Corticosteroids were used to treat two children with presumed idiopathic thrombocytopenic purpura and one with juvenile rheumatoid arthritis without examination of the bone marrow. Of the two with presumed idiopathic thrombocytopenic purpura, one had Fanconi's anaemia and the other may have had aplastic anaemia. The third child had acute lymphoblastic leukaemia. The diagnosis of Fanconi's anaemia was delayed. A diagnostic and therapeutic dilemma was caused in the second case. In the third, delayed diagnosis and, perhaps, compromised outlook resulted. These three cases re-emphasize the well aired caveats about the diagnosis of idiopathic thrombocytopenic purpura and juvenile rheumatoid arthritis and provide further support for the arguments of those who believe that if corticosteroids are to be used to treat such children, their bone marrow should be examined first.