Large habitat fragments are generally thought to host more species and to offer more diverse and/or better quality habitats than small fragments. However, the importance of small fragments for population dynamics in general and for reproductive performance in particular is highly controversial. Using an information-theoretic approach, we examined reproductive performance and probability of local recruitment of color-banded reed buntings Emberiza schoeniclus in relation to the size of 18 wetland fragments in northeastern Switzerland over 4 years. We also investigated if reproductive performance and recruitment probability were density-dependent. None of the four measures of reproductive performance (laying date, nest failure probability, fledgling production per territory, fledgling condition) nor recruitment probability were found to be related to wetland fragment size. In terms of fledgling production, however, fragment size interacted with year, indicating that small fragments were better reproductive grounds in some years than large fragments. Reproductive performance and recruitment probability were not density-dependent. Our results suggest that small fragments are equally suited as breeding grounds for the reed bunting as large fragments and should therefore be managed to provide a habitat for this and other specialists occurring in the same habitat. Moreover, large fragments may represent sinks in specific years because a substantial percentage of all breeding pairs in our study area breed in large fragments, and reproductive failure in these fragments due to the regularly occurring floods may have a much stronger impact on regional population dynamics than comparable events in small fragments.