Tim Kunze | How models of canonical microcircuits implement cognitive functions

Major cognitive functions are thought to rely on distributed networks of a large number of fundamental neural elements, called canonical microcircuits. Mechanistic insight into the interaction of these canonical microcircuits promises a better comprehension of cognitive functions, their potential disorders, and corresponding treatment techniques. In this talk I will present a computational modeling framework that rests on canonical microcircuits and serves the investigation of conceivable mechanisms for syntax parsing.

We presumed that higher neural operations emerge from the combination of basic information processing operations, namely signal flow gating and working memory. A canonical microcircuit model, represented by a biologically plausible neural mass model (Deco et al., 2008, PLoS CB), exhibited these basic information processing operations by assessing an input’s salience in terms of intensity and temporal transiency. Time simulations and bifurcation analysis showed that the local network balance, in regard of synaptic gains, is one means to constrain this functionality.

At the level of hierarchically interacting canonical microcircuits facilitative feedback information modified the retention of sensory feedforward information. Consequently, meta-circuits of two interacting canonical microcircuits enabled state-dependent (adaptive) processing operations, such as priming and structure-building (i.e., the sequential, but selective, activation of neural circuits). Latter was identified as essential mechanism in a neural network for syntax parsing, proving the modeling framework's potential in neurocognitive research.

Our results showed that higher processing operations emerge from the recombination of minimal processing elements and suggested canonical microcircuits to be nodes of associative network models.

Poster