Discussion
The experimental shelter sites were often used by both gammarid species under continuous light conditions. D. villosus was the stronger competitor for shelter sites, which eventually resulted in an increased mortality of G. roeselii in the presence of a predatory fish. Single individuals of D. villosussheltered more than single individuals of G. roeselii. These results match those of the intraspecific experiment, where individuals of D. villosus also sheltered more than large males and large females of G. roeselii. It seems that D. villosus is not affected by intraspecific competition, in contrast to G. roeselii, where small individuals showed higher sheltering rates than large individuals in intraspecific competition. Since smaller individuals are more vulnerable to cannibalism and predation by (larger) predators, they hide more likely in shelters (Platvoet et al. 2009a). The small males and small females of G. roeselii sheltered significantly more than large males and large females in the separate as well as the mixed size classes. The sheltering individualsin oursingle individual experiments were notsubject to competition by other individuals or species. Probably because shelter sites are only efficiently used when there is an increased predation risk (Magalhaes, Janssen, Hanna, & Sabelis, 2002), fewer single individuals of G. roeselii sheltered in comparison with sheltering individuals of G. roeselii in the intraspecific experiment.