Research into doctor patient communication and their roles is reviewed. Two series of semi-structured interviews with general practitioners (e.p.e) and patients separately are reported. In the first, 4 g. p. s each selected 6 patients for 3 groups; terminally ill with knowledge, terminally ill without, and chronic controls. Content analysis of patient transcripts revealed differences in frequency of utterances classed as showing knowledge. Differences between groups and qualitative analysis of all interviews, verified g.p.s’ categorisation. Knowledge was not simply absent or present, since all groups show awareness, but differ in the degree to which the nature of illness is openly acknowledged. One g.p. reported routinely informing patients, but others did not, although agreed this is occasionally appropriate. A typology of methods g.p.s and patients use to control information exchange is proposed. g.p.s' role formulations appear to underpin their reported communicative behaviour. 22 g.p.s were interviewed in the second series. Again they were asked to select patients, but few did this within the time allotted: therefore analysis focused upon g.p.s. Content analysis of interviews splits g.p.s into "informers" and "non informers". Qualitative analysis of g.p.s’ reported communication strategies suggests that underlying this simple dichotomy is a continuum from revelation through passive response to concealment. Strategies are implemented by various tactics; e.g. euphemism, uncertainty, denial, evasion, prompting questions. Differences in reported intentions and behaviour are interpreted by recourse to g.p.s' descriptions of their role during terminal care. Those who conceal attempt to maintain the Parsonian curative physician role and are reluctant to classify patients as terminal. Those who reveal try to abandon this role for one modelled on that of counsellor. Nearly all experience role ambiguity and attempt to resolve this in ways predicted by role theory.