Tracing the evolution of knowledge structures

The goal of the research described here is to understand the evolution of knowledge structures as novices learn to use text editors. Two groups of individuals were trained to use one of two text editing systems: the IBM Displaywriter and UCSD-Pascal. Individuals rated the similarity among pairs of editing commands before and after training. Before training, subjects’ similarity ratings are characterized by heavy reliance on natural language denotations of command names. After training, subjects’ similarity ratings are explainable in terms of a small set of identifiable properties of underlying knowledge structures that are typical of experts’ knowledge about a text editor. These properties reflect functional features of command actions, and the psychological salience of a property for a given individual is related to editing performance.