Philosophie du langage et de l'esprit - François Recanati

François Recanati

Philosophie du langage et de l'esprit

Collège de France

Année 2024-2025

Colloque - Discourse Files

The Vanished Jar: Children's Problems with Non-Verbal Identity Information

Gergely Csibra

Central European University, Vienne

Résumé

Résumé

Children become able to process verbal identity statements as they learn to attribute false beliefs. The mental files explanation points to the need for linked indexed files for solving both tasks, which children achieve around 3 to 5 years. Clearly, an indexed file is needed for attributing belief; but an identity statement could just result in merging two regular files. I argue that indexed files are necessary for understanding the statement. This opens the possibility that children might cope with non-verbal, perceptual identity information well before they pass false belief tests. In our study, children had to figure out which jar is identical with a jar that had vanished. The results dampen any expectations of earlier competence. Without the help of an identity statement children operate exclusively with regular files, which are subject to Strawson's Constraint (only one file per object). Consequently, children displayed strong mutual exclusivity. They could not get themselves to identify the vanished jar with the (for us) obvious candidate, which was already tracked by another file. In sum, deployment of indexed files is associated with language use. This is also reflected in the strong dependence of passing false belief tests on language development.

Josef Perner

Josef Perner received his PhD in Psychology from the University of Toronto. He was Professor in Experimental Psychology at the University of Sussex and is now Professor emeritus of Psychology and member of the Centre for Cognitive Neuroscience at the University of Salzburg.

François Recanati

Philosophie du langage et de l'esprit

Collège de France

Année 2024-2025

Colloque - Discourse Files

Tracking Discourse Referents in Symbolic Depictions

Gergely Csibra

Central European University, Vienne

Résumé

Scene constructions, or more broadly, episodic simulations, are internal representations of states, relations, and events in the spatial navigational system of the brain. These constructions serve various purposes, including action planning, imagination, and memory retrieval. The format of these constructions is hybrid: they arrange discrete elements (internal symbols of objects, agents, abstract entities) in analog space and time to represent relations among them iconically. Such constructions can also represent not inherently spatial relations and operate both in humans and in non-human species. However, humans also create external representations in a format analogous to scene constructions, which I call symbolic depictions. Symbolic depictions are composed of discrete elements (objects) that function as local physical symbols standing for their internally represented counterparts in scene constructions. The arrangement of these symbols in physical space and time depicts relations or events among the entities they stand for. When entities represented internally in a constructed scene are tied to object-symbols in external displays, I consider them functioning as discourse referents. I will outline the cognitive operations required to establish and track such discourse referents and argue that some of these operations are available in early ontogeny, even before capacities for tracking discourse referents linguistically emerge.

Gergely Csibra

Gergely Csibra is a professor at the Deparment of Cognitive Science at Central European University, Vienna, and a professor emeritus at Birkbeck, University of London.

François Recanati

Philosophie du langage et de l'esprit

Collège de France

Année 2024-2025

Colloque - Discourse Files

Stand-For Relations between Objects and Discourse Referents in Early Development

Barbu Revencu

Postdoctorant, Collège de France

Résumé

From static diagrams to dynamic animations, humans routinely convey information through depictions—representational stimuli in which visual objects are arranged in spatiotemporal configurations for communicative ends. First, I will argue that depictions recruit two representational layers: one that tracks object symbols in space, and one that represents the entities under discussion through discourse referents. The two layers are linked by an assignment function that provides pointers from the symbols to their corresponding discourse referents. This link constitutes a stand-for relation, which allows interpreters to gather information about the discourse referents from the object symbols. Second, I will present empirical evidence that stand-for relations are available in early human development. By 15 months, infants can interpret geometric shapes as symbols of familiar animals and objects: Pointing to a triangle and calling it "a duck" is enough for infants to interpret the triangle as a symbol of a duck. Infants use these stand-for relations to interpret how shapes move and interact, while recognizing that these relations do not extend outside the current communicative context. Finally, I will argue that positing stand-for relations in early ontogeny provides a better explanation of pretend play than alternative accounts.

