Heterogeneity in direct recovery rates is an important area of study because deductions about survival rates and hunting mortality are made from recovery data. We tested the hypothesis of condition bias in hunter-killed lesser scaup (Aythya affinis) and tested differences among direct recovery rates (f = proportion of birds banded that are killed by a hunter ???3 months after release and reported) of scaup staging on Thief Lake Wildlife Management Area (TLWMA) and Agassiz National Wildlife Refuge (ANWR) in northwestern Minnesota. Scaup were trapped, held for 24 hr, weighed, banded with standard aluminum leg bands, and released during 4 October-8 November 1983-86 at TLWMA, and 1984-86 at ANWR. During 1984-86, we also sprayed areas of white plumage with fluorescent paint, colorcoded by age and location. Recovery data obtained from cooperative hunters and the U.S. Geological Survey Bird Banding Laboratory were subjected to 3 sets of analyses. In an analysis of variance (ANOVA), after adjusting for sex, age, year, location of capture, and date of capture, body mass at capture of recovered birds was lower, on average, than those not recovered. Logistic regression indicated (1) probability of recovering a bird decreased with increasing body mass; (2) within years, probability of recovery increased with date of first capture; (3) scaup at TLWMA were recovered at greater rates than scaup at ANWR; (4) hatching-year (HY) scaup were recovered at greater rates than were after-hatching-year (AHY) birds; and (5) recovery was more likely during 1984-86 than in 1983. Finally, in a crude test of recovery distribution, recoveries of birds banded at ANWR were about half as likely as those banded on TLWMA to occur in Minnesota. We conclude that cohort membership (age, capture site) and condition, as measured by body mass, influence susceptibility to hunting mortality. These and other sources of heterogeneity in recovery rates should be considered when modeling survival of scaup.