Variability in disease presentation and course is a hallmark of cancer. Variability is seen among similarly diagnosed cancers in different patients or animal hosts and in the same cancer at different periods of time. This latter type of variability, termed "tumor progression," was defined by Foulds in a series of six rules that describe the independent behavior of individual cancers and the independent evolution of different cancer characteristics. Tumor progression is believed to result from variability among subpopulations of tumor cells within individual cancers and from selection of these subpopulations by conditions within the cancer environment, such that different subpopulations come to prominence over the course of cancer development and growth. Interactions among subpopulations, however, modulate tumor behavior as well as tumor evolution. The leading hypothesis for the origin of tumor subpopulations is the genetic instability of cancer cells. There are a number of possible mechanisms of genetic instability, some internal to cancer cells (mutation, amplification, mutator phenotypes, DNA repair deficiencies) and some present in the tumor microenvironment (endogenous mutagens). There are also potential epigenetic mechanisms of variability, including alterations in gene regulation, differentiation, adaptation, and cell fusion. Regardless of mechanism, the heterogeneity of tumor subpopulations poses a number of challenges to the practice of cancer research, including the design of reproducible and meaningful experiments. Tumor heterogeneity also has significant consequences for the clinical assessment of tumor prognosis and the development of effective treatment regimens.