Studies of doctor-patient communication, although leading to diverse findings, have not lent themselves to replication and also have not captured important features of medical discourse. Quantitative methods alone do not deal with the complexities of medical encounters, usually are not helpful in analyzing the social context of discourse, do not clarify underlying themes and structures, and are costly and tedious to use. With qualitative methods, the selection of discourse for analysis is not straightforward, quality of interpretation is difficult to evaluate, and textual presentation is not clear-cut. Several criteria of an appropriate method offer reasonable compromises in dealing with medical discourse: 1) discourse should be selected through a sampling procedure, preferably a randomized technique; 2) recordings of sampled discourse should be available for review by other observers; 3) standardized rules of transcription should be used; 4) the reliability of transcription should be assessed by multiple observers; 5) procedures of interpretation should be decided in advance, should be validated in relation to theory, and should address both content and structure of texts; 6) the reliability of applying interpretive procedures should be assessed by multiple observers; 7) a summary and excerpts from transcripts should accompany the interpretation, but full transcripts should also be available for review; and 8) texts and interpretations should convey the variability of content and structure across sampled texts. An ongoing study applies these criteria to research on ideology and social control in medical encounters.