Distributed cell assemblies for general lexical and category-specific semantic processing as revealed by fMRI cluster analysis.

Here, we ask whether frontotemporal cortex is functionally dissociated into distributed lexical and category-specific semantic networks. To this end, fMRI activation patterns elicited during the processing of words from different semantic categories were categorized using k-means cluster algorithms. Results showed a distributed pattern of inferiorfrontal, superiortemporal, and fusiform activation shared by different word categories. This shared activation contrasted with patterns of category-specific semantic activation in widely distributed neural systems. Clustering revealed congruent functional specificity of focal area activations in frontal and temporal cortex; thus suggesting a correspondence between functional partitionings of frontocentral mirror neuron systems and those of inferiortemporal lexical and semantic circuits. Action words related to the face, arms, and legs specifically activated the motor system in a somatotopic manner, whereas form-related words activated prefrontal areas. Similar functional specificity was evident in temporal cortex, where a different semantic topography emerged for form- and action-related words. Results were replicated in a separate data set, therefore recommending fMRI cluster analysis as a reliable method for scrutinizing the brain basis of lexical, semantic, and conceptual systems in humans. As focal modules do not explain the distributed character of functionally specific clusters and their distinct topographies are at variance with general distributed processing accounts, the functionally-homogenous distributed clusters specific to semantic types are best explained by specifically-distributed cortical circuits which, similar to Hebbian cell assemblies, represent functional units with specific roles in cognitive processing, especially in lexical and semantic access and memory. 2009 Wiley-Liss, Inc.