Although even "spontaneous" tumours are immunogenic and are commonly infiltrated by tumour antigen-specific T cells (at least in melanoma), most tumours are not completely rejected by the host, and cancer progresses. There is a growing realisation that many responses defined as antitumour effector mechanisms act as double-edged swords and under different conditions either become ineffective or even protumorigenic. Examples are interleukin 2 (also proapoptotic for activated T cells), interferon gamma (by induction of ligands for T and NK cell inhibitory receptors), angiogenesis inhibition (by hypoxia-mediated induction of growth factors promoting metastasis), and macrophage free radical-mediated cytotoxicity (by inhibiting T cells). Immune selection pressure itself, resulting in outgrowth of resistant tumour variants could also be viewed in this light. On the other hand, knowledge of the many tumour escape pathways offers the theoretical possibility of reconstituting antitumour immunity. Tumour escape from immunosurveillance represents the last series of hurdles to be overcome in formulating truly effective cancer immunotherapy, but given the immense plasticity of the tumour cell, and the complex balance between pro- and antitumour activity of the very same effector pathways, this remains a major challenge.