Models of sexual selection propose that exaggerated secondary sexual ornaments indicate a male's own fitness and the fitness of his offspring. These hypotheses have rarely been thoroughly tested in free-living individuals because overall fitness, as opposed to fitness components, is difficult to measure. We used 20 years of data from song sparrows (Melospiza melodia) inhabiting Mandarte Island, British Columbia, Canada, to test whether a male's song repertoire size, a secondary sexual trait, predicted overall measures of male or offspring fitness. Males with larger song repertoires contributed more independent and recruited offspring, and independent and recruited grandoffspring, to Mandarte's population. This was because these males lived longer and reared a greater proportion of hatched chicks to independence from parental care, not because females mated to males with larger repertoires laid or hatched more eggs. Furthermore, independent offspring of males with larger repertoires were more likely to recruit and then to leave more grandoffspring than were offspring of males with small repertoires. Although we cannot distinguish whether observed fitness variation reflected genetic or environmental effects on males or their offspring, these data suggest that female song sparrows would gain immediate and intergenerational fitness benefits by pairing with males with large song repertoires.