Real-time RT-PCR is a powerful technique for the measurement of gene expression, but its accuracy depends on the stability of the internal reference gene(s) used for data normalization. Tobacco (Nicotiana tabacum) is an important model in studies of plant gene expression, but stable reference genes have not been well-studied in the tobacco system. We address this problem by analysing the expression stability of eight potential tobacco reference genes. Primers targeting each gene (18S rRNA, EF-1alpha, Ntubc2, alpha- and beta-tubulin, PP2A, L25 and actin) were developed and optimized. The expression of each gene was then measured by real-time PCR in a diverse set of 22 tobacco cDNA samples derived from developmentally distinct tissues and from plants exposed to several abiotic stresses. L25 and EF-1alpha demonstrated the highest expression stability, followed by Ntubc2. Measurement of L25 and EF-1alpha was sufficient for accurate normalization in either the developmental or stress-treated samples, but Ntubc2 was also required when considering the entire sample set. Analysis of a tobacco circadian gene (NTCP-23) verified these reference genes in an additional context, and all techniques were optimized to enable a high-throughput approach. These results provide a foundation for the more accurate and widespread use of real-time RT-PCR in tobacco.