An animal's reproduction and survival depend critically on the choices it makes when selecting a habitat for breeding. Therefore, we expect animals to select breeding habitats that confer fitness benefits. For mallards (Anas platyrhynchos), previous research has shown that populations are most sensitive to variation in nest survival rates; thus, we expect strong selection lot safe nest sites. We used patterns of nesting success to predict nest distributions. We compared distributions of nests from radio-marked mallard females (n = 1,710) to random locations using logistic regression and modeled how nest-site locations were related to a number of landscape metrics measured at several spatial scales. Consistent with predictions derived from patterns of nesting success, mallard nest sites were best modeled using habitat attributes measured at the scale of the vegetation patch containing the nest. though ultimately habitat features measured at multiple scales improved predictions. Similarly, for most habitat types mallards tended to nest away from habitat edges, as predicted. Conversely. opposite to predictions, mallards nested in smaller habitat patches and nearer to wetlands than expected by chance. Hence, mallards only provide equivocal evidence that nest-site selection is adaptive. Several possible explanations for this lack of correspondence between patterns of nesting success and nest-site selection include the following: insufficient time to adequately sample available habitats, unpredictable environments, nest sites selected to maximize a different vital rate (e.g., adult female survival), and. finally, anthropogenic changes to the landscape may have resulted in a decoupling of evolved cues used to select nest sites and current predation processes. [ABSTRACT FROM AUTHOR]