In human glioma research, quantitative real-time reverse-transcription PCR is a frequently used tool. Considering the broad variation in the expression of candidate reference genes among tumor stages and normal brain, studies using quantitative RT-PCR require strict definition of adequate endogenous controls. This study aimed at testing a panel of nine reference genes [beta-2-microglobulin, cytochrome c-1 (CYC1), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), hydroxymethylbilane synthase, hypoxanthine guanine phosphoribosyl transferase 1, ribosomal protein L13a (RPL13A), succinate dehydrogenase, TATA-box binding protein and 14-3-3 protein zeta] to identify and validate the most suitable reference genes for expression studies in human glioma of different grades (World Health Organization grades II-IV). After analysis of the stability values calculated using geNorm, NormFinder, and BestKeeper algorithms, GAPDH, RPL13A, and CYC1 can be indicated as reference genes applicable for accurate normalization of gene expression in glioma compared with normal brain and anaplastic astrocytoma or glioblastoma alone within this experimental setting. Generally, there are no differences in expression levels and variability of candidate genes in glioma tissue compared to normal brain. But stability analyses revealed just a small number of genes suitable for normalization in each of the tumor subgroups and across these groups. Nevertheless, our data show the importance of validation of adequate reference genes prior to every study.