Background: The association between body size and head and neck cancers (HNCA) is unclear, partly because of the biases in case–control studies. Methods: In the prospective NIH–AARP cohort study, 218,854 participants (132,288 men and 86,566 women), aged 50 to 71 years, were cancer free at baseline (1995 and 1996), and had valid anthropometric data. Cox proportional hazards regression was used to examine the associations between body size and HNCA, adjusted for current and past smoking habits, alcohol intake, education, race, and fruit and vegetable consumption, and reported as HR and 95% confidence intervals (CI). Results: Until December 31, 2006, 779 incident HNCAs occurred: 342 in the oral cavity, 120 in the oro- and hypopharynx, 265 in the larynx, 12 in the nasopharynx, and 40 at overlapping sites. There was an inverse association between HNCA and body mass index, which was almost exclusively among current smokers (HR = 0.76 per each 5 U increase; 95% CI, 0.63–0.93), and diminished as initial years of follow-up were excluded. We observed a direct association with waist-to-hip ratio (HR = 1.16 per 0.1 U increase; 95% CI, 1.03–1.31), particularly for cancers of the oral cavity (HR, 1.40; 95% CI, 1.17–1.67). Height was also directly associated with total HNCAs (P = 0.02), and oro- and hypopharyngeal cancers (P < 0.01). Conclusions: The risk of HNCAs was associated inversely with leanness among current smokers, and directly with abdominal obesity and height. Impact: Our study provides evidence that the association between leanness and risk of HNCAs may be due to effect modification by smoking. Cancer Epidemiol Biomarkers Prev; 23(11); 2422–9. ©2014 AACR.