Abstract
Similar to objects, actions can be described at different hierarchical levels, ranging from very broad (e.g., locomotion) to very specific (e.g., breaststroke) information. Here we aimed to determine distinct representations of observed actions at three different levels of abstraction (superordinate, basic, and subordinate) in the human brain. To address this question, we conducted an fMRI study (3T; voxel resolution 2.5*2.5*2.5, TR= 2s, multiband sequence, acceleration factor 3) in which we presented N = 23 participants with static images of twelve different actions (six exemplars each) that were divided into three superordinate, six basic and twelve subordinate action categories. Participants were instructed to view the images, and to perform a category verification task during occasional catch trials, with an equal proportion of questions for each of the three taxonomic levels. Multivariate pattern analysis was carried out on t-values resulting from a general-linear model analysis, using a linear discriminate analysis (LDA) classifier and independent exemplar cross validation. To be able to compare results between the three taxonomic levels, decoding accuracy was normalized to account for the differences in chance level. A ROI-based analysis revealed that normalized decoding accuracy for the distinction between observed actions was higher at the subordinate in comparison to the superordinate level in V1, right superior parietal lobule (SPL) and right premotor cortex. By contrast, decoding accuracy in the right lateral occipitotemporal cortex (LOTC) and the left SPL was higher at the basic level than the superordinate level. Furthermore, the whole-brain searchlight analysis revealed peaks in the right inferior lateral occipital cortex (LOC), the left temporal occipital fusiform cortex and the right superior LOC for the subordinate, basic and superordinate level, respectively. Together, our results are in line with the view that observed actions can be decoded at all three taxonomic levels in high-level visual cortex.