Marrow relapse is the major obstacle to cure for 10-15% of young patients with acute lymphoblastic leukaemia (ALL). Recent investigations into the biology of minimal residual disease indicate that many early relapses derive from residual cells present at first diagnosis, but some late relapses might represent new mutations in leukaemic cells not eliminated by conventional therapy. Treatment of marrow relapse involves higher doses and more intensive schedules of the drugs used for initial therapy with or without haemopoietic stem cell transplantation. In most reports, transplantation is better than continuation chemotherapy in early marrow relapse, but its role in later relapse is less clear. Current therapy cures 10% of patients with early marrow relapses and 50% of those with late relapses, but outcomes have changed little in the past two decades. Understanding the molecular biology of ALL underlies development of improved risk stratification and new therapies. Although better drugs are needed, introduction of new agents into clinical trials in paediatric disease has been difficult. Innovative trial designs and use of valid surrogate endpoints may expedite this process.