Even though continued smoking by cancer patients adversely affects survival and quality of life, about one third of patients who smoked prior to their diagnosis continue to smoke after their diagnosis. The implementation of smoking cessation treatments for cancer patients has been slowed by the lack of data on correlates of tobacco use in this population. Thus, this longitudinal study assessed demographic, medical, addiction, and psychological predictors of tobacco use among 74 head, neck, and lung cancer patients. Multivariable binary logistic regression analyses, with outcome categorized as smoker or nonsmoker, indicated that the likelihood that patients would be a smoker was associated with lower levels of perceived risk and a higher level of quitting cons. Multivariable nominal logistic regression, with outcome classified as continuous smoker, continuous quitter, relapser, or follow-up quitter, indicated that: (a). patients categorized as continuous smokers reported significantly lower quitting self-efficacy than follow-up quitters and continuous quitters, (b). relapsers reported a significantly lower level of quitting self-efficacy than either follow-up quitters or continuous quitters, and (c). continuous smokers exhibited a significantly lower level of risk perceptions than continuous abstainers. These findings can be useful for the development and evaluation of treatments to promote smoking cessation among cancer patients.