Background Bipolar Disorder (BD) is a complex and highly heritable mood disorder associated with a range of cognitive deficits across all illness states. Previous research has demonstrated associations between the dopamine D2 receptor gene (DRD2) and BD diagnosis, structural and functional brain changes, and cognitive dysfunction. Specific SNPs impacting DRD2 function have been identified: rs1800497 (Taq1A1, minor T allele) is associated with reduced striatal D2R dopamine binding; and rs1076560 and rs2283265 (minor A/T alleles) are associated with alternative splicing of DRD2 exon 6, resulting in increased expression of DRD2-short mRNA compared to DRD2-long isoform, and reduced inhibition of glutamate release. The present study adopted a cognitive-neurogenic approach to explore the mediating role of brain structure in the relationship between DRD2 functional haplotypes and cognitive outcomes. Methods Participants comprised 26 BD cases (BD-I or BD-II), 62 at-risk subjects (unaffected first-degree relatives of BD probands), and 55 control subjects with 3 T MRI, genetic and cognitive data. DRD2 haplotypes were defined by three functional SNPs (rs1800497, rs1076560, rs2283265). Cortical thickness, surface area and volume were extracted for three functionally defined regions of interest (ROIs) that were parcellated on the basis of resting state functional connectivity data – Dorsal Attention (DA), Ventral Attention (VA), and Fronto-Parietal (FP). Working memory and attention were assessed using Repeatable Battery for Assessment of Neuropsychological Status (RBANS), and intelligence (IQ) was determined using Wechsler Abbreviated Scale of Intelligence (WASI). A series of simple and multiple mediation models were implemented using SPSS PROCESS macro. These models predicted the effects of the functional DRD2 haplotype on working memory, attention and intelligence, as independently and jointly mediated by cortical area, thickness and volume in the three functionally defined ROIs. Results The frequency of CCG homozygotes versus TAA carriers was significantly higher in at-risk subjects compared to controls (freq=0.758 vs 0.564, p=0.013) with a trend in BD (freq=0.692, p=0.1). In the control group, CCG homozygotes exhibited an increase in RBANS attention scores as a result of the mediating effects of DA area and volume (of 1.6 and 1.7 standard deviations). Simple mediation effects were also observed in the control group for IQ and RBANS immediate memory in the absence of a significant total effect. In contrast, simple mediation effects were not apparent in the slightly larger at-risk group nor the BD group. While multiple mediation models reinforced many simple mediation effects observed in controls, an additional mediation effect was observed in the BD group – TAT haplotype carriers exhibited a 0.6 SD decrease in RBANS attention scores as a result of the mediating effects of FP volume. Discussion The DRD2 mediation effects observed in controls were absent in the at-risk and BD groups, suggesting that the normally ameliorative effect of the CCG haplotype on cognitive outcomes is either negated by other genetic and familial risk factors for BD, or is mediated by alternate brain structures. Additionally, evidence is provided for the reversal of typically observed DRD2 effects on cognitive outcomes by the significant mediation effect for the BD group in the multiple mediation model of attention, where carriers of the TAT haplotype demonstrated improved attention.