Although much research has investigated children's use of metaphoric language, methodological concerns raise questions about the conclusions, and it remains unclear whether preschoolers can produce metaphors. These studies employed a new methodology to test children's ability to produce metaphors incorporated into metaphoric compounds. In two studies, 59 children aged 2;8–4;3, 63 children aged 4;4–6;1, and 34 adults participated in elicited production tasks. In Study 1, subjects in the COMPOUND condition corrected a puppet's incorrect compound labels for pictures that had metaphoric resemblances to other objects (e.g. ‘leaf-bug’ for a bug shaped like a stick). Subjects in the NON-METAPHORIC condition heard incorrect compounds describing pictures without obvious metaphoric resemblance (e.g. ‘leaf-bug’ for a round black beetle). Children in the REVERSAL condition heard compounds with nouns reversed (e.g. ‘bug-leaf’ for the stick-bug) to discover whether children distinguished between the literal and metaphoric labels. Study 2 provided an additional test of children's metaphoric–literal distinction. Results showed that children as young as 3;0 produced intentional, appropriate metaphors incorporated into compound nouns when the stimuli and puppet's labels primed recognition of metaphoric similarity and compound production. Moreover, children showed evidence of a distinction between literal and metaphoric labels. The data show that preschool children have an early ability to use metaphoric language but that significant developmental change occurs between the ages of 3;0 and 5;0 as well as beyond 5;0. Additionally, metaphoric language in preschoolers is not limited to single-word renamings.