The infiltration of breast carcinomas with lymphoid cells has often been interpreted as an indication of an active immune response against the tumor and thus a favorable prognostic sign. Several studies have, however, cast doubt on this assumption. In situ breast carcinomas are more common than invasive cancers and it may be speculated that immune surveillance plays a role in preventing some localized cancers from becoming invasive. A secondary type of immune surveillance might be implicated in the long persistence of dormant breast carcinoma cells in the bone marrow. Breast cancer cells can carry tumor-associated antigens, particularly MUC1. These may elicit specific antibody responses but there is less evidence for a CTL response. There are indications that professional antigen presenting cells may be present and active at the edge breast tumours. Breast cancer cells may also interact directly with macrophages and NK cells. In terms of immune effector mechanisms in breast cancer, the communication with potential effector cells is likely to be often faulty because of altered expression of HLA class I molecules. Pleiotrophic cytokines are frequently present and could have a variety of effects ranging from growth inhibition to stimulated proliferation, loss of cell adhesion and activation of matrix degrading enzymes. Fas ligand is unlikely to play a role in the immune evasion of breast cancer. There is thus evidence for a variety of immune reactions to breast cancer. It is possible that they mediate some form surveillance, but growing, invasive tumors have escape routes and may even use cytokines to their advantage.