Neural correlation dynamics during a cognitive task are central to study brain information encoding, transfer and computation. In particular, we studied the problem of information transfer during task performance by analyzing spike trains that were simultaneously recorded in sensory (S1, S2), premotor (MPC, DPC), and motor (M1) cortical areas of two monkeys during a somatosensory discrimination task. Upon modeling spike-trains as binary time series we used a non-parametric Bayesian method to estimate pairwise directional correlations between many pairs of neurons at different time delays and throughout different stages of the task, namely, perception, working memory, decision making and motor report. We found that solving this decision-making task involves modulated feed-forward and feedback correlation paths linking sensory and motor areas that are particularly active among single encoding neurons at certain areas and task intervals. Crucially, when sensory comparison is no longer requested for task performance (e.g., during control tasks), a great proportion of directional correlations modulated by task variables consistently vanish across all cortical areas and task periods. These modulated correlations are typically manifested by the existence (ON) of a neural correlation in the presence of a given stimulus/response feature and the absence (OFF) for the rest, which indicates that task variable values are internally mapped onto disjoint neural interactions. Further, modulated correlations are significantly faster when they distribute information across somatosensory areas (S1 and S2) during working memory periods than when they link S2, premotor and motor areas during decision-making.

[1] Tauste Campo et al. PNAS. 112(15):4761-66 (2015).