Leaf pathogens are limiting factors in banana (Musa spp.) production, with Pseudocercospora spp. responsible for the important Sigatoka disease complex. In order to investigate cellular processes and genes involved in host defence responses, quantitative real-time PCR (RT-qPCR) is an analytical technique for gene expression quantification. Reliable RT-qPCR data, however, requires that reference genes for normalization of mRNA levels in samples are validated under the conditions employed for expression analysis of target genes. We evaluated the stability of potential reference genes ACT1, α-TUB, UBQ1, UBQ2, GAPDH, EF1α, APT and RAN. Total RNA was extracted from leaf tissues of Musa acuminata genotypes Calcutta 4 (resistant) and Cavendish Grande Naine (susceptible), both subjected to P. musae infection. Expression stability was determined with NormFinder, BestKeeper, geNorm and RefFinder algorithms. UBQ2 and RAN were the most stable across all M. acuminata samples, whereas when considering inoculated and non-inoculated leaf samples, APT and UBQ2 were appropriate for normalization in Calcutta 4, with RAN and α-TUB most stable in Cavendish Grande Naine. This first study of reference genes for relative quantification of target gene expression in the M. acuminata-P. musae interaction will enable reliable analysis of gene expression in this pathosystem, benefiting elucidation of disease resistance mechanisms.