Cancer stem cells undoubtedly exist in many tumour types, including the prostate. This hypothesis can explain both the heterogeneity of prostate tumours and their variable responses to several conventional therapies. In the longer term, therapies directed against tumour stem cells should offer a real possibility of long-term cure, rather than current palliative therapy. Identifying specific tumour stem-cell markers will enhance this process, but the scarcity of these cells within the mass of more differentiated amplifying progeny that comprise >99.9% of most cancers makes this a severe technical challenge. In addition, many tumour stem-cell markers are probably shared with normal stem cells, both in prostate and in stem cells from other tissues, but tumour-specific patterns of gene expression, probably designed to allow the tumour stem cell to survive outside its protective 'niche' in normal tissues, will be the best initial targets for new therapeutic agents.