Abstract This study focuses on errors in estimations of age at which alcohol and tobacco are used for the first time. The data come from a 5-year longitudinal study with three measurements. Self-reports about age of first use at the baseline measurement were compared with similar self-reports at two follow-up surveys. Adolescents were more likely to report a higher age of first use at follow-up measurements. Those who at the baseline measurement reported having smoked ( n = 338) or consumed alcohol ( n = 523) 61.7% and 89%, respectively, underestimated their years of use. By comparison with estimations at the first and third measurement, 13.6% for smoking and 4.6% for drinking were consistent about their age of first use. Self-reports about the age of onset at the baseline measurement were correlated with frequency and intensity of tobacco and alcohol use 5 years later to assess the predictive power of age of onset for later use. With one exception (correlation with intensity of alcohol use 5 years later, r = .14) no significant correlation was found. The results show that the concept age of first use should be utilized with caution for two reasons: 1. (1) the reliability of assessment is insufficient, and 2. (2) correlations of different estimates with actual frequency and intensity of consumption at the third wave are inconsistent. Explanations for errors in measurement, and recommendations for improvement, are discussed.