Proceedings of the Annual Meeting of the Cognitive Science Society

Inductive generalization involves extending knowledge fromsparse samples of evidence to arrive at broad conclusions.Most of the research in this area has focused on generalizationfrom sparse samples of positive evidence (cases known to shareproperties with known cases; e.g., birds have hollow bones).Much less is known about generalization from samples ofnegative evidence (cases known to lack the propertiesattributed to known cases; e.g., bats do not have hollow bones).This paper reports the results from three experiments thatexamined factors that were believed to influence adults’evaluation of negative evidence. Experiment 1 showed thatwhen selecting among samples most useful for teaching abouta particular category, participants (N=36) preferred sampleswith negative evidence rather than those with single, oradditional, positive evidence. Experiment 2 revealed thatparticipants (N=25) preferred samples with negative evidencethat included a closer (rather than more distant) taxonomicmatch with the category in question. Finally, Experiment 3revealed that adults (N=52) only preferred samples thatprovided a close match when evidence was provided by acompetent informant. Overall these results emphasize theimportant role of pragmatic expectations when reasoning aboutsamples that include negative evidence.