The distribution of individuals among habitats and habitat-specific breeding output are basic elements for understanding population limitation and regulation. We studied the connection between habitat selection and population limitation in breeding mallards on boreal lakes in Sweden and Finland with experimental and long-term observational data. Wingclipped mallards were introduced on breeding lakes before migratory wild mallards arrived to test 2 alternative hypotheses of habitat selection, namely ideal preemption and conspecific attraction. The ideal preemptive rule was rejected while the conspecific attraction rule was to some extent supported. However, by combining the results of the introduction experiment with those from an experiment done with mallard ducklings to measure habitat quality, we found that only good-quality lakes attracted wild mallards, whereas poor-quality lakes did not. Long-term observational data from 35 lakes, classified into rich and poor based on shore vegetation, revealed that breeding pairs prefer rich lakes over poor, breeding success also being better there than on the poor lakes. In accordance with the experimental findings, density on the rich lakes increased with overall population density while that on the poor did not. However, breeding success did not show clear density dependence. Our results indicate habitat-specific limitation of breeding numbers, though this may not translate to limitation at the population level.