There is now a large amount of evidence showing that mammographic densities are an indicator of increased risk of breast cancer. There is as yet no generally agreed upon and recognized method of classifying these densities, although the available evidence shows that quantitative description of densities creates larger gradients of risk than Wolfe's classification and larger risk gradients than most other risk factors for breast cancer. It seems likely that improved methods of describing densities quantitatively, and possibly other methods of characterizing the tissue changes that are responsible for the densities, will allow greater discrimination. However, it is already clear that breast cancer develops in a large number of women who do not have radiologic changes indicating increased risk, and that it is unlikely that mammographic pattern, or any other risk factor for breast cancer identified to date, will be useful for the selection of women for mammographic screening. Although mammographic densities are associated with an increased risk of developing histologic changes that are risk factors for breast cancer, the histologic feature most consistently associated with mammographic densities is stromal fibrosis. We suggest that the relation between stromal fibrosis and risk of breast cancer can be explained by the known actions of a variety of growth factors that are thought to play a role in a number of aspects of breast development and carcinogenesis. The association between mammographic densities and several other risk factors for breast cancer suggests that these factors may also modulate the activity of growth factors in breast tissue, and that this may be the means by which they influence breast cancer risk. Further research is needed to determine whether differences in the activity of growth factors in breast tissue can be found in association with radiologic and other risk factors for breast cancer. The available evidence indicates, therefore, that mammographic parenchymal patterns do, at least in part, meet the criteria outlined in the introduction of this paper. Some mammographic appearances are associated with a substantial increase in the risk of breast cancer, and, as shown by observations on the effects of hormone use, are capable of change. Mammographic densities have also been found to be associated with biochemical characteristics of possible relevance to carcinogenesis. The appearances that are related to risk may, therefore, be most useful as a means of investigating the etiology of breast cancer and of testing hypotheses about potential preventive strategies.