Two laboratories equipped with CAS 200 (Becton Dickinson Image Cytometry Systems, San Jose, CA) instruments participated in this study of variability of DNA analysis of bladder tumor specimens. Formalin fixed paraffin embedded specimens were disaggregated and centrifuged onto microscope slides from ten bladder tumor specimens and two specimens of normal urothelium. Sources of variability considered were Specimen, Slide, Run, Laboratory, and Error. Slides were systematically scanned and 200 cells measured followed by the operator selecting 100 nuclei with abnormal morphology. DNA index (DI) and hyperdiploid fraction (HDF) were calculated from the DNA frequency distributions. For systematic sampling, 92% of the variability was due to Specimen indicating that differences in HDF values between specimens reflect biological differences. With selective sampling, only 67% of the variability in HDF is due to Specimen differences. Other factors, Laboratory, Error, and Laboratory 3 Specimen interaction each accounted for approximately 10% of the variability. Similarly variability of DI with selective sampling was also higher, and less specimen dependent than systematic sampling. It is important that sampling schemes and selection criteria be carefully documented in order to control variability. Enriched (or selective) sampling for abnormal cells has the potential to increase sensitivity but specimen classification based on these measurements must depend on determination of the frequency of such cells in the total population. Cytometry 26:176-180, 1997.