The assumption that question-answer sequences consist of two moves was examined using data from recordings of second-grade classrooms in Austin, Texas. Interactions initiated by children's spontaneous questions in informal, small group settings were evaluated. Twenty-nine dialogue samples are presented and analyzed. Four issues in the study of interaction are discussed: the selection of data for analysis, the use of "null set" in discourse, recognition of a "notable absence," and the usefulness of a two-move unit for describing events in conversation. The data indicate that answers are negotiated, and that three moves, rather than two, are required to accomplish an inquiry sequence: a question, a reply, and an evaluation. The status of a reply as answer or nonanswer is not determined until a questioner contributes his evaluation. A replier therefore does not know how his reply fits into an on-going conversation until the questioner reacts to it. (SW)