This Note compares self-reported tobacco use by adolescents with laboratory test results that measured levels of cotinine in respondents' saliva. The validity of self-administered questionnaires to investigate sensitive topics, such as substance abuse, is controversial. Investigators worry that respondents may intentionally underreport or overreport their actions or that they may conceal relevant behaviors. Thus, two major goals in substance use studies are to maximize the validity of self-reported use and to confirm that validity. In Project ALERT, a drug use prevention experiment for adolescents, the authors took several steps to encourage honest and accurate self-reports, including guaranteeing the confidentiality of responses and collecting samples of saliva from students. They found that the magnitude of underreporting is exceedingly low — less than 1 percent of the total responses at each of the four waves of data collection considered. An unexpected result was the substantial level of apparent overreporting of tobacco use that emerged. Further investigation suggests that the overreporting phenomenon is an artifact of the lack of sensitivity of the lab test to the low levels of tobacco use characteristic of many adolescents. The authors conclude that when proper data collection procedures are followed, students will provide accurate and valid reports of their tobacco use.