The smoking behavior of 365 sixth grade children was examined in a pretest-treatment-posttest quasi-experimental investigation. Subjects in treatment conditions were presented information about the immediate consequences of smoking and, as is suggested by social learning theory, subjects were also presented with several forms of direct and vicarious feedback about the levels of carbon monoxide found in their breath. After viewing a film about how carbon monoxide immediately enters the smoker's body and can be measured in the breath, subjects were given vicarious feedback via a live demonstration of the detection process using smoking and non-smoking adult models. Subjects in the first feedback condition then had their breath tested for carbon monoxide and received immediate feedback on the results. A second group received feedback on the carbon monoxide content of their breath after a one-week delay and subjects in a third condition produced breath specimens but were never given feedback on the results. A fourth feedback group was not monitored for carbon monoxide but did view the vicarious feedback demonstration. An additional treatment group only viewed the carbon monoxide film while a control group only completed pretest and posttest questionnaires. Statistical analysis of the numbers of smokers in the treatment and control groups indicated that the proportion of subjects who reported smoking in the information only and the delayed feedback conditions were respectively reduced and maintained while subjects in the groups that received immediate and withheld feedback reported substantial increases in the numbers of smoking subjects. Although low rates of increase prohibited statistical analysis, a comparison of the proportion of first-time smokers revealed that subjects who received immediate feedback had the lowest rate of recruitment. In contrast, subjects in the delayed feedback and information only conditions had rather large increases in first-time smoking. Increases in new smokers were associated with an increased awareness of peers who smoked by former non-smokers which was attributed to social comparison as a means of satisfying increased interest produced by either temporarily or permanently withholding feedback. Further analysis revealed that subjects in the immediate and vicarious feedback conditions made significant decreases in their estimates of how soon large amounts of carbon monoxide could be found in their bodies. The vicarious feedback subjects also made significant decreases in their estimates of how soon carbon monoxide would get into others' bodies. Carbon monoxide analyses performed on subjects did not provide a means of discriminating smoking from non-smoking subjects. Subjects were found to consistently underestimate the occurrence of long-term consequences (cancer, and heart trouble) and overestimate the occurrence of short-term consequences (carbon monoxide and increased heart rates). They also estimated consequences as occurring later for generalized others than for themselves. Beliefs about the occurrence of consequences were not predictive of subjects' behavior. However, distant past behavior was meaningfully related to subjects' current smoking patterns, suggesting that even by the sixth grade, significant patterns of behavior had developed. Overall, this study suggests a social-psychological technique for presenting meaningful information to children which may be used to reduce or deter experimenting with the use of cigarettes.