It is now well accepted that cancers are heterogeneous and contain multiple subpopulations of cancer cells that differ in important behavioral properties, such as growth rate, ability to metastasize and sensitivity to treatment. This review describes results using a mouse mammary tumor system to model the effect of heterogeneity on tumor growth and response to chemotherapy. Tumor subpopulations that differ in growth and therapy parameters are mixed together to simulate tumor heterogeneity. The mixtures are studied in a variety of protocols designed to detect interactions between the subpopulations and to determine the mechanisms of these interactions. The central conclusion of this work is that tumor subpopulations do not behave independently of each other but rather form a society of cells in which they influence each other's growth and treatment response. The mechanisms by which they do so depend on the unique characteristics of the interacting subpopulations, on the behavior being affected and on the circumstances of their interaction. These observations suggest that the maintenance of tumor heterogeneity is a consequence of the cancer cell society and that cancer behavior and progression depend on interactions among all the members of that society, not just on its most deviant variants.