The primary unit of analysis was the mean startle amplitude (peak), computed for each participant, within each session for each test block by spray by valence combination. Peak startle amplitude (orbicularis EMG) was integrated and rectified and defined as the largest peak from 20–120 ms following picture onset. A mixed models regression approach (Littell, Milliken, Stroup, Wolfinger, & Schabenberfer, 2006) was the primary analytic strategy for estimating the effects of 5-HTTLPR genotype, nasal spray, smoking deprivation, and their interaction terms on the startle response in Caucasian and African Americans separately. Subject was included as a random effect. Covariates included deprivation status and slide valence. Where appropriate, significant interactions were further evaluated using least-square means and univariate t-tests for all relevant pair-wise comparisons. Likelihood ratio tests were conducted to evaluate model fit after the inclusion of additional variables and only models with significant increases were retained. An additional model then examined the variables of interest within the entire sample (African Americans and Caucasians) including race as a covariate. Subsequent iterations of the model involved including other potential variables of interest such as craving and negative affect, though none were significant and will not be reported here.
