Analysis of fMRI and behavioural response data showing how neuroimaging provides a uniquely direct window into our thoughts (vs. methods in behavioural cognitive science).
Code snippets of clustering analyses (agglomerative k-means and hierarchical) revealing the semantic dimensions of our thoughts about natural concepts (animal concepts). fMRI provides a direct window into our thoughts and reveals that our thoughts of animals (and other concepts more generally) are less sophisticated than behavioural response data otherwise indicates. In the case of animal concepts, clustering analyses of fMRI data indicate that our thoughts of animals concern a small number of dimensions, e.g. an animal's habitat, ferocity, etc.
Brain reading and behavioral methods offer complementary perspectives on the representation of concepts
The advent of neuroimaging and brain-reading techniques has enabled new approaches to the study of knowledge representations, based on multivoxel analysis of the brain activation patterns evoked by contemplation of concepts such as animal concepts. The present fMRI study characterized the content and organization of 30 animal concepts. A factor analysis of the multivoxel activation patterns underlying the individual concepts indicated that the semantic building blocks of the brain’s representations of the animals were ferocity, intelligence, and body size. These findings can be compared to behavioral studies of knowledge representation, which have typically collected pairwise similarity ratings between two concepts. The main semantic components inferred from the fMRI data generally resembled the semantic components inferred from the behavioral data from a prominent previous study of the same animal concepts. But despite the similarity in semantic content, hierarchical clustering analysis of the two datasets revealed differences in the semantic organization observed between the two paradigms. The behavioral similarity judgments indicated that the animals were organized into taxonomically defined groups (e.g. canine, feline, equine), consistent with other behavioral studies. By contrast, the neural representations of the animals were organized to a greater extent by thematic relations that cut across taxonomic groups (e.g. animal personality, body size, habitat). The difference in the results might derive from differences in cognitive processing during judging similarities versus contemplating one animal at a time. The results highlight the unique perspective afforded by neuroimaging, and suggest that knowledge is fundamentally more thematically organized than previously thought.