We believe several factors may have contributed to our failure to replicate SNP associations identified in earlier meta-analyses of smoking data. First, and we think most significant, our nicotine phenotype is defined differently from how smoking was defined in these previous studies. Specifically, earlier studies implemented an exposure screen for having smoked at least 100 cigarettes in a lifetime before being considered a non-smoking control. In this way, they ensured that everyone in the sample would have had sufficient exposure to nicotine to have had an opportunity to develop a regular smoking habit. We applied no similar screen because exposure to nicotine is by itself an indicator of behavioral disinhibition (Reynolds & Fields, 2012) and we did not want to eliminate from the analysis individuals who had smoked fewer than 100 cigarettes as these individuals may be especially low on the behavioral disinhibition dimension. Second, in studies that have found significant SNP associations with smoking, the sample size has been substantially larger than the size of the sample in this study. For example, the ENGAGE consortium includes more than 30,000
