We have developed a mathematical model for the regulation of the growth of autoreactive T cells (the T cells responsible for autoimmunity). The model is very simple in that it is based only on the fundamental properties of T cells. However, despite this simplicity, it can account for a variety of phenomena referred to as T-cell vaccination. The purpose of T-cell vaccination is to create resistance to autoimmunity. This can be achieved by injecting either a subpathogenic quantity of autoreactive T cells, or attenuated autoreactive cells, or cells that recognize the autoreactive cells. The results of our model are based on the assumption that the self antigens involved in T-cell vaccination are normally not expressed; thus the autoreactive T lymphocytes are neither activated nor negatively selected. Self tolerance, therefore, corresponds to a 'passive' state. T-cell vaccination induces a transition from this passive state of tolerance to an active state of tolerance. In this state the autoreactive cells are controlled by regulator cells which recognize the autoreactive cells. The model predicts a qualitative difference between vaccination with normal autoreactive cells and vaccination with attenuated autoreactive cells. Normal cells may give rise to a permanent switch to the vaccinated state; attenuated cells, however, can provide only transient protection, which is dose dependent. Preliminary experimental data confirm this prediction. Finally, we propose a speculative explanation for relapsing autoimmune disease.