Aims. To examine the reliability of self-reported age of first substance use experiences among national samples of adult and child respondents. Design. Survey responses from seven waves of the National Longitudinal Survey of Youth (NLSY) were examined. Participants. Adult and child NLSY respondents reporting age of first tobacco, alcohol, marijuana, cocaine and/or crack use during two or more survey interviews. Measurements. Four indicators of reliability: intraclass correlation coefficients (ICC), mean and absolute mean differences in reported age and reports consistent within 1 year. Findings. The adjusted mean ICC for all comparisons was 0.69. The adjusted mean difference in self-reported age of first substance use was - 0.52 years and the adjusted absolute mean difference was 2.00 years. The adjusted percentage of all comparisons reporting ages consistent within 1 year was 55.28%. More consistent reports were provided by adults, and in response to questions posed over 2 years as opposed to longer time intervals. Respondent answers to questions concerned with first use of marijuana were generally found to be most reliable; questions concerned with first use of crack were least reliable and reports of tobacco, alcohol and cocaine were intermediate. Logistic regression analyses also identified age, race, gender, education and poverty status as predictors of consistent reporting. Conclusions. Self-reports of age of first substance use experiences, as currently collected via survey questionnaires, are of sufficient reliability for most current epidemiological applications. For inquiries where age of substance use onset is itself a research focus, however, researchers should invest additional effort in improving the reliability of measurement.