Reverse transcription quantitative polymerase chain reaction (qRT-PCR) has rapidly become the most sensitive and accurate method for the quantification of gene expression. To facilitate gene expression studies and obtain more accurate qRT-PCR data, normalization relative to stable housekeeping genes is required. These housekeeping genes need to show stable expression under the given experimental conditions for the qRT-PCR results to be accurate. Unfortunately, there are no studies on the stability of housekeeping genes used in Spodoptera litura. In this study, eight candidate reference genes, elongation factor 1 alpha (EF1), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), ribosomal protein L10 (RPL10), ribosomal protein S3 (RPS3), beta actin (ACTB), beta FTZ-F1 (FTZF1), ubiquinol-cytochrome c reductase (UCCR), and arginine kinase (AK), were evaluated for their suitability as normalization genes under different experimental conditions using the statistical software programs, BestKeeper, geNorm and Normfinder, and the comparative ΔCt method. We determined the expression levels of the candidate reference genes for three biotic factors (developmental stage, tissue and population), and four abiotic treatments (temperature, insecticide, food and starvation). The results indicated that the best sets of candidates as reference genes were as follows: GAPDH and UCCR for developmental stages; RPL10, AK and EF1 for different tissues; RPL10 and EF1 for different populations in China; GAPDH and EF1 for temperature-stressed larvae; AK and ACTB for larvae treated with different insecticides; RPL10, GAPDH and UCCR for larvae fed different diets; RPS3 and ACTB for starved larvae. We believe that these results make an important contribution to gene analysis studies in S. litura and form the basis of further research on stable reference genes in S. litura and other organisms.