Perception-action circuits for word learning and semantic grounding: a neurocomputational model and neuroimaging study

A neurocomputational architecture of the left-hemispheric areas of the brain is presented which was used to simulate and explain neural correlates of word learning and semantic grounding. The model’s main distinguishing features are that (i) it replicates connectivity and anatomical structure of the relevant brain areas, and (ii) it implements only functional mechanisms reflecting known cellular- and synaptic-level properties of the cerebral cortex. Stimulation of the “sensorimotor” model areas (mimicking early stages of word acquisition) leads to the spontaneous formation of cell assemblies (CAs), network correlates of memory traces for meaningful words. Preliminary results of a recent functional Magnetic Resonance Imaging study confirm the model's predictions, and, for the first time, localise the neural correlates of semantic grounding of novel spoken items in primary visual cortex. Taken together, these results provide strong support for perceptual accounts of word meaning acquisition in the brain, and point to a unifying theory of cognition based on action-perception circuits whose emergence, dynamics and interactions are grounded in known neuroanatomy and neurobiological learning mechanisms.