The Dog that Didn’t Bark

Inductive reasoning involves generalizing from samples of evidence to novel cases. Previous work in this field has focused on how sample contents guide the inductive process. This chapter reviews a more recent and complementary line of research that emphasizes the role of the sampling process in induction. In line with a Bayesian model of induction, beliefs about how a sample was generated are shown to have a profound effect on the inferences that people draw. This is first illustrated in research on beliefs about sampling intentions: was the sample generated to illustrate a concept or was it generated randomly? A related body of work examines the effects of sampling frames: beliefs about selection mechanisms that cause some instances to appear in a sample and others to be excluded. The chapter describes key empirical findings from these research programs and highlights emerging issues such as the effect of timing of information about sample generation (i.e., whether it comes before or after the observed sample) and individual differences in inductive reasoning. The concluding section examines how this work can be extended to more complex reasoning problems where observed data are subject to selection biases.