previously mentioned, such an argument can be deductively valid, such as one having the form of deductive modus ponens. But as shown by the examples, schemes such as argument from witness testimony and argument from practical reasoning are important. However the argument map shown in Fig.  4, and the argument maps in the rest of this paper, are meant to be abstract representations of typical kinds of argumentations with an abstract representation of the distance comprised of the intervening arguments between the evidence and the ultimate conclusion. For this purpose we abstract from the argumentation schemes and other details.

Assessing Dialectical Relevance Using Argument Distance


Fig. 4  Step 1 in abstract example 1

First Abstract Example Figure 4 displays three arguments, a1, a2 and a3, in a graph structure characteristic of the Carneades Argumentation System to help explain the difference between relevant and irrelevant arguments. As with the previous argument maps, a graph is a set of points called nodes and a set of lines (pairs of nodes) that join some nodes to others. In Fig. 4, the square nodes represent propositions that can be premises or conclusions of arguments while the round nodes represent arguments. For this reason, such a graph is called a bipartite graph. There are four arguments and six propositions in Fig.  4. Conclusion UC (the ultimate claim or ultimate probandum) is the proposition that is supposed to be proved or justified by the argumentation in a given case, to the standard of proof required for the context of dialogue. Two of the propositions, P1 and P2 are colored with a green (light grey in print) background in the rectangles, indicating that both have been accepted by the audience prior to the beginning of the argumentation stage in the dialogue. To take the next step we need to recall that following the rules of Carneades, an argument is said to carry probative weight in support of its conclusion if and only if the argument is valid and its premises are accepted. We assume for purposes of illustration that all the arguments shown in Fig. 5 are valid. Making the transition from Fig. 4, since arguments a1 and a2 are valid, and premises P1 and P2 are accepted, then two things will be done automatically by Carneades: conclusion P3 will be colored in green and conclusion P4 will be colored in green. The reason is that each of a1 and a2 is a single argument with only one premise. Now for example for argument a1, which is assumed to be valid, its only premise P1 is colored green. Therefore Carneades will automatically color P3 green. The reason is that this argument fits the criteria since it is a valid argument with all its premises accepted, and therefore its conclusion will be judged to be accepted by Carneades. Once we have reached this outcome, we need to go on to the next step, shown in Fig. 6. Argument a3, which is assumed to be valid, is a linked argument with two premises P3 and P4, both of which are accepted. The reader will recall that in a linked argument all the premises need to be accepted before the conclusion can be drawn as accepted. In this case, the requirement is met and so Carneades automatically colors the conclusion P5 with a green background. Finally we need to look at the sequence of argumentation shown in Fig. 7.


D. Walton

Fig. 5  Step 2 in abstract example 1

Fig. 6  Step 3 in abstract example 1

Fig. 7  Step 4 in abstract example 1

As shown in Fig. 7, Carneades has automatically drawn the ultimate conclusion (UC) in a rectangle with a green background. What has happened through the whole sequence of argumentation then is that starting from the two evidential premises P1 and P2 the sequence of argumentation over the distance between these two premises and UC has propagated from right to left, sequentially moving towards UC and eventually showing that UC is proved. Now we have represented how inferential distance can be displayed on an argument graph tracking the distance between evidential propositions and the ultimate conclusion in a case as part of the evidence to show that this sequence of argumentation and its parts are relevant. The next example illustrates how to apply the same technique to show that an argument is irrelevant.

Second Abstract Example In the approach taken in the first abstract example, an argument is relevant if and only if it is an argument that appears in a sequence of argumentation in an argument graph structure where the sequence moves forward from the propositions initially accepted as evidential to reach an ultimate conclusion. By this means the argument carries probative weight from evidence in support of (or against) the ultimate

Assessing Dialectical Relevance Using Argument Distance


Fig. 8  Step 1 in abstract example 2

Fig. 9  Step 2 in abstract example 2

conclusion to be proved (or disproved) in a given dialogue. Let us now apply this approach to another example, shown in Fig. 8. Next an abstract example is presented to illustrate an irrelevant sequence of argumentation. As the first step in this demonstration, consider the bipartite argument graph shown in Fig. 9. As with abstract example 1 (Fig. 5), Carneades automatically colors P3 and P4 green, on the basis that each of them is a single premised argument and in each instance the single premise is accepted, as shown by its being colored green. Carneades will automatically then color both premises P3 and P4 green, since each has been drawn from a single accepted premise by a valid argument. Next let’s turn to Fig. 10. Since argument a3 is a single premised argument, and its only premise P3 has been accepted, Carneades therefore automatically colors its conclusion P5 in green. Now consider argument a4. a4 is taken to be valid, but it is a linked argument, and a linked argument requires all of its premises to be accepted in order to prove its conclusion by a valid argument. For this reason, Carneades will not color UC in green. Therefore, what has been shown is that the ultimate conclusion UC cannot be proved even though the two evidential propositions P1 and P2 were originally accepted by the audience. Therefore the whole sequence of argumentation from P1 and P2 to UC is judged to be irrelevant. But there is one other thing to be considered with respect to this example. There is something in law called conditional relevance, defined in the Federal Rules of Evidence in Article 1, in the General Provisions of Rule 104,1 under Preliminary Questions. Conditional relevance, also called “Relevance That Depends on a Fact”, is explained by the following quoted comment. When the relevance of evidence depends on whether a fact exists, proof must be introduced sufficient to support a finding that the fact does exist. The court may admit the proposed evidence on the condition that the proof be introduced later.



D. Walton

Fig. 10  Step 3 in abstract example 2

To return to Fig. 10, it could be said, by using the terminology of conditional relevance that the entire sequence of argumentation shown in this figure could be ruled as conditionally relevant on the condition that an argument sufficient to warrant the acceptance of P6 could be introduced later in the sequence of argumentation once it has been expanded to include additional evidence. However, here are two things need to be noted. First, conditional relevance is a contested concept that is easy to get confused about. Second, it needs to be made clear that if a proposition or argument in an argumentation graph is conditionally relevant in relation to a dialogue that is underway where future arguments could come in that would provide evidence sufficient towards the acceptance of that proposition or argument, that does not prove that the argument really is relevant at the given stage of the dialogue we are in right now. In other words, just because an argument or proposition is conditionally relevant, in this sense of the term, it does not follow that it is relevant in the logical sense of the term. Once an argument has been concluded, it may be relatively easy, in hindsight, using the methods in this paper to see which arguments were relevant and which were not. But the basic problem is that in the middle of a discussion, it may not be easy to see or predict where a line of argument may be leading. It may sometimes be difficult to judge whether the argument will turn out to be relevant or not if it is not clear in the given case whether all the evidence is in or where that evidence may lead, as shown by the examples in this paper.

Conclusions The problem of making judgments of probative relevance or irrelevance, as shown by Macagno (2018) depends on three factors: argumentation schemes, missing premises and the factor of distance. The three real examples, along with the two abstract examples in this paper, have supported this approach by making the notion of distance more precise, showing how in these examples it can be applied to these examples by using argument maps having the form of bipartite graphs after the fashion of Carneades. It needs to be emphasized that the goal of this paper was not to build a formal model of relevance using Carneades. Carneades cannot do that yet. The purpose of

Assessing Dialectical Relevance Using Argument Distance


the paper was to show how the notion of distance can be extended using argumentation tools that are currently available in computational argumentation systems. It can be concluded that there are at least the beginnings of a technology for resolving relevance disputes here, and that the way forward is to exploit the notion of distance as represented by the graph structures of the abstract models of relevance and irrelevance. Much more work needs to be done with examples, but the paper has shown by three real examples what direction needs to be taken, and by providing two abstract examples, how it can be taken a little further. The bank robbery example was especially indicative in showing how the technique of mapping distance with graphs should properly work by seeing it as part of a dispute resolution method to resolve conflicts of opinions about relevance or irrelevance of a proposition or argument that works by weighing the pro arguments with the con arguments. This case shows how these problems can be resolved by ascending to a meta-dialogue level, where the arguments for and against relevance can themselves be evaluated using the tools outlined in this paper: argument maps taking the form of graphs, argumentation schemes and the elicitation of implicit premises in argument graphs. The second abstract example is particularly useful for illustrating how continuing the discussion at a meta-dialogue level could be assisted by a dialogue in which one of the parties in the discussion uses an argument map to pinpoint a particular proposition in an irrelevant sequence of argumentation that needs to be proved, or at least supported by evidence, in order to make this line of argumentation relevant. This discussant could ask the person who put forward the irrelevant argument to provide evidence to support this proposition, shifting the burden of proof to the arguer. If the arguer fails to reply by furnishing the requested evidence, then it can be declared that his argument is irrelevant, and the specific fault or error has now been identified to back up this evaluation. In the tuna fish example there would seem to be more evidence for classifying the argumentation in the case as an instance of the red herring fallacy, because there was an element of diversion involved in the line of argumentation to the tuna scandal issue that was moving away from, and distracting from, the ultimate conclusion it was supposed to be directed towards. However, there is some evidence for also classifying it as a fallacy of wrong conclusion, if you interpret the conclusion the MP was arguing for was the proposition that the government was at fault for the tuna scandal. Thus if you compare this example with the nuclear waste example, the underlying fault seems to be different in the two cases, even though there is some overlap in the tactics used. To me there still seems to be some merit in recognizing the distinction between the wrong conclusion variant and the red herring variant of this general kind of failure of relevance in argumentation, even though in practice it can be difficult in real cases to always separate cleanly between the two. It seems best to leave the discussion on how apply the distinction between these two variants as further work for those researching on informal fallacies.


D. Walton

References Anderson, A. R., & Belnap, N. (1975). Entailment: The logic of relevance and necessity. Princeton, NJ: Princeton University Press. Blakemore, D. (2002). Relevance and linguistic meaning: The semantics and pragmatics of discourse markers. Cambridge, UK: Cambridge University Press. Carston, R. (2004). Relevance theory and the saying/implicating distinction. In L. Horn & G. Ward (Eds.), The handbook of pragmatics (pp. 633–656). Oxford, UK: Blackwell. Epstein, R. (1979). Relatedness and implication. Philosophical Studies, 36, 137–173. Epstein, R. (1990). The semantic foundations of logic. Vol. 1, propositional logics. Dordrecht, Netherlands: Kluwer. Erduran, S. (2008). Methodological foundation of the study of argumentation in science classroom. In S.  Erduran & M.  Pilar Jimenez-Aleixandre (Eds.), Argumentation in science education: Perspectives from classroom-based research (pp. 47–69). Dordrecht, Netherlands: Springer. Giora, R. (1985). Notes towards a theory of text coherence. Poetics Today, 6(4), 699–715. https:// Giora, R. (1997). Discourse coherence and theory of relevance: Stumbling blocks in search of a unified theory. Journal of Pragmatics, 27(1), 17–34. 0378-2166(95)00065-8. Gordon, T. (2010). An overview of the Carneades argumentation support system. In C. Reed & C. Tindale (Eds.), Dialectics, dialogue and argumentation. An examination of Douglas Walton’s theories of reasoning and argument (pp. 145–156). London, UK: College Publications. Grice, P. (1975). Logic and conversation. In P. Cole & J. Morgan (Eds.), Syntax and semantics 3: Speech acts (pp. 41–58). New York, NY: Academic Press. Hobbs, J.  R. (1979). Coherence and coreference. Cognitive Science, 3, 67–90. https://doi. org/10.1207/s15516709cog0301_4. Hurley, P. (2003). A concise introduction to logic. Belmont, CA: Wadsworth. Lascarides, A., & Asher, N. (1993). Temporal interpretation, discourse relations and commonsense entailment. Linguistics and Philosophy, 16, 437–493. Macagno, F. (2016). Argument relevance and structure. Assessing and developing students’ uses of evidence. International Journal of Educational Research, 79, 180–194. https://doi. org/10.1016/j.ijer.2016.07.002. Macagno, F. (2018). Assessing relevance. Lingua, 210–211, 42–64. lingua.2018.04.007. Macagno, F., Mayweg-Paus, E., & Kuhn, D. (2015). Argumentation theory in education studies: Coding and improving students’ argumentative strategies. Topoi, 34(2), 523–537. https://doi. org/10.1007/s11245-014-9271-6. Mauet, T. (2005). Trials: Strategy, skills, and the new power of persuasion. New York, NY: Aspen. Nussbaum, M., & Edwards, O. V. (2011). Critical questions and argument stratagems: A framework for enhancing and analyzing students’ reasoning practices. Journal of the Learning Sciences, 20(3), 443–488. Rapanta, C., Garcia-Mila, M., & Gilabert, S. (2013). What is meant by argumentative competence? An integrative review of methods of analysis and assessment in education. Review of Educational Research, 83(4), 483–520. Sperber, D., & Wilson, D. (1995). Relevance: Communication and cognition. Oxford, UK: Blackwell Publishing Ltd. Taboada, M. (2009). Implicit and explicit coherence relations. In J. Renkema (Ed.), Discourse, of course: An overview of research in discourse studies (pp. 127–140). Amsterdam, Netherlands/ Philadelphia, PA: John Benjamins Publishing. Walton, D. (1989). Question-reply argumentation. Westport, CT: Greenwood Press. Walton, D. (2003a). Defining conditional relevance using linked arguments and argumentation schemes: A commentary on professor Callen’s article, rationality and relevancy: Conditional relevancy and constrained resources. Michigan State Law Review, 4(4), 1305–1314.

