Breeding success in ground-nesting birds is primarily determined by nest survival, which may be density-dependent, but the generality of this pattern remains untested. In a replicated crossover experiment conducted on 30 wetlands, survival of simulated mallard nests was related to "biome" (n=14 mediterranean and 16 boreal wetlands), breeding "phenology" (early vs late nests), and "density" (2 vs 8 nests per 225 m shoreline). Local abundances of "waterfowl", "other waterbirds", and "avian predators" were used as covariates. We used an information-theoretic approach and Program MARK to select among competing models. Nest survival was lower in late nests compared with early ones, and it was lower in the mediterranean than in the boreal study region. High-density treatment nests suffered higher depredation rates than low-density nests during days 1-4 of each experimental period. Nest survival was negatively associated with local abundance of "waterfowl" in the boreal but not in the mediterranean biome. Effect estimates from the highest-ranked model showed that nest "density" (d 1-4) had the strongest impact on model fit; i.e. three times that of "biome" and 1.5 times that of "phenology". The latter's effect, in turn, was twice that of "biome". We argue that our study supports the idea that density-dependent nest predation may be temporally and spatially widespread in waterfowl. We also see an urgent need for research of how waterfowl nesting phenology is matched to that of prey and vegetation.