To identify reliable reference genes for toxicological studies, 16 commonly-used reference genes were selected as candidates to evaluate their expression stabilities under experimental conditions in Caenorhabditis elegans. Sixteen candidates were composed of 12 protein-coding genes and 4 non-coding RNAs, they were act-2, ama-1, arp-6, cdc-42, csq-1, eif-3.C, idhg-1, mdh, pmp-3, rbd-1, tba-1, Y45F10D.4, 18S rRNA, Ce234, U18, and U6. Larval stage 1 synchronized hermaphrodites were exposed to benzo-α-pyrene (BαP), chlorpyrifos, diazinon, gossypol, zinc oxide nanoparticles, and the vehicle control DMSO for 30 h, respectively. Expression stabilities of candidate genes were analyzed using 4 independent evaluating approaches (BestKeeper, the delta Ct approach, geNorm, and NormFinder) followed by a comprehensive method. Results showed that there were slight differences in ranking order between evaluation methods due to their different assumptions and computations. The results also showed that responses of candidate genes to different chemicals were distinct, 18S rRNA was the best for BαP and chlorpyrifos, tba-1 was the most stable gene for diazinon and gossypol treatments, while pmp-3 was more stable for zinc oxide exposure. Additionally, results demonstrated that combinations of multiple genes were more reliable than individual gene, suggesting selecting two or more candidates as reference genes may generate more reliable results for toxicological studies.