Previous studies have shown that quantitative, histopathologic features obtained from a carefully selected area in the tumor section ("selective" approach) have a strong prognostic value in breast cancer. On the other hand, it was found that mean nuclear volume estimation in the whole area of the tumor section by means of "unbiased" stereologic techniques is of great value in predicting the clinical outcome as well. In the present study the results of the two different (ie, selective and random, systematic) sampling methods in assessing mean nuclear volume have been compared as to their intraobserver and interobserver reproducibility in 22 invasive breast cancer cases. The mean nuclear volume (nuclear vv) was assessed both in the most atypical area (AREA) (selected on morphologic criteria) and in the whole tumor section (TOTAL). Furthermore, the correlation with mean nuclear (profile) area (MNA) was studied. Mean nuclear (profile) area was determined in the AREA only. With bivariate correlation analysis the two sampling methods showed high correlation for the nuclear vv values (range of the correlation coefficient, 0.92 to 0.97). There were no systematic intraobserver differences between the different sampling methods. The results of observer 1 showed higher values, both with the selective and random systematic sampling methods. However, these systematic interobserver differences were small (< 9% of the average value of nuclear vv), much smaller than the variation between the tumors (which was > 60%). The time required for assessments in the AREA was less than that required for the determinations in the TOTAL (average, 10 v 20 minutes) in spite of the similar sample size. This is understandable, as in a sclerotic tumor many fields of vision do not contain cancer nuclei. The time required for MNA determinations in the AREA was longer than for nuclear vv assessments in the AREA (15 v 10 minutes). Nuclear vv and MNA (both assessed in the AREA) were (log distributed) significantly correlated (r = .77). Thus, nuclear vv determination in the AREA is the fastest method, and it is also well reproducible and strongly correlated with nuclear vv assessed in the TOTAL. In invasive breast cancer assessments in the whole tumor section can be used if delineation of the measurement area cannot be done easily. In small areas with a limited number of nuclei (eg, microinvasive parts) MNA can be easier to assess than nuclear vv. Further studies are required to compare and evaluate the prognostic value of nuclear vv and MNA.