Age at the onset of a disorder is one of its key descriptive characteristics. Early onset may indicate increased risk of a severe course and increased genetic liability. However, retrospectively reported onset in surveys is subject to forward telescoping, a bias in which respondents report events closer to the time of interview than is true. We examined the effect of this bias on age of onset for smoking initiation and daily smoking. Data came from the 1966-77 birth cohorts interviewed in the cross-sectional National Household Surveys on Drug Abuse (NHDSA) 1979-98 (N = 82,122). An association between age at onset and age at interview, within birth year, was found for experimenters and for daily smokers. This was indicative of forward telescoping. As age at interview increased from 12 to 25 the probability of reporting early onset dropped by half. An association was also found between early onset of daily smoking and age at interview. This response bias differed significantly by sex and race, created significant misclassification of smokers as late instead of early onset cases, and biased cohort comparisons from cross-sectional data. These results suggest efforts need to be made to limit the effect of forward telescoping in epidemiological studies by survey question and sampling design.