In this study, we explored two mechanisms through which parental smoking may influence their children's smoking behavior and found evidence for both mechanisms. Consistent with previous research [22–24], we found that the odds for smoking increased with the number of parents who currently smoked. Compared to participants whose parents did not currently smoke, participants who reported that one parent currently smoked had a 1.3 times increased risk for ever smoking, and those who reported that both parents currently smoked had a 2.2 times increased risk. The p for trend was significant, suggesting a dose-response relationship. Growing up in a family in which one or both parents smoke not only provides children with the opportunity to imitate the behavior, living with two parents who smoke as compared to one or none, increases access to cigarettes [25, 26]. In addition, for many children, living in a household where one or both parents smoke means that they have been second-hand smokers, are used to breathing smoke, and therefore may not have a physically aversive reaction to tobacco when they first try.
We also found that the association between children's smoking attitudes and ever smoking was stronger when at least one parent currently smoked. Among participants whose parents did not currently smoke, children's smoking attitudes were associated with a 1.7 times increased risk for ever smoking, whereas among participants who reported that at least one parent currently smoked, children's smoking attitudes were associated with a 2.5 times increased risk. Our results suggest that parental smoking influences children's attitudes toward smoking, which in turn affect the likelihood of the child smoking. To the best of our knowledge, although one previous study examined whether parental smoking mediates (is an intermediate step in the casual pathway) the relationship between children's attitudes toward smoking and smoking behavior  and another examined the relationship between children's implicit and explicit attitudes toward smoking and parental smoking , no studies have examined if parental smoking status moderates (modifies) the relationship between children's attitudes toward smoking and their ever smoking. However, because moderation is best determined prospectively, our results lend support to the hypothesis and need to be confirmed using longitudinal data.
Our results differ from those of the Chassin et al.  study, which found that neither the children's implicit nor their explicit attitudes toward smoking were associated with parental smoking status. Our results, however, are consistent with, and extend the earlier findings of Flay et al. , who noted that parental smoking was associated with changes in adolescent's explicit attitudes. Not only did we observe an overall association between parental smoking and children's explicit attitudes, we found that the relationship between children's attitudes toward smoking and their ever smoking was weaker among children of parents who did not currently smoke compared to children whose parents currently smoke. Again, because moderation is best determined prospectively, our results lend support to the hypothesis that parental attitudes moderate this relationship.
It is of concern that the relationship between attitudes toward smoking and ever smoking is stronger when at least one parent smokes because positive attitudes toward smoking among youth are associated with increased susceptibility to smoking , and youth who have experimented with cigarettes or who currently smoke [27, 29, 30] hold significantly more positive attitudes toward smoking than do their nonsmoking peers. Moreover, Chassin et al.  found that mothers who smoke report significantly more positive implicit and explicit attitudes toward smoking than mothers who have quit or never smoked, underscoring the possibility that positive attitudes toward smoking are learned and sanctioned at home.
The participants in the Chassin et al.  were predominantly non-Hispanic white, as were the majority of the participants in Ridner's study . However, the participants in the Chalela study were all Latino, while the sample in the Flay et al. study  was multi-ethnic. Although the goals of these four studies and ours were different and all used different analytic approaches, all studies examined the relationship between attitudes toward smoking and behavior and all reported consistent results: positive attitudes were associated with ever smoking. Taken as a whole, this suggests that consistent with the conclusions drawn from other studies based on multi-ethnic samples [9, 10], predictors of smoking behavior may be universal.
While our study focused on the family context, other factors may impact the relationship between children's attitudes towards smoking and their ever-smoking. For example, many studies have documented the role that peer influence [31, 32] and perceptions of peer norms  play on smoking initiation. While outside the scope of the current study, examining peer influence, both separate from and in conjunction with parental influence, would refine our understanding of the relationship between attitudes and smoking.
This study has some limitations. First, the analysis is based on self-reported cross-sectional survey, limiting our ability to draw causal conclusions and test for moderation. Second, while most contemporary approaches to assessing family structure tend to compare differences between single and two parent families without regard to marital status, the data collected in this study did not permit such a distinction. The response categories probing parental marital status included married, separated, divorced, single, and widowed. Therefore we examined the influence of reporting married parents, which may serve as a proxy for two parent households, compared to reporting that parents are separated, divorced, single, or widowed. In future research we intend to fully examine the relationship between household structure, living arrangements and number of parents who smoke. Third, we do not know how long the participants were exposed to parental smoking, which limits our ability to determine if there is a threshold of exposure required to influence children. Fourth, we did not ask the ever smokers where they obtained the cigarettes they smoked. Therefore we cannot determine if current parental smoking directly increases access, and we cannot control for its potential influence in the analysis. Fifth, active consent was required of all students to participate in this study; more girls than boys returned their consent form resulting in the differential participation rates. Finally, we did not examine the influence of exposure to a parent who quit smoking while the participant was growing up, which has been shown to increase the likelihood of smoking . However, the net result from the lack of information about past parental smoking would have underestimated the effect sizes observed in our study (biased the results toward the null), suggesting our findings may have been more pronounced had we adjusted for past parental smoking.