This paper examines the evidence used to support the claim that children initially do not encode new referents like adults do (e.g., Maratsos 1974; Warden 1976; Emslie and Stevenson 1981; Hickmann et al. 1996). It argues that a better understanding of the information structure of the target language forces a reinterpretation of previous experimental results in the sense that children comply with the adult requirements more than has been assumed. The discussion is based on a detailed examination of how new information is encoded in spoken French. (Contains 11 notes).