Quantitative real-time polymerase chain reaction (qRT-PCR) is the most important tool in measuring levels of gene expression due to its accuracy, specificity, and sensitivity. However, the accuracy of qRT-PCR analysis strongly depends on transcript normalization using stably expressed reference genes. The aim of this study was to find internal reference genes for qRT-PCR analysis in various experimental conditions for seed, adventitious underground bud, and other organs of leafy spurge. Eleven candidate reference genes (BAM4, PU1, TRP-like, FRO1, ORE9, BAM1, SEU, ARF2, KAPP, ZTL, and MPK4) were selected from among 171 genes based on expression stabilities during seed germination and bud growth. The other ten candidate reference genes were selected from three different sources: (1) 3 stably expressed leafy spurge genes (60S, bZIP21, and MD-100) identified from the analyses of leafy spurge microarray data; (2) 3 orthologs of Arabidopsis "general purpose" traditional reference genes (GAPDH_1, GAPDH_2, and UBC); and (3) 4 orthologs of Arabidopsis stably expressed genes (UBC9, SAND, PTB, and F-box) identified from Affymetrix ATH1 whole-genome GeneChip studies. The expression stabilities of these 21 genes were ranked based on the C(T) values of 72 samples using four different computation programs including geNorm, Normfinder, BestKeeper, and the comparative ΔC(T) method. Our analyses revealed SAND, PTB, ORE9, and ARF2 to be the most appropriate reference genes for accurate normalization of gene expression data. Since SAND and PTB were obtained from 4 orthologs of Arabidopsis, while ORE9 and ARF2 were selected from 171 leafy spurge genes, it was more efficient to identify good reference genes from the orthologs of other plant species that were known to be stably expressed than that of randomly testing endogenous genes. Nevertheless, the two newly identified leafy spurge genes, ORE9 and ARF2, can serve as orthologous candidates in the search for reference genes from other plant species.