What features make samples seem representative?

In 3 experiments, Ss were 1,794 high school students, university students, Israeli army recruits, and Americans studying Hebrew. When asked to judge which of 2 observations was more likely to be sampled from a given bell-shaped distribution, Ss correctly judged the more extreme to be less likely. When asked to judge which of 2 samples was more likely to be randomly drawn from that population, Ss did not rely on extremity but appeared to view the sample as a gestalt and to order it in terms of configural properties, which were only coincidentally related to the extremeness of the members. Thus, a sample with no variance was judged less likely than one that was variable; a sample representing only the upper half of the population was judged less likely than one representing both halves; and samples appeared more likely the closer their range and mean were to some ideal. These properties seem to have been considered hierarchically. As a result, an almost uniquely defined sample, the "prototypical," can be constructed, which is judged at least as likely as any other. Findings yield a description of representativeness in terms of a hierarchy of simple sample features. (6 ref) (PsycINFO Database Record (c) 2006 APA, all rights reserved).