How to t(r)ap user's mental models

Abstract The paper discusses current approaches to tapping users' mental models of complex devices, and argues that most experimental studies in the area suffer from two methodological shortcomings. The first problem described is over-interpretation of performance data - often the only dependent variables in the design. The second problem is lack of ecological validity : the devices used are often rather artificial, and one-off studies with a short time of interaction between users and system tend to concentrate on static aspects of one model the user is supposed to hold. Five experimental studies on users' mental models, designed and conducted by the author, are described to illustrate how ecological validity can be increased, and how videotapes of interaction can be analysed to obtain a range of quantitative and qualitative data on the users' model.