Naturalistic data are a useful source for language acquisition research. Recently, the emphasis has been placed on using denser corpora, in order to capture a more accurate picture of child language development. However, working with large amounts of data raises resource issues, since it is time-consuming to record and to transcribe. In this article, I focus on what would be the ideal duration of a naturalistic recording for it to be considered a representative enough sample of children's linguistic behaviours to observe the acquisition of words and sounds. Some of the results suggest that 30 minutes of recording may be enough to capture these specific developments, but these results are discussed in the perspective of what an ideal session could be.

This article focuses on whether determiner omission by two-year-old children isconstrained at the level of the prosodic foot or whether it is a function of thedifferent levels of the prosodic hierarchy. Nine French-speaking children aged2;0 to 2;7 were asked to repeat 54 four- or five-word sentences of the form“Pronoun V NP” with three conditions: a) det + monosyllabicnoun; b) det + bisyllabic noun; c) det + monosyllabic adjective+ monosyllabic noun. The results show 1) more determiner omission incondition b than in a; 2) more determiner omission in c than in b. It is shownthat determiner omission is not accounted for by a low-level stress-alternationconstraint and that the level of prosodic structure to which the determiner isattached plays a role in determiner omission.