You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

 
Poster B63, Wednesday, November 8, 3:00 – 4:15 pm, Harborview and Loch Raven Ballrooms

Uncovering the cascade of computations involved in ambiguity resolution using MEG decoding

Laura Gwilliams1,2, Jean-Rémi King1;1New York University, 2NYUAD Institute

Ambiguity is ubiquitous in language — common to phonetic features, part of speech, lexical meaning and syntactic constructions. The Hierarchical Bayesian Inference (HBI) framework provides a formal description of the computations required to resolve such ambiguity. The aim of the present study is to test whether the human brain implements an architecture like this framework in deriving a stable interpretation of ambiguous input. Seventeen participants discriminated between 1960 ambiguous visual symbols, constructed from 8-step morphs of letter/digit pairs, while magneto-encaphalography (MEG) was recorded concurrently. Six putative computations derived from the HBI framework were modelled: stimulus location, number of visual edges, stimulus identity, stimulus contrast, stimulus ambiguity and response button. Linear multivariate pattern analyses (MVPA) were used to decode, track and source-localise the dynamics of each of these computations. Our analyses uncover three main findings. First, sensory features associated with each computation are sequentially decodable from the MEG signal, and continue to be maintained in parallel thereafter. Second, neural activity correlates with objective sensory information up to ~300 ms; after, neural responses become progressively categorical and correlated with subjective reports. Third, these computations are generated by a cortical hierarchy spanning across the visual, parietal and motor cortices respectively. Together, our findings reveal that a cascade of neural computations unfolds within and across cortical regions to resolve perceptual ambiguity. This study provides a stepping stone to link computational accounts of inference to the neural bases of perceptual decision making at the meso- and macroscopic level, and provide insight into how sensory information is extracted, maintained and integrated over time.

Topic Area: Perception: Orthographic and Other Visual Processes

Back to Poster Schedule