By combining and reanalysing data from two independent field experiments we explore whether food limitation at the brood stage affects habitat selection in nesting mallards (Anas platyrhynchos). In an introduction experiment we found that, independent of treatment, some study lakes remained empty of wild mallard pairs ("empty lakes"), whereas on other lakes introduced birds attracted wild mallards ("attractive lakes"). In the other experiment we used mallard ducklings to address brood-stage food limitation by studying mass change of ducklings, We found that ducklings foraging on lakes that did not attract wild mallard pairs in the introduction experiment gained much less mass than those foraging on attractive lakes. In most cases ducklings even lost mass in the empty-lake foraging trials, providing strong evidence for food limitation. Therefore. lakes that remained empty of wild mallard pairs in the introduction experiment proved to be inferior brood habitats, particularly in terms of food. Our results give insight into the mechanisms underlying the general habitat selection hypotheses, specifically the ideal preemptive and conspecific attraction rules. The results further support our earlier conclusion that mallards do not use the ideal preemptive rule when selecting nesting lakes. However, conspecific attraction may not be generally applicable either, because. independent of the presence of introduced conspecifics, wild mallards somehow anticipated the low quality of the empty lakes as brood-rearing habitats and made their habitat-selection decision accordingly.