Tail elongation in the polygynous widowbirds (Euplectes spp.) has evoked both adaptive and non–adaptive explanations. Female choice has been shown in the three longest tailed species (20–50 cm), whereas an agonistic function was proposed for a medium–tailed (10 cm) widowbird. To test the generality and directionality of sexual selection on tail length in widowbirds, we experimentally investigated selection in the relatively short–tailed (7 cm) red–shouldered widowbirds (E. axillaris). Prior to territory establishment, males were assigned to four tail–treatment groups; control, short, long and supernormal (similar to a sympatric long–tailed congener). No effects on male competition were detected as the groups were equally successful in acquiring territories of similar size and quality. However, mating success among the 92 territorial males was strongly skewed in favour of supernormal–tailed males (62% of active nests; 5.2 ± 1.3 nests per territory). Long–tailed males also acquired more nests (1.9 ± 0.7) than control (0.7 ± 0.5) and short–tailed (0.5 ± 0.3) males, while the latter two groups did not differ significantly. These results support a general, open–ended female preference for long tails in widowbirds and may represent a receiver bias that arose early in their divergence from the short–tailed weaverbirds (Ploceinae).