The difficulties and obstacles encountered in research in the analysis and interpretation of writing conference talk are theoretically important. These dilemmas may serve as markers to help researchers see how the routines of research and the interpretations of findings are integrally related. The ways in which talk "rearranges" problems should be of interest to those who concern themselves with the relationship of oral discourse to learning to write. Discourse analysis of classroom interaction is comprised of two phases of research: data collection, and data analysis. The use of recording equipment, and the kind of equipment used, will affect the kind of information obtained from classroom settings, as student comments demonstrate. Related to this is the need to gather contextualizing information, since student-teacher talk can be misunderstood outside of context. Often, teacher and student motives can only be attained through personal interviews, as examples show. The timing of interviews, furthermore, affects what is discovered in the data. The dilemmas of when and how to collect data affect all aspects of a study and affect what the data can teach. Other phenomena of importance when considering conferences are "adjacency pairs" and the question of who is "steering," (invaribly the teacher, as illustrated in one scenario). In sum, all of these obstacles make the responsibility of interpretation a challenging process replete with methodological paradoxes. (HB)