4 studies investigate how differential input affects preschoolers' abilities to learn novel color words. 3-, 4-, and 5-year-old children saw objects in novel shapes and colors and heard a novel color label for the object. Labels were presented through ostensive definition (e.g., "See, it's mauve"), corrective linguistic contrast (e.g., "See, it's not purple; it's mauve"), or an inclusion statement (e.g., "See, it's mauve; it's a kind of purple"). 4- and 5-year-old children interpreted the novel word as a shape term when ostensive information was provided but as a color term when additional information, either contrastive or inclusive, specified a relation between the novel term and a known label for that color. Furthermore, children who consistently interpreted the novel word as a color word tended to treat the novel and known labels as mutually exclusive color terms if they heard contrastive information, whereas they tended to treat the words as hierarchically related if they heard inclusion information. 3-year-olds generally did not make use of either type of information in determining the semantic domain of the novel word or the relation between terms.