In two experiments with adults (N = 126), we examined the influence of sampling procedure on inductivegeneralization. In predicate sampling, participants learned the category identity of individuals known to pos-sess some property. In subject sampling, individuals selected for category identity were discovered to possessa novel property. In both experiments, sampling procedure influenced induction. Predicate sampling resultedin very narrow generalization, whereas subject sampling yielded a fairly high and constant rate of projection.Differences in confidence of generalizations were also observed. Conditions in which evidence was describedas randomly sampled from a collection of animals yielded a consistent decrease in projections as predicted bysimilarity-based models. The results are presented as support for an evidence-based view of induction.
展开▼