Summary Previous investigators have shown that there is a strong association between the fraction of fibroglandular tissue within the breast as determined by X-ray mammography (per cent density) and breast cancer risk. In this study, the quantitative correlation between per cent density and two objective magnetic resonance (MR) parameters of breast tissue, relative water content and mean T2 relaxation time, as investigated for 42 asymptomatic subjects. Using newly developed, rapid techniques MR measurements were performed on a volume-of-interest incorporating equal, representative portions of both breasts. X-ray mammograms of each subject were digitised and analysed semiautomatically to determine per cent density. (Kelsey et al., 1983).More sensitive risk assessment is provided, however, by quantitative methods of classification based on the relative fraction of fibrogandular breast tissue, known as the per cent density (PD), and also by increasing the number of classification categories to define individuals at highest risk more precisely (Brisson et al., 1982;Warner et al., 1992). A quantitative classification scheme using six categories with PD values of 0%, 0-10%, 10-25%, 25-50%, 50-75% and >75% has been applied recently in a nested case-control study of 332 pairs of women selected from the cohort of women in the mammography arm of the Canadian National Breast Screening Study (Miller et al., 1992a,b). A 6-fold difference in risk was found between the highest and lowest density categories, and the two upper categories accounted for 44% of incident cancers . This result is particularly significant because breast cancer susceptibility genes (Miki et al., 1994; Wooster et al., 1994) account only for approximately 5-10% of breast cancer incidence, and the established risk factors in combination are known to account for only 25-30% of breast cancer incidence (Seidman et al., 1982). The use of quantitative classification schemes indicates clearly that PD is a major risk factor for breast cancer.Semiautomated methods have been developed to calculate PD and other quantitative parameters as continuous variables from digitised mammograms (Byng et al., 1994). These methods substantially reduce the intra-and interobserver variability traditionally involved in classification, are quite insensitive to variations in mammography technique and are strongly correlated with the six-category classification of mammograms by radiologists.The ability to assess breast cancer risk by parenchymal pattern has several applications, e.g. as a tool to study breast cancer aetiology, or to monitor changes in breast cancer risk during interventional studies. Although most investigations of breast parenchymal patterns have used X-ray mammography to date, other non-hazardous imaging modalities can provide similar, and possibly complementary, risk assessment. For example, a qualitative four-category classification scheme has been developed for breast ultrasonography that correlates with the area of the breast occupied by mammographic density ...