Difficulties of vaccination documentation in developing countries lead to a dependence on maternal recollection of vaccination to evaluate tetanus toxoid (TT) vaccination. To date, the validity of this approach has not been assessed. Recollection of TT immunization in 1477 women was compared with carefully maintained records of immunization in Matlab, rural Bangladesh. Sensitivity, specificity and predictive values for maternal recall were calculated in a highly vaccinated population of women of reproductive age. Maternal recollection of the second TT dose was good, but considerably reduced if they had received more than two doses. Recollection was also reduced if the doses were administered more than a year prior to questioning. Up to 40% of women did not remember doses received 12-23 months earlier. Logistic regression evaluated the effect of confounding variables on recalling the number of TT doses received. Younger women, with fewer children and with a good vaccination status, tended to have a better recollection of the number of doses received. Maternal education, household characteristics, religion or family planning resumption within 1 year of the last delivery had no effect on the accuracy of recall. These results suggest that maternal recall may underestimate TT doses received a year before the date of questioning. Current methods to determine TT status by asking mothers at infant vaccination contacts may result in underestimating coverage.