Initial conditions combine with sensory evidence to induce decision-related dynamics in PMd

We employed a dynamical systems perspective to bridge decision-related neural activity and decision-making behavior, a fundamentally unresolved problem. The dynamical systems approach posits that neural dynamics (X) are parameterized by a state equation (dX/dt=F(X)+U) with varying initial conditions (x0) and that evolves in time by combining at each time step, recurrent activity (F) and inputs (U).
We considered different dynamical systems that make predictions about whether 1) initial conditions substantially predict subsequent dynamics and reaction time (RT) and/or choice, and 2) whether inputs are combined with initial conditions to lead to differences in choice-related dynamics. We evaluated which of the various dynamical mechanisms best described neural population activity by investigating neural population dynamics in the dorsal premotor cortex (PMd) of monkeys performing a red-green RT checkerboard discrimination task.
Prestimulus neural state, a proxy for the initial condition, predicted poststimulus neural trajectories and showed organized covariation with RT. Furthermore, faster RTs were associated with faster pre- and poststimulus dynamics as compared to slower RTs, with these effects observed within a stimulus difficulty. Poststimulus dynamics depended on both the sensory evidence and initial condition, with easier stimuli and “fast’’ initial conditions leading to the fastest choice-related dynamics whereas harder stimuli and “slow” initial conditions led to the slowest dynamics. Finally, changes in initial condition were related to the outcome of the previous trial, with slower pre- and poststimulus population dynamics and RTs on trials following an error as compared to trials following a correct response.
Models with initial conditions that are random or biased towards one choice, or with sensory encoding delays are inconsistent with these findings. Instead, these results suggest that decision-related activity in PMd are most consistent with a dynamical system where inputs combine with outcome-sensitive initial conditions that covary with eventual RT to induce decision-related dynamics.