In hematoxylin-eosin-stained sections of 20 pediatric sarcomas the mitotic index was assessed by four experienced pathologists and four less-experienced observers without prior instructions. In adjacent sections immunolabeled for MIB-1, the proliferation index was assessed as the estimated percentage of labeled cells in the tumor cell population. ANOVA revealed that the variation between tumors explained 77% of the variation in mitotic indices in the group of experienced observers compared with 49% in the less experienced group. The variation between tumors explained 64% of the variation in proliferation indices in the experienced group and 71% in the less experienced group. The proliferation indices showed less variation between observers than the mitotic indices. No correlation was found between mitotic and proliferation indices. The results suggest that training is an important factor in the reliability of mitotic counting. The use of proliferation markers has a higher reproducibility, especially in less-experienced observers. However, for clinical use it has the disadvantage of being the more expensive, more time-consuming technique; moreover, the biological significance of proliferation has not been established and may differ from that of mitotic activity.