In patients with idiopathic aplastic anaemia (n = 34) and Fanconi's anaemia (n = 8), sampled once or on several occasions, serum erythropoietin (Epo) increased with increasing severity of anaemia with apparently similar rates of increase in each group. However, after adjustment for Hb, log Epo values for the Fanconi's anaemics tended to be greater than those for the idiopathic aplastic anaemics (P < 0.01). Erythropoietin concentrations in serum samples from patients with Fanconi's and idiopathic aplastic anaemias tended to be greater than in samples from patients with anaemias from protein energy malnutrition, myelodysplasia and iron deficiency. The results suggest that there is no deficiency of erythropoietin in Fanconi's and idiopathic aplastic anaemias and that if exogenous erythropoietin is of any benefit it would need to be administered in doses large enough to induce a significant increase in log Epo. Results of the study illustrate the need to take account of the assumptions which underlie interpretation of the statistical analysis. Use of erythropoietin values in place of log Epo gives misleading conclusions demonstrable as invalid as the conditions for normality of distribution of the data and homogeneity of variances were not satisfied.