Assessing Dialectical Relevance Using Argument Distance


Walton, D. (2003b). Relevance in argumentation. Amsterdam, Netherlands/Philadelphia, PA: Routledge. Walton, D. (2004). Relevance in argumentation. Amsterdam, Netherlands/Philadelphia, PA: Routledge. Walton, D., & Gordon, T. (2012). The Carneades model of argument invention. Pragmatics & Cognition, 20(1), 1–26. Walton, D., & Krabbe, E. (1995). Commitment in dialogue. Albany, NY: State University of New York Press. Walton, D., & Macagno, F. (2016). Profiles of dialogue for relevance. Informal Logic, 36(4), 523–556. Walton, D., & Macagno, F. (2019). Diagnosing Misattribution of Commitments: A Normative and Pragmatic Model of for Assessing Straw Man. In A. Capone, M. Carapezza, & F. Lo Piparo (Eds.), Further Advances in Pragmatics and Philosophy: Part 2 Theories and Applications (pp. 111-136). Cham, Switzerland: Springer. Walton, D., Reed, C., & Macagno, F. (2008). Argumentation schemes. New York, NY: Cambridge University Press.

The Communicative Functions of Metaphors Between Explanation and Persuasion Fabrizio Macagno and Maria Grazia Rossi

Metaphors: Between Persuading and Explaining In the literature, the pragmatic dimension of metaphors has been clearly acknowledged. Metaphors have been described as tools that can be used for pursuing a wide range of goals including explaining, summarizing, supporting a viewpoint, illustrating, clarifying or persuading (Semino 2008; Cameron 2003; Goatly 2011, 148-167). These goals, however, are heterogeneous. Summarizing can be considered as a stylistic strategy, pursuing different types of communicative goals (such as persuading or making a decision), while illustrating is normally considered as an instrument for explaining. In contrast, clarifying is commonly considered as a specific type of explanation (Walton 2011), and persuading presupposes supporting a viewpoint. Explaining and persuading are also the two fundamental purposes that most of the works focusing on the communicative goals of metaphors have focused on (Ortony 1975, 45; Jaszczolt 2002, 358; Ottati et al. 1999; Sopory and Dillard 2002; Burgers et al. 2016; Ervas et al. 2015). An analysis of the pragmatics of metaphors needs to start from analyzing these two basic functions, and see how they can be pursued through metaphorical utterances.

M. G. Rossi (*) Instituto de Filosofia da Nova (IFILNOVA), Facultade de Ciências Sociais e Humanas, Universidade Nova de Lisboa, Lisbon, Portugal e-mail: [email protected] F. Macagno (*) Universidade Nova de Lisboa, Lisbon, Portugal e-mail: [email protected] © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 F. Macagno, A. Capone (eds.), Inquiries in Philosophical Pragmatics, Perspectives in Pragmatics, Philosophy & Psychology 27,



F. Macagno and M. G. Rossi

Explanation Versus Persuasion Explanation and argumentation (and a fortiori persuasion) are commonly considered as two distinct “speech acts” (Snoeck Henkemans 2001, 232) or “pieces of discourse” (Walton 2015, 102), both characterized by an inferential relation between one or more premises and a conclusion, which in case of explanations is normally a causal one (Berland and Reiser 2009, 27). These two speech acts are related, as an explanation, to be accepted by the interlocutor, needs to be justified, and thus supported by arguments and evidence, bearing out the claim that the explanation is the “best one” (Bex and Walton 2016; Berland and Reiser 2009). However, the crucial problem is to define the boundaries of the two concepts, and provide criteria for distinguishing them. The distinction is commonly drawn at distinct levels. At the level of the intended consequences, explanations are provided to increase the listener’s comprehension (performing a clarifying function when successful), while argumentation aims at increasing the acceptability of a claim (thus removing the doubt) (Hahn 2011; Snoeck Henkemans 2001). At the goal level, the purpose of an argument is to give a reason to support a claim, while an explanation is intended to help the questioner to improve his or her understanding (Walton 2006, 76). At a formal level, they can be considered as two distinct types of responses, which presuppose two distinct dialogical situations and needs (Walton 2015, 104): […] there are two different kinds of why questions. In one type of why question, the speaker asks the hearer to prove some claim made by giving evidence to support it. This type of why question asks for an argument. In another type of why question, the speaker asks the hearer to help him understand something that he presently does not understand.

An explanation is thus not only an answer (Van Fraassen 1980, 134), but an answer to a specific need that leads to a transfer of understanding (Walton 2011, 353). This definition has two dimensions: a cause-effect relation (transfer), and a presupposition (need). The first dimension concerns how understanding is transferred, as explanation is understanding (Schank 1986, 22–24). According to Schank, explanations are repairs of anomalies that occur when a fact or an event fails to make sense, namely it cannot fit in the connections between events that is part of an individual’s common knowledge (scripts, i.e. frozen inference chains stored in memory, Schank et al. 2014, 77). An explanation is thus a repair process that is used to help account for the anomaly by using other scripts or modifying the one that failed (Schank, 1986, Chap. 3; Schank et al. 2014, 77–80). The notion of scripts underscores the relationship between explanation and common ground. The need of an explanation results from the lack of a specific type of information that connects an event to the “story” in which it can make sense. For this reason, an explanation is always relative to the “why-question” that requested it, namely not only the causal relation that provides a reason for an event, but more importantly the “story” of the interlocutor that needs to be integrated through the explanation. As Van Fraassen put it (Van Fraassen 1980, 129):

The Communicative Functions of Metaphors


[…] an explanation often consists in listing salient factors, which point to a complete story of how the event happened. The effect of this is to eliminate various alternative hypotheses about how this event did come about, and/or eliminate puzzlement concerning how the event could have come about. But salience is context-dependent, and the selection of the correct ‘important’ factor depends on the range of alternatives contemplated in that context. In N. R. Hanson’s example, the barrister wants this sort of weeding out of hypotheses about the death relevant to the question of legal accountability; the carriage-builder, a weeding out of hypotheses about structural defects or structural limitations under various sorts of strain. The context, in other words, determines relevance in a way that goes well beyond the statistical relevance about which our scientific theories give information.

An explanation needs to be relevant not only to the why-question, but more specifically to the interlocutor’s needs in the given context (Van Fraassen 1980, 126). In the context of medical communication, we can easily identify very different needs behind patients’ and providers’ explanations. While commonly providers use explanations to share information about the diagnosis, disease management and prognosis, patients need to share information about symptoms also explaining the impact they have in terms of concerns and quality of life. The crucial problem that emerges is the interpretation of the “why” that leads to an explanation. The distinction between persuasion and explanation is thus pragmatic for distinct reasons. First, if we exclude the definition by consequences (resulting from a successful explanation), we can define explanation as an act aimed at providing a specific type of information, namely a causal relationship that allows the interlocutor to make sense of an event. Second, since it is related to the interlocutor’s understanding, it is contextual, and needs to be based on the presumptions about the interlocutor’s background knowledge. Third, as an explanation is a response to a request, it essentially depends on the purpose of the request, namely the object and the reason of the need of understanding. This definition brings the two concepts of explanation and persuasion extremely close. They are both responses to a why question that needs to be interpreted considering the dialogical setting (the context), and they both need to draw on what the interlocutor already accepts or can take for granted. However, in case of persuasion, the question results from a potential difference of interpretation or conflict; moreover, it is not a request of information concerning a (causal) relationship, but of a reason for accepting a doubtful claim, which can be not only opinions, but also problems, hypotheses, or decisions (Walton 1990, 411). Considering this distinction, we notice that the two acts are not only compatible, but often intertwined. Persuasion can occur only if the reasons are more acceptable than the conclusion, and the concepts of understanding and acceptance are extremely interrelated, and at best hard to distinguish (Slovic and Tversky 1974; Stanovich and West 1999).


F. Macagno and M. G. Rossi

Persuasive and Explanatory Metaphors Two functions of metaphorical utterances have been analyzed in the literature as the most important or common ones, namely explaining and persuading. However, these two functions are often considered separately, as two goals that are pursued independently of each other. The persuasive use of metaphors is one of the crucial topics in communication and rhetoric, as metaphors characterize every persuasive context. Metaphors are regarded as instruments for increasing the persuasiveness of a communication (Sopory and Dillard 2002; Boeynaems et al. 2017; Burgers et al. 2016; Ervas et al. 2018; Ottati et al. 1999), but this effect has been analyzed from distinct perspectives, focusing in particular on their emotive and argumentative role. Metaphorical utterances are regarded as having an emotive function (Ricoeur 1976, 49), as they can be used for evoking emotions and interpersonal intimacy (Bowes and Katz 2015; Hopper et al. 1981; Gibbs 2006; Casarett et al. 2010). They can be used to influence or affect the evaluation of a state of affairs (Hendricks et al. 2018), and convey emotional information, thus creating emotional stimuli (Barchard et  al. 2013, 333) that can motivate action (Flusberg et al. 2018). The persuasive effects have been also explained from a cognitive perspective. Metaphorical utterances have been commonly considered as more difficult to process (Searle 1979), thus affecting cognitive elaboration in different ways (Carston 2002; Gibbs 1992). This processing has been explained as resulting in different effects on the central processing, related to greater persuasive effects (Ottati and Renstrom 2010; Sopory and Dillard 2002, 385-387). The persuasive effects of metaphors have been also explained in terms of their argumentative role. On the one hand, they have been regarded from the perspective of argument “delivery,” namely as tools for structuring the arguments of a message, evoking semantic associations that are used for connecting more coherently the arguments in a text, resulting in higher comprehension and thus persuasion (Read et al. 1990; Sopory and Dillard 2002, 387). On the other hand, metaphorical utterances have been analyzed as conveying arguments or elements thereof (Santibáñez 2010; Ervas et al. 2018; Macagno and Zavatta 2014), acting as instruments for mapping arguments (Xu and Wu 2014) or as triggers of analogical inferences for supporting a conclusion (Musolff 2004; Fischer 2014; Van Eemeren and Garssen 2014, 52). The explanatory function of metaphorical utterances has been analyzed especially in relation to their use as teaching devices (Jaszczolt 2002; Ortony 1975), namely in contexts characterized by an “epistemic imbalance,” in which the interlocutors have different knowledge about a subject matter. In these contexts, metaphors are conceived as framing strategies that help understand something that is less familiar and less known in terms of something that is more familiar and already known (Semino et  al. 2016; Schiappa 2003). Framing is a strategy of selection (Kittay 1989, 128–131), which allows highlighting some properties of the target while hiding others that are not relevant to the intended purpose of the utterance. This selection can result in bringing to light the explanatory principle that the

The Communicative Functions of Metaphors


speaker intends to convey; however, this process has also another side, namely the interpretation. A metaphor needs to be interpreted, and depending on the interlocutor’s background, the property or the properties transferred through the metaphorical utterance (the emergent meaning) can be different from the speaker’s intended ones (Gibbs 1992, 587). To these two crucial functions of metaphors, a third can be added, related to the explanatory one, namely theorizing. While explanatory metaphors make a concept or issue easier to understand to an interlocutor, “constitutive” metaphors are part of the development of a theory concerning an issue (Ungerer and Schmid 2006). They are used to select properties that allow a new explanation, in this case not for the interlocutor but also for the speaker. The distinction between the two metaphorical functions of explaining and persuading is, however, problematic in two respects. First, persuasion and explanation are defined based on a pragmatic distinction, but this distinction is captured only partially by the standard theories of speech acts, as in both cases we are dealing with assertive speech acts (Searle and Vanderveken 1985, 112) or expositives, defined as acts aimed at clarifying of reasons, arguments, and communications (Austin 1962, 162). Second, the distinction between these two functions becomes extremely complex to draw when we analyze metaphorical utterances. A metaphorical explanation involves a selection of properties and can trigger emotional effects, which can be used for persuasive goals (Ungerer and Schmid 2006, 153). Metaphors provide a new way of looking at a state of affairs (Ungerer and Schmid 2006, 118), which can affect not only understanding, but also our attitude towards it or the further decisions to be made involving or concerning it. These two problems will be addressed in the following sections.

 etaphors and the Blurred Distinction Between Explanation M and Persuasion According to the Aristotelian account, metaphors are defined as a type of “transference” (epiphora), a process of change of meaning that concerns the semantic kernel (Ricoeur 1993, 18), and based on logic-semantic relations (Aristotle, Poetics 1457b, 7–10): Metaphor consists in giving the thing a name that belongs to something else; the transference being either from genus to species, or from species to genus, or from species to species, or on grounds of analogy.

