The nature of habits in the nonhuman primate : the formation of sequences of eye movements and neural activity in the frontal eye field

Abstract

The nature of habits, their formation, expression, and underlying causes have been pondered for centuries. Early definitions, still in use today, posited that habits are actions associated with outcomes that, when repeated, become stimulus-response associations that can be performed automatically, or without the reinforcement of a rewarding stimulus. A prominent theory of what drives the process is reinforcement learning (RL). This definition and underlying theory may be inadequate to describe the complicated series of actions that we form and express as habits in every day life. We designed a task that would test the limits of RL by providing a nearly infinite number of action choices and no clear association with reward. We recorded using ~100 chronically implanted independently moveable electrodes from the frontal eye fields (FEF), prefrontal cortex (PFC), and caudate nucleus (CN) simultaneously as naïve monkeys performed a free-viewing scan task. Neural recordings began on the first day of this task where a random dot on a grid of targets was chosen to be baited with reward on every trial and the monkeys were free to look around until they captured the baited target. We found that monkeys formed selfguided and uninstructed sequences of eye movements that gradually evolved over months of task performance and did not appear to be driven by overall reward or cost measures. Only on a much smaller, trial-by-trial, time scale were we able to detect the RL forces at work and that the monkeys were minimizing cost on an extremely local level. We also found that neural units in the FEF showed standard single direction and non-standard multiple direction tuning very early in task acquisition. We also found a disproportionately high number of units whose tuning directions were selective for those eye movement combinations that were members of the monkeys' habitual sequences. This suggested that the FEF very rapidly adapts to the task at hand and the neural representation becomes biased towards those sequences that are repeated. Together these findings lay the foundation to understand natural habit formation and the neural mechanisms that underlie it.