The aim of this study was describe a training and calibration exercise for dental caries and to discuss the use of validity and reliability statistical tests. Eight dentists examined seventeen 12 year-old schoolchildren using World Health Organization (1997) criteria. Data analysis included calculation of the DMF-T index, the percent agreement, and the kappa scores on a tooth-by-tooth basis. Inter-examiner agreement was tested against a gold standard examiner. Scores for measures of agreement calculated on a tooth-by-tooth basis were lower than those calculated for the whole mouth. The training and calibration exercise is a crucial step in cross-sectional epidemiological surveys, and the kappa test calculated on a tooth-by-tooth basis is the appropriate measurement to test agreement between examiners for dental caries.