Incidental learning of predictive temporal context within cortical representations of visual shape.

Abstract

Incidental learning of spatio-temporal regularities and consistencies - also termed statistical learning - may be important for discovering the causal principles gov-erning the world. We studied statistical learning of temporal structure at two temporal scales: the presentation of rotating visual objects (3 s) and predictive temporal context (30 s) in a sequence of such objects. Observers viewed fifteen initially unfamiliar synthetic objects recurring many times each, intermixed with other similar objects that appeared only once, while whole-brain BOLD activity was recorded. Analyzing multivariate BOLD activity, we located view-independent responses to synthetic objects throughout the ventral occipitotemporal cortex, revealing a view-invariant representation of object shape. Predictive temporal context modified representations of object shape at early and middle levels of the visual processing hierarchy. Approximately 20% of the sites representing shape also exhibited a significant representation of temporal context. A few higher-level sites represented temporal context, without also representing object shape. Additionally, objects marking transitions between communities were distinctly represented at many sites. We conclude that, at the sensitivity afforded by fMRI, the cortical represen-tations of invariant visual shape and of predictive temporal context are largely coextensive. Our results suggest that incidental learning of causal principles and structures operating at different timescales ("structural learning") may involve vi-sually responsive areas in the ventral occipitotemporal cortex.