EXTRAPAIR FERTILIZATIONS IN RED BISHOPS (EUPLECTES ORIX): DO FEMALES FOLLOW CONDITIONAL EXTRAPAIR STRATEGIES?

The Auk 122(1):57 70, 2005 The American Ornithologists Union, 2005. Printed in USA. EXTRAPAIR FERTILIZATIