Previous studies have suggested that the bacille Calmette-Guérin (BCG) vaccine may have a non-specific beneficial effect on childhood survival in areas with high mortality. We examined whether BCG-vaccinated children with a BCG scar or a positive tuberculin reaction had better survival than children without such reactions. As part of an ongoing two-dose measles vaccine trial for which children were recruited at 6 months of age, we examined 1813 children for BCG scar at 6 months of age and 813 BCG-vaccinated children were skin-tested for delayed hypersensitivity to tuberculin, tetanus and diphtheria. We found that BCG-vaccinated children with a BCG scar had significantly lower mortality compared with BCG scar-negative children, the mortality ratio in the first 12 months of follow-up being 0.41 (0.25-0.67). BCG-vaccinated children with a positive tuberculin test had a mortality ratio of 0.45 (0.24-0.85) compared with tuberculin negative children. These results were unchanged by control for potential confounders or using different cut-off points for a tuberculin-positive response. Exclusion of dead children who had HIV antibodies did not modify the estimate (mortality rate (MR)=0.46 (0.23-0.94)). After censoring for tuberculosis (TB) exposure at home, the mortality ratios for having a scar and being tuberculin-positive were 0.46 (0.27-0.79) or 0.42 (0.21-0.84), respectively. Children positive to tetanus or diphtheria in the skin test had the same mortality as children not responding to these vaccine-related antigens. Thus, BCG scar and a positive tuberculin reaction were associated with better survival in early childhood in an area with high mortality. Since nothing similar was found for responders to diphtheria-tetanus-pertussis (DTP) vaccine, and the effect could not be explained by protection against tuberculosis, the effect of BCG vaccination could be due to non-specific immune-stimulation protecting against other infections.