The standard tool for assessing tobacco dependence is the Fagerström Tolerance Questionnaire (FTQ) or its more recent variant, the Fagerström Test for Nicotine Dependence (FTND). Although both of these scales reportedly assess physiological dependence on nicotine, they might not tap some facets of dependence, particularly psychosocial factors. To determine whether tobacco dependence exhibits multidimensional properties, we examined two existing, independent data sets, one from SRI International (n=443) and another from the University of Michigan (n=445). Based on our knowledge from existing literature, standard psychometric statistical analyses, and results from exploratory factor analysis using SRI's data set, we identified two competing models for dependence representing a hybrid of the FTQ/FTND and the Smoking Motives Questionnaire. We then examined these models using confirmatory factor analysis with data from the University of Michigan. We characterized the final model by five first-order factors, each consisting of two to four items, and one higher-order factor. The first-order factors were termed stimulation, automaticity, sedation, psychosocial motives, and morning smoking; the higher-order factor, tobacco dependence, underlay each of the first-order factors. The ranges of interitem correlations and Cronbach's alpha estimates of internal consistency for the first-order factors were .34 - .68 and .64 - .81, respectively. Results of these analyses support the hypothesis that tobacco dependence is multidimensional.