Seven years after the initial studies of the prognostic value of the proto-oncogene c-erbB-2 in breast cancer, its role is still being defined. The interpretation of studies on the use of this gene and its protein product in prognostic and predictive tests for breast cancer is complicated by multiple methodologies and the inherent difficulties in the studies. The work has moved beyond the stage at which small studies with short follow-up (useful for hypothesis generation) are of value, to the stage in which large studies with sufficient statistical power to find significant correlations are central. These larger studies do not lend support for the use of c-erbB-2 in the evaluation of axillary-node-negative patients, the group of breast cancer patients for whom refinement of prognostic estimates is now most important. There are, however, hints that c-erbB-2 may have value in predicting response to certain treatments, though the studies so far are too few, often too small and too conflicting to reliably confirm this.