Learning by chunking : a production-system model of practice

The power law of practice states that the time to perform a task decreases as a power-law function of the number of times the task has been performed. One possible explanation for this ubiquitous regularity is the chunking theory of learning. It proposes that the acquisition and use of chunks is the basis for these improvements. In this article we describe a first attempt at implementing a learning mechanism based on the chunking theory of learning. This work includes: (1) filling out the details of the chunking theory; (2) showing that it can form the basis of a production-system learning mechanism; (3) showing that the implemented mechanism produces power-law practice curves; and (4) investigating the implications of the theory for production system architectures in general. The approach we take is to implement and analyze a production-system model of the chunking theory in the context of a specific task — a 1023-choice reactiontime task. In the process, we develop a control structure for the task; describe the three components of the implemented model — the Xaps2 production-system architecture, the performance model for the task, and the chunking mechanism; and analyze simulations to verify that the implemented model does generate power-law practice curves. This research was sponsored by the Defense Advanced Research Projects Agency (DOD), ARPA Order No. 3597, monitored by the Air Force Avionics Laboratory Under Contract F33615-78-C-1551. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the Defense Advanced Research Projects Agency or the US Government LEARNING BY CHUNKING I Table of