Most studies of children’s spontaneous speech and of parent-child interaction have collected data in dyadic contexts, those in which activities designed to elicit dyadic talk (book reading, toy play) were prescribed or encouraged and then observed. These activities are constrained in a number of ways: They may not be common occurrences in some homes, where parent-child relations do not revolve around play; and they are by nature biased toward child interests-the books and the toys typically chosen are designed for children and are limited to child-appropriate topics and entities. Although

some work has been done on adult-child inter-action in more ecologically valid care-taking situations such as bathing, dressing, and diapering (e.g., Aukrust, 1995, 1996; Masur, 1987, 1989), this too has tended to be dyadic in participation structure. Although dyadic talk is relatively easy to observe, transcribe, and code, it has obvious limitations-both as a source of information about children’s language environments, which are often multiparty, and as a context for observing certain kinds of interactive phenomena that may only emerge in larger groups.