This definition is based on two crucial concepts, namely “transference” and “strangeness” (allotrios, that is, “a name that belongs to something else”). The first dimension is the one more discussed. Metaphors are regarded as instruments that bring about a conceptual reorganization (Black 1955, 280-288), extending the boundaries of a concept (Leech 1981, 37). This reorganization is a redefinition of reality, a new organization of a state of affairs that is in itself informative (Ricoeur 1993, 24):


F. Macagno and M. G. Rossi

[…] should we not say that metaphor destroys an order only to invent a new one; and that the category-mistake is nothing but the complement of a logic of discovery? […] Pushing this thought to the limit, one must say that metaphor bears information because it ‘redescribes’ reality. Thus, the category-mistake is the de-constructive intermediary phase between description and redescription.

The second dimension concerns the emotive dimension of metaphors. Metaphors are riddles (Aristotle, Rhetoric, 1415b, 9), but more importantly they are paradoxes, in the sense of deviation from the ordinary use (Ricoeur 1993, 29). They are based on strangeness, in the sense that the vehicle “is temporarily employed outside its home context, to which it continually returns” (Moran 2017, 51). Strangeness generates marvel, and thus pleasure, which attracts attention and desire to understand new ideas. This relationship between pleasure (strangeness) and learning is stressed in the Rhetoric (Moran 2017, 51), in which metaphors are described as instruments for making a concept easier to understand by generating pleasure and thus interest (Aristotle, Rhetoric 1410b, 10–27): We will begin by remarking that we all naturally find it agreeable to get hold of new ideas easily: words express ideas, and therefore those words are the most agreeable that enable us to get hold of new ideas. Now strange words simply puzzle us; ordinary words convey only what we know already; it is from metaphor that we can best get hold of something fresh. […] We see, then, that both speech and enthymemes are lively in proportion as they make us seize a new idea promptly. For this reason people are not much taken either by obvious arguments (using the word ‘obvious’ to mean what is plain to everybody and needs no investigation), nor by those which puzzle us when we hear them stated, but only by those which convey their information to us as soon as we hear them, provided we had not the information already; or which the mind only just fails to keep up with.

This passage expresses a crucial principle, namely that “the way in which a thing is said does affect its intelligibility” (Aristotle, Rhetoric 1404a, 9–10), as through the manifestation of our thinking (lexis) it is possible to teach (Ricoeur 1993, 35). The key concept in this excerpt is “lively” (energeia) (Newman 2002, 4): an apt metaphor sets something vividly before the eyes of the audience (pro ommaton poiein) (Aristotle, Rhetoric 1411b, 27), making the interlocutor to see things in a specific way. Metaphors are thus at the same time instruments of understanding and believing, as they can lead the audience to see, and consequently experience, feel and believe, a certain state of affairs under a specific perspective (Moran 1989, 91). In this sense, they are linguistic instruments for providing proofs strategically, visualizing a situation in a state of activity and actuality, as a living being (Schneck 2011, 28), so that it can trigger emotions (Aristotle, Rhetoric, 1408a, 10). The strategic dimension of energeia is represented by the framing effect, the specific view that metaphors impose (Moran 1989, 108): It is almost as if the imagistic “seeing” of metaphor comprehension really involved one’s seeing things that way, that is, believing them to be that way, which would give us no way to distinguish between understanding a metaphorical assertion and believing it. Speaking of the adoption of a perspective is useful precisely because it is neutral with respect to belief and assertion. But if this were the only dimension of metaphor, then we couldn’t explain, among other things, what denying the statement comes to. For the denial is not the refusal to adopt the perspective, any more than the original statement is simply the invitation to

The Communicative Functions of Metaphors


take up that perspective. The speaker does not say, “Imagine Juliet as the sun”; rather he makes a statement about Juliet. She is the object of his thinking, and various beliefs of his about her are necessary to account for his original adoption of this perspective. And someone who denies that statement will be sharing that perspective for the moment, but disagreeing with some of what Romeo believes. To sum up, then: there are two dimensions of metaphor, the dimension of effects, which I’ve referred to in terms of framing or the adoption of a perspective, and the dimension of the beliefs that prompt the comparison in the first place, and which are necessary for the framing-effect to be something other than mere juxtaposition.

The strict interrelation between the explanatory and persuasive functions of metaphors leads to the problem of interpreting the purposes of an explanation, and more importantly the request thereof. To illustrate this issue, we consider an example from one of the less “rhetorical” or persuasive contexts in which explanations are prototypically used, namely medical interviews. In this context, metaphors are employed commonly for sharing information on medical concepts or causal relations. A typical example is the following, in which multiple metaphors clarify a technical concept that could be hardly understood by the patient if technical terminology was used (emphasis added):1 Example 1 Physician: This value, madam, needs to be taken care of. Patient: What is it? Physician: The glycated hemoglobin is a value very important for us. Now, let us say that, ehh, you know it, we repeated it many times, it is like to have a river mm? the blood with polluting substances, which we need to keep under control. The glycaemia during the day tells me how I am doing at that specific moment. The glycated hemoglobin tells me the global trend of diabetes. If I go to buy a dress, the glycated hemoglobin is the size, and glycaemia is the model. Ok? The size tells me my condition; I can then customize the model. In this case, the doctor replies to a request of understanding: the patient does not understand a specific concept, glycated hemoglobin, and the physician provides a vivid description thereof that can be accessible to the interlocutor. However, if we consider the metaphors used (“river,” “pollutants,” “size,” and “model”), we notice  Medico: “È questa qua signora. mi raccomando eh.” Paziente: “Che cosa è?” Medico: “L’emoglobina glicata è un valore per noi molto importante. Allora diciamo che eh:::: la signora lo sa, gliel’abbiamo detto tante volte, ehm è come avere un fiume mm? il sangue con degli inquinanti che noi dobbiamo tenere sotto controllo. allora la glicemia nel corso della giornata mi dà in quel momento specifico, in quella fascia oraria, come sto funzionando. L’emoglobina glicata mi dice l’andamento generale del diabete. come dire: se vado a comprare un vestito, l’emoglobina glicata è la taglia, la glicemia è il modello” . Medico: “Ok? quindi io la taglia mi dice più o meno come sto, e il modello me lo  personalizzo.” This example has been already discussed in (Rossi 2016, 42; Ervas et al. 2016, 106). Hereafter, when it is not further specified the examples are part of the corpus collected by Sarah Bigi (2014). 1


F. Macagno and M. G. Rossi

that their role is not limited to providing a response to a request of understanding. If we try to describe what these metaphors have in common with their targets (the blood, the level of glucose, the glycated hemoglobin, and the glycaemia), we come to a system of associations that can be described tentatively as follows:

Target Blood High level of glucose Glycated hemoglobin Glycaemia

1 (description of the relevant aspect of the target) Bodily fluid Dangerous substance

2 (description of the relevant Vehicle aspect of the vehicle) River Stream of a fluid Pollutants Toxic substance

Trend of the condition


Stable condition

Temporary condition


Customizable aspect

These associations are not only aimed at setting out a general description of the causal relations underlying the notion of glycated hemoglobin. Rather, they are intended to provide the understanding that the patient needs in the specific context, namely for the specific goal that he is presumed (and in this conversational setting is also expected) to have, i.e. making appropriate decisions concerning the monitoring and the control of his values. The physician is not answering a general request of understanding, but a goal-directed need of understanding. The notion of explanation seems to become more complex, and more related to a pragmatic dimension of discourse that involves the types of dialogue that the interlocutors propose to engage in.

 escribing the Conversational Goals D of Metaphorical Utterances As underscored above, metaphors are considered as instruments of persuasion for their emotive effects that depend on  the cognitive efforts and effects that  they involve. Such effects, however, are have also an explanatory function, as they are used for understanding and making understand better the target. Persuasion is thus regarded as related to understanding, which is in turn the result of the actuality (or vividness) of metaphors, which can bring about the other component of persuasion, namely emotive reactions. In all these cases, however, the function of a metaphor is described based on a sign of its successful use. Both persuasion and explanation are defined based on one of the effects that a metaphor has achieved on the interlocutor, namely modifying his or her attitude towards a doubtful viewpoint or making an unclear concept clearer. This description can be problematic for two reasons. First, metaphors failing to persuade the interlocutor or to explain a concept still have been used for a communicative purpose. Second,  explanations and arguments can be used to pursue different communicative goals. An argument can be used to support

The Communicative Functions of Metaphors


not only a doubt in difference of opinions, but also a proposal, an offer, a doubtful piece of information, or the acceptability of a hypothesis (Walton 1990). Similarly, an explanation is an act aimed at transferring understanding (Walton 2004), but this understanding can concern the content of different types of communicative acts, such as the sharing of information, a proposal, a premise in an argument, a hypothesis, etc. In this sense, the categories of persuasion and explanation need to be framed within the bigger picture of the conversational goals of metaphors to explain what metaphors are used for. On this view, the explanatory and persuasive effects of metaphors need to be analyzed in a dialogical perspective taking into account the conversational goals that they are intended to serve, namely their relevance (Clark 1987; Kittay 1989, 131; Goatly 2011). However, this endeavor faces two issues. The first concerns the representation of such conversational goals in a way that capture the interactive linguistic nature of the relation between the interlocutors (Gu 1993, 427–28). In the literature, the problem of identifying the goals of metaphors (Semino 2008; Cameron 2003; Goatly 2011, 148-167) has not been addressed systematically, nor are the unit of analysis and the theoretical principles of classification clearly established. The issues that can arise from the detection of the conversational goals of a metaphorical utterance can be explained by considering the following case (emphasis added):2 Example 2 Physician: If I know that my blood pressure is, let’s say, dancing, I measure it. In this excerpt, taken from a healthcare provider-patient interview in the context of diabetes care, the doctor is providing instructions to the patient, and she is using a metaphorical utterance to achieve her goal. However, if we  analyze this example  through the categories of the speech act theory, we cannot fully capture the interactive purpose of the move (Streeck 1980). The utterance can be considered as an assertive; if assessed as an indirect speech act, it can be classified as a directive. However, in both cases the speech acts are unilateral (van Eemeren and Grootendorst 1984, 28), while this move is uttered for a specific dialogical purpose (defined by the conversational setting of a medical interview, Bigi and Lamiani 2016), namely making a joint decision. Similarly, our Example 1 can be considered as constituted by assertives, failing to capture the dialogical goal of the utterances, namely the dialogical reason why they are uttered. From an analytical perspective, this challenge corresponds to the problem of determining the joint communicative intention underlying an utterance (Leech 1983, 35). The second problem concerns the boundaries of the reconstruction of the communicative purpose. As Leech pointed out, “all illocutions are ‘indirect’ in that their force is derived by implicature” from the sense of an utterance (Leech 1983, 33). However, the degrees of indirectness can vary. Some illocutions can be interpreted 2  Dottore: “Se io so di avere la pressione diciamo ballerina, me la misuro.” This example has been already discussed in (Rossi et al. submitted).


F. Macagno and M. G. Rossi

defaultively, namely applying the most likely pragmatic hypothesis. In other cases, it is necessary to reconstruct it through more complex inferences. To these two situations, we can add a third one, namely the possibility of detecting both a default interpretation and a more indirect one, or two indirect interpretations (Strawson 1964), both plausible and compatible hypotheses in the given context, even though one is more “central” or with a higher degree of probability (Leech 1983, 34–43). An illustration is our Example 2 above. The physician’s utterance can have two conversational goals, namely proposing an action (“you should measure your blood pressure”) and more indirectly expressing a reason in support thereof (“ if I know that my blood pressure is, let’s say, dancing, I measure it ”). Also our Example 1 shows this twofold communicative intention. The physician intends to provide information but does it in a way that can be interpreted as a reason to regard a certain value as important, or even more implicitly, to act in a certain way (monitoring the glycated hemoglobin regularly). These two issues can be addressed by considering the concept of dialogue move, and consequently the definition of metaphorical move.

Metaphorical Moves The problems that arise from considering a metaphorical utterance as the unit of analysis lie in the limits that the concept of speech act has in the analysis of dialogical interactions. The crucial problem is, however, the development of an analytical instrument that can serve the goal of capturing the interlocutors’ conversational goals without considering the conversational settings that can lead to a potentially infinite number of joint acts (Mey 2016; Capone 2010).

