BACKGROUND: In 1992, the Tomlinson Report recommended a shift from secondary to primary care, including specific primary care provision in accident and emergency (A&E) departments. Availability of short-term so-called Tomlinson moneys allowed a number of experimental services. A study of the experience of A&E-based staff is reported to assist general practitioners (GPs) and purchasers and identify areas for further research. AIMS: To find the number and scope of primary care facilities in A&E services in North Thames; to find factors encouraging or inhibiting the setting-up of a successful service; to examine the views of a range of A&E staff including GPs, consultants, and nurses; and to suggest directions for more specific research. METHOD: A postal questionnaire was sent to all North Thames A&E departments, and an interview study of staff in one unit was arranged, leading to a questionnaire study of all GPs employed in North Thames primary care services in A&E. This was followed by interviews of staff members in five contrasting primary care units in A&E. RESULTS: By mid-1995, at least 16 of the 33 North Thames A&E departments ran a primary care service. Seven mainly employed GPs, the others employed nurse practitioners (NPs). Problems for GPs included unclear role definition and their non-availability at times of highest patient demand. GPs' reasons for working in A&E sometimes differed from the aims of primary care in an A&E service. Staff interviews revealed differing views about their role and about use of triage protocols. Ethnicity data were being collected, but not yet being used, to improve service to patients. CONCLUSIONS: A number of benefits follow the introduction of primary care practitioners into A&E. Different models have evolved, with a variety of GP and NP staffing arrangements according to local ideas and priorities. There is some confusion over whether these services aim to improve A&E-based care or to divert it to general practice. Cost information is inadequate so far, though the use of GPs has shown the possibility of economy. Appropriate location of services requires clearer identification of costs. This may be possible for the proposed primary care groups.