Uncertainty differentially shapes premotor and primary motor activity during movement planning

Precise movement execution often relies on adequately planning the upcoming action, a process that is largely mediated by the dorsal premotor (PMd) and primary motor (M1) cortices. Due to their anatomical and functional similarities, PMd and M1 are commonly assumed to contain similar planning- related information. Here we show that when planning movements under uncertainty, the activity of PMd and M1 neural populations show distinct features, suggestive of area-specific computations.
We analyzed simultaneous recordings from populations of PMd and M1 neurons from monkeys performing both a standard center-out reaching task, and a task requiring them to plan reaches based on uncertain visual information about the target's location (Dekleva et al. eLife 2016). We investigated how uncertainty shaped planning activity by computing “latent dynamics” during the instructed delay epoch for each population.
In the uncertainty task, both PMd and M1 latent dynamics contained information about reach direction, as well as the associated uncertainty level. However, the relative importance of these signals varied across areas: M1 activity was mainly organized by reach direction, while in PMd uncertainty was dominant. For both regions, the axes needed to linearly decode uncertainty and reach direction were orthogonal to each other, supporting the view that orthogonal subspaces may enable separate computations. Another intriguing difference between PMd and M1 was that their activity contained qualitatively different information about the intended movement. In agreement with their relative position along the neuraxis, M1 might integrate PMd activity to formulate a final motor plan: decoding performance using PMd activity integrated through time resembled that of instantaneous decoding from M1 latent dynamics, even after subsampling neurons to match firing rate statistics between areas. Thus, despite their apparent similarities, when target information is not clearly specified, PMd and M1 may become primarily involved in different planning-related computations.