Is it possible to construct valid explanations of behavioral phenomena from differences found among naturally existing groups? Many of psychology's cherished findings in such domains as intellectual performance and temperament rest on differences of this kind. A venerable and rich literature on selection and selection effects both reminds us that naturally existing groups might or might not derive from a common parent population and warns that valid decisions between these alternatives are often jeopardized by the very methods on which those decisions rest. We examine these matters within the context of intellectual performance, illustrate key points with an analysis of data, and examine the implications for current research on differences in ability levels.