We tested whether predation on duck nests (Anas spp.) was density dependent at three spatial scales using artificial and natural nests in the Suisun Marsh, California, USA. At the largest spatial scale, we used 5 years (1998–2002) of data to examine the natural variation in duck nest success and nest densities among 8–16 fields per year, each 5–33 ha in size (n=62 fields). At an intermediate spatial scale, we deployed artificial nests (2000, n=280) within 1-ha plots at three experimental densities (5, 10, and 20 nests ha⁻¹) in a complete randomized block design and examined differences in nest predation. At the smallest spatial scale, we examined nest success in relation to nearest-neighbor fates and distances for artificial (2000, n=280) and natural nests (2000, n=507). We detected no relationship between nest success and the density of natural nests among fields in any year, nor when we pooled data for all years after controlling for year effects. The proportion of artificial nests that survived also did not depend on experimental nest densities within 1-ha plots. Overall, 15.0±12.4%, 15.0±11.0%, and 6.2±4.3% of artificial nests survived the 32-day exposure period in the low, intermediate, and high nest densities, respectively. Additionally, we detected no consistent effect of nearest-neighbor fate or distance on the success of artificial or natural nests. Thus, our results provide no evidence of density-dependent predation on duck nests at any scale of analysis, in contrast to a number of previous studies. Variation among geographical locations in the degree to which predation is density-dependent may reflect the composition of the predator community and the availability of alternate prey.