The importance of reinfection as a cause for recurrence of tuberculosis is unclear and has potential public-health implications. We systematically searched published material for studies using DNA fingerprinting to provide data on the issue. Very few studies were designed for that particular research objective and/or report on a sufficient number of observations. Differences in methods--eg, case-definitions--seriously hamper comparisons between studies. The proportion of recurrences due to reinfection ranged between 0% and 100%; however, this figure cannot be a useful indicator since the two causes of recurrence--relapse and reinfection--are essentially independent. Only one study provides an estimate of the incidence of recurrence due to reinfection, indicating its importance for HIV-infected patients in an environment with an unusually high tuberculosis incidence. We argue that apart from extreme situations like this one the problem of recurrence of tuberculosis due to reinfection has few implications for tuberculosis-control programmes.