Youth who first smoke cigarettes during childhood are a high risk for habitual smoking. Evaluating the reliability of children's smoking initiation is essential to research efforts to explain or prevent smoking onset. The present study is the first to establish reliability of self-reported smoking behavior with questionnaire data from elementary school children (N = 1,184). Data from a longitudinal investigation are used to examine the consistency of children's self-reported smoking across items and over time. Both cross-sectional and longitudinal analyses demonstrate that children report having tried smoking and lifetime use remarkably consistently. However, only about half the children reliably estimated their grade at first use. The study results suggest that some but not all standard questionnaire items yield reliable self-report data about initial smoking behavior from respondents as young as 8 to 11 years.