Learning Styles and Learning Behaviour in the Hypermedia Environment RACE - Analysing Computer Log files and Questionnaire Data

The following Paper describes an experiment which analyses in how far specific learner, set by questionnaire, types show congruent behaviour in a concrete learning situation. The undertaken a priori subdivision in surfaceand deep-strategic learner types based on established models (Schmeck, 1988; Mandl & Friedrich, 1992; Krapp, 1993). First a group of 52 learners worked with the hypermedia program RACE (Rockmann & Thielke, 2000a; 2000b; 2001) for one hour. Then a general questionnaire about their learning behaviour in self-regulated learning situations was given. The questionnaire data (trait measurement) and the process data from the log files (state measurement) were evaluated by cluster-analysis and compared with each other. The experiment shows that both analyses result in a two-cluster solution with clear allocation of surfaceand deep-strategic variables. The congruency between these solutions amounts to approx. 70%, and a significant difference could not be determined (χ = 581, 1, p = 446). The comparison of the z-transformed means for the cluster variables points out that the analyses of the process data may be of a higher selectivity for determining these learning types. Interesting findings concerning the cluster composition can be shown when you include further descriptive variables (expert-novice status, state of training and computer experience).