Real time quantitative PCR (qPCR) is the method of choice for miRNA expression studies. For relative quantification of miRNAs, normalization to proper reference genes is mandatory. Currently, no validated reference genes for miRNA qPCR in prostate cancer are available. In this study, the expression of four putative reference genes (hsa-miR-16, hsa-miR-130b, RNU6-2, SNORD7) was examined with regard to their use as normalizer. After SNORD7 was already shown an inappropriate reference gene in preliminary experiments using total RNA pools, we studied the expression of the putative reference genes in tissue and normal adjacent tissue sample pairs from 76 men with untreated prostate carcinoma collected after radical prostatectomy. hsa-miR-130b and RNU6-2 showed no significantly different expression between the matched malignant and non-malignant tissue samples, whereas hsa-miR-16 was significantly underexpressed in malignant tissue. Softwares geNorm and Normfinder predicted hsamiR-130b and the geometric mean of hsa-miR-130b and RNU6-2 as the most stable reference genes. Normalization of the four miRNAs hsa-miR-96, hsamiR-125b, hsa-miR-205, and hsa-miR-375, which were previously shown to be regulated, shows that normalization to hsa-mir-16 can lead to biased results. We recommend using hsa-miR-130b or the geometric mean of hsa-miR-130b and small RNA RNU6-2 for normalization in miRNA expression studies of prostate cancer.