A spiking network model of prefrontal cortex activity during strategic behavior

The ability to adjust behavior according to context is essential to survival. In turn, this may require the ability to respond to identical stimuli with different behaviors. In the task analyzed here (Genovesio et al, Neuron 47:307–320, 2005), macaque monkeys must select a response based on an instruction stimulus and previous behavior. Depending on past trial type and behavior, the same instruction stimulus may require different responses. Specifically, if the instruction stimulus is the same as the previous trial, the same response is required (repeat-stay strategy); otherwise, a new response is required among the remaining two possible answers (change-shift strategy). If the latter response is not the required one, a second-chance change-shift trial is offered, and the monkey must select the remaining available response. We have preliminary evidence that prefrontal neural activity is metastable in this task, raising the question of how metastable neural dynamics are linked to the ability to produce strategic behavior. Here, we analyze under what conditions a metastable network of spiking neurons can exhibit strategic behavior. We found that a hierarchical metastable network can solve tasks with first-order contextual dependency (e.g., in change-shift trials), while additional mechanisms are required for higher-order dependency (e.g., in second-chance shift trials).