Genome-wide association studies using SNP genotype data have had limited success in the identification of variants associated with major depressive disorder (MDD). Haplotype data provide an alternative method for detecting associations between variants in weak linkage disequilibrium with genotyped variants and a given trait of interest. A genome-wide haplotype association study for MDD was undertaken utilising a family-based population cohort, Generation Scotland: Scottish Family Health Study, as a discovery sample with a population-based cohort, UK Biobank, used as a replication sample. Fine mapping of haplotype boundaries was used to account for overlapping haplotypes tagging causal variants. Within the discovery cohort, two haplotypes exceeded genome-wide significance (P<5 x 10-8) for an association for MDD. One of these haplotypes was located in 6q21, in a region which has been previously associated with bipolar disorder, a psychiatric disorder that is phenotypically and genetically correlated with MDD. The detection of associated haplotypes potentially allows the causal stratification of MDD into biologically informative aetiological subtypes.