The neural mechanisms underlying decision-making are typically inferred from the average activity from sequentially recorded single neurons, which obscures important aspects of decision-making dynamics. I will show that covert decision variables (DV) can be tracked dynamically on single behavioral trials via simultaneous recording of large neural populations in primate motor cortex. I will also show—in nonhuman primates under the conditions of our experiments—that decisions are encoded by relatively stationary populations of neurons, not by sequences of activity passed from neuron-to-neuron.