Mixed model regression (SAS Proc Mixed v9.2; SAS Institute Inc., Cary, NC) was used to assess all models where physiology or in-session ratings were the dependent variables. The mixed model approach is a form of the generalized linear model (GLM) that allows for more specific estimation of the correlation structure of the residuals and that will not exclude cases with missing observations, thus allowing the use of all available data (Bageilla, Sloan, & Heitjan, 2000; Kristjansson, Kircher, & Webb, 2007). All models included subject as a random effect. Models run on postquit time points included baseline dependent measures and session abstinence status (abstinent vs. smoking) as covariates. All descriptions of differences between means following a significant mixed model effect were the result of contrast comparisons of least-square means (LSM) and standard errors (SE) of fixed effects using tests of simple effects (Winer, 1971). To correct for the effects of multiple comparisons on Type I error rate, the family-wise α levels of post hoc contrasts were adjusted using the Holm-Bonferroni correction (Holm, 1979). For interactions of continuous (e.g., FTND) with categorical
