Distributed Neural Systems Support Flexible Attention Updating during Category Learning

摘要

To accurately categorize items, humans learn to selectivelyattend to the stimulus dimensions that are most relevant tothe task. Models of category learning describe how attentionchanges across trials as labeled stimuli are progressivelyobserved. The Adaptive Attention Representation Model(AARM), for example, provides an account in which categorizationdecisions are based on the perceptual similarity of a newstimulus to stored exemplars, and dimension-wise attentionis updated on every trial in the direction of a feedback-basederror gradient. As such, attention modulation as described byAARM requires interactions among processes of orienting,visual perception, memory retrieval, prediction error, and goalmaintenance to facilitate learning. The current study exploredthe neural bases of attention mechanisms using quantitativepredictions from AARM to analyze behavioral and fMRIdata collected while participants learned novel categories.Generalized linear model analyses revealed patterns of BOLDactivation in the parietal cortex (orienting), visual cortex(perception), medial temporal lobe (memory retrieval), basalganglia (prediction error), and pFC (goal maintenance) thatcovaried with the magnitude of model-predicted attentionaltuning. Results are consistent with AARM’s specification ofattention modulation as a dynamic property of distributed cognitivesystems.