We use the assignment technique and a new approach, the ‘novel allele technique’, to detect sex-biased dispersal in great reed warblers Acrocephalus arundinaceus. The data set consisted of immigrants and philopatric birds in a semi-isolated population in Sweden scored at 21 microsatellite loci. Fourteen cohorts were represented of which the four earliest were used to define a reference population. Female immigrants had lower assignment probability than males (i.e. were less likely to have been sampled in the reference population), and carried the majority of ‘novel alleles’ (i.e. alleles observed in the population for the first time). The difference in number of novel alleles between sexes was caused by a strong over-representation of females among the few individuals that carried several novel alleles, and there was a tendency for a corresponding female bias among individuals with low assignment probabilities. Immigrant males had similar or lower reproductive success than females. These results lead us to conclude that important interregional gene flow in great reed warblers depends on relatively few dispersing females, and that the novel allele technique may be a useful complement to the assignment technique when evaluating dispersal patterns from temporally structured data.