UCL Logo



Information flow through local cortical networks is not democratic


John Beggs



Information flow through local cortical networks is not democratic



The average pyramidal neuron in cortex makes and receives ~7,000 synaptic contacts, suggesting that local cortical networks are connected in a fairly equal manner. The pattern of information flow in such networks, however, is poorly understood and can not be inferred from anatomy alone.

Interestingly, theory indicates that an unequal distribution of flows can actually contribute to network efficiency and robustness. Accordingly, we sought to examine the distribution of information flow in recordings from cortical slice cultures and monkey motor cortex (n = 1) containing 100 +/-

25 (mean +/- s.d.) identified neurons. We used transfer entropy to quantify information flow, as validation tests revealed that this measure could reliably distinguish true from spurious flows in a variety of realistic conditions. Information flow was distributed significantly more unevenly in the networks extracted from the data than in random control networks. This was evident in the distribution of information flow strengths, the distribution of total information flow into and out of each neuron, and in the distribution of connections with significant information flow per neuron. Simulations indicated the observed cortical information flow networks were significantly more efficient in routing signals, could form significantly more combinations among inputs per node, and were significantly more robust than random control networks. To our knowledge, this was the first study of information flow in local cortical networks. We conclude that the highly unequal distribution of information flow among cortical neurons contributes to the efficiency and robustness of information processing in cortex.