The reproducibility in quantitation of proliferation activity, determined using the monoclonal antibody 19A2 to proliferating cell nuclear antigen (PCNA), was tested in visual and computer-assisted analyses of brain tumor material. The PCNA labeling index was scored using count (PCNA-LI, visual and computer analyses) and area (PCNA-LIa, computer analysis) estimates of immunopositivity. The quality of immunostaining was the most important reason for variation in the assessment results. Other significant variation sources in the assessment were experience in selecting microscopic fields and distinguishing immunopositive nuclei from immunonegative ones. Computer-assisted analysis improved the reproducibility of quantitation between different observers (visual rPCNA-LI = 0.624 versus computer assisted rPCNA-LI = 0.904). Also, the use of PCNA-LIa improved the intraobserver and interobserver reproducibility in different stainings (observer 1:rPCNA-LI = 0.857 versus rPCNA-LIa = 0.874; observers 1 and 2: rPCNA-LI = 0.904 versus rPCNA-LIa = 0.927; observers 3 and 4: rPCNA-LI = 0.848 versus rPCNA-LIa = 0.906). PCNA-LIa by computerized image analysis improves accuracy in the evaluation of the granularly expressed PCNA level. Furthermore, the effect of tumor heterogeneity on the assessment results can be diminished with the computerized method because large tissue areas can be analyzed faster.