Reference genes are used to normalize target genes for relative quantification in gene expression studies. However, different experimental conditions may affect the expression of reference genes, which could lead to erroneous quantitative results. In this study, we performed real-time polymerase chain to investigate the expression of eight reference genes (rpoN, rpoD, dbhA, phaF, 16S rRNA, gst, lexA, and atkA) in Pseudomonas putida mt-2 during degradation of p-xylene. According to their expression stability, geNorm software analysis revealed that rpoN, rpoD, 16S rRNA, and atkA were suitable reference genes with highly stable expression, whereas phaF and dbhA were not suitable due to unstable expression. When normalized either to phaF or dbhA, xylA and xylE expression were significantly different compared to the expression levels normalized with the normalization factor (NF(4)) obtained from the four most stable reference genes (rpoN, -rpoD, -16S rRNA, and -atkA). The use of unstably expressing reference genes resulted in an over- or underestimation of target gene expression, a delay in maximal gene expression, and an increase in gene expression in the absence of inducer. While experimental results indicated that the relative maximum expression of xylA and xylE occurred at different times, unstable reference genes indicated that the maximum expression occurred at the same time. Our study indicates that a valid set of reference genes covering a broad expression range is recommended to accurately normalize and quantify the relative expression levels of the target gene(s) transcripts in many microbial processes.