Scientists estimate seed abundances to calculate seasonal carrying capacities and assess wetland management actions for waterfowl and other wildlife using soil core samples. We evaluated recovery of known quantities of moist-soil seeds from whole and subsampled experimental core samples containing 12 seed taxa representing small, medium, and large size classes. We recovered 86.3% (SE = 1.8) of all seeds added to experimental cores; 8.3% (SE = 1.2) of seeds were destroyed during the sieving process and 5.4% (SE = 1.2) were not recovered by observers. Recovery rates varied by seed size, but not seed quantity or disproportionate ratios of seed-size classes. Overall seed recovery rates were similar between subsampled ( = 81.2%, SE = 3.6) and whole–processed core samples ( = 86.3%, SE = 1.8). We used recovery rates to generate size-specific, taxon-specific, and constant correction factors and applied each to actual core sample data. Size-specific correction factors increased seed mass estimates in the Mississippi Alluvial Valley ( = 10.1%, SE = 0.32), upper Midwest ( = 21.2%, SE = 0.61), and both regions combined ( = 15.7%, SE = 0.51) differently, as seed composition in core samples varied regionally. We suggest scientists consider using size-specific correction factors to account for seed recovery bias in core samples because these factors may be applied to a variety of taxa and produced similar mass estimates as taxon-specific correction factors. However, if data from core samples are unavailable at the resolution of seed size classes, we suggest increasing seed mass estimates by 16% to account for seed recovery bias. © 2011 The Wildlife Society.