SummaryHuman immunodeficiency virus (HIV)-induced destruction of follicular dendritic cells (FDCs), which are important in immunological memory, may be a major pathway of HIV pathogenesis. We use a mathematical model to investigate this hypothesis and conclude that a low level of FDC destruction could ultimately result in loss of control of HIV. Their slow turnover makes them good candidates for the part of the immune system that fails during the long period of HIV infection. As FDC destruction is essentially a misdirected immune response, too much immunotherapy may be detrimental. Our model shows how to estimate this critical level of immunotherapy. We derive an expression for the time taken to the loss of immune control. Transient changes in the viral growth rate before the immune system fails do not affect this time, providing a possible explanation for the results of the Concorde trial. We suggest that inducible B cell function is a good potential marker of disease progression, indicating the functional ability of the FDC network. Finally, we rereview data in the light of the FDC theory, paying particular attention to data on CD4+numbers and function that are inconsistent with the classical view of HIV pathogenesis.