This paper examines some characteristics of the learning process in a model of skill learning (Miyata, 1987) in which performance of executing sequential actions becomes increasingly more efficient as a skill is practiced. The model is a hierarchy of sequential PDP networks which was designed to model a shift from a slow, serial performance of a novice to a fast, parallel performance of an expert in tasks such as typing. The network develops representation of a set of sequences as it tries to produce the sequences faster. The model was found to yield the power law of learning (Newell and Rosenbloom, 1981). In addition, it exhibited a frequency effect on substitution errors similar to what was found in typing (Grudin, 1983).
Learning has intrinsic importance to the study of skilled performance because the nature of performance dramatically changes as a skill develops. However, study of skill learning is difficult because one has to explain not only what processing structure underlies skilled performance in a particular task domain, but also what mechanism enables us to build such structures as a result of experience in many different tasks. The approach taken in this work is to look for phenomena that are observed across a wide range of tasks and to try to develop a model of action learning that attempts to account for what seem to be quite general phenomena.
I have previously proposed a model of skill learning in which performance of sequential actions becomes faster as a skill is practiced (Miyata, 1987). This model successfully accounted for some effects of presentation frequency in typing, specifically the effect on speed (Grudin &Larochelle, 1982) and on a class of execution errors (Sellen, 1986). This paper reports on some additional experiments which revealed some interesting characteristics of the learning process in the model. In particular, the model is shown to exhibit the power law of learning. In addition, it exhibited a frequency effect on error patterns similar to typing errors at the keystroke level as well as at the sequence level as previously shown (Miyata, 1987). I will start by describing an example of the power law to illustrate the kind of skills being modeled in this work.
The Power Law
Probably the most general phenomenon we know about learning is that practice makes performance faster. However, more specific regularities seem to exist. For a wide variety of tasks, the learning curve (i.e., a plot of the time to perform the task versus the number of trials) produces approximately a straight line in log-log coordinates (Newell &Rosenbloom, 1981). This has been generally called the power law . . .