The majority of genetic risk factors for major depressive disorder (MDD) still await identification. Since copy number variants (CNVs) have been implicated in various neuropsychiatric disorders, the question arises as to whether CNVs also play a role in MDD. We performed a genome-wide CNV study using Illumina's SNP array data from 604 MDD patients and 1,643 controls. Putative CNVs were detected with the CNV algorithms QuantiSNP and PennCNV. CNVs with ≥30 consecutive SNPs and a log Bayes Factor/confidence value of ≥30 were statistically analyzed using PLINK. Further analyses and technical verification were only performed in the case of regions for which CNV calls from both programs showed nominal significance. Set-based tests were used to test whether common variants in the CNV regions showed association in two GWAS datasets of MDD. CNVs from four chromosomal regions were associated with MDD. The following were more frequent in patients than controls: microdeletions in 7p21.3 (P = 0.033) and 18p11.32 (P = 0.030); microduplications in 15q26.3 (P = 0.033); and the combination of microdeletion/duplications in 16p11.2 (P ≤ 0.018). SNPs in CNV region 16p11.2 showed significant association in a set-based test (P = 0.026). Microdeletions/duplications in 16p11.2 are the most promising CNVs, since these affect genes and CNVs in this region have been implicated in other neuropsychiatric disorders. The association finding for common SNPs provides further support for the hypothesis that this region is involved in the development of MDD. © 2012 Wiley Periodicals, Inc.