Progress in tumor immunology has not been translated to effective immunotherapies for cancer. Most of the current effort in basic and clinical research concentrates on generating effective immune responses against model or well characterized antigens, yet vaccines targeting defined antigens have been less clinically successful than those based on whole tumor cells or their extracts. This review considers characteristics of proteins that determine how effectively they might serve as targets of immune control, and how different sources of antigens have fared in clinical trials.