A 9-month study of a 1-year-old child's acquisition of the pronunciation of "camera" is presented. The data show that while the child can articulate and perceive all the phonological segments of the adult form, she uses an idiosyncratic child-based form when she spontaneously draws from her lexicon to produce an utterance, systematically modifying it when she perceives the necessity to bring it closer to the adult convention. This suggests that she is using two forms: the dynamic, child-based form used for production and the stable adult-based form she uses for comprehension. This evidence supports the theory that children use both auditory and articulatory underlying representations. However, the theory that the child's variant forms are subject to output rules is rejected in favor of the idea that they are drawn straight from the child's lexicon. (MSE)