Accurate measures of fitness are important for both basic research on sexual selection and applied conservation actions to promote genetic diversity. For polygynous mammals, good estimates of male reproductive success are often critically important, but especially difficult to obtain. Because the genetic contribution of males is impossible to directly measure in the field, investigators have developed surrogate measures of fitness based on behavioral observations. Such measures are founded on the assumption that observed mating success can reliably predict reproductive success, yet only a few studies have been in a position to validate the accuracy of this assumption. We studied the bison herd at Fort Niobrara National Wildlife Refuge for 8 years, conducting intensive behavioral observations on breeding behavior during the rut (2003–2009) and collecting tissue samples of calves born the following year (2004–2010) for genetic paternity analysis. Our results reveal 2 major trends also observed in other studies: Estimates of mating success were positively correlated with reproductive success when we pooled the entire herd across age classes or years. However, copulatory success did a poor job of predicting the actual number of offspring sired by individual males. For example, 44% of observed matings did not result in the birth of offspring, and 60% of the copulations that did produce a calf did not accurately predict the sire bull. Generalized linear mixed model analysis revealed that observation of mating by a given bull in itself had no predictive power regarding likelihood of paternity, whereas total copulations per season, dominance status, and age of bull or dam significantly influenced the probability of siring offspring. Although use of behavioral data was unable to predict the sire for particular cows, it did give insights into patterns of reproductive success that use of genetic data alone could not provide, such as the role of alternate mating strategies and sperm competition for male reproductive success. We conclude that both behavioral and genetic measures of fitness are needed to understand sexual selection and meet the challenges faced by species of conservation concern.