From Metaphorical Utterances to Metaphorical Moves The starting point for our analysis is the concept of metaphorical utterance. Black (Black 1955, 255-257) introduced the idea of metaphorical utterance to underscore the dimension of the use of a metaphorical expression, the relationship between the focus and the frame in an utterance, and the role of the circumstances and more importantly the speaker’s intentions in identifying and interpreting a metaphor. The notion of utterance represented the “carrier of ‘complete and finished meaning’ (according to Fontanier’s own expression), in the production of metaphorical meaning” (Ricoeur, 1993, 74). In this sense, metaphorical utterances can be defined as “utterances that are either intended to be understood metaphorically or it is possible to attribute to it a metaphorical interpretation” (Kittay 1989, 148). The notion of utterance, however, does not provide any specific information on the type of joint intention that the speakers intend to pursue through their metaphor,

The Communicative Functions of Metaphors


namely the “joint action” (Kovecses 2015, 178-180; Goatly 2011, 292) or “local context” (Kovecses 2015, 188) that guide the process of metaphorical interpretation. To capture the conversational goal that a metaphorical utterance pursues, it is necessary to move from the notion of utterance to the one of move. Dialogue moves can be defined as sequences (Grosz and Sidner 1986, 177), which correspond to utterances or aggregates or parts thereof, aimed at proposing a dialogue to the interlocutor (Macagno and Bigi 2017). For this reason, a move is the representation of a dialogical intention, namely an interactional (or, more precisely, communicative) goal that a speaker has (Ruhi 2007, 109; Haugh 2015, 95-97). This definition leads to the problem of capturing the types of dialogical interactions. Such a typology needs to be general enough to allow a justification of each category, but at the same time specifiable, namely adaptable to the various types of institutional or conversational settings and analyses that require the introduction of context-specific subcategories (Macagno and Bigi 2020a). An effective classification can be made by considering the literature on dialogue games and types of dialogue (Walton 1989, 1998; Walton and Krabbe 1995; Macagno 2008; Dunin-Keplicz and Verbrugge 2001; McBurney and Parsons 2009; Walton 1990). Walton (1989, 1990, 1998; Walton and Krabbe 1995, 66; Macagno 2008) defined  seven basic “types of dialogue,” namely abstract dialogical intentions that the interlocutors can pursue in their interactions. Such types of dialogue, originally designed for representing formal dialogues of the kind used in artificial intelligence, are summarized in the following Table 1 (Walton 2010; Walton and Krabbe 1995, 66). Such dialogue types, to which can be translated into categories of conversational demands, namely communicative goals pursued by the speaker and proposed to the interlocutor, which affect the interlocutor’s response in a specific fashion (Dascal 1992; Levin and Moore 1977). In particular, the category of eristic dialogue is broadened to include moves aimed at establishing the relationship necessary for Table 1  Types of dialogue and their characteristics Type Initial situation 1. Persuasion Dialogue Conflicting points of view 2. Negotiation

3. Inquiry 4. Discovery

Conflict of interests and need for cooperation General ignorance

Main goal Resolution of such conflicts by verbal means Making a deal

Growth of knowledge and agreement Choose best hypothesis for testing

Need to find an explanation of facts 5. Deliberation Need for action Reach a decision 6. Information-­seeking Personal ignorance Spreading knowledge and revealing positions 7. Eristics Conflict and Reaching a antagonism (provisional) accommodation in a relationship

Participants’ aims Persuade the other(s)

Get the best out of it for oneself Find a “proof” or destroy one Find and defend a suitable hypothesis Influence outcome Gain, pass on, show, or hide personal knowledge Strike the other party and win in the eyes of onlookers


F. Macagno and M. G. Rossi Utterance


Dialogical Negotiating a relation (ground level of joint action)

Acting together

Eristic (rapport building) moves

Joint action


Practical Quality of the possible actions

Information Information known by one interlocutor



Unknown information (explanation) to be found by both the interlocutors


Based on arguments


Interests of the interlocutors


Based on assessment of proofs and evidence



Fig. 1  Classification of dialogical intentions

further dialogues – a kind of basic level of communication necessary for the other types of dialogue. These categories define the types of dialogue moves that can be represented in (Fig. 1) (Macagno and Bigi 2017, 2020a). Dialogue moves can be thus used as units of analysis as they represent – even though abstractly and generally – the dialogical intentions of the interlocutors. On this view, metaphors need to be conceived in terms of metaphorical moves: they are the result of the interpretation of a sequence of discourse that is produced to pursue a specific conversational goal, which can be reconstructed based on different types of clues, including the co-text, the conversational setting, and the context (Levinson 1992, 2012; Sanders 2013, 1987).

Goals of Metaphorical Explanations The different types of metaphorical moves can be used to describe the most important and general goals that the interlocutors can pursue using their metaphors. In this framework, the distinction between the cognitive function of explanation and the dialogical persuasive goal pursued is not made at the utterance level but looking at the structural organization of the interactions determined through an analysis of the dialogue moves constituting them. In (Rossi et al. submitted), we used this theoretical framework as a coding scheme for detecting the dialogical functions of metaphorical moves, showing how the aforementioned system of dialogue moves (Macagno and Bigi 2017, 2020a) can be used to codify metaphorical moves to clarify the strict interrelation between the conversational setting, the move, and the metaphorical meaning.

The Communicative Functions of Metaphors


Explanatory metaphorical moves will be shown below to be instruments for various dialogical goals. Through a selection of representative examples – drawn from a corpus of diabetes care interviews collected in Northern Italy (Bigi 2014) – the twofold nature of metaphorical explanations will be illustrated. The most prototypical dialogical function of explanations – and thus explanatory metaphorical moves – is information sharing, which is the most frequent type of metaphorical move that can be found in this corpus (Rossi et al. submitted). A clear illustration of the metaphorical move of information sharing is given by Example 1 above, in which a series of metaphors (“pollutants;” “river;” “size;” “model”) are used to provide information to the patient on the concepts of glycated haemoglobin and the functioning of diabetes. The information-sharing goal can concern also other types of information. In particular, metaphors are frequently used for referring to symptoms that the patient needs to recognize. In these cases, the healthcare provider employs metaphors to request information that would be otherwise complex to understand, making it accessible to a patient who does not know a more technical vocabulary. Patients also use metaphorical utterances for sharing information with the health providers on their conditions, symptoms, or experiences. In lack of a more specific terminology, and without the necessary medical experience for identifying symptoms and describing them, the use of metaphors makes it possible the communication of complex phenomena and the perceptions thereof, such as in the following case (emphasis added):3 Example 3 Patient: […] then I noticed that if I eat gnocchi, it [the glycaemia] empties itself quickly. I love gnocchi so much. Nurse: How does it empty itself quickly? Patient: Eh it goes down, goes down. In this case, the metaphorical expression “empties itself quickly” is used to explain what happens to the patient’s glycaemia after she eats gnocchi. The transfer of understanding is achieved through a metaphor, which allows the sharing of important information. Metaphors are crucial instruments for patients in sharing symptoms especially when they are not able to interpret them or to assess their clinical relevance. The classic rhetorical function of metaphors, persuasion, is the second most frequent in the corpus (Rossi et al. submitted). In Example 4 below (emphasis added), the dietician is instructing the patient on how to control better the diabetes, especially on vacation. To support her advice (proposal) of keeping healthy habits, she uses a series of metaphors to represent the patient’s unreasonable reasoning:4

3  Paziente: “Poi ho notato che se mangio gli gnocchi, mi si svuota in fretta. A me piacciono tantissimo.” Infermiera: “Come si svuota in fretta?” Paziente: “Eh va giù, va giù.” This example has been already discussed in (Macagno and Rossi 2019, 110) 4  Dietista: “Quando si va in ferie il diabete si porta dietro, non si chiude a Milano e si parte. Rimane con lei.”


F. Macagno and M. G. Rossi

Example 4 Dietician: When you go on vacation, you carry the diabetes with you, you don’t lock it up in Milan when you leave. The diabetes stays with you. Here, metaphors play both a cognitive and a persuasive function. The dietician uses them to explain what diabetes is, and more importantly the persistence of the condition. However, at the same time they express a complex argument (Walton 1990), in which the patient’s behavior and understanding is interpreted based on the commitments underlying it, which are then attacked by a metaphorical reductio ad absurdum. The metaphorical explanation of what happens to diabetes when one leaves is an instrument of persuasion. Metaphorical utterances can be used at the same time for transferring understanding at a cognitive level, and for pursuing a decision-making dialogical goal at a pragmatic level. Some metaphorical moves at the same time explain a concept and advance a proposal, namely suggest a course of action that the interlocutor can accept or refuse. A clear case is Example 2 above, in which the metaphor “dancing” is used to refer to the condition that is presented as requiring action. In that example, the metaphorical expression is used within a move that has a twofold purpose: the cognitive one of explaining the importance of measuring the pressure to prevent further complication, but also the pragmatic/dialogical one of securing the patient’s commitment to the action described, namely proposing a course of action (and often persuading him to engage in that action). Another complex use of a metaphorical expression that combines a metaphorical explanation with a complex combination of a decision-making and persuasive purpose is the following (emphasis added):5 Example 5 Doctor: Let’s say, the three levers of diabetes care are physical exercise, diet, and medications. I am already switching up the medications lever, so it would be better to agree on a strategy to improve the other two levers. Just one, or both, partly one and partly the other one, it is up to you to come up with suggestions or ideas. For instance: what is your plan? In this example, the doctor is proposing different options to the patient (e.g. to start working on physical exercise and/or diet) to better manage his diabetes (a decision-making move). Within this move, the metaphorical expression “lever” works as an explanation of the possibilities offered. However, this explanation (“the three levers of diabetes care are physical exercise, diet, and medications”) is aimed not only at proposing a course of action, but also framing the possibilities in a strategic way. In this sense, the explanation is part of a persuasive move intended to justify the need for an individual and personalized choice (we can express it as an  Dottore: “Se i tre bracci della cura sono l’attività fisica la dieta e il farmaco, io sul farmaco sto già ad un livello molto elevato. quindi ci conviene interagire su questi due livelli. uno solo, tutti e due, un po’ di qua un po’ di là, me lo deve dire lei. come: come pensa di riuscire ad organizzarsi?” This example has been already discussed in (Ervas et al. 2016, 106). 5

The Communicative Functions of Metaphors


implicit conclusion of the kind, “it would be better to agree on a strategy to improve the other two levers”). The following Example 6 can be analyzed in a similar way (emphasis added):6 Example 6 Dietician: You must try to reach, to get close to the ideal weight. Not gain weight. That’s because diabetes and weight generally go hand in hand, like an engaged couple. So, if you gain weight, also diabetes tends to increase a bit. The dietician explains the relationship between changes in weight and changes in diabetes by introducing a metaphorical frame developed from the idea that diabetes and weight act as an engaged couple. Once again, this metaphorical explanation is included in a dialogical context characterized also by a strong persuasive argument, aimed at leading the patient to follow the medical recommendation (in this case, to lose weight). The metaphorical explanation is not argumentatively neutral, namely is not a simple sharing of information. The physician does not only secure the patient’s understanding of a concept, but also his acceptance of the principle on which the conditional premise of the argument is based (“if you gain weight, also diabetes tends to increase a bit”). Using Toulmin’s terminology (Toulmin 1958, 95–98), the explanation acts as a backing of the warrant of an argument, namely it is an integral part of the latter. Metaphors can be used for explaining a concept or a correlation for the dialogical purposes of discovery, namely providing hypotheses, for new or unknown phenomena, or inquiry, namely assessing evidence. Discovery moves are not frequently pursued in the corpus, but they can be crucial for bringing to light the medical misconceptions of the patients (Rapanta 2019). A clear case is the following (emphasis added): Example 77 Wife:

Itold him, you have reached almost the maximum levels, now careful, as I blamed the melon, the apricot. Patient: The fuit, the sugar. Wife: The fruit, yes, the peach, nothing fried as they are not liked at our place.

