In a category-based induction, knowing that a property is true of some category members leads one to conclude that the property is true of other category members. An example is: Cardinals have ulnar arteries. Therefore hawks have ulnar arteries. Recently, Osherson et al. (1990) demonstrated a number of phenomena involving category-based inductions, and proposed that these phenomena can be explained by variations in 2 processes: (a) the similarity between the premise category (cardinals in the above example) and the conclusion category (hawks in the above example), and (b) the degree to which the premise category “covers” (roughly, is similar to) instances of the lowest-level category that includes both the premise and conclusion categories (birds in the above example). The present paper traces the developmental course of the relevant phenomena and of the similarity and coverage processes that presumably underlie them. Studies 1 and 2 demonstrated that the inductions made by kindergartners are sensitive only to the similarity between the premise and conclusion categories. Studies 3 and 4 showed that second graders' inductions are sensitive to both premise-conclusion similarity and coverage, as long as there is no need actually to use a generated category that includes both the premise and conclusion categories. These developmental findings reveal an orderly process in the growth of category-based inductions, and also decompose the Osherson et al. model into 3 basic components that have not previously been explicitly distinguished.