Barbu Revencu

Barbu Revencu is a postdoctoral researcher at Collège de France in the ERC Mental Files project to François Recanati. His work investigates the cognitive mechanisms underlying depictions: representational stimuli, ranging from diagrams to animations, in which visual objects function as symbols and are arranged in spatial configurations for communicative ends.

François Recanati

Philosophie du langage et de l'esprit

Collège de France

Année 2024-2025

Colloque - Discourse Files

Guises of 'I'

Sam Cumming

UCLA

Résumé

I propose that communication with the indexical pronoun 'I' requires coordination on a familiar discourse referent, just like communication with an anaphoric (or deictic) use of 'she'. This proposal improves on a contrasting approach descended from Kaplan (1989) on which 'I' is interpreted by bridging anaphora to the event of speaking (which eliminates discretion from the interpretation by binding the pronoun to the speech event). I use the proposal to address a puzzle about communication with 'I' due to Andrea Onofri (2022).

Sam Cumming

Sam Cumming works on diverse topics in the philosophy of language and semantics, from attitude reports, noun phrases and the foundations of meaning to the semantics of edits in film. Along with Josh Armstrong and Gabe Greenberg, he runs the interdisciplinary workshop SLIME. His PhD was from Rutgers, and he is professor of philosophy at the University of California, Los Angeles.

François Recanati

Philosophie du langage et de l'esprit

Collège de France

Année 2024-2025

Colloque - Discourse Files

Variables Keep Them(selves) separate

Gabor Brody

Yale University

Résumé

Discourse referents are "mental files" insofar as they are mental representations that function like variables: they architecturally separate addresses from contents. I argue for a construal of discourse referents in which they are not mental files in other respects. First, discourse referents are not addressed by sortal, kind, or noun-type descriptive information; consequently, they cannot account for developmental phenomena such as Mutual Exclusivity. Second, unlike mental files, discourse referents are components of the grammatical system rather than the general belief system. Third, unlike mental files, discourse referents cannot be merged or linked via identity statements.

Gabor Brody

Gabor Brody is a Research Scientist and Lecturer at Yale University, working at the the Department of Psychology, affiliated with the Department of Linguistics and the Program in Cognitive Science. His work explores infants', children's, and adults' ability to keep track of and communicate about objects, concepts, and agents.

François Recanati

Philosophie du langage et de l'esprit

Collège de France

Année 2024-2025

Colloque - Discourse Files

Discourse Referents as Mental Files

Karen Lewis

Barnard College, Columbia University

Résumé

In previous work, I argue that discourse referents—devices for tracking objects under discussion in discourse—are best understood as mental files: vehicles for tracking the same object in cognition independently of any particular descriptive content. Mental files are private by nature; I treat discourse referents in the public, shared conversational context as intersubjective, coordinated, temporary files. On the proposed view, discourse referents are ways of cognizing content rather than contents themselves. Discourse referent mental files are (among other things) characterized by their role in non-satisfactional or pseudo-singular way of thinking. However, unlike the mental files of singular thought, discourse referents are not typically satisfied by a single, specific object (though they can be). Accordingly, I treat the contents of indefinite sentences and sentences containing expressions anaphoric on indefinites as having existential content. In this talk, I address the question of how to make sense of treating existential, descriptive content as being cognized in a pseudo-singular fashion. I argue that while sentences containing indefinite descriptions typically trigger the creation of a discourse referent, they are not themselves cognized in a pseudo-singular way. By contrast, sentences containing anaphoric expressions (definites and pronouns) contain pseudo-singular restrictors that bridge the gap between the pseudo-singular form and the descriptive content.

Karen Lewis

Karen Lewis is an associate professor in the Department of Philosophy at Barnard College, Columbia University. She works primarily in the philosophy of language, on topics such as context-sensitivity, discourse referents, anaphora, dynamic pragmatics, counterfactual conditionals, and the pragmatics of non-paradigmatic conversations, among other things.

