This dissertation introduces a biologically-detailed computational model of how rule-guided behaviors become automatic. The model assumes that initially, rule-guided behaviors are controlled by a distributed neural network centered in the prefrontal cortex, and that in addition to initiating behavior, this network also trains a faster and more direct network that includes projections from sensory association cortex directly to rule-sensitive neurons in premotor cortex. After much practice, the direct network is sufficient to control the behavior, without prefrontal involvement. The model is implemented as a biologically-detailed neural network constructed from spiking neurons and displaying a biologically plausible form of Hebbian learning. The model successfully accounts for single-unit recordings and human behavioral data that are problematic for other models of automaticity.

The dissertation also presents the results from two experiments investigating the nature of what is automatized after lengthy practice with a rule-guided behavior. The results of both experiments suggest that an abstract rule, if interpreted as a verbal-based strategy, was not automatized during training, but rather the automatization linked a set of stimuli with similar values on one visual dimension to a common motor response. The experiments were designed to test the Cortex Automatizes Rules Model. The present results support this model and suggest that the projections from visual cortex to prefrontal and premotor cortex are restricted to visual representations of the relevant stimulus dimension only.

One defining characteristic of human behavior is the ability to select an appropriate action in an entirely novel situation. How abstract rules are represented in the brain, and how these representations operate on internal models of the world to generate flexible rapidly optimized behaviors remains an open question in neuroscience as well as computer science. The hallmark of quickly optimized flexibility inherent to explicit behavioral rules has lead to the assumption that these rules are based on high level abstractions. In the recent past, behavioral measurements in the human psychophysics literature were linked to predictions at the processing level and the convenient mathematical construct of a decision criterion has precipitated in several cognitive process models. However, the assumption that higher order decision processes involve comparisons to some internal criterion is not trivial and was investigated. This thesis provides evidence to falsify the criterion as a processing component in human decision making and offers fundamental insights to reverse engineer decision processes in the brain. Cognitive flexibility and rule guided behavior appear to rely on phylogenetically advanced extrapolation processes that are mediated by dynamic feed-forward and feed-back circuits, which continually update internal and external information to support goal directed behavior.