ABSTRACT
The connection between the theory of knowledge spaces and formal models of cognitive problem solving processes is examined. This connection is described for a model of the cognitive processes of subjects in solving letter series completion problems, which is based on ideas from Simon and Kotovsky (1963). The model is formulated as an algorithm, depending on its ability to solve letter series completion problems on two parameters. Different abilities of subjects in solving such problems can be represented within the model by different choices for these parameters. It is shown that the model determines surmise relations on sets of letter series completion problems. These surmise relations, respectively the corresponding quasi-ordinal knowledge spaces, are used in two experimental in-vestigations to test the underlying process model empirically. The results of these experiments show that the surmise relations derived from the process model are able to predict the difficulty of letter series completion problems in a satisfactory manner.
