Using a tissue microarray cohort of 300 breast cancers and 84 samples of normal breast epithelium, we analyzed HER2/neu expression and compared traditional clinical (manual) scoring with a recently developed system for the quantitative measurement of immunohistochemical stains (AQUA). As expected, both methods identified a population (10-15%) of high-HER2-expressing tumors with poor 30-year disease-related survival. Using AQUA analysis, we found that normal epithelium expresses a low but detectable level of HER2 and that 17.5% of tumors exhibit similar low-level HER2 expression. This low group was not definable by manual scoring. Surprisingly, HER2-normal tumors were as aggressive as HER2-overexpressing tumors. Our studies suggest that in situ quantitative measurement of HER2 stratifies breast tumors into three expression levels: normal, intermediate, and high, where both normal and high levels are associated with a worse outcome.