The cellular and molecular mechanisms underlying the formation of distinct central, effector, and exhausted CD8+ T-cell memory subsets were first described in the setting of acute and chronic viral diseases. The role of these T-cell memory subsets are now being illuminated as relevant to the tumor-bearing state. The generation and persistence of productive CD8+ T-cell memory subsets is determined, in part, by antigen clearance, costimulation, responsiveness to homeostatic cytokines, and CD4+ T-helper cells. By contrast, chronic exposure to antigen, negative costimulation, and immunomodulation by CD4+ T regulatory cells corrupt productive CD8+ T memory formation. It has become clear from human and mouse studies that the mere generation of CD8+ T-cell memory is not a 'surrogate marker' for cancer vaccine efficacy. Some current cancer vaccine strategies may fail because they amplify, rather than correct or reset, the corrupted CD8+ memory population. Thus, much of the present effort in the development of vaccines for cancer and chronic infectious diseases is aimed at creating effective memory responses. Therapeutic vaccines for cancer and chronic infectious diseases may achieve consistent efficacy by ablation of the dysfunctional immune state and the provision of newly generated, non-corrupted memory cells by adoptive cell transfer.