A Delphi study was conducted with a multi-disciplinary group of experts attending a conference on mental illness in primary care settings. Respondents' priorities for research which could maximize benefits for service developments in this field were elicited by asking them to distribute hypothetical annual research budgets among (a) 6 mental health priority areas, and (b) 7 subjects 'needing' investigation within the primary care priority area. The respondents' three top-ranked priority areas for research were: (1) primary care; (2) the elderly mentally ill; and (3) chronically disabling mental illness. Their three top-ranked subjects for investigation within the primary care priority area were: (1) the effectiveness of treatment measures; (2) the problems of mental health presenting to primary care services; and (3) the training requirements for family doctors in psychiatric skills. The results are discussed with reference to the method used and to the economic concepts of programme budgeting and marginal analysis.