Skip to main content

The role of connectivity in shaping sensory processing and memory in network models of the cerebral cortex

Final Activity Report Summary - NETCONN (The role of connectivity in shaping sensory processing and memory in network models of the cerebral cortex)

The cerebral cortex processes sensory information in order to direct behaviour. This may entail diverse types of processing, such as extracting a task-salient feature from the stimulus in order to make a decision or making use of intrinsic states of the cortex itself, e.g. memories, in order to categorise a stimulus. In any case, the relevant cortical dynamics are strongly shaped by the patterns of synaptic connectivity found therein. The fellow used analytical and numerical tools to study the role of network connectivity both on sensory processing and on the generation of intrinsic cortical states in models of cortical dynamics.

In the context of simple decision making tasks, when a monkey or human must make a binary decision based on ambiguous sensory stimulus, the fellow showed that the symmetry in the mutually inhibitory connections between cells in lateral parietal cortex led naturally to a canonical description of the decision making process. The resulting model described decision making behaviour surprisingly well and related the relevant behavioural measures to the underlying physiological parameters.

In investigating the role of connectivity in generating spontaneously emergent cortical states, the fellow adopted a two-pronged approach. On the one hand, he showed, through numerical and analytical analysis of simplified rate and network models, that the interplay between patterns of synaptic connectivity and delays in the neuronal interactions was critical in determining the type of cortical activity seen in oscillations, waves etc. He fully characterised how this states might emerge from the non-patterned asynchronous state analytically. At the same time he analysed experimental data from cortical slices in which spontaneous cortical activity consisted of repeating patterns of neuronal activation, reminiscent of attractor states. He also fit a simple, stochastic model to this data and succeeded in reproducing the repeating patterns while at the same time extracting the effective connections between the neurons. Subsequent analysis of the extracted connectivities revealed highly non-trivial network motifs, consistent with what was characterised in intracellular studies in vitro.