Preplay of Successor States in V1

Keywords

Loading...
Thumbnail Image

Issue Date

2021-07-31

Language

en

Document type

Journal Title

Journal ISSN

Volume Title

Publisher

Title

ISSN

Volume

Issue

Startpage

Endpage

DOI

Abstract

Learning and anticipating sequential visual events is essential in daily life. Resulting visual expectations have been shown to strongly in uence perception. However, the underlying neurocomputational representation of such sensory, relational knowledge is less well understood. Drawing on insights from predictive representations in reinforcement learning, we propose that the successor representation (SR) might be a suitable model to explain sequential anticipatory activity in the visual cortex, called preplay. To test this hypothesis, we derived predictions from the SR theory and investigated whether preplay activity patterns in V1 are in line with those predictions. Indeed, using functional magnetic resonance imaging (fMRI), we found that after learning a non-linear sequence of four successively presented dots, presenting any of the dots individually elicits anticipatory activity solely at the retinotopic locations of the future dots but not at those preceding the shown dot. At the same time, we did not nd any evidence for preplay in V2 and argue that low-level explanations are unlikely to cause the absence of preplay. Our results indicate that preplay in V1 is consistent with the hypotheses derived from the SR and that there might be a di erential involvement of V1 and V2 in predictively representing a visual sequences.

Description

Citation

Faculty

Faculteit der Sociale Wetenschappen