The creation of new words through the novel combination of English words or morphemes (e.g., "map ball" to refer to a globe) was studied and compared in 40 preschool children, 40 grade school children, and 40 adults. These made-up words or lexical innovations were collected while subjects attempted to name pictured objects and were evaluated in terms of incidence, communicative effectiveness, novelty, semantic accuracy, and certain linguistic characteristics. Communicative effectiveness was established by asking naive judges to guess the intended referent of the made-up word. It was expected that the preschool children's innovations would be wild and referentially opaque. Instead, the preschool children's innovations were as frequent and as communicatively effective as those of the grade school children and adults. Grade school children produced the highest proportion of innovations with semantic inaccuracy and contamination, whereas preschool children constructed more innovations containing redundant elements. This finding suggests that preschoolers may construct made-up words from a limited set of highly familiar terms, whereas grade schoolers may rely more on partially known terms. Both groups of children are able to circumvent gaps in their lexicons by creating communicatively effective lexical innovations as often and as well as adults. (Author/MSE)