TP53 is mutated in most types of human cancers and is one of the most popular genes in cancer research. The p53 protein is a sensor of multiple forms of genotoxic, oncogenic and non-genotoxic stress. It suppresses growth and controls survival of stressed cells, and as such, is the focal point of selection pressures in tissues exposed to carcinogens or to oncogenic changes. Thus, the clonal expansion of cells with mutations in TP53 may be seen as the result of a selection process intrinsic to the natural history of cancer. In this review, we discuss the nature of these various forms of selection pressure. We present a hypothesis to explain why TP53 is often mutated as either an early or a late event in cancer. Furthermore, we also summarise current knowledge on the molecular consequences of mutation for loss of wild-type protein function, dominant-negative activity, and a possible gain of oncogenic function.