Previous research has suggested that very young children learning English adhere quite rigidly to a grammatical constraint on the possible contexts for contraction of "want" and "to" into the reduced form "wanna". Two elicited production studies reported here suggest that young children do produce "wanna" in illicit contexts. One study identifies one factor that offers a partial explanation for the conflicting results, but this factor alone does not completely explain the earlier nearly categorical results observed for children. Our findings cast doubt on the claim that children have innate knowledge of a categorical constraint against contraction over a trace or of any absolute constraint on the occurrence of "wanna", and highlight the need for further developmental work. (Contains 3 figures, 3 footnotes and 5 tables.)