Authors:

GARAGNANI, M. & PULVERMULLER, F.

Reference:

Year of publication:

2011

CBU number:

7271

Abstract:

The neurobiological basis of working memory has been elucidated by the discovery of “memory cells”, neurons that exhibit several seconds of persistent activity when animals have to keep in mind stimulus information in view of future action. These memory cells, and the corresponding neurometabolic changes in the human brain, are typically found in prefrontal cortex and higher sensory areas, distant from primary cortices. However, assuming that the formation of distributed memory circuits is the result of correlated activity in motor and perceptual systems, why should these circuits emerge far away from their antecedent activations in sensorimotor areas, leading to what could be called “disembodiment” of memory? We use a six-area neuronal-network architecture (modelling primary motor and sensory cortex, secondary and higher association areas in frontal and temporal lobes) to simulate formation of memory circuits and memory cells. The long-range, between-area connections implemented reflect known neuroanatomical links between corresponding cortices. We report results in terms of recorded network activity at single-cell, circuit-, and area-specific level. Similarly to previous models, we observed the spontaneous emergence of memory circuits (cell assemblies, CAs) and cells replicating various realistic behaviours. Due to the intrinsic between-areas connectivity of the model, however, the formation of strong synaptic links holding the CA together was facilitated in higher, associative (but not in primary) areas, leading to more memory cells in prefrontal and anterior-temporal than in primary cortices. As these findings answer the “where” question of cortical working memory in terms of neuroanatomy, they further strengthen a distributed-network memory account.