François Recanati

Philosophie du langage et de l'esprit

Collège de France

Année 2024-2025

Colloque - Discourse Files

Types of Entity Representations in MSDRT

Hans Kamp

University of Stuttgart

Résumé

MSDRT (Mental State Discourse Representation Theory) has developed a number of different Types of Entity Representations (its term for what is referred to by others as Mental Files). The different ER Types are for different kinds of entities, e.g. fictional entities as opposed to entities in the real world, or for events, states and actions of various kinds as distinct from physical objects. The talk will be primarily concerned with the different forms of these different Types of ERs and the motivations for them.

Hans Kamp

Hans Kamp was born, and got his education up to and including the MA, in the Netherlands. After a PhD in Philosophy at UCLA on formal properties of Priorean Tense Logic, his work turned increasingly to model-theoretic semantics of natural language, following his Ph.D. supervisor Richard Montague. In the late seventies and early eighties Kamp developed Discourse Representation Theory (DRT), which, together with Heim's File Change Semantics, was the first formally worked out approach to the semantics of discourse. A primary current interest is the development of Mental State DRT, an extension of DRT in which files play a central part.

François Recanati

Philosophie du langage et de l'esprit

Collège de France

Année 2024-2025

Colloque - Discourse Files

Discourse Files and Human Memory Systems

Ellen Lau

University of Maryland

Résumé

Two ideas characterize many discussions of discourse files or discourse referents. One is that they are only temporary, serving as mental proxies for individuals only across some segment of discourse. Dissociations observed in many forms of amnesia are consistent with the idea that distinct circuits support indexing of individual representations within the current context vs. long-term. Neighboring regions of the inferior parietal cortex show fMRI responses during both visual working memory tasks and language comprehension tasks, and might support a common structured representation of the current scene or situation, to which individual object representations are temporarily indexed. Solving the different problems of long-term indexing requires different regions, including the hippocampus. The second idea is that discourse referents are in some way not actually referential, and that this explains certain linguistic patterns observed with common nouns, as well as their use in discussing hypothetical or fictional objects. I will dispute this second idea. Many classic puzzles of common noun reference disappear if nouns are taken to function as nonsingular names, rather than predicates (Geach, 1962). And the theory of instance-of-kind conceptual structure (Prasada 2016, in press) suggests that an instance-of-kind concept affords reference to the (potentially infinite) instances of a kind, independent of physical acquaintance with them, or whether they are hypothetical vs. actual. With such a non-predicative theory of noun meaning, we can therefore adopt a view where the way in which individual representations are indexed differs from the current context to long-term knowledge, but where the logic of reference is the same throughout.

Ellen Lau

Ellen Lau is an Associate Professor in Linguistics and Neuroscience and Cognitive Science at the University of Maryland, where she has been on the faculty since 2011. She co-directs the KIT-UMD Magnetoencephalography lab, and her research uses behavioral and neurophysiological measures to investigate language processes and representations. In recent years her work has explored the architecture of lexical knowledge, and mental representations of individuals, in psychological and neural models of language.

François Recanati

Philosophie du langage et de l'esprit

Collège de France

Année 2024-2025

Colloque - Discourse Files

François Recanati : Introduction

François Recanati

Philosophie du langage et de l'esprit

Collège de France

Année 2025-2026

Conférence - Gergely Csibra : Non-Verbal Generics

Gergely Csibra

Central European University, Vienne

Résumé :

Humans learn most of their knowledge from each other, and much of this learning relies on communication. While language is the primary medium of human communication, it is not the only one, and, importantly, it is of limited use for infants and young children. We have been investigating how human communication is adapted to the need of transmitting knowledge to others, especially across generations, from adults to young children. In linguistic communication, special types of assertions, namely generic expressions, can directly convey knowledge about classes of entities. We propose that non-verbal communication, especially communication via and about objects, can also express generic knowledge. In a series of studies, we have found that young children quickly learn properties of objects demonstrated to them in an ostensive-communicative manner. We have also observed that knowledge acquired this way is more rigid than knowledge accumulated from individual experience. Non-verbal expressions can transfer knowledge and elicit the same kind of confirmation bias that is characteristic of knowledge acquired through linguistic communication.