Occipitotemporal Representations Reflect Individual Differences in Conceptual Knowledge

Abstract

Through selective attention, decision-makers can learn to ignore behaviorally-irrelevant stimulus dimensions. This can improve learning and can increase the perceptual discriminability of relevant stimulus information. Across cognitive models of categorization, this is typically accomplished through the inclusion of attentional parameters, which provide information about the importance of each stimulus dimension during decision-making. These parameters are often described geometrically, such that perceptual differences over relevant psychological dimensions are accentuated (or stretched), and differences over irrelevant dimensions are down-weighted (or compressed). An open question is whether these attention parameters are reflected in the brain. We examined two published fMRI datasets in which participants learned to categorize stimuli with features that differed in their relevancy. Attentional parameters were derived for each participants using formal categorization models fit to behavioral data. Using multivoxel pattern analysis (MVPA), we found that when greater attentional weight was devoted to a particular visual feature (e.g., "color"), its value (e.g., "red") was more easily decoded from occipitotemporal cortex. This effect was sufficiently sensitive to reflect individual differences in conceptual knowledge, indicating that occipitotemporal stimulus representations are embedded within a space similar to that formalized by classic categorization theory.

Subjective value and decision entropy are jointly encoded by aligned gradients across the human brain

“…Furthermore, fits to individuals' behaviour yielded measures of risk aversion that reflect individual differences in brain response (see SI). In effect, the cognitive model is demonstrating a reality at both the behavioral and neural level for individual participants, which mirrors recent findings in the concept learning literature on attentional shifts (Braunlich & Love, 2018;Mack et al, 2020). Our results support the claim that cognitive models can reveal intricate facets of behaviour and brain response.…”

Ventromedial prefrontal cortex compression during concept learning

“…To evaluate whether vmPFC compression tracked changes in attentional allocation, we characterized the participant-specific attentional weights given to each stimulus feature across the three problems using a computational learning model 8 . Attention weight compression indexed changes in attentional allocation based on model fits to behavior; low attention compression indicates equivalent weighting to all three features, whereas high attention compression indicates attention directed to only one feature 18 . We found that attention compression varied with the interaction of learning block and conceptual complexity (Bayesian-estimated mixed effects linear regression: β mean = −0.028, 95% HDI = [−0.035, −0.020], P < 0.001).…”

Measures of Neural Similarity

“…First, across regions, the signal and type of information represented can differ (Ahlheim and Love 2018;Bracci and de Beeck 2016;Diedrichsen et al 2011), which might lead the accompanying similarity operations to also differ. Second, task differences, such as those that shift attention (Braunlich and Love 2018;Mack et al 2013;Mack et al 2016), lead to changes in the brain's similarity space which may reflect basic changes in the underlying similarity computation. Our interest is in describing similarity computations that could, in principle, be used for behavioral output, focusing on a necessary but not sufficient condition for producing behavior from neural representations.…”