The significance of stated preference data obtained using portable computers was empirically supported by comparing this with the data from a conventional paper-based questionnaire survey. The internal validity of mode choice models based on Mulitinomial and Rank logit types was tested for these two different data sets which were obtained in Hiroshima, 1988 and 1989. As a result, the computerized interview data represented a higher level of fit in model specification than the manually completed data, indicating the effectiveness of the former survey method.