 Dietologa: “Bisogna cercare di arrivare, avvicinarsi più che possiamo al peso ideale. Non aumentare. Perché generalmente diabete e peso viaggiano come due fidanzati, mano nella mano. Allora, se lei mi aumenta di peso, anche il diabete tende un pochino a salire.” This example has been already discussed in (Rossi forthcoming). 7  MO: “A fare quella. no dico sei arrivato quasi al massimo, ora attenzione, perché io davo la colpa appunto al melone, all’albic- sì a-“ P: “Ai frutte::: zuccherini insomma” MO: “Ai frutti insomma, sì la pesca, fritti niente perché quelli a casa nostra non è che piaccia.” MO: “No. e quindi niente da fare. davo la colpa a quelle cose, ma se quelle cose non sono, sarà perché cammina poco, non lo so.” D: “Ma che cosa signora intende?” MO: “Eh no dicevo davo la colpa quando tornavamo dalle vacanze.” 6


F. Macagno and M. G. Rossi


There is nothing to do. I blamed those things, but if these things are not the cause, it can be the fact that he walks very little, I do not know. Doctor: But what do you mean? Wife: I was telling what I was blaming when we came back from holiday. The patient’s wife is formulating a hypothesis on the possible causes of the worsening of the husband’s conditions, and the metaphorical moves explain the vague causal relation between the type of food and the possible effect on health. Through the metaphorical expression “to blame,” the wife is humanizing the possible causes of the increase of the levels of glycaemia. This verb attributes agency to inanimate entities, conveys the vaguest possible causal relation. The wife explains her discovery of a possible cause of the husband’s condition as an attribution of responsibility, which does not presuppose a specific type of cause. Inquiry moves (Rapanta 2019, 45-48) consist in searching and eliciting evidence from the interlocutor and interpreting it in order to explain a phenomenon. In the case of chronic care interactions, the physician does not only collect evidence from the patient to make a decision, but more often and importantly explains to the interlocutor how to interpret it, in a type of educational dialogue (Macagno and Bigi 2020b). An example in which metaphorical explanatory moves play an inquiry function is the following (emphasis added):8 Example 8 Nurse:

Eh, yes. Because it has a value in itself – a 110 before lunch, ok it is good, I am saying that it is a very good start. However, what happened two hours after eating? Was it however – have you find a target glycaemia? Patient: Yes, here it was 109 and 120. Nurse: Yes, but it would be interesting, on the day in which you do it. Patient: To have both of them. Nurse: Instead of doing it before lunch and dinner, I do it before lunch and two hours after lunch, or before dinner and two hours after dinner. Because for me it is good also in couples, the important is that it touches the different hour phases – breakfast, lunch, and dinner. Here, the nurse is interpreting the evidence provided by the patient and demanding the data needed for assessing her condition. The search for evidence here is not purely information sharing, as the nurse is interpreting the evidence and teaching the patient how to collect meaningful data that can be used for evaluating the 8  I: “Eh sì. perché ha un valore di per sé- di un centodieci prima di pranzo, ok mi va bene sto dicendo che è un’ottima partenza. Però poi cosa è successo due ore dopo aver mangiato? É stata comunque un- ha trovato la glicemia a target.” P: “Sì [qua era] un centonove e un centoventi]” I: “Sì però il giorno che la fa sarebbe interessante.” P: “Averli tutti e due.” I: “Invece di farlo prima di pranzo e prima di cena, lo faccio prima di pranzo e due ore dopo [pranzo] o prima di cena e due ore dopo cena. […] perché a me va bene anche così a coppia, l’importante è che mi tocca le varie fasce orarie. quelle della colazione, del pranzo e della cena.”

The Communicative Functions of Metaphors


glycaemia. The metaphors play a crucial role, as more complex medical concepts concerning the types of evidence  and data collection procedures are explained through the use of metaphors (“in couples;” “touches”). Here, metaphors are explanatory – they clarify concepts by using ordinary language –, but the explanation is an instrument aimed at the specific dialogical goal of collecting and interpreting evidence – and instructing the other on how to interpret it.

Conclusion Metaphors have been acknowledged to have several different uses, as they can be used to pursue distinct dialogical purposes, regardless of whether they are knowingly used as metaphors as such. In this paper, we have addressed the problem of determining what goals metaphors are used to pursue, especially taking care of the distinction between explanation and persuasion. Discussing the distinction between the two metaphorical functions of explaining and persuading, we have shown the main theoretical problems behind this distinction and we have built up a theoretical model that puts in connection intuitions from speech act theory, argumentation, and metaphor theory. We have observed that explanations have a cognitive and a pragmatic dimension; however, while the cognitive function of transferring understanding defines what an explanation is, different purposes can characterize its use in discourse, namely what it is used for. The current approaches mix the two levels, identifying explanation with the common goal of sharing information. However, the transference of understanding can have also other purposes, such as making a proposal, supporting a conclusion or justifying or describing a hypothesis. Metaphorical explanations clearly illustrate this twofold nature of explanations, as they are used frequently to pursue a persuasive goal, and in many cases to propose a course of action, share information, and advance a hypothesis. The distinction between the two dimensions led us to addressing the pragmatic functions of metaphors using an instrument that merges pragmatics with dialogue analysis and argumentation, namely dialogue moves. We have used this model to detect distinct pragmatic goals that interlocutors intend to achieve in dialogues through different types of explanatory metaphorical moves. We have discussed examples from an Italian corpus of diabetes interviews to describe not only the relationship between metaphors and explanations, but also between metaphorical explanations and dialogical purposes. Future research is needed for highlighting the implications of our model for the practical context of health, stressing its importance for the improvement of the quality of the communication between patients and providers. Acknowledgments  This work was supported by the Fundação para a Ciência e a Tecnologia (research grants no. SFRH/BPD/115073/2016 and PTDC/FER-FIL/28278/2017).


F. Macagno and M. G. Rossi

References Aristotle. (1991a). Poetics. In J. Barnes (Ed.), The complete works of Aristotle, Vol. II. Princeton, NJ: Princeton University Press. Aristotle. (1991b). Rhetoric. In J. Barnes (Ed.), The complete works of Aristotle, Vol. II. Princeton, NJ: Princeton University Press. Austin, J. L. (1962). How to do things with words. Oxford, UK: Oxford University Press. Barchard, K. A., Hensley, S., Anderson, E. D., & Walker, H. E. (2013). Measuring the ability to perceive the emotional connotations of written language. Journal of Personality Assessment, 95(4), 332–342. Berland, L. K., & Reiser, B. J. (2009). Making sense of argumentation and explanation. Science Education, 93(1), 26–55. Bex, F., & Walton, D. (2016). Combining explanation and argumentation in dialogue. Argument and Computation, 7(1), 55–68. Bigi, S. (2014). Healthy reasoning: The role of effective argumentation for enhancing elderly patients’ self-management abilities in chronic care. In G.  Riva, P.  A. Marsan, & C.  Grassi (Eds.), Active ageing and healthy living: A human Centered approach in research and innovation as source of quality of life (pp. 193–203). Amsterdam, Netherlands: IOS Press. Bigi, S., & Lamiani, G. (2016). The power of words: Deliberation dialogue as a model to favor patient engagement in chronic care. In G. Graffigna (Ed.), Transformative Healthcare. Practice through Patient Engagement (pp. 66–92). Hershey PA: IGI Global. Black, M. (1955). Metaphor. Proceedings of the Aristotelian Society, New Series, 55, 273–294. Boeynaems, A., Burgers, C., Konijn, E.  A., & Steen, G.  J. (2017). The effects of metaphorical framing on political persuasion: A systematic literature review. Metaphor and Symbol, 32(2), 118–134. Bowes, A., & Katz, A. (2015). Metaphor creates intimacy and temporarily enhances theory of mind. Memory and Cognition, 43(6), 953–963. Burgers, C., Konijn, E. A., & Steen, G. J. (2016). Figurative framing: Shaping public discourse through metaphor, hyperbole, and irony. Communication Theory, 26(4), 410–430. https://doi. org/10.1111/comt.12096. Cameron, L. (2003). Metaphor in educational discourse. London, UK: Continuum. Capone, A. (2010). Barack Obama’s South Carolina speech. Journal of Pragmatics, 42(11), 2964–2977. Carston, R. (2002). Metaphor, ad hoc concepts and word meaning - more questions than answers. UCL Working Papers in Linguistics, 14(2002), 83–105. Casarett, D., Pickard, A., Fishman, J. M., Alexander, S. C., Arnold, R. M., Pollak, K. I., & Tulsky, J. A. (2010). Can metaphors and analogies improve communication with seriously ill patients? Journal of Palliative Medicine, 13(3), 255–260. Clark, H. (1987). Relevance to what? Behavioral and Brain Sciences, 10(4), 714–715. Dascal, M. (1992). On the pragmatic structure of conversation. In H.  Parret & J.  Verschueren (Eds.), (On) Searle on conversation (pp. 35–57). Amsterdam, Netherlands/Philadelphia, PA: John Benjamins Publishing. Dunin-Keplicz, B., & Verbrugge, R. (2001). The role of dialogue in cooperative problem solving. In E.  Davis, J.  McCarthy, L.  Morgenstern, & R.  Reiter (Eds.), Proceedings of the 5th international symposium on logical formalizations of Commonsense reasoning (pp. 89–104). New York, NY: Courant Institute of Mathematical Sciences, New York University. Ervas, F., Gola, E., & Rossi, M. G. (2015). Metaphors and emotions as framing strategies in argumentation. In G.Airenti, B. Bara, & G. Sandini (Eds.), Proceedings of the EuroAsianPacific Joint Conference on Cognitive Science (pp. 645–650). Torino, Italy: CEUR. Ervas, F., Gola, E., & Rossi, M.  G. (2018). Argumentation as a bridge between metaphor and reasoning. In S.  Oswald, T.  Herman, & J.  Jacquin (Eds.), Argumentation and language— Linguistic, cognitive and discursive explorations (pp. 153–170). Cham, Switzerland: Springer.

The Communicative Functions of Metaphors


Ervas, F., Montibeller, M., Rossi, M. G., & Salis, P. (2016). Expertise and metaphors in health communication. Medicina and Storia, 16(9–10), 91–108. Fischer, E. (2014). Messing up the mind? Analogical reasoning with metaphors. In H. J. Ribeiro (Ed.), Systematic approaches to argument by analogy (pp.  129–148). Cham, Switzerland: Springer. Flusberg, S., Matlock, T., & Thibodeau, P. (2018). War metaphors in public discourse. Metaphor and Symbol, 33(1), 1–18. Gibbs, R. (1992). When is metaphor? The idea of understanding in theories of metaphor. Poetics Today, 13(4), 575–606. Gibbs, R. (2006). Metaphor interpretation as embodied simulation. Mind and Language, 21(3), 434–458. Goatly, A. (2011). The language of metaphors. London, UK: Routledge. Grosz, B., & Sidner, C. (1986). Attention, intentions, and the structure of discourse. Computational Linguistics, 12(3), 175–204. Gu, Y. (1993). The impasse of Perlocution. Journal of Pragmatics, 20(5), 405–432. https://doi. org/10.1016/0378-2166(93)90038-Q. Hahn, U. (2011). The problem of circularity in evidence, argument, and explanation. Perspectives on Psychological Science, 6(2), 172–182. Haugh, M. (2015). Im/politeness Implicatures. Berlin, Germany: De Gruyter. Hendricks, R.  K., Demjén, Z., Semino, E., & Boroditsky, L. (2018). Emotional implications of metaphor: Consequences of metaphor framing for Mindset about cancer. Metaphor and Symbol, 33(4), 267–279. Hopper, R., Knapp, M. L., & Scott, L. (1981). Couples’ personal idioms: Exploring intimate talk. Journal of Communication, 31(1), 23–33. Jaszczolt, K. (2002). Semantics and pragmatics: Meaning in language and discourse. Harlow, UK: Pearson, 358. Kittay, E. F. (1989). Metaphor: Its cognitive force and linguistic structure. Oxford, UK: Oxford University Press. Kovecses, Z. (2015). Where metaphors come from: Reconsidering context in metaphor. Oxford, UK: Oxford University Pres. Leech, G. (1981). Semantics: The study of meaning. Harmondsworth, UK: Penguin. Leech, G. (1983). Principles of pragmatics. London, UK: Longman. Levin, J., & Moore, J. (1977). Dialogue-games: Metacommunication structures for natural language interaction. Cognitive Science, 1(4), 395–420. Levinson, S. (1992). Activity types and language. In P. Drew & J. Heritage (Eds.), Talk at work: Interaction in institutional settings (pp. 66–100). Cambridge, UK: Cambridge University Press. Levinson, S. (2012). Action formation and ascription. In J. Sidnell & T. Stivers (Eds.), The handbook of conversation analysis (pp. 101–130). Chichester, UK: Wiley. Macagno, F. (2008). Dialectical relevance and dialogical context in Walton’s pragmatic theory. Informal Logic, 28(2), 102–128. Macagno, F., & Bigi, S. (2017). Analyzing the pragmatic structure of dialogues. Discourse Studies, 19(2), 148–168. Macagno, F., & Bigi, S. (2020a). Analyzing dialogue moves in chronic care communication. Dialogical intentions and customization of recommendations for the assessment of medical deliberation. Journal of Argumentation in Context, 9(2),167–198. ­ jaic.18044.mac. Macagno, F., & Bigi, S. (2020b). The role of evidence in chronic care decision-making. Topoi. Macagno, F., & Rossi, M.  G. (2019). Metaphors and problematic understanding in chronic care communication. Journal of Pragmatics, 151, 103–117. pragma.2019.03.010. Macagno, F., & Zavatta, B. (2014). Reconstructing metaphorical meaning. Argumentation, 28(4), 453–488.


