The purpose of this paper is to examine critically the evidence that atypical hyperplasia (AH) is a risk factor for breast cancer. First, we appraised studies that have examined the association between AH and breast cancer risk for their adherence to widely accepted standards for the conduct of research. Second, we examined the available evidence to determine the plausibility of an association between AH and breast cancer risk using the guidelines proposed by Bradford Hill. A total of 18 studies (11 cohort studies, two case-control studies, and five cross-sectional studies) were found that were published in the English language from January 1960 to March 1992 that examined the association of AH as a distinct entity and breast cancer risk. A systematic approach was adopted to examine the collected studies for their adherence to methodologic standards, which showed wide variation among studies. A meta-analysis was carried out, based on a total sample size of 182,980 women. Of 16 studies that gave point estimates of risk, 14 exceeded unity and 12 were significantly different from unity. The pooled estimate from all studies of the association between AH and breast cancer, gave an overall odds ratio (OR) of 3.67 (95 percent confidence interval = 3.16-4.26). The test of the hypothesis of homogeneous association was rejected (chi 2 = 151.6, df = 14, P < 0.0001), indicating significant variability among the ORs of individual studies. The conclusions from the application of the Bradford Hill criteria indicated strongly that AH is a risk factor for breast cancer.