The efficiency and sensitivity of fluorescence in situ hybridization (FISH) and G-banding for the detection of male cells in mixtures of male and female cells derived from peripheral blood (PB) and bone marrow (BM) are compared. Zero false positive and negative rates were obtained with G-banding in all studies. False positive rates obtained with FISH were low (0-2.5%) in both PB and BM. False negative rates varied from 0-13.3%. Variations due to differences in fluorescent counterstain or scorers were not significant and in some cases the percentage of Y-positive cells as determined by FISH agreed more closely with the estimated frequency of male cells than that obtained by G-banding. The minimum limits of detection for FISH were 5% male cells on counting 500 PB or BM interphases; 10% male cells on counting 100 PB metaphases; and 20% male cells on counting 50 BM metaphases. This is in agreement with the theoretical limits of detection [95% confidence limit (CL)]. However, minimum levels of detection for female cells were below the 95% CL. It was not possible to reliably distinguish 95% female cells from 100% on counting 500 PB or BM interphases; 90% from 100% on counting 100 PB metaphases; or 80% from 100% on counting 50 BM metaphases.