The zebra finch is the model species for the study of the evolution of birdsong, female song preferences, song learning and the neural processes underlying song learning and production. Despite this, almost all work to date in these fields has focused on domesticated zebra finches in captivity and remarkably little is known about song preferences, or the reproductive success of males with different songs, in wild populations. In this study we tested, for the first time in a wild zebra finch population, whether a male's song structure predicts his reproductive success. We recorded male songs in a nestbox population. Males from this wild population sang longer songs, with a higher peak frequency, than domesticated males. The number of each male's offspring that survived until day 12 posthatching (a proxy for fledging success) was used as a measure of reproductive success. Nestlings were partially cross-fostered, allowing us to disentangle the indirect effects of male genetic quality or maternal effects from those of direct benefits such as parental care. Male song structure predicted the number of genetic offspring surviving, as well as hatching success, but not the number or size of eggs in a clutch. Song structure did not predict the number of unrelated foster-offspring that survived. These results provide the first evidence that differences in male song can predict differences in reproductive success in the zebra finch, and suggest that differences in genetic quality are responsible, rather than differences in parental care or maternal investment in the eggs.