A cohort of 1088 couples in Xinjiang, PR China, were recruited to study infant feeding practices and paternal smoking. Mothers were interviewed in hospital and at 0.5, 1.5, 2.5, 3.5, 4.5, and 6 months. Survival analysis was used to calculate breastfeeding rates in smokers and nonsmokers. The paternal smoking rate was 64.8% and maternal smoking rate 1.7% (P < .01). The rates for any breastfeeding in the smoking group were significantly lower than in the nonsmoking group from 3.5 months to 6 months (P < .05). The rates for exclusive breastfeeding in the paternal smoking group were lower than in the nonsmoking group from discharge to 6 months (P < .05). The median duration of exclusive breastfeeding in the paternal smoking group was shorter than in the nonsmoking group. Paternal smoking was a risk factor for stopping any breastfeeding (hazard risk, 1.84; 95% confidence interval, 1.11-3.04) and exclusive breastfeeding (hazard risk, 1.33; 95% CI, 1.09-1.64) compared with nonsmokers.