Sampling with UV fluorescent light tubes is a commonly used technique both in applied and basic insect studies. Our study compares the performance of two such methods: manual sampling (light towers) and automatic sampling (funnel light traps). The abundance, diversity, and body size of moths representing two species-rich families (Lepidoptera: Arctiidae, Geometridae) were analyzed in a lowland rain forest in Costa Rica (La Selva Biological Station, 10.4° N, 84.0° W) during 2003 and 2004. Light towers were equipped with two 15-W UV fluorescent tubes and were operated for 3 h in 16 nights. Traps equipped with single 8-W fluorescent tubes were run throughout 20 nights in the understory of the forest. In addition, parallel trap sampling was carried out in the canopy. A total of 1,238 arctiid moths representing 162 species and 1,769 geometrid moths representing 196 species were collected. In Geometridae, tower samples were significantly larger than trap samples. Towers also attracted a higher overall number of species. Very small geometrids (particularly of the subfamily Sterrhinae) were under-represented in trap samples, suggesting that this method is biased toward larger species. In arctiid moths, there were no significant differences in either the sample sizes, the number of species or in the size of the individuals sampled. Diversity calculated as Fisher’s α was similar for towers and understory traps in both families. A major component of diversity was added with canopy trap samples for arctiid moths, but not for geometrid moths. In conclusion, ground-based tower sampling proved to be the most suitable method for geometrid moths, and trap sampling including both understory and canopy for arctiid moths. For full moth species inventories, a combination of both approaches is recommended.