Routine flow cytometric DNA analysis was compared in two laboratories by using matched fresh-frozen breast cancer and soft tissue sarcoma biopsy specimens. Laboratory I applied the Vindelöv preparation method and an exponential background subtraction algorithm in the cell cycle calculation. Laboratory II used the Formalin-protease preparation technique and the sliced-nuclei background model. The results of the ploidy analysis showed good agreement between the two laboratories; however, the results of the cell cycle analysis showed considerable systematic differences between labs. Laboratory I obtained significantly lower values of S-phase fraction and higher values of G2-phase fraction than laboratory II. To explain these discrepancies, the effects of differences in the preparation methods and background subtraction algorithms were studied. The Vindelöv preparation method yielded higher debris and aggregation levels than the Formalin-protease technique and tended to give higher %S and %G2 values. When the two background models were used in the same histograms, the exponential background model tended to give %S values distinctly lower than and %G2 values almost identical to those obtained with the sliced-nuclei algorithm. The sum of these effects accounts for the observed inter-laboratory discrepancies. Different from the sliced-nuclei fit, the exponential background fit often did not accommodate to the original data in the <2c histogram region and resulted in a considerable inter-operator variability of %S calculation in histograms with <5% S. When aggregate correction was added to the sliced-nuclei algorithm, the differences between %S values in histograms from the two laboratories almost disappeared.