Pragmatic Inference Without Words? Evaluating RSA in Great Ape Communication

09:00 - 10:30

Part I: Linguistic and Non-Linguistic Cognition - Comparative and Developmental Perspectives

Room: HSZ - N3

Chair/s:

Barbara Kaup, David Dignath

This symposium examines the interplay between linguistic and non-linguistic cognition. While some cognitive functions appear to depend on language, others seem rather independent of it and many more integrate both aspects. In psychology, however, the distinction between linguistic and non-linguistic cognition is rarely made explicitly and there is currently no consensus on how language may shape, enable or constrain thought.

The symposium brings together perspectives from cognitive research, developmental psychology and animal cognition to address three questions:

(1) How are language and thought related?
(2) Which cognitive functions are inherently linguistic, and which are not?
(3) To what extend can language modulate domains traditionally considered non-linguistic?

Part 1 brings together comparative and ontogenetic perspectives, focusing on animal cognition and human development. Lena Veit will speak about vocal communication in birds. Marlen Fröhlich's contribution deals with pragmatic inference abilities in orangutans. Paul Gallenkemper studies expectation formation in conceptual and non-conceptual contexts. Krisztina Orban looks at pointing in human and non-human primates as well as from a development perspective, concluding that pointing is a proto-linguistic behavior bridging the gap between non-linguistic behavior and fully developed language. Claudio Tennie discusses the hypothesis that human culture requires language and language in turn requires know-how copying abilities, that are nearly or completely absent in non-human apes.

Part 2 adopts a cognitive psychology perspective (see detailed description there).

Submission 283

Pragmatic Inference Without Words? Evaluating RSA in Great Ape Communication

SymposiumTalk-02

Presented by: Marlen Fröhlich

Marlen Fröhlich ¹, Manuel Bohn ², Michael Franke ¹

¹ University of Tübingen, Germany

² Leuphana University Lüneburg, Germany

Studies of meaning in human and primate communication face, in principle, similar methodological challenges. In both cases, meaning is not directly observable but must be inferred from indirect sources, such as observable behaviour. Recent work building on the Rational Speech Act (RSA) framework has developed methods for inferring latent semantic meaning through probabilistic models of language use. Here, we explore how such an approach can be adapted for the study of primate communication. To this end, we developed a bespoke probabilistic model of the processes that generate communicative behaviour, using functionally specified latent meaning representations. As a proof of concept, we applied this novel Lexical Association Model (LAM) to a richly annotated dataset of dyadic communicative interactions (>8,000 communicative acts) observed in wild and captive orangutans (Pongo abelii and P. pygmaeus). Specifically, we inferred the latent meaning of gestures by integrating observations of the signal (the gesture itself), the state (the presumed social goal), and the recipient’s response (the outcome). Our preliminary results indicate that latent signal meanings vary with social relationship (e.g., mother–offspring vs. others) and research setting (wild vs. captive). A critical next step will be to extend this model to multimodal communication, including vocalisations and facial expressions. We conclude that explicit probabilistic modelling can provide valuable insights into animal communication, particularly regarding the context-dependent nature of signals and the gradual evolution of human communication systems.

Bookmark