The commute impedance model (e.g. Stokols, Novaco, Stokols and Campbell, 1978) is evaluated through a literature review and a survey of commuters who drive to work (N=418). A new construct for the study of commuting—commute variability—is introduced and tested in the framework of the survey. Findings from the literature review indicated (a) partial support for the first proposition of the impedance model namely, that impedance—measured as commute distance or time—causes strain (e.g. employee health deficits), but (b) no support for the second proposition of the model namely, that the impedance–strain is moderated by control or the availability of choice. Similarly, the survey showed that (a) commute length (impedance) is consistently correlated with strain, (b) choice is weakly and negatively related to strain, but (c) there is no interactive relationship of choice and length with strain. Furthermore, the results suggest that (a) commute variability is the strongest correlate of commute strain, and that (b) once commute variability is partialled out, most of the partial correlations between impedance and strain are not significant. The paper concludes with an alternative commuting–strain model which includes the new commute-variability construct. © 1998 John Wiley & Sons, Ltd.