The evidence considered here reinforces the conclusion that T-cell responses to tumours involve complex cellular interactions. An attempt to summarize some of these interactions is shown. This emphasizes that not only are the interactions between the effector cell populations complicated, but that the target cell surface is also subject to variation and modification as a result of the immune response. A feature that also emerges from these studies is that most cells apparently responding to or infiltrating a tumour do not necessarily participate in its destruction, and it is in this area that experimental tumour systems have particular value. This also perhaps explains the preoccupation of experimentalists with the identification of 'the' effector cell crucial to tumour rejection. However, there is heterogeneity between systems in terms of the type of rejection response induced, but a logical basis for this heterogeneity is not established. If experimental studies could define the nature of the immune response generated by a tumour in the context of the biological features of the tumour itself, this could lead to the prediction of the immunogenicity and potential for induction of a rejection response for that tumor. Clearly, experimental tumour systems do not provide an exact reflection of the situation with human tumours. However, they may provide systems that illuminate particular aspects of the human response, and give precedents to guide the interpretation of data derived from human systems. This form of assessment is still at an early stage, but developments in the experimental field should provide a framework for the development and exploitation of T-cell responses to tumours.