This paper reports on an attempt to introduce the rough sets (RS) technique into an activity-based modelling approach. Since their introduction in 1981 by Pawlak, rough sets have already been successfully applied in a series of research domains. Surprisingly, though the approach has never been applied in modelling travel choice patterns. In order to investigate the opportunities and constraints of the rough sets approach in activity-based travel modelling, we use data from a Belgian travel diary survey, and analyse these data by means of the ROSETTA-software. This software package was originally developed by Ohrn [Discernibility and Rough Sets in Medicine: Tools and Applications (1999)]. In total, two case studies are conducted through which the performance of the rough sets approach for travel pattern generation, classification and choice prediction will be ascertained. Although the overall results may seem not that promising, interesting conclusions for future research can be drawn. (C) 2004 Elsevier Ltd. All rights reserved.