The treatment of cancer during any period has been based not on the whim of a clinician but on the therapeutic consequences of the dominant biologic model of the disease. Until the 1960s, the dominant model of breast cancer was of a disease that spread centrifugally along anatomic pathways, with time being the only determinant of prognosis. An alternative model, that of biologic determinism, posits that the outcomes of treatment are determined by the extent of microscopic dissemination that occurred before the tumor became detectable. This model, too, has flaws, and the author suggests that it is time for a Kuhnian paradigm shift. Breast cancer exhibits a heterogeneity of phenotypes resulting from one or perhaps two mutations. The multiple prognostic variables may be epiphenomena, expressing different degrees of amplification of a limited domain of the genome.