The role of the frontoparietal attention network during rule-based categorization

Category learning enables humans to predict the relevance of new visual information. We previously found that such learning is supported by localized perceptual enhancement of relevant sensory signals within early visual cortex, particularly for feature values near the perceptual boundary that separates one category from another (O’Bryan et al., 2024). We next hypothesized that the sensory effects are likely governed by top-down feature-based attention, generally localized to the frontoparietal attentional control network. The current neuroimaging study therefore tested the role of topographically organized intraparietal brain areas while two groups of male and female human participants learned to categorize a single set of gratings based on either an orientation (N=10) or spatial frequency rule (N = 11). First, a generalized psychophysiological interaction (gPPI) analysis revealed significant functional connectivity between three subregions of the intraparietal sulcus (IPS0/1, IPS2/3, and IPS4/5) and visual cortex for both participant groups during category learning compared to an orthogonal contrast discrimination task. Next, we employed an inverted encoding model to reveal enhanced reconstructed representations of stimulus orientation within IPS during category learning compared to discrimination. Within posterior (IPS0/1) and middle (IPS2/3) subregions, this enhancement was observed across all tested orientation values only when orientation was task relevant; within anterior IPS (IPS4/5), orientation representations were equally enhanced regardless of rule. These results provide converging evidence of the critical role top-down attention plays in category learning, while also highlighting the ways in which stimulus representations gradually change across the visual system in response to task demands.