F. Macagno and M. G. Rossi

McBurney, P., & Parsons, S. (2009). Dialogue games for agent argumentation. In G.  Simari & I.  Rahwan (Eds.), Argumentation in Artificial Intelligence (pp.  261–280). Heidelberg, Germany: Springer. Mey, J. (2016). Why we need the Pragmeme, or: Speech acting and its Peripeties. In K. Allan, A.  Capone, & I.  Kecskes (Eds.), Pragmemes and theories of language use (pp.  133–140). Cham, Switzerland: Springer. Moran, R. (1989). Seeing and believing: Metaphor, image, and force. Critical Inquiry, 16(1), 87–112. Moran, R. (2017). Artifice and persuasion: The work of metaphor in the rhetoric. In R.  Moran (Ed.), The Philosophical Imagination: Selected Essays (pp.  49–60). Oxford, UK: Oxford University Press. Musolff, A. (2004). Metaphor and political discourse: Analogical reasoning in debates about Europe. Houndmills, UK: Palgrave Macmillan. Newman, S. (2002). Aristotle’s notion of ‘bringing-before-the-eyes’: Its contributions to Aristotelian and contemporary conceptualizations of metaphor, style, and audience. Rhetorica: A Journal of the History of Rhetoric, 20(1), 1–23.  Ortony, A. (1975). Why metaphors are necessary and not just nice. Educational Theory, 25(1), 45–53. Ottati, V., & Renstrom, R. (2010). Metaphor and persuasive communication: A multifunctional approach. Social and Personality Psychology Compass, 4(9), 783–794. https://doi. org/10.1111/j.1751-9004.2010.00292.x. Ottati, V., Rhoads, S., & Graesser, A. C. (1999). The effect of metaphor on processing style in a persuasion task: A motivational resonance model. Journal of Personality and Social Psychology, 77(4), 688–697. Rapanta, C. (2019). Argumentation strategies in the classroom. Wilmington, DE: Vernon Press. Read, S. J., Cesa, I. L., Jones, D. K., & Collins, N. L. (1990). When is the federal budget like a baby? Metaphor in political rhetoric. Metaphor and Symbol, 5(3), 125–149. https://doi. org/10.1207/s15327868ms0503_1. Ricoeur, P. (1976). Interpretation theory: Discourse and the surplus of meaning. Fort Worth, TX: Texas Christian University Press. Ricoeur, P. (1993). The rule of metaphor: Multi-disciplinary studies of the creation of meaning in language. Toronto, ON: University of Toronto Press. Rossi, M. G. (2016). Metaphors for patient education. A pragmatic-argumentative approach applying to the case of diabetes care. Rivista Italiana Di Filosofia Del Linguaggio, 10(2), 34–48. Rossi, M. G. (forthcoming). Metaphors and persuasion in Healthcare communication. Language. Rossi, M. G., Macagno, F., & Bigi, S. (submitted). Dialogical functions of metaphors in medical communication. Ruhi, Ş. (2007). Higher-order intentions and self-politeness in evaluations of (Im)politeness: The relevance of compliment responses. Australian Journal of Linguistics, 27(2), 107–145. https:// Sanders, R. (1987). Cognitive foundations of calculated speech: Controlling understandings in conversation and persuasion. Albany, NY: Suny Press. Sanders, R. (2013). The duality of speaker meaning: What makes self-repair, insincerity, and sarcasm possible. Journal of Pragmatics, 48(1), 112–122. pragma.2012.11.020. Santibáñez, C. (2010). Metaphors and argumentation: The case of Chilean parliamentarian media participation. Journal of Pragmatics, 42(4), 973–989. pragma.2009.08.019. Schank, R. (1986). Explanation patterns: Understanding mechanically and creatively. Hillsdale, NJ: Lawrence Erlbaum Associates. Schank, R., Kass, A., & Riesbeck, C. (2014). Inside case-based explanation. New  York, NY: Psychology Press. Schiappa, E. (2003). Defining reality. Definitions and the politics of meaning. Carbondale and Edwardsville, IL: Southern Illinois University Press.

The Communicative Functions of Metaphors


Schneck, P. (2011). Rhetoric and evidence: Legal conflict and literary representation in US American culture (Vol. 1). Berlin, Germany/New York, NY: Mouton de Gruyter. Searle, J. (1979). Metaphor. In A. Ortony (Ed.), Metaphor and thought (pp. 93–123). New York, NY: Cambridge University Press. Searle, J., & Vanderveken, D. (1985). Foundations of illocutionary logic. Cambridge, UK: Cambridge University Press. Semino, E. (2008). Metaphor in discourse. Cambridge, UK: Cambridge University Press. Semino, E., Demjén, Z., & Demmen, J. (2016). An integrated approach to metaphor and framing in cognition, discourse, and practice, with an application to metaphors for cancer. Applied Linguistics, 39(5), 625–645. Slovic, P., & Tversky, A. (1974). Who accepts Savage’s axiom? Behavioral Science, 19(6), 368–373. Snoeck Henkemans, F. (2001). Argumentation, explanation and causality: An exploration of current linguistic approaches to textual relations. In T.  Sanders, J.  Schilperood, & W.  Spooren (Eds.), Text representation: Linguistic and psycholinguistic aspects (pp. 231–246). Amsterdam, Netherlands: John Benjamins Publishing. Sopory, P., & Dillard, J. P. (2002). The persuasive effects of metaphor a meta-analysis. Human Communication Research, 28(3), 382–419. Stanovich, K. E., & West, R. F. (1999). Discrepancies between normative and descriptive models of decision making and the understanding/acceptance principle. Cognitive Psychology, 38(3), 349–385. Strawson, P. (1964). Intention and convention in speech acts. The Philosophical Review, 73(4), 439–460. Streeck, J. (1980). Speech acts in interaction: A critique of Searle. Discourse Processes, 3(2), 133–153. Toulmin, S. (1958). The uses of argument. Cambridge, UK: Cambridge University Press. Ungerer, F., & Schmid, H.-J. (2006). An introduction to cognitive linguistics. Harlow, UK: Pearson. Van Eemeren, F., & Garssen, B. (2014). Argumentation by analogy in stereotypical argumentative patterns. In H. J. Ribeiro (Ed.), Systematic approaches to argument by analogy (pp. 41–56). Cham, Switzerland: Springer. van Eemeren, F., & Grootendorst, R. (1984). Speech acts in argumentative discussions: A theoretical model for the analysis of discussions directed towards solving conflicts of opinion. Dordrecht, Netherlands: Floris Publications. Van Fraassen, B. (1980). The scientific image. Oxford, UK: Oxford University Press. Walton, D. (1989). Informal logic. New York, NY: Cambridge University Press. Walton, D. (1990). What is reasoning? What is an argument? Journal of Philosophy, 87, 399–419. Walton, D. (1998). The new dialectic. Conversational contexts of argument. Toronto, ON: University of Toronto Press. Walton, D. (2004). A new dialectical theory of explanation. Philosophical Explorations, 7(1), 71–89. Walton, D. (2006). Fundamentals of critical argumentation. New  York, NY: Cambridge University Press. Walton, D. (2010). Types of dialogue and burdens of proof. In P. Baroni, F. Cerutti, M. Giacomin, & G. Simari (Eds.), Computational models of argument (COMMA) (pp. 13–24). Amsterdam, Netherlands: IOS Press.  Walton, D. (2011). A dialogue system specification for explanation. Synthese, 182(3), 349–374. Walton, D. (2015). Goal-based reasoning for argumentation. Cambridge, MA: Cambridge University Press. Walton, D., & Krabbe, E. (1995). Commitment in dialogue. Albany, NY: State University of New York Press. Xu, C., & Wu, Y. (2014). Metaphors in the perspective of argumentation. Journal of Pragmatics, 62, 68–76.

Stereotypes Favour Implicatures and Implicatures Smuggle Stereotypes: The Case of Propaganda Edoardo Lombardi Vallauri

Introduction The paper builds on the assumption that conveying some information implicitly causes reduced critical attention on that content by addressees, as compared to the overt assertion of the same content. In section “The Evolution of Critical Attention: Implicit Strategies as Persuasive Devices” it will be argued that this may have precise evolutionary reasons, connected to the core functions for which language has evolved, and to the fact that – in particular – implicatures partly entrust the construction of the message content to the addressee. Section “Stereotypes in the Semantics of Advertising” will recall that stereotypes are very useful and as a consequence frequently used in persuasive communication, such as for instance in advertising, but when a stereotype is too bold, too simplistic or exaggerate, it conveys the risk of causing irritation on the part of (some) addressees, and rejection of the message. In section “The Combined Effects of Stereotypes and Implicatures in Advertising and Propaganda”, stereotypes will be regarded as complex pieces of information that can be used to provide addressees with the necessary conceptual background for the intended implicatures to arise. At the same time, it will be shown that questionable or exaggerate stereotypes are more likely to escape critical rejection if they are recalled via some implicature, than if directly stated. Examples from advertising and political propaganda will be given.

E. Lombardi Vallauri (*) Dipartimento di Lingue, Letterature e Culture Straniere, Università Roma Tre, Rome, Italy e-mail: [email protected] © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 F. Macagno, A. Capone (eds.), Inquiries in Philosophical Pragmatics, Perspectives in Pragmatics, Philosophy & Psychology 27,



E. Lombardi Vallauri

 he Evolution of Critical Attention: Implicit Strategies T as Persuasive Devices In a well-known paper dealing with animal signals, Krebs and Dawkins maintained that communication evolved to ensure not only cooperation, but especially manipulation, i.e. to actively change the other individuals’ behaviors (Krebs and Dawkins 1984, 383). In the same paper, they recognized that the ability to manipulate others cannot have evolved separately from the specular ability to resist it, i.e. to detect attempts at manipulating us and to consequently avoid the behaviours aimed at by the manipulator (Krebs and Dawkins 1984, 390–92). Crucial in this detection task is the fact of becoming aware of the manipulator’s intentions. This could explain the way many manipulating messages are designed. Take the following advertising clip, broadcast by a widely diffused whisky brand (Fig. 1): The ad (in its subsequent versions, during a couple of decades) displays a group of rich, handsome, young and happy people drinking that brandy in a very elegant mansion (Fig. 2). The aim of these images is to convey the idea that if you drink the same whisky, you will be (at least a little bit more) like those people: and it actually works, otherwise it would make no sense to spend huge amounts of money to produce the clip and broadcast it world- or nationwide. Interestingly the same content, if stated explicitly, would not work. Imagine a voice telling you from the screen: “If you buy and drink this whisky, you will be (or at least feel) richer, more handsome, more elegant, happier, end even younger!”. This would rather trigger hostile reactions than convince anybody. But, if presented in a visual, not explicitly stated form, the same idea works pretty well.

Fig. 1  Images from a whisky advertising videoclip

Stereotypes and Implicatures in Propaganda


Fig. 2  Images from a later version of the advertisement of Fig. 1 

This demands explanation, which can be derived from the evolutionary assumptions we have just summarized. In other animals, reactions to manipulative attempts take their specific forms described by Krebs and Dawkins, all accounted for by the general statement that “interactions are likely to be characterized by coevolution between persuasion and sales-resistance” (Krebs and Dawkins 1984, 394). We try to manipulate others and when, in turn, they try to manipulate us, if we become aware of that, we resist by critically challenging their messages and – in case they are not completely convincing – possibly rejecting them. As a consequence of this, the particular kind of animal interaction represented by linguistic communication between humans seems to obey the following rule: Awareness that the source is trying to modify our beliefs raises critical reaction.

Now, this is the very essence of assertion: to propose some content as something by which one wants to modify your beliefs. Statements overtly show that their source endorses their content and wants for us to get convinced about it. On the contrary, images and non-linguistic sounds give us the impression that we are free to give them any value we want. And the feeling that someone is trying to modify our set of beliefs is what typically triggers our critical reaction, while the feeling that we are free to form our opinion autonomously reasonably reduces our tendency to challenge the information we are exposed to. This makes assertion particularly unsuitable for persuasion. So, when trying to persuade of some doubtful content, if that content cannot be conveyed by non-­ linguistic means and really needs to be encoded in a text, that text should better convey it in a different, non-assertive form.


