A program is describedforfitting a regression model in which the relationship between the dependent and the independent variables is described by two regression equaThe last decade has seen substantial interest in the application of cognitive-process models of task performance to the analysis and description of individual differences in performance (see Lohman & Ippel, 1993;Snow & Lohman, 1993). One ofthe determinants of task performance that has emerged from these investigations is the strategy-shift phenomenon, whereby a subject may apply different strategies for solving different items of the same task. Studies that have identified strategies employed by subjects have used a variety of measures, including retrospective reports, eye-fixation records, errors, and measures of response latencies (e.g., Cooper, 1980;Glushko & Cooper, 1978;Kyllonen, Lohman, & Snow, 1984;Snow, 1980). Kyllonen, Lohman, and Woltz (1984) systematically studied subjects' strategy shifts on different items of the same spatial task as a function of item characteristics. They fitted several regression models for all subjects in order to identify the strategies that subjects used and the possible strategy shifts that occurred. Each model postulated different processes underlying performance, and some models postulated strategy shifts (of different Requests for reprints should be sent to A. L. Beem, Centre for the Study of Education and Instruction, Leiden University, Wassenaarseweg 52, P.O. Box 9555, 2300 RB Leiden, The Netherlands (e-mail: beem@rulfsw.leidenuniv.nl). kinds). The strategy-shift models were fitted with a few different shift points that were fixed a priori and that were the same for all subjects, rather than being estimated from each subject's data separately. The best-fitting process model for each subject was determined by comparing adjusted R-square statistics from different models. The results suggested not only that different subjects use different strategies, but also that the same subjects may use different strategies on different items of the same task, as well as that subjects differ in the point at which they shift from one strategy to another.Ippel and Beem (1987) fitted latencies of subjects on a mental-rotation task. In mental-rotation tasks, subjects must determine whether two figures are the same in shape or are mirror images. The figures can differ in orientation. Latencies typically suggest that subjects mentally rotate one of the figures into congruence with the other. On each trial, rotation can be achieved by either a clockwise or a counterclockwise routine. Depending on the angular disparity ofthe figures, one ofthese routines is objectively the most efficient, whereas 180 0 of angular disparity is an objectively optimal point for shifting from one rotation routine to the other. Ippel and Beem fitted subjects' latencies as a linear function of angular disparity with a model for which the shift point was not assumed to be known a priori, but was estimated from each subject's data separately. Latencies' lin...