AIMS--To evaluate the reliability of fluorescence in situ hybridisation (FISH) in the retrospective cytogenetic assessment of old bone marrow smears stored for periods of up to 20 years. METHODS--A series of bone marrow smears either Romanowsky stained, or frozen and unstained, and aged from one month to 20 years were hybridised with biotin labelled probes specific for the centromeric regions of human chromosomes X, 6, and 18. Sites of hybridisation were detected with fluoresceinated avidin. One hundred to 400 cells from each preparation were examined and the number of signals observed was recorded. RESULTS--All smears exhibited signals in most cells examined. In cytogenetically normal cases, an average 67.6% of cells (range 36%-90%) demonstrated the appropriate number of X centromere signals. In those samples known to contain extra chromosomes X, 6, or 18 the presence of cells with the abnormal copy number was clearly detected in each case. CONCLUSION--When applied in the way described, FISH can give consistent and accurate results with a variety of archival bone marrow smears, including aged prestained material. This will permit retrospective assessment of specific cytogenetic abnormalities in patients with leukaemia using their initial diagnostic slides even where these are several years old.