The Revised National Tuberculosis Control Programme (RNTCP), which incorporated the WHO DOTS strategy was introduced in India in the mid-1990s. An operational research project was conducted between 1996 and 1998 to assess the needs and perspectives of patients and providers in two chest clinics in Delhi, Moti Nagar and Nehru Nagar, during the introduction of the new strategy. This paper reports on the findings of the project, concentrating on information collected from 40 in-depth interviews with patient defaulters and from non-participant observations in clinics and directly observed treatment centres. In Moti Nagar chest clinic, 117 of 1786 (6.5%) patients and 195 of 1890 (10%) patients in Nehru Nagar left care before their treatment was complete. It was argued that the reasons for default stem from a poor correlation between patient and programme needs and priorities, and from particular characteristics of the disease and its treatment. Patient needs that were not met by the health system included convenient clinic timings, arrangements for the provision for treatment in the event of a family emergency and provision for complicated cases like alcoholics. The problems facing the provider were poor interpersonal communication with the health staff, lack of attention and support at the clinic, difficulty for patients to re-enter the system if they missed treatment and, in certain areas, long distances to the clinic. Problems related to diseases were inability of the staff to deal with drug side-effects, and patients' conception of equating well-being with cure. Simple, practical measures could improve the provision of tuberculosis (TB) treatment: more flexible hours, allowances for poor patients to reach the clinics and training health care staff for respectful communication and monitoring drug side-effects. The findings indicate a need to rethink the label of 'defaulter' often given to the patients. The important areas for future operational research is also highlighted.