The bank vole has become an important species for the study of gene expression changes underlying evolutionary adaptation, within-host dynamics and resistance to pathogens, or response to pollutants. RT-qPCR is an optimized method for the rapid and accurate measuring gene expression, but it relies on the use of reference genes (RGs) as endogenous controls to normalize mRNA levels of the target genes. Contrary to the common practice, the expression level of the housekeeping genes traditionally used as RGs cannot be assumed stable across different species and experimental settings. Furthermore, compared to model laboratory species, there are potentially additional sources of variation when collecting gene expression data from natural populations with unknown genetic and environmental backgrounds. Thus, rigorous determination of RGs in natural populations of the bank vole was necessary to facilitate gene expression studies in this emerging model species. Therefore, we evaluated the expression variation of 10 RG candidates in spleen samples of bank voles spanning a broad latitudinal range across Europe, using four approaches. Ube2d2a, Ppia, and Tbp were consistently identified as the least variable genes, followed by Rn18s, Ywhaz, and Rplp0. The combinations Ube2d2a and Ppia or Ube2d2a and Tbp were identified to be sufficient for the normalization. In contrast, the traditional housekeeping genes Actb, Gapdh, Sdha, and Tuba1a displayed large expression variation and are not recommended as internal controls. Our results underscore the need of a systematic evaluation of appropriate RGs to each particular experimental system and provide a useful starting point for further studies.