Several conclusions about measuring adherence can be drawn. Probably the best approach is to use multiple measures, including some combination of urine assays, pill counts, and detailed patient interviews. Careful monitoring of patient behavior early in the regimen will help predict whether adherence is likely to be a problem. Microelectronic devices in pill boxes or bottle caps have been used for measuring adherence among patients with tuberculosis, but their effectiveness has not been established. The use of these devices may be particularly troublesome for some groups such as the elderly, or precluded for those whose life styles might interfere with their use such as the homeless or migrant farm workers. Carefully designed patient interviews should be tested to determine whether they can be used to predict adherence. Probably the best predictor of adherence is the patient's previous history of adherence. However, adherence is not a personality trait, but a task-specific behavior. For example, someone who misses many doses of antituberculosis medication may successfully use prescribed eye drops or follow dietary recommendations. Providers need to monitor adherence to antituberculosis medications early in treatment in order to anticipate future problems and to ask patients about specific adherence tasks. Ongoing monitoring is essential for patients taking medicine for active tuberculosis. These patients typically feel well after a few weeks and either may believe that the drugs are no longer necessary or may forget to take medication because there are no longer physical cues of illness. Demographic factors, though easy to measure, do not predict adherence well. Tending to be surrogates for other causal factors, they are not amenable to interventions for behavior change. Placing emphasis on demographic characteristics may lead to discriminatory practices. Patients with social support networks have been more adherent in some studies, and patients who believe in the seriousness of their problems with tuberculosis are more likely to be adherent. Additional research on adherence predictors is needed, but it should reflect the complexity of the problem. This research requires a theory-based approach, which has been essentially missing from studies on adherence and tuberculosis. Research also needs to target predictors for specific groups of patients. There is clear evidence of the effect on adherence of culturally influenced beliefs and attitudes about tuberculosis and its treatment. Cultural factors are associated with misinformation about the medical aspects of the disease and the stigmatization of persons with tuberculosis. Culturally sensitive, targeted information is needed, and some has been developed by local tuberculosis programs.(ABSTRACT TRUNCATED AT 400 WORDS)