1. The ability of hepatic microsomal metabolic stability assessments to predict in vivo clearance in rat has been retrospectively evaluated for 1,163 compounds from 48 programmes of chemistry. Using a simple binary classification system, the in vivo clearances of approximately 64% of the compounds were correctly classified. 2. About 24% of compounds were metalbolically stable yet had clearance greater than half of liver blood flow in vivo. This might be expected as microsomes only contain a limited number of fully functioning drug-metabolizing enzymes and cannot be expected to account for extrahepatic or non-metabolic clearance processes. 3. About 13% of compounds had in vivo clearances of less than half liver blood flow despite being classified as metabolically unstable. Despite overcoming metabolic instability, these compounds had other undesirable properties and were generally more highly bound to plasma proteins, had smaller volumes of distribution (and shorter half-lives despite their clearance) and were more inhibitory against the major human cytochrome P450s. 4. Taking plasma protein binding into consideration reduced the proportion of misclassified low-clearance compounds but did not improve the overall success appreciably. Somewhat surprisingly, human microsomes were nearly as effective as rat microsomes at classifying rat in vivo clearance.