Neural Dynamics of Learning and Performance of Fixed Sequences: Latency Pattern Reorganizations and the N-STREAMS Model
MetadataShow full item record
Fixed sequences performed from memory play a key role in human cultural behavior, especially in music and in rapid communication through speaking, handwriting, and typing. Upon first performance, fixed sequences are often produced slowly, but extensive practice leads to performance that is both fluid and as rapid as allowed by constraints inherent in the task or the performer. The experimental study of fixed sequence learning and production has generated a large database with some challenging findings, including practice-related reorganizations of temporal properties of performance. In this paper, we analyze this literature and identify a coherent set of robust experimental effects. Among these are both the sequence length effect on latency, a dependence of reaction time on sequence length, and practice-dependent lost of the lengths effect on latency. We then introduce a neural network architecture capable of explaining these effects. Called the NSTREAMS model, this multi-module architecture embodies the hypothesis that the brain uses several substrates for serial order representation and learning. The theory describes three such substrates and how learning autonomously modifies their interaction over the course of practice. A key feature of the architecture is the co-operation of a 'competitive queuing' performance mechanism with both fundamentally parallel ('priority-tagged') and fundamentally sequential ('chain-like') representations of serial order. A neurobiological interpretation of the architecture suggests how different parts of the brain divide the labor for serial learning and performance. Rhodes (1999) presents a complete mathematical model as implementation of the architecture, and reports successful simulations of the major experimental effects. It also highlights how the network mechanisms incorporated in the architecture compare and contrast with earlier substrates proposed for competitive queuing, priority tagging and response chaining.