HER-2 has a well-established role as a prognostic indicator in breast cancer and as a predictor for response to trastuzumab. Recent studies have also suggested that it may serve as a predictor of response to anthracycline-based therapies. This article argues that the data are insufficient to accept this hypothesis as scientifically established. The argument is developed along several lines: first, that the trials used to support a predictive role for HER-2 have real flaws with regard to this hypothesis; second, that HER-2 is a remarkably inconsistent predictor of anthracycline response when examined in a broader context that includes preoperative and metastatic disease; third, that preclinical data fail to support the hypothesis; and finally, that even if accepted, the hypothesis is difficult to extrapolate to the everyday world of breast cancer.