Darwin's 1 hypothesis that male secondary sexual ornaments evolve through female preferences is theoretically plausible 2-7, but there is little experimental field evidence that such preferences exist. 8-10 I have studied female choice in relation to male tail length in the long-tailed widowbird, Euplectes progne, and report here that males in which the tail was experimentally elongated showed higher mating success than males having normal or reduced tails. The possibility that intrasexual competition among males maintains the long tail was not supported: males with shortened tails held their territories as long as did other males. These results suggest that the extreme tail length in male long-tailed widowbirds is maintained by female mating preferences.
Male long-tailed widows have the most extreme sexual ornaments among Euplectes, an African genus of polygynous weaverbirds (the Ploceidae) 11. Reproductive adult males are black except for a red epaulet on the wing, but the most conspicuous feature is the tail: 6-8 of the 12 tail feathers are ~0.5 m long, the rest being one- to two-thirds as long. The tail during flight display is expanded vertically into a deep, long keel below the male as he flies with slow wingbeats 0.5-2 m above the territory. Displaying long-tailed widows are visible from over 1 km distance on their 0.5-3-hectare territories. The territories lie in open grassland on the Kinangop plateau, Kenya, where E. progne is one of the most common birds, and where the present study was performed between November 1981 and March 1982. Females are inconspicuous, being mottled brown, with short tails (≈7 cm). They build their nests on the territories of the males, in the upper third of the 0.5- 0.8 m-high grass, Eleusine jaegeri, and raise their young (2-3) unaided by the male. These features make the long-tailed widow suitable for a test of the theory of intersexual selection of male ornaments.
Darwin 1 and Fisher 2 suggested that further evolution of an ornament ceases when it becomes so large that it reduces survival enough to exactly balance the mating advantage. For this to be so, females must prefer larger than normal-sized ornaments; otherwise there can be no balance between the two selection pressures. In the present experiment, females chose from males with shortened, normal or elongated tails. The Darwin-Fisher theory therefore predicts that mating success should be highest among males with elongated tails, and lowest among males with shortened tails.
The experiment included nine groups, each containing four individually colour-ringed males, of similar initial tail length and territory quality. Territory boundaries were determined by plotting on maps the locations of male displays and attacks, using cattle fences, streams and vegetation features as landmarks. In each matched group, the following treatments were randomly allocated among the four males. The tail was cut to ~14 cm in one of them; each removed feather was then attached with rapidly (~1 s) hardening cyanoacrylate glue to the corresponding feather in another male, whose tail was thus prolonged by an average of 25 cm. About 3 cm of each removed feather was first cut off and glued back on to its counterpart in the 'shortened' male, which hence was manipulated in a similar manner to the 'elongated' male. The two remaining males were controls; one was ringed only. To check whether the cut-and-glue operation influenced male behaviour or female choice, the tail of the second control male was cut off at the midpoint; each feather was then glued back on again. This operation shortened the tail by only 1 cm (≈2%), which is probably not noticeable by females. Uneven joints or ends of glued feathers were trimmed with a scalpel. The joints were difficult to see from more than 1 m.
As capture and manipulation might influence subsequent behaviour, which could confound interpretation, flight displays and territorial disputes were counted in each male for 30 min 1-5 days before, and 10-14 days after the treatment.
Male mating success was estimated by the number of active nests (containing eggs or young) on each territory, for which I searched for about 1 h just after treatment of the male, and at weekly intervals for 1 month afterwards. No new clutches were laid after early January. To avoid bias, I searched each territory in proportion to its area of nesting habitat (tall, rank grass). The first count provided a standard with which to compare male mating success after tail treatment, estimated by the number of clutches laid during the remainder of the breeding season, after the day on which the male was manipulated. This use of each male as his own control reduces the importance of differences in territory quality, which influences female choice of mate in the long-tailed widow (M.A., in preparation). Counts spanned 1 month, so a few nests might represent re-laying, but this should not bias the result as treatments were randomized within each group.
Euplectes females usually seem to mate with the male on whose territory they nest 12, 13. The possibility that a female might mate with one male and nest on the territory of another should not produce bias favouring the Darwin-Fisher prediction but should reduce the likelihood of detecting female mating preferences in the present study. Whereas attractive males in such cases receive more matings, some of their females may build on the less crowded territories of less attractive males. Group defence against nest predators is poorly developed, and there is no other evidence that female long-tailed widows would benefit from clumped breeding; the nests are well dispersed within a territory.
Females will be selected to respond to a character only if it varies among potential mates 10. This is the case in long-tailed widows: the fully grown tails (no blood quills) of seven territorial males had a mean length of 49.6 cm, with c.v. = 9.4%. There was no significant correlation between male tail length and number of nests on the territory before the experiment.
However, the present randomized block design is unsuitable for detecting such a correlation; it was used to make the experiment maximally sensitive to any importance of tail length.
The two types of control males (I, cut and restored; II, only ringed—see Fig. 1) did not differ significantly in tail length, display rate or any other measured character. Therefore, they should be equivalent from a female perspective, so I have treated them as one category below, with two representatives in each group of four males.
Before tail treatments, there were only minor differences in mating success between 'shortened', control and 'elongated' males (Fig. la). After treatment, however, mating success changed as predicted by the female choice hypothesis, with lowest success for males having shortened tails, and highest success for males with elongated tails (Fig. 1b). Hence, tail length apparently did influence mate choice: females preferred those males having the longest tails. As the main difference was between 'elongated' males and other males (including controls), the difference did not result from possible destruction of species-specific features in males having shortened tails.
FIGURE 1 - see the PDF version of the paper (online or on SAKAI) to see a larger version of the figure
Fig. 1 Mating success in male long-tailed widows subjected to different tail treatments. a, Mean number of active nests per territory for the nine males of the four treatment categories, before the experiment. Numbers of nests for the nine males in each of the four categories are given at the bottom of the bars, always in the same order of matched 4-group. b, Number of new active nests in each territory after treatment of males. The following procedure was used to test for differences in mating success between the treatment categories. For each male I calculated the difference between the number of new active nests on his territory during the remainder of the breeding season after tail treatment, and the original number of nests on the territory before tail treatment. (Subtraction of the original number of nests reduces the influence of initial variation among males and territories.) These differences were used for matched comparisons (with respect to each group of four males) between shortened, control and elongated males. As predicted by the Darwin-Fisher theory of sexual selection, males with elongated tails became significantly more successful (as measured above, compared with before the tail treatment) than shortened males and control males II (P < 0.05 in both cases, paired randomization tests). For the four categories tested together, there was a significant trend of increasing mating success as tail length increased from shortened via control to elongated males (P = 0.03, Pitman randomization test 18 adapted to the present experimental design; no standard test fits this situation, with two different treatments (shortened and prolonged tails) and two control categories, and an alternative hypothesis which predicts a specified trend).
