A randomized controlled trial was carried out in Pakistan in 1999 to establish the effectiveness of the direct observation component of DOTS programmes. It found no significant differences in cure rates for patients directly observed by health facility workers, community health workers or by family members, as compared with the control group who had self-administered treatment. This paper reports on the social studies which were carried out during and after this trial, to explain these results. They consisted of a survey of all patients (64% response rate); in-depth interviews with a smaller sample of different types of patients; and focus group discussions with patients and providers. One finding was that of the 32 in-depth interview patients, 13 (mainly from the health facility observation group) failed to comply with their allocated DOT approach during the trial, citing the inconvenience of the method of observation. Another finding was that while patients found the overall TB care approach efficient and economical in general, they faced numerous barriers to regular attendance for the direct observation of drug-taking (most especially, time, travel costs, ill health and need to pursue their occupation). This may be one of the reasons why there was no overall benefit from direct observation in the trial. Provider attitudes were also poor: health facility workers expressed cynical and uncaring views; community health workers were more positive, but still arranged direct observation to suit their, rather than patients', schedules. The article concludes that direct observation, if used, should be flexible and convenient, whether at a health facility close to the patient's home or in the community. The emphasis should shift in practice from tablet watching towards treatment support, together with education and other adherence measures.