E. Lombardi Vallauri

Languages provide users with a partial solution to this problem. As a matter of fact, implicit constructions such as presuppositions and implicatures can reduce addressees’ critical attention on the encoded content, as it has been noticed by several scholars (see Ducrot 1972; Givón 1982; Kerbrat-Orecchioni 1986; Lombardi Vallauri 1993, 1995; Rigotti 1988; Sbisà 2007) and specifically relative to texts with persuasive aims (Lombardi Vallauri 1995, 2009b, 2016, 2019; Sbisà 2007; Reboul 2011; De Saussure 2013; Lombardi Vallauri and Masia 2014). As suggested by Reboul (2011, 10), “implicit communication evolved to facilitate manipulation by allowing communicators to hide their (manipulative) intentions”. By partly concealing the communicator’s intention in producing a speech act, implicit strategies “circumvent” critical judgment. In particular, as claimed by Mercier (Mercier 2009, 117), when someone tells us something, when we accept it and make inferences based on this information, the resulting conclusions are considered as ours, and will be more easily accepted than if they had been explicitly communicated. In other words, “the less important the communicator’s role in the formation of the conclusion by the addressee, the more the addressee will accept the conclusion” (Mercier 2009, 118, translated). This has sometimes been called the “egocentric bias”: The egocentric bias leads to a preference for one’s own beliefs and will induce a preference for beliefs which one has reached by oneself; this explains why it may be advantageous for the communicator to use implicit communication: it allows him to induce in his addressee beliefs (i.e., reasons and conclusions) which the addressee having reached them by himself will be more prone to accept and to hide his ultimate intentions regarding the conclusion he wants the addressee to reach as to the future course of her action. (Reboul 2011, 17)

This seems related to what we know about the general cognitive processes by which humans reach their judgments on surrounding reality. In particular, to the fact that: people are nearly-incorrigible “cognitive optimists”. They take for granted that their spontaneous cognitive processes are highly reliable, and that the output of these processes does not need re-checking. (Sperber et al. 1995, 90)

Importantly, this should not be regarded as a defect in the system though, as Tversky & Kahneman (1974), or more recently Gigerenzer (2008) have shown: heuristics are the results of an evolutionary drive in optimising cognitive efficiency as they offer the best balance between speedy derivation of new knowledge and costly inferential thorough evaluation processes. As such they offer fast and reasonably robust means of acquiring new knowledge at a fraction of the cognitive cost. [...] our cognitive system tends to privilege fast and economical processes over reflective ones, thereby giving prevalence to cognitive illusions. (Oswald et al. 2016, 524)

As concerns language, since its progression in communicative situations is fast (Christiansen and Chater 2016), processing must be done quickly, and attention cannot be full on all contents. As a consequence, we are led to lend thorough critical attention to contents asserted by others, and to be more “optimistic”, i.e. to process with lesser epistemic vigilance (Sperber et al. 2010) those contents we have at least in part built by ourselves. We tend not to double-check what comes from us.

Stereotypes and Implicatures in Propaganda


As a consequence, linguistic implicit strategies are the best approximation offered by language to visual (and musical) signals, in that they reduce the addressees’ awareness that they are being convinced of some content. Among them, implicatures work by entrusting to addressees the building of part of the semantic content of the message, having recourse to contextual cues and to their general knowledge of the world. Addressees may feel implicated contents more as something they have arrived at by themselves than as something the speaker is imposing on them; or if so, not the speaker alone (Lombardi Vallauri 2009a). This makes implicated contents less prone to be challenged than asserted ones. We will see how this can affect the difference between presenting stereotypes explicitly and making allusion to them by means of some implicature.

Stereotypes in the Semantics of Advertising Stereotypes, as explained by Putnam (1975), determine the social nature of meaning. As pointed out e.g. by Lakoff (1987), Levinson (2000, 37–38; 112–34) and Wilson (2003), they guide semantic interpretation, typically in semantic narrowing. The meanings of some lexical items, if they occur in utterances such as those in examples (A) and (B), are interpreted in a narrow sense which is guided by stereotypes: (A) All doctors drink (B) Mary is a working mother In (A) “drink might convey not the encoded sense ‘drink liquid’ but, more specifically, ‘drink alcohol’, or ‘drink significant amounts of alcohol’” (Wilson 2003: 274). In (B) it is not just said that Mary is a mother and works, but that she is a stereotypical working mother, who must bring up small children alone, working outside the home, etc. (Lakoff 1987, 80–82). In this perspective, semantic narrowing is regarded as a default inference guided by an informativeness heuristic according to which what is expressed simply must be interpreted as stereotypically exemplified, i.e. having as its meaning the stereotypical meaning associated to that expression. This is not valid for all cases of semantic narrowing (Barsalou 1987); but more in general, as noted already by Shegloff (1972, 102) and by Levinson (1998, 566–67), minimal expressions require to be attributed a meaning via some implicature, which typically involves recourse to the stereotype associated to that expression: “she’s gone to the supermarket” or “she’s gone to school” also mean that she will do the typical activities usually carried out respectively in supermarkets and schools. We will try to shed light on what the consequences of this mechanism are, when applied in strongly indirect communication: essentially, conversational implicatures used with persuasive purposes. We will show that in persuasive texts such as advertisements, (i) the interpretation of implicatures often relies on stereotypes, and (ii) the persuasiveness of stereotypes often relies on their being recalled via implicature.


E. Lombardi Vallauri

In other words, implicatures would not be successful without appealing to stereotypes, and stereotypes would not be successful without being disguised by implicatures. Stereotypes are widely used in advertising. Ads typically exploit assumptions like “Men are strong and ambitious”, “Women like to shop” or “Children love to play with toys”. This is due to multiple factors, among which their being able to convey complex contents very quickly (Sheehan 2014, 75): Advertisements present brief dramatic stories with a message (i.e., a selling message) in a very short time period. For example, television commercials take place in 30 seconds and most print advertisements fit into a single page of a magazine. [...] Using stereotypes in advertising messages quickly sets the stage for the messages: Stereotypes convey characters and images quickly and clearly.

Moreover, stereotypes have a specifically persuasive effectiveness (Sheehan 2014, 79): Studies have shown that if you have knowledge of how others sterotype people who are part of your own demographic group, you will start acting that way. [...] The real power of stereotypes is their ability to change the behavior of the person holding the stereotype.

This means that if an ad stereotypically portrays a person about your age and social role, you will be strongly pushed to imitate her/his behavior. Still, though clearly invited by their capacity to catalyze the target’s agreement, the use of stereotypes contains some risks. Consumers are becoming more and more defensive. Decades of deceptive advertising have engendered distrust, which negatively affects people’s responses to subsequent advertising from any source. The initial deception induces negative beliefs about advertising and marketing in general, thus undermining the credibility of further advertising (Darke and Ritchie 2007). As a consequence (Sheehan 2014, 81–82), also stereotypes must be used with caution: Do stereotypes work in advertising? The answer is, It depends. If people in the target audience relate positively to a stereotypical portrayal, the portrayal may help sell the product. If people in the target audience create opinions about groups of people other than the group that they are in, and these opinions put the stereotyped group into a negative light, a stereotyped portrayal could be problematic. Although we can all recognize stereotypes that are blatantly offensive, how we view other portrayals depends on the taste and sensitivity of members of the target audience. For example, many advertisements targeted to women portray husbands and fathers as incompetent around the home. Women may laugh at these portrayals, but men viewing the same commercials may be less amused. [...] Advertisers have an obligation to be aware that many consumers may find the use of certain stereotypes questionable and possibly offensive. [...] In general, advertisers need to routinely examine and evaluate imagery and language that they use to determine how consumers are affected by the messages, both in terms of the meaning consumers derive from the messages and the overall appropriateness of the strategic approach. In particular, any messages that may cross the fuzzy boundaries of good taste should be closely evaluated.

As any other content whose encoding may cause critical reactions, stereotypes whose boldness makes them difficult to convey by means of overt assertion may take advantage from being encoded indirectly. For instance, by means of implicature. We will see how this can happen, considering that stereotypes are the typical

Stereotypes and Implicatures in Propaganda


kind of information that can easily be recovered although not explicitly stated, and present themselves among the first ideas available to the mind of the addressee when (s)he starts an implicational process.

 he Combined Effects of Stereotypes and Implicatures T in Advertising and Propaganda Implicatures are more easily drawn if the content to be recovered is a stereotype, as compared to less expectable information. And, at the same time, stereotypes are more easily accepted and less probably challenged if they are presented implicitly, via implicature. Advertising and political propaganda make intense use of this synergy. We will now show some examples from Italian public communication of the last 30 years. The unofficial catholic ad in (Fig. 3) suggests addressees to pray for each other and for the raising of fraternity among the people in the world. The message can only make sense if prayers are effective, which is a very important content for religious propaganda and, at the same time, a diffused, mentally highly available stereotype. As a consequence, the message raises the implicature that if we pray for fraternity, fraternity will happen. At the same time, it implicitly proposes and, so to say, “refreshes” the stereotype that prayers effect results. It may even create the

Fig. 3  (1) Let us always pray for us, for each other, let us pray for the world, for a great fraternity!


E. Lombardi Vallauri

stereotype in the addressees’ minds, if they do not possess it already. Now, crucially, though it can help the implicature to arise, this stereotype is by far too simplistic to be believed if one really pays attention to it. Even the most official versions of religions consider it incorrect. Direct assertion that prayers actually cause preferred things to happen would rather trigger critical evaluation and rejection by average addressees; still, if just alluded at implicitly, the same content – being reconstructed by the addressee himself – can exploit the egocentric bias, escape critical judgment and get reinforced by the ad. The ad in (Fig. 4, ad no. 2) exploits the stereotype that “broadcasting companies distort information”. The headline, apparently referred to a “rectified” Pisa Tower, effectively appeals to some interpretation maximizing relevance, and – due to the existence of the mentioned stereotype – triggers the implicature that other channels show reality different from how it is, while Discovery Channel doesn’t. Both claims would be found more questionable by the target if they were asserted explicitly: TV channels always show reality different from how it is; and On the contrary, Discovery Channel doesn’t, would be too bold assertions to convince as such. The “lighter” critical checking the same content receives when introduced implicitly helps them escape possible awareness of their being exaggerate. The stereotype exploited in (Fig. 4, ad no. 3) is that “sugar is good”, that is to say a perfect example of pleasant taste. At the time the ad was diffused (the Nineteen-­ eighties), aspartame was a new product, and it immediately suffered from the suspect of causing cancer. The aim of the ad was to demolish this opinion. But obviously

Fig. 4  (2) Is it really necessary to correct everything? (3) As good as sugar. As safe as aspartame

Stereotypes and Implicatures in Propaganda


the unrequested, direct statement that aspartame was healthy and did not cause cancer would have obtained the effect of a self-accusation. So, the copywriters chose to propose aspartame as the very prototype of a healthy substance: in other words, they tried to establish a stereotype according to which aspartame is the typical healthy stuff. Since such a content would have been immediately rejected if proposed explicitly, they were bound to introduce it by way of implicature. The effect was strengthened by means of sophisticated rhetoric: the juxtaposition of two formally identical hemistichs enforces the impression that they are also identical in truth value. In the first, sugar is mentioned, actually a widely recognised prototype of a well-tasting food. This triggers a “similarity-effect” in the interpretation of the second hemistich, implicitly inducing the reader to identify aspartame as no less than a universally recognised prototype of safe food. Interestingly, the stereotypical conception about sugar can induce the parallel conception about aspartame exactly because, instead of being mentioned directly, both are evoked implicitly by way of implicature. In particular, the direct statement that aspartame is the typical healthy substance, which is even stronger than its being simply healthy, would have been believed by nobody, and would rather have raised strong critical evaluation and rejection. In other words, what we observe here is that a stereotype provides the conditions for the establishment of an implicature, and the fact of being implicated rather than asserted provides the conditions for the acceptance of another stereotype. Implicit is the key to the persuasive success of a content. The ads no. (4) and (5) (Fig. 5) encode some suggestions on the part of the pope. They are very easy to accept. Interestingly, addressees who consciously agree with them will also (less consciously) implicate that their felicity conditions are true, namely, that: (implicature i) There are people who try to steal our hope.

Fig. 5  (4) Don’t let them steal your hope (5) We must not be afraid of goodness and tenderness


E. Lombardi Vallauri

(implicature ii) We fear goodness and tenderness. Such implicatures, in association with the images of the pope, evoke at least three further implicit contents, namely three stereotypes which one would hardly state without risking to sound excessively bold, but which nevertheless significantly help the implicatures themselves to be drawn: (stereotype i′) Those who steal your hope are typically not Catholic (stereotype i″) Being a Catholic makes it easier to experience hope. (stereotype ii′) It is easier to experience goodness and tenderness if one follows the Catholic religion. In other words, the stereotypes (i′), (i″) and (ii′) are useful for the implicatures (i) and (ii) to be drawn, but at the same time the fact that they are not explicitly stated, but just evoked within the process of drawing the implicature, exploits the egocentric bias and so prevents addressees from becoming aware of their being at least exaggerate, if not completely false. Reverting to commercial advertisements, the fruit juice ad shown in (Fig.  6) exploits the widely accepted stereotype that the most desirable product in this domain is made 100% of fruit. (In this case, the stereotype doesn’t need particular reinforcing because it is kept alive by the usual claim “100% fruit”, which is to be found virtually on all cartons.) As a consequence, “100% Yoga” in (6) conveys by implicature something which tries to present itself as a further stereotype, but one that would be hardly assertable, namely that Yoga is the most desirable content of a fruit juice carton. Stereotypes and implicatures interact in a strong way when it comes to political propaganda, probably because ideology is one of the domains where stereotypes play a greater role. We will exemplify this from the 2006 national political campaign in Italy. On that occasion, the announcements on both political sides consisted in accusing the opposite side of planning measures that would certainly sound unwelcome to undecided voters, who are the target of any political campaign. Explicitly accusing the opponents of bad intentions would have been unpleasant and counterproductive: perhaps even illegal, since such measures were not included in the electoral programs of neither side. So, the accusations had to be indirect, i.e. encoded by means of implicatures. This allowed for them to convince lots of people without raising their critical reaction (Fig. 7): As it can be seen, announcements systematically exploited the same strategy (ads no. 7–12 of Fig. 7 represent all the posters diffused by the Right coalition in 2006): by explicitly refusing a possible political action, the message triggers a typical Gricean path. Namely, each public refusal of a given unwelcome measure would violate the Maxim of Relation in case there is no threat that the rejected thing may be become real. As a consequence, the uttering of the refusal raises the implicature that the Left will actually do that threatening thing if they win the election (Table 1): Since the implicature is drawn by every person in the target audience “autonomously”, the Right does not count (in the target’s mind) as responsible for throwing unpleasant and largely false accusations at the Left. And, moreover, the very

Stereotypes and Implicatures in Propaganda


Fig. 6  (6) Hundred per cent Yoga. No sugar, no hurry, just fruit

contents of such accusations are not evaluated under the full light of a critical attitude, as they would if, instead of being constructed by the target themselves, they were presented to them as direct assertions on the part of the Right. Imagine the most probable reactions (on the part of undecided, equidistant voters) to posters boldly saying: The Left will re-introduce the inheritance tax!, or The Left will/wants to welcome immigrants without any regulation!, etc. The consequence of implicit encoding is that such far-from-true contents were more likely to be believed.1 1  We do not dwell here upon the presence of vagueness, which is among the reasons why these messages include exaggeration in their meanings. For a thorough account of the issue, cf. Lombardi


E. Lombardi Vallauri

Fig. 7  (7) Inheritance tax again? No, thanks. (8) The “no globals” in the government? No, thanks (9) Illegal immigrants at will? No, thanks (10) More taxes on your savings? No, thanks (11) More taxes on your house? No, thanks (12) Halting major works? No, thanks Table 1  The Gricean path from refusing statements to accusing implicatures in Italian 2006 political ads of the Right coalition  Statement (by the right) We are against the inheritance tax We are against the “no globals” in the government We are against illegal immigrants without regulation We are against more taxes on your savings We are against more taxes on your house We are against halting major works

Gricean path It makes no sense to mention this threat, if there is no danger that it comes true. Consequently,

Implicature (drawn by the target) the Left will introduce the inheritance tax the Left will put the “no globals” in the government the Left will accept illegal immigrants without regulation the Left will put more taxes on my savings the Left will put more taxes on my house the Left will halt major works

Interestingly, the implicature is always guided by a stereotypical concept of the Left: stereotype a. – The Left takes sides with all “irregular people”: immigrants, no globals etc. Vallauri (2009a), (2016) and (2019).

Stereotypes and Implicatures in Propaganda


stereotype b. – The Left is against major economical enterprises, including big construction work. stereotype c.  – The Left strongly taxes incomes, properties, real estate and even savings. The stereotypical assumptions above guide the implicature process, and at the same time they are re-inforced by that process. If asserted, such stereotypes would appear as too simplistic, and would probably be rejected, as signalled e.g. by Sheehan (2014); but since they are only evoked implicitly, they exploit the protecting effect of the egocentric bias, getting some chances to bypass the addressees’ critical reaction and to be transferred into their set of beliefs. Addressees who read, say, ads no. (8) and (9) will be led to implicate that the Left will act in a way that favors the mentioned categories of lawbreakers. Inattentively evoking the underlying stereotype, they don’t process its details, and conceive of it in its general significance: “Political Lefts are less strict on irregular people, and as a consequence complicit with them”. Average addressees, in order to draw the implicature by which the single messages can be interpreted as cooperative, accept the general stereotype as valid, without verifying to which extent it applies to the specific Left at issue (the 2006 Ulivo/Margherita). If they applied their critical attention more deeply, they would become aware that at least in that case the stereotype is actually exaggerate, and ultimately false. The propaganda of the Left coalition in the same elections was strikingly similar, which confirms that what we are describing here is not the result of casuality or coincidence. Here follow the ads by the Left (Fig. 8): Posters no. (13–15) of (Fig.  8) elegantly assert truisms, namely that nursery schools are an important help to families, that a good public health service frees the people from having to worry about what will happen if they get sick, and that the constant threat of remaining without a job is more or less the opposite of having good hopes for the future. The aim of the campaign was more than that. Such statements, if uttered during an electoral campaign, since they have to be understood as cooperative (in the Gricean sense), raise the implicatures that nursery schools are likely to be reduced, that the public health service may be cut, and that firms may be allowed to increase forms of temporary work, thus damaging (especially young) workers; crucially, all this will be done by the Right (Table 2): Now, if it is implied that the threats come from the Right, it is because these indirect accusations are carefully chosen among things that belong in the stereotype of a Right party: stereotype a. – The Right cuts on welfare, such as public schools, health services and the like. stereotype b. – The Right takes sides with the (big) firms and, in general, against the interests of workers. Once again, it can be observed that the accusing statements about the future actions of the Right would hardly be accepted by (undecided and equidistant) addressees if presented in the posters in the form of assertions, and in any case would put the


E. Lombardi Vallauri

Fig. 8  (13) Without nursery schools, families can’t grow. (14) A public health service that works means more freedom. (15) Temporary work clamps down your hopes Table 2  The Gricean path from obvious statements to accusing implicatures in Italian 2006 political ads of the Left coalition  Statement (by the left) Without nursery schools, families can’t grow A public health service that works means freedom Temporary work clamps down your hopes

Gricean path It makes no sense to mention this truism, if there is no danger about it. Consequently,

Implicature (drawn by the target) the right will cut on nursery schools the right will cut on the public health service the right will favour forms of temporary work

accusing Left in a bad light. However, the fact that they result from an inferential process carried out by the addressee himself excludes the Left from being responsible for them, and locates the accusations among the contents which the addressee has come at by himself; thus including them in the egocentric bias and protecting them from being critically challenged. The same holds for the general, stereotypical assumptions themselves, mentioned in (a) and (b). They are understood in their most generic sense: “The Right Parties are the allies of the rich, and the enemies of workers”. In order to draw the implicature by which the message can be interpreted as cooperative, addressees accept this assumption as valid, and they do not verify to which extent it actually applies to Forza Italia in 2006. If they did so, they would become aware that, at least in that case, they find the stereotype exaggerate and incorrect.

Stereotypes and Implicatures in Propaganda


In other words, each ad convinces addressees about the implicated content by exploiting the existence of a stereotype, and at the same time it reinforces and confirms the (persuasively useful) stereotype in the mind of addressees by evoking it in an implicit manner, which allows for it to make the most of the egocentric bias.

Conclusions In the examples of persuasive communication we have commented on, stereotypes and implicatures reinforce each other. On the one side, stereotypes facilitate addressees to draw implicatures, thus allowing for questionable contents to be conveyed in an implicit way, which makes them fall under the egocentric bias, thus being more likely to avoid critical evaluation and rejection. On the other side, stereotypes themselves, being not asserted but just evoked during the inferential process, benefit from the egocentric bias and are less likely to be recognised as simplistic and exaggerate representations of reality. This can become particularly important, and significantly affect democratic cohabitation, when it is done in messages influencing political life, by reducing critical evaluation of simplistic stereotypes that stem from some ideological background.

References Barsalou, L. (1987). The instability of graded structure: Implications for the nature of concepts. In U. Neisser (Ed.), Concepts and conceptual development: Ecological and intellectual factors in categorization (pp. 101–139). Cambridge, UK: Cambridge University Press. Christiansen, M., & Chater, N. (2016). The now-or-never bottleneck: A fundamental constraint on language. Behavioral and Brain Sciences, 39(April), e62. S0140525X1500031X. Darke, P. R., & Ritchie, R. J. B. (2007). The defensive consumer: Advertising deception, defensive processing, and distrust. Journal of Marketing Research, 44(1), 114–127. https://doi. org/10.1509/jmkr.44.1.114. Ducrot, O. (1972). Dire et Ne Pas Dire. Paris, France: Hermann. Givón, T. (1982). Evidentiality and epistemic space. Studies in, 6(1), 23–49. https://doi. org/10.1075/sl.6.1.03giv. Kerbrat-Orecchioni, C. (1986). L’Implicite. Paris, France: Armand Colin. Krebs, J., & Dawkins, R. (1984). Animal signals: Mind-Reading and manipulation. In J. Krebs & N.  B. Davies (Eds.), Behavioural ecology: An evolutionary approach (pp.  380–402). Sunderland, MA: Sinauer Associates. Lakoff, G. (1987). Women, fire, and dangerous things. Chicago, IL: University of Chicago Press. Levinson, S. (1998). Minimization and conversational inference. In A. Kasher (Ed.), Pragmatics. Critical concepts (pp. 545–612). London, UK/New York, NY: Routledge. Levinson, S. (2000). Presumptive meanings: The theory of generalized conversational Implicature. Cambridge, MA: MIT Press. Lombardi Vallauri, E. (1993). Clausole a Contenuto Presupposto e Loro Funzione Discorsiva in Italiano Antico. Quaderni Del Dipartimento Di Linguistica Dell’Università Di Firenze, 4, 71–95.


E. Lombardi Vallauri

Lombardi Vallauri, E. (1995). Tratti Linguistici Della Persuasione in Pubblicità. Lingua Nostra, 2(3), 41–51. Lombardi Vallauri, E. (2009a). Grice Elettorale. In M.  Fatigante, L.  Mariottini, & E.  Sciubba (Eds.), Linguistica e Società. Studi in Onore Di Franca Orletti (pp. 172–184). Milano, Italy: Franco Angeli. Lombardi Vallauri, E. (2009b). La Struttura Informativa. Forma e Funzione Negli Enunciati Linguistici. Roma, Italy: Carocci. Lombardi Vallauri, E., & Masia, V. (2014). Implicitness impact: Measuring texts. Journal of Pragmatics, 61, 161–184. Lombardi Vallauri, E. (2019). La lingua disonesta. Bologna, Italy: Il Mulino. Lombardi Vallauri, E. (2016). The ‘exaptation’ of linguistic implicit strategies. Springer Plus, 5(1), 1106–1130. Mercier, H. (2009). La Théorie Argumentative Du Raisonnement. Paris, France: E.H.E.S.S. Oswald, S., Maillat, D., & de Saussure, L. (2016). Deceptive and uncooperative verbal communication. In L. de Saussure & A. Rocci (Eds.), Verbal communication (pp. 509–534). Berlin, Germany: Mouton de Gruyter. Putnam, H. (1975). The meaning of ‘meaning.’. In H. Putnam (Ed.), Philosophical papers (Vol. 2, pp. 215–271). Cambridge, UK: Cambridge University Press. Reboul, A. (2011). A relevance–theoretic account of the evolution of implicit communication. Studies in Pragmatics, 13, 1–19. Rigotti, E. (1988). Significato e Senso. In E. Rigotti & C. Cipolli (Eds.), Ricerche Di Semantica Testuale (pp. 71–120). Brescia, Italy: La Scuola. De Saussure, L. (2013). Background relevance. Journal of Pragmatics, 59, 178–189. https://doi. org/10.1016/j.pragma.2013.08.009. Sbisà, M. (2007). Detto Non Detto. Le Forme Della Comunicazione Implicita. Roma-Bari, Italy: Laterza. Sheehan, K.  B. (2014). Controversies in contemporary advertising. Thousand Oaks, CA: Sage Publications. Shegloff, E. (1972). Notes on a conversational practice: Formulating place. In D. Sudnow (Ed.), Studies in social interaction (pp. 75–119). New York, NY: Free Press. Sperber, D., Cara, F., & Girotto, V. (1995). Relevance theory explains the selection task. Cognition, 57(1), 31–95. Sperber, D., Clément, F., Heintz, C., Mascaro, O., Mercier, H., Origgi, G., & Wilson, D. (2010). Epistemic Vigilance. Mind & Language, 25(4), 359–393. https://doi. org/10.1111/j.1468-0017.2010.01394.x. Wilson, D. (2003). Relevance and lexical pragmatics. Italian Journal of Linguistics, 15, 273–292.