Abstract

lt is hypothesized that people use formally incorrect deductive procedures, and this is sometimes advisable. The particular "prototypicaiity heuristic" investigated is: to determine validity of a proof, (a) work out an example, and (b) pick a "good" rather than arbitrary example. An interaction was predicted betv;een validity of inference and proto-typicality of example. Experiment 1, although quite sensitive to "calibration' variables, does not reveal the interaction in reaction times. However, Experiment 2, in which subjects' time was limited, seems to elicit the interaction in subjects' first-trial confidence judgments; the explanation proposed is that subjects then decide they must shift to a "quick and dirty" heuristic. Experiment 3 provides some preliminary evidence that subjects' 'metaheuristic"