Over the last decade, paternity analysis using molecular markers has revealed that observed mating systems do not necessarily correlate with reproductive systems and thus cannot provide reliable information about male reproductive success (RS). This is especially true for primate species with a complex multimale-multifemale social organization, such as mandrills (Mandrillus sphinx). Using molecular markers for the measurement of individual RS and a comprehensive data set comprising 193 offspring from 27 birth cohorts over a 20-year period of sampling, we investigated the social, genetic, and demographic factors that may influence the probability of paternity by dominant male mandrills, living in a semi–free-ranging colony. We observed a significant skew in RS towards dominant males, with their probability of paternity increasing as the number of adult males in the group increased, and when they were closely related to subordinate adolescent males. Conversely, the probability of dominant males siring infants decreased when the number of simultaneously tumescent females increased. Fewer offspring were sired by dominant males when female partners were closely related to them and when the relatedness between dominant and subordinate adult males increased. These two last points suggests particularly that mechanisms of kin recognition are operating to avoid the costs of inbreeding and may also reflect the lower costs to dominant males of losing conception opportunities to more closely related subordinate adult males. This study is, to our knowledge, one of the first in primates to use an integrative approach and multivariate analysis to show that multiple factors are involved in determining the probability of paternity by dominant males.