We propose inductive confidences as a metacognitive concept assessing the degree of the certainty of inductive inferences in property induction, and further distinguish between relevance-based and similarity-based property inductions according to the difference in inductive confidences. We predict that relevance strength will affect inductive confidences and inferences in different ways. We conducted two experiments to examine the influence of relevance strength on inductive confidences and inferences in both property inductions. The prediction is confirmed by the following findings. Relevance-based inductions showed higher inductive confidences than similarity-based inductions. In both relevance-based and similarity-based inductions, inductive inferences increased with relevance similarity or overall similarity, whereas inductive confidences accorded with relevance strength regardless of similarity. Thus, there was the distinction between inductive confidences and inferences. Moreover, in relevance-based inductions, the influence of relevance similarity on inductive inferences depended on relevance strength, and relevance strength affected inductive inferences and confidences in the different ways. These findings will arouse further research on metacognition in inductive reasoning in the future.Keywords: property induction; relevance strength; relevance similarity; inductive inference; inductive confidence.The distinction between inductive confidences and inferences 3 Inductive reasoning is that people project information from known cases to the unknown (Heit, 2000;Hayes, Heit, & Swendsen, 2010;Kemp & Jern, 2013). One main research topic in inductive reasoning is category-based property induction (Hayes, Heit, & Swendsen, 2010;Kemp & Jern, 2013;Medin, et al., 2003;Osherson, et al., 1990;Sloman & Lagnado, 2005). This typically involves making an inference about the properties of some conclusion category, based on knowledge of the properties of some premise category or set of categories (Hayes, Heit, & Swendsen, 2010). A main function of categories is to support inductive inferences (Anderson, 1991;Bright, 2010;Hayes, Heit, & Swendsen, 2010;Rosch & Mervis, 1975). In category-based property induction, according to that some premise category has a property (e.g. Polar bears have thick subcutaneous fat), people infer the probability that a conclusion category has the property (e.g.Penguins have thick subcutaneous fat). The inferred property is the generalized property, and the inferred probability is inductive inference or strength. There are three kinds of main theoretical accounts for category-based property inductions: the similarity accounts, the knowledge accounts, and the integration accounts such as the Bayesian model and structured statistical models (Bright, 2010;Bright & Feeney, 2014;Feeney, Coley, & Crisp, 2010;Hayes, Heit, & Swendsen, 2010;Heit, 1998Heit, , 2000Kemp & Tenenbaum, 2009;Medin, et al., 2003;Sloman & Lagnado, 2005). All these accounts concern only inductive inferences, but neglect ...