Mud-dauber wasps are major predators of spiders, as female wasps provision their nests with paralyzed spiders for their young. Whereas three-dimensional web-building spiders (3D) are more prevalent and speciose than are two-dimensional web-building spiders (2D), 2D spiders are more common in wasp nests, perhaps because they are easier for the wasps to capture (Blackledge et al. 2003). When we offered Sceliphron caementarium (Sphecidae) wasps a choice of Larinioides sclopetarius (Araneidae, 2D) and Achaearanea tepidariorum (Theridiidae, 3D) spiders on their respective webs, wasps took significantly more 2D than 3D spiders, a result that initially seemed to support the importance of web architecture. However, when offered a choice of freshly killed 2D and 3D spiders without their webs, wasps rarely took 3D spiders. Furthermore, wasps spent more time on filter papers containing silk and chemical residues from 2D than 3D spiders, suggesting that prey recognition involves chemo-tactile cues. Analysis of cuticular extracts of two 2D and one 3D species by Gas Chromatography/ Mass Spectrometry (GC/MS) revealed distinct differences in hydrocarbon profiles between 2D and 3D spiders. We found that wasps rejected hexane-washed 2D spiders, suggesting that a cuticular chemical cue is necessary for prey recognition. Furthermore, when we offered individual wasps a choice of paper balls treated with cuticular extracts of 2D spiders, 3D spiders, or hexane controls, the wasps spent significantly more time antennating 2D extract-coated balls. These results provide the first evidence that cuticular cues are involved in recognition and differentiation of prey species by sphecid wasps.