Taking the speaker’s perspective? Investigating quantifier and implicature processing in an AI-language training paradigm using EEG-VR
Abstract
Interpreting speaker meaning requires taking into account the speaker’s epistemic perspective, i.e., their access to relevant information. The extent to which listeners integrate this perspective during comprehension is known as the perspective-taking problem. We present the first EEG study to investigate the role of perspective-taking in the processing of quantifiers and scalar implicatures. Using immersive virtual reality, we asked participants to assist an AI, represented by a virtual agent, in learning to use quantifiers appropriately. Critically, the agent had limited access to the reference domain compared to the participant.
Our results show that when the speaker’s information access is restricted, listeners initially process the semantic content of quantified statements from an egocentric perspective. Effects of the speaker’s perspective emerge only later, reflected in a sustained anterior negativity. For pragmatically ambiguous sentences containing ``some", we observe neither an effect of speaker perspective nor evidence for scalar implicature processing. We argue that contexts requiring consideration of the speaker’s knowledge primarily support Gricean, weak implicatures, which are computed relatively late as post-propositional inferences.
Related
- Investigating lexical knowledge and lexical processing in order to explore their role in second language utterance fluency
- SNAzzy: an image processing pipeline for investigating global Synchronous Network Activity
- Fluently Grooving: Investigating neural and behavioural indices of processing fluency in complex auditory rhythms
- Investigating the mediating effects of working memory and processing speed on age-related differences in visual path integration
- Connectionist models of human language processing : Progress and prospects
BibTeX
@misc{spychalska2026taking,
title = {{Taking the speaker’s perspective? Investigating quantifier and implicature processing in an AI-language training paradigm using EEG-VR}},
author = {Maria Spychalska and Ludmila Reimer and Marjolijn Dijkhuis and Peter Hagoort},
howpublished = {PsyArXiv},
year = {2026},
month = sep,
doi = {10.31234/osf.io/u4hcy_v2},
url = {https://osf.io/preprints/psyarxiv/u4hcy_v2/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Taking the speaker’s perspective? Investigating quantifier and implicature processing in an AI-language training paradigm using EEG-VR
AU - Maria Spychalska
AU - Ludmila Reimer
AU - Marjolijn Dijkhuis
AU - Peter Hagoort
PY - 2026
DA - 2026/09/16
PB - PsyArXiv
DO - 10.31234/osf.io/u4hcy_v2
UR - https://osf.io/preprints/psyarxiv/u4hcy_v2/
AB - Interpreting speaker meaning requires taking into account the speaker’s epistemic perspective, i.e., their access to relevant information. The extent to which listeners integrate this perspective during comprehension is known as the perspective-taking problem. We present the first EEG study to investigate the role of perspective-taking in the processing of quantifiers and scalar implicatures. Using immersive virtual reality, we asked participants to assist an AI, represented by a virtual agent, in learning to use quantifiers appropriately. Critically, the agent had limited access to the reference domain compared to the participant. Our results show that when the speaker’s information access is restricted, listeners initially process the semantic content of quantified statements from an egocentric perspective. Effects of the speaker’s perspective emerge only later, reflected in a sustained anterior negativity. For pragmatically ambiguous sentences containing ``some", we observe neither an effect of speaker perspective nor evidence for scalar implicature processing. We argue that contexts requiring consideration of the speaker’s knowledge primarily support Gricean, weak implicatures, which are computed relatively late as post-propositional inferences.
ER -