A sample lacking variance was judged less probable than a variable sample and a sample representing the upper half of the population was judged less likely than one representing both. As range and mean approached an ideal, samples appeared more probable. A hierarchical model of sample cues seems appropriate. (CPT)