Bipolar disorder (BPD) is a brain illness characterized by dramatic shifts in mood and activity levels, affecting ~60 million people worldwide. Despite the severity of these symptoms, management of BPD can be successful if diagnosis and treatment are initiated early. Consequently, research has focused on identifying genetic factors that facilitate early detection and/or treatment of this illness. In particular, genes in the serotonin pathway have been under intense investigation given their strong association with other behavior-related disorders with similar symptoms. However, little is still known about patterns of genetic variation in the human serotonin system in natural populations, which can impact treatment. To address this current gap in knowledge, we examined 20 genes in the serotonin pathway, totaling >1.6 million bases of sequence data, in ~1400 individuals from worldwide populations. Here, we report striking patterns of diversity in the SLC6A4 gene, including an excess of high-frequency polymorphisms and long-range haplotypes, consistent with a model of positive selection in African populations. We also inferred that non-coding polymorphisms at SLC6A4 are likely the targets of selection, raising the possibility that these variants may play a role in gene expression. Although the precise function of these polymorphisms is currently unknown, we argue that they are or have been selectively advantageous during evolutionary history, representing new candidate loci for further study. Because serotonin transport inhibitors are first-line treatments for BPD, our study, the largest of its kind to date, will be informative for the development of targeted interventions based on more “personalized” genomic information.