| Recent experimental work has begun to characterize activity in local cortical networks containing hundreds to thousands of neurons. One finding is that neocortical circuits can produce cascades of electrical activity whose sizes follow a power law distribution. Another finding is that these cascades of activity form spatio-temporal patterns that re-occur significantly more often than expected by chance. A simple critical branching model can account for both of these findings from multielectrode recording data. Interestingly, this critical branching model also suggests that local cortical networks are poised to operate near a critical point where information processing would be optimal. In this thesis, we experimentally test three key predictions of the critical branching model.;The first prediction is that local cortical networks operate at a critical point. Although a power law may suggest that these circuits operate near a critical phase transition point, many other non-critical mechanisms can also produce power laws. To explore the origin of this distribution, we recorded cascade sizes and then perturbed activity. Deviations from a power law distribution varied systematically with a control parameter, as expected in a continuous phase transition. We also performed a data collapse analysis, showing that both avalanche size and length distributions exhibited scaling relationships. These results strongly suggest that neocortical circuits belong to the class of critical phenomena.;The second prediction is that changes in the connection weights within the network will alter the trajectories produced by spatio-temporal patterns of activity. Recently there has been an explosion of work on connectivity in networks of all types. It would seem natural then to explore the influence of connectivity on dynamics at the local network level. Dynamics can be measured as a trajectory in state space. We recorded significant changes in connectivity after a pharmaceutical agent was applied. As predicted, agents that changed the network weight structure also altered spatio-temporal trajectories in a manner that was consistent with the critical branching model. Importantly, trajectories at the critical point are dynamically neutral, allowing flexibility without introducing instabilities.;The third prediction is that a skewed distribution of connection weights will optimize the number of significantly repeating spatio-temporal patterns retained in the network. We found that when the model weight distribution was appropriately skewed, it correctly matched the distribution of repeating patterns observed in the data. In addition, the skewed distribution of weights maximized the capacity of the network model to retain stable activity patterns. We conclude that living cortical networks are very likely to use the skewed weight distributions predicted by theory to optimize information retention.;Taken together, these results suggest that the critical branching model can capture important features of the data. In addition, these results support the hypothesis that local cortical networks operate near a critical point where they may optimize information processing. |
