Role of features and categories in the organization of object knowledge: Evidence from adaptation fMRI

Abstract

There are two general views regarding the organization of object knowledge. The feature-based view assumes that object knowledge is grounded in a widely distributed neural network in terms of sensory/function features (e.g., Warrington & Shallice, 1984), while the category-based view assumes in addition that object knowledge is organized by taxonomic and thematic categories (e.g., Schwartz et al., 2011). Using a functional magnetic resonance imaging (fMRI) adaptation paradigm, we compared predictions from the feature- and category-based views by examining the neural substrates recruited as subjects read word pairs that were identical, taxonomically related, thematically related or unrelated while controlling for the function features involved across the two categories. We improved upon previous study designs and employed an fMRI adaptation task, obtaining results overall consistent with both the category-based and feature-based views. Consistent with the category-based view, we observed for both hypothesized regions of interest (ROI) and exploratory (whole-brain analyses) reduced activity in the left anterior temporal lobe (ATL) for taxonomically related versus unrelated word pairs, and for the exploratory analysis only, reduced activity in the right ATL. In addition, the exploratory analyses revealed reduced activity in the left temporo-parietal junction (TPJ) for thematically related versus unrelated word pairs. Consistent with the feature-based view, we found in the exploratory analyses that activity reduced in the bilateral precentral gyri (i.e., function regions) including part of premotor cortex as the function relatedness ratings increased. However, we did not find a relationship between adaptation effects in the bilateral ATLs and left TPJ and corresponding ratings of taxonomic/thematic relationships suggesting that the adaptation effects may potentially not reflect aspects of taxonomy that have been traditionally assumed. Together, our findings indicate that both feature and category information are important for the organization of object knowledge although the exact nature of those organization principles is an important question for future research.