MIB-1 labeling index has been used as a complementary method to differentiate better and worse prognostic groups of astrocytic tumors. However, its reproducibility has been doubted because of the interlaboratory and interobserver variability. We studied the interobserver reproducibility of four manual counting methods of MIB-1 labeling index, including direct counting, area fraction, line intersection with a graticule, and line intersection without a graticule to estimate the total number of the cells. Using line intersection estimate, either with or without a graticule, yielded a worse result. Those MIB-1 labeling indices of astrocytomas were overlapping with those of anaplastic astrocytomas and glioblastomas. Besides, the intraclass correlation coefficient was smaller, and the coefficient of variation was greater than that attained when using a graticule and counting the total number of cells either directly or by area-fraction estimate. The level of interobserver agreement using an MIB-1 labeling index cutoff value of 11.0 was moderate to substantial when applying the line-intersection estimate, and it was almost perfect when applying the direct counting or the area-fraction estimate. The interobserver reproducibility and agreement of MIB-1 labeling index in astrocytic tumors were different depending on the counting methods. Direct counting with a graticule yielded the best result. Before establishing reference standards for staining and counting, the reproducibility of MIB-1 labeling index should be verified.