Cortical and basal ganglia contributions to habit learning and automaticity

Abstract

The 20th century was dominated by the view that novel behaviors are mediated primarily in cortex and that the development of automaticity is a process of transferring control to subcortical structures. Much evidence suggests however, that subcortical structures, such as the striatum, make significant contributions to initial learning. More recently, evidence has been accumulating that neurons in the associative striatum are selectively activated during early learning, whereas those in the sensorimotor striatum are more active after automaticity has developed. At the same time, other recent reports suggest that automatic behaviors are striatum- and dopamine-independent, and may be mediated entirely within cortex. Resolving this apparent conflict should be a major goal of future research.

Habits as action sequences: hierarchical action control and changes in outcome value

“…Along similar lines, the hypothesis that insensitivity to devaluation is rooted in the formation of action sequences is consistent with a body of evidence demonstrating that a similar neural substrate mediates both habit development, as measured by outcome devaluation, and action sequence learning (for a review, see [10,15]). For example, lesion or inactivation of sensorimotor striatum restores sensitivity to outcome devaluation in over-trained animals [50,51].…”

“…These two aspects of automaticity share a similar neural structure (e.g. [10]), however, computationally, they have been attributed to two different models: insensitivity to changes in outcome value has often been interpreted as evidence for a model-free reinforcement learning (RL) account of instrumental conditioning [11,12], whereas the development of action sequences has been linked to hierarchical RL [13]. Here, building on previous work [13][14][15], we demonstrate that the insensitivity of specific actions to changes in outcome value can be a consequence of developing action sequences and, therefore, that both types of automaticity can be reconciled within hierarchical model-based RL.…”

Probabilistic inference under time pressure leads to a cortical-to-subcortical shift in decision evidence integration

“…This experimental design, however, raises a concern that changes in behavior and cortical activity across experimental phases we report could be influenced by task practice. At the behavioral level, practice effects are defined as an increase in accuracy and decrease in response time (Ashby et al, 2010;Kelly and Garavan, 2005;Schneider and Shiffrin, 1977). As reported in the previous section (see Task performance), both decision accuracy and RT remained constant within each the LowP and HighP phase during the scan (ps > 0:05).…”

“…Some studies, additionally, have reported enhanced post-learning subcortical processing (Doyon et al, 2009;Lehericy et al, 2005;Van Turennout et al, 2003), although the results are rather mixed (see Ashby et al, 2010 for a detailed discussion). To examine changes in cortical and subcortical activity over time, we divided our experiment into four different sets (120 trials/set, 2 sets/phase), each consisting of a complete set of unique stimulus combinations, and conducted an additional GLM analysis.…”

Deep Brain Stimulation for Movement Disorders of Basal Ganglia Origin: Restoring Function or Functionality?

“…In terms of procedural learning, the Bassociative^portions of the caudate nucleus appear to be involved in early phases of learning, while the Bmotor^puta-men is more prominently engaged when animals execute previously learned movement sequences [109][110][111][112][113][114][115][116][117][118]. Hypotheses about the role of dopamine in reinforcement learning are closely tied to the finding that dopamine neurons fire in relation to (positive) prediction errors in rewarded tasks [119][120][121][122][123][124][125].…”