After establishing the importance of knowledge of settlement rates, this article first shows that different research questions can yield different settlement rates. Using data gathered from about 3,300 federal cases in the Eastern District of Pennsylvania (EDPA) and the Northern District of Georgia (NDGA), differing measures of settlement emerge depending on whether one is interested in (1) settlement as a proxy for plaintiffs' litigation success, or (2) settlement as a measure of litigated disputes resolved without final adjudication. Using settlement as a proxy for plaintiff success, we estimate the aggregate settlement rate across case categories in the two districts to have been 66.9 percent in 2001–2002. Regardless of the method of computing settlement rates, no reasonable estimate of settlement rates supports an aggregate rate of over 90 percent of filed cases, despite frequent references to 90 percent or higher settlement rates. The aggregate rate for the EDPA alone was 71.6 percent and for the NDGA alone was 57.8 percent, suggesting significant interdistrict variation, which persists even within case categories. We report separate settlement rates for employment discrimination, constitutional tort, contract, and tort cases in the two districts. The highest settlement rate was 87.2 percent for tort cases in the EDPA and the lowest was 27.3 percent for constitutional tort cases in the NDGA. Our results suggest a hierarchy of settlement rates. Of major case categories, tort cases tend to have the highest settlement rates, then contract cases, then employment discrimination cases, followed by constitutional tort cases. Attorney fee structure and the nature of the parties may explain settlement rate variation. Our findings provide no evidence of a material change in aggregate settlement rates over time.