Technical Program

Paper Detail

Paper: PS-1B.77
Session: Poster Session 1B
Location: H Fläche 1.OG
Session Time: Saturday, September 14, 16:30 - 19:30
Presentation Time:Saturday, September 14, 16:30 - 19:30
Presentation: Poster
Publication: 2019 Conference on Cognitive Computational Neuroscience, 13-16 September 2019, Berlin, Germany
Paper Title: Cognition as inference: a unifying account of some neural effects associated with mental imagery and attention
Manuscript:  Click here to view manuscript
License: Creative Commons License
This work is licensed under a Creative Commons Attribution 3.0 Unported License.
DOI: https://doi.org/10.32470/CCN.2019.1388-0
Authors: Ghislain St-Yves, Thomas Naselaris, Medical university of South Carolina, United States
Abstract: Mental imagery and attention are difficult to disentangle, suggesting a shared computational mechanism. We propose that imagery and attention are both inferences about the visual world conditioned on retinal input and a high-level anticipatory representation. Neural effects that have been previously associated with imagery occur when the anticipatory representation is determined by a memory and the retinal input is uninformative. Neural effects associated with spatial attention arise when the anticipatory representation is biased toward a location that is experimentally manipulated into or out of alignment with the presented stimuli. We tested the feasibility of this proposal with \emph{in silico} experiments in a deep generative model that roughly analogizes the hierarchy of functionally distinct visual areas. We show that such a model jointly characterizes some of the ways that imagery and attention modulate activity levels, tuning to visual features, and the location and size of receptive fields. Based on these results, we consider the possibility that top-down volitional spatial attention is essentially equivalent to imagining the stimuli at the attended location, and imagining a stimuli involves reinstating the activity evoked by that stimulus near the top of a representational hierarchy.