The results for colony forming unit granulocyte-macrophage (CFU-GM) assays vary substantially between centres. It is possible that colony counting is largely responsible for this discrepancy. In order to examine this exclusively from the many factors that make up the CFU-GM assay, we performed a colony counting exercise involving 11 laboratories. Two-way analysis of variance showed a highly significant difference (P =0 .0001) in the counts obtained from the centres. One centre was found to score consistently high and two others scored consistently low numbers of colonies. This suggests that identification of colonies is a major source of variation between centres.