Abstract Background Oat (Avena sativa L.), a hexaploid crop with unknown genome, has valuable nutritional, medicinal and pharmaceutical uses. However, no suitable RGs (reference genes) for qPCR (quantitative real-time PCR) has been documented for oat yet. Single-copy gene is often selected as RG, which is challengeable or impactable in unexplored polyploids. Results In this study, eleven candidate RGs, including four duplicated genes, were selected from oat transcriptome. The stability and the optimal combination of these candidate RGs were assessed in 18 oat samples by using four statistical algorithms including the ΔCt method, geNorm, NormFinder and BestKeeper. The most stable RGs for “all samples”, “shoots and roots of seedlings”, “developing seeds” and “developing endosperms” were EIF4A (Eukaryotic initiation factor 4A-3), UBC21 (Ubiquitin-Conjugating Enzyme 21), EP (Expressed protein) and EIF4A respectively. Among these RGs, UBC21 was a four-copy duplicated gene. The reliability was validated by the expression patterns of four various genes normalized to the most and the least stable RGs in different sample sets. Conclusions Results provide a proof of concept that the duplicated RG is feasible for qPCR in polyploids. To our knowledge, this study is the first systematic research on the optimal RGs for accurate qPCR normalization of gene expression in different organs and tissues of oat.