Reports from the Department of Psychology, Stockholm University, 1996, No. 826.

ABSTRACT
The goal of this project is to explain learning when insight is the key feature. To account for characteristic twists of knowledge occurring in situations like these I have adopted a framework of deconstruction and reconstruction of knowledge.
The present study builds on observations of naive, unguided subjects learning an out of sight property of a text editor. A previous report presumed that knowledge of the editor would advance in logical steps, as experience accumulated. An informal analysis of verbal and manual data blurred this picture. Subjects' consecutive constructions of the editor seemed inconsistent. This result evoked a series of questions. Is this an effect of poor data, or is it a fact of the subject? Does it demonstrate that an explanation lies outside the logic of information processing systems (axiomatic systems)? Or, is the problem of a methodological nature?
In the present study I address these questions by a renewed analysis of data, based on computer simulations of one subject. This sharpens the method by requiring formal accounts of constructions inferred from subject data. As a result the previous report is qualified. From this I argue that we choose to interpret the present observations of inconsistency as facts of subjects, and not as artefacts of method.

Key words: Learning, problem solving, information processing, insight, creativity, restructuring, conceptual change, knowledge reconstruction, deconstruction, protocol analysis, meaning analysis, verbal data, think aloud, computer simulation, production system.