Predicting and binding: interacting algorithms supporting the consolidation of sequential motor skills

The efficient execution of serially ordered actions is crucial for many everyday tasks. Rather than emerge from a singular learning process, a growing body of evidence in both cognitive science and neuroscience suggests that the acquisition of habitual motor sequences relies on a multitude of learning systems that fall under two general classes of computation: fast prediction of transition probabilities between events and slower binding of serial actions into unified sets. Here we review the emerging empirical support for this multi-system model of sequential skill acquisition and show how these systems coordinate together to foster the crystallization of complex skills across time.