The evolution of sexual cannibalism as the most extreme form of nuptial feeding is still poorly understood. Although increasing evidence suggests that female aggressiveness is related to other aspects of foraging behaviour, it is not clear whether the nutritional value of a male is sufficient to provide an adaptive significance for sexual cannibalism. A widely cited though rarely tested explanation is based on a paternal investment model, and predicts that consumption of a male results in increased female fecundity. The available evidence is either correlational or restricted to species with relatively large and potentially nutritious males, and different studies have come to different conclusions. Here we present a test of the paternal investment hypothesis using the very cannibalistic and highly size-dimorphic spider Argiope bruennichi. After a preset schedule, we had females consume none, one or two males independent of the females cannibalistic behaviour. Consumption of male bodies did not result in any detectable fitness benefit for the female: neither the number of clutches, nor clutch size or hatching success were affected by consumption of males. The frequency of cannibalism was around 80%, independent of the female mating status. We did not observe male complicity, but cannibalism was associated with prolonged copulation. This suggests a sexually selected benefit of cannibalism for males. We conclude that the paternal investment hypothesis does not explain the existence of sexual cannibalism in A. bruennichi and probably not in other spider species with a pronounced sexual size dimorphism.