Anti-androgens used in prostate cancer therapy inhibit AR (androgen receptor) activity via largely unknown mechanisms. Although initially successful in most cases, they eventually fail and the disease progresses. We need to elucidate how anti-androgens work to understand why they fail, and prolong their effects or design further therapies. Using a cellular model, we found different anti-androgens have diverse effects on subcellular localization of AR, revealing that they work via different mechanisms and suggesting that an informed sequential treatment regime may benefit patients. In the presence of the anti-androgens bicalutamide and hydroxyflutamide, a significant proportion of the AR is translocated to the nucleus but remains inactive. Receptor inhibition under these conditions is likely to involve recruitment of co-repressor proteins, which interact with antagonist-occupied receptor but inhibit receptor-dependent transcription. Which co-repressors are required in vivo for AR repression by anti-androgens is not clear, but one candidate is the Notch effector Hey1. This inhibits ligand-dependent activity of the AR but not other steroid receptors. Further, it is excluded from the nucleus in most human prostate cancers, suggesting that abnormal subcellular distribution of co-repressors may contribute to the aberrant hormonal responses observed in prostate cancer. A decrease in co-repressor function is one possible explanation for the development of anti-androgen-resistant prostate cancer, and this suggests that it may not occur at the gross level of protein expression.