Kari Stefansson and colleagues report genome-wide meta-analyses for association to smoking behaviors within the population cohorts of the ENGAGE consortium. They report three loci associated to cigarettes smoked per day. Smoking is a common risk factor for many diseases1. We conducted genome-wide association meta-analyses for the number of cigarettes smoked per day (CPD) in smokers (n = 31,266) and smoking initiation (n = 46,481) using samples from the ENGAGE Consortium. In a second stage, we tested selected SNPs with in silico replication in the Tobacco and Genetics (TAG) and Glaxo Smith Kline (Ox-GSK) consortia cohorts (n = 45,691 smokers) and assessed some of those in a third sample of European ancestry (n = 9,040). Variants in three genomic regions associated with CPD (P < 5 × 10−8), including previously identified SNPs at 15q25 represented by rs1051730[A] (effect size = 0.80 CPD, P = 2.4 × 10−69), and SNPs at 19q13 and 8p11, represented by rs4105144[C] (effect size = 0.39 CPD, P = 2.2 × 10−12) and rs6474412-T (effect size = 0.29 CPD, P = 1.4 × 10−8), respectively. Among the genes at the two newly associated loci are genes encoding nicotine-metabolizing enzymes (CYP2A6 and CYP2B6) and nicotinic acetylcholine receptor subunits (CHRNB3 and CHRNA6), all of which have been highlighted in previous studies of smoking and nicotine dependence2,3,4. Nominal associations with lung cancer were observed at both 8p11 (rs6474412[T], odds ratio (OR) = 1.09, P = 0.04) and 19q13 (rs4105144[C], OR = 1.12, P = 0.0006).