A considerable amount of data has been collected showing the association of high-grade prostatic intraepithelial neoplasia (HGPIN) with adenocarcinoma of the prostate, and many studies have yielded results that suggest that HGPIN is a precursor of carcinoma. A few studies have indicated that HGPIN may, in some cases, be a sequela of prostatic adenocarcinoma. We examined the proliferative indices of HGPIN, carcinoma, and benign prostatic epithelium by computer-aided counting of Ki-67-positive nuclei in 15 cases in which HGPIN and carcinoma were in close proximity. There were 13 radical prostatectomy specimens with prostate cancer and two cystoprostatectomy specimens with both transitional cell carcinoma and prostatic adenocarcinoma. First, we showed the accuracy of the computer-aided counting method compared with direct counting through the binoculars of the microscope. Then proliferative activity was assessed for each case by picking the two areas of carcinoma, the two areas of HGPIN, and the one area of benign epithelium with the greatest density of carcinomatous, dysplastic, and benign Ki-67-positive nuclei, respectively. The total number of nuclei and the number of positive nuclei were counted. Basal cells were not counted. The mean proliferative index was higher for cancer (caindex, average 0.054) than for HGPIN (pinindex, average 0.048) (P < .05). We found that the 15 cases fell into two distinct groups. The average ratio of pinindex to caindex (pinindex/caindex) was lower in group 1 (0.72) than in group 2 (1.54) (P=.17), and when the results were corrected for the nonzero gamma-intercepts of the regression lines of pinindex versus caindex, the ranges were widely separated, and the difference between the means was statistically significant (0.15 v 0.62; P < .0001). A greater subjective similarity between the nuclear features in the HGPIN and those of the corresponding carcinoma was noted for the cases in group 2. The average value of bngnindx was 0.014. The value of bngnindx did not correlate with either caindex or pinindex. We conclude that there may be two types of lesions with the morphological appearance of HGPIN and that they may have different relationships to carcinoma. Computer-aided counting of digitized microscopic images is both labor-saving and as accurate as enumeration directly through the binoculars of the microscope.