In determining relative gene expression by quantitative measurements of mRNA levels using real-time quantitative PCR, internal standards such as reference genes are essential. Large-scale studies evaluating (candidate) reference genes for veterinary research have not been conducted as thoroughly as for human research, although they are equally important. Our goal was to design and evaluate a genome-wide panel of reference genes from different functional classes. First, primers were optimized using mRNA from canine cell lines and from 30 tissues of one dog as template and SYBR green as fluorescent probe. Second, the expression variation and stability of a gene within one specific tissue were determined. Prostate, kidney, mammary gland, left ventricle, and liver tissues from five to nine dogs of different breeds, sexes, ages, body weights, and disease status were used. Averaging relative stabilities over these tissues revealed the usefulness of individual genes as reference genes. Furthermore, according to expression variation of a reference gene within a specific tissue, usually two to four reference genes are sufficient. Taken together, ribosomal protein S19 (RPS19), ribosomal protein S5 (RPS5), beta-2-microglobulin (B2M), and hypoxanthine phosphoribosyltransferase (HPRT) are advocated. However, the optimal set of reference genes depends on the tissue and should be selected and evaluated for each series of experiments.