Barbary macaques (Macaca sylvanus), now restricted in the wild to a few isolated forested areas of Morocco and Algeria, are present in a free-ranging colony on Gibraltar. For many decades, the Gibraltar colony was exposed to multiple bottlenecks due to highly nonrandom removal of animals, followed by repeated introductions of animals from North Africa. Moreover, because of complete isolation, Gibraltar's several social groups of macaques provide an ideal system to study the genetic consequences of dispersal in cercopithecines in situ. Predictions of genetic consequences due to male-biased dispersal in cercopithecines will be different for autosomal and maternally inherited genetic markers, such as the control region of the mitochondrial DNA. We used a panel of 14 highly polymorphic microsatellite loci and part of the hypervariable region I of the mitochondrial control region to estimate genetic structure between five social groups in Gibraltar. Surprisingly, for autosomal markers, both classical summary statistics and an individual-based method using a Bayesian framework detected significant genetic structure between social groups in Gibraltar, despite much closer proximity than wild Algerian and Moroccan populations. Mitochondrial data support this finding, as a very substantial portion of the total genetic variation (70.2%) was found between social groups. Using two Bayesian approaches, we likewise identified not only a small number of male first-generation immigrants (albeit less than expected for cercopithecines) but also unexpectedly a few females. We hypothesize that the culling of males that are more likely to disperse might slow down genetic homogenization among neighbouring groups, but may also and more perversely produce selection on certain behavioural traits. This may have important repercussions for conservation, as it could lead to evolutionary changes that are not due to inbreeding or genetic drift.