In a cross sectional survey within a community trial of BCG efficacy evaluation in Brazil, trained teams inspected children's upper arms and obtained information on BCG vaccination from guardian letters and vaccination cards. Nurses re-examined the sub-sample of children blindly. High agreement was found between the two scar readings (Kappa = 0.839). High sensitivity and low specificity was observed when guardian or card information was the gold standard. Sensitivity remained high when guardian and card information agreed. When disagreement occurred, sensitivity remained high and specificity was very low. BCG scar is a good indicator of BCG vaccination.