The Natural History of a Habit

Theresa M. Desrochers

MIT


Everyone has habits. Despite our familiarity with them, little is knownabout how our habits are formed. A dominant theory of the rules governing habit formation comes from reinforcement learning (RL). We designed anexperiment and developed a recording technique to push the boundaries of our understanding of naturalistic habit formation and test the explanatory powerof RL while recording neural data from awake, behaving naïve macaquemonkeys.  Our task paradigm required the monkey to freely scan a grid of 4or 9 targets until capturing the target that had been pseudorandomly baitedwith reward. Thus, in this task reward was unpredictable, there were a large number of choices, and no explicit training was given. We found, under these uncertain conditions, that naïve monkeys nevertheless developed highly repeated, habitual scan paths that gradually evolved throughout task performance. Surprisingly, these scan paths continued to shift even aftersession-averaged reward and distance traveled by the eye (cost) parametershad reached asymptote. Consequently, these scan path shifts could have been driven by an intrinsic drive, e.g. towards efficiency, rather than the extrinsic optimization of reward. Alternatively, the driving factors were too subtle to be detected in the session averages and trial-by-trial analyses were necessary. We found evidence for both these alternatives, suggesting that along with directly observing the formation of habitual, uninstructed, repetitive behaviors, we have observed what could be a novel, driving force favoring the expression of such behaviors. Further, the chronic, simultaneous neural recordings we made from two neocortical area sand the striatum during the formation of these habits add the opportunity to illuminate the mechanisms underlying such behaviors.