It was hypothesized that smoking's calming effects are cognitively mediated and depend on the presence of a benign distractor, smoking narrows the focus of attention, thereby reducing anxiety by facilitating distraction from stressful cognitions (cf. C. M. Steele & R. A. Josephs's [1988] attention-allocation model of alcohol reinforcement). This notion was tested by examining the effect of smoking (vs. not smoking) on anxiety with and without a concurrent distraction in 82 smokers; distraction effects were also assessed in 42 nonsmokers. As predicted, smoking reduced anxiety only when paired with a distractor. Further, these findings could not be explained by direct nicotine effects or nicotine withdrawal. Several measures of attention allocation failed to confirm the hypothesized cognitive mechanisms, however. Implications of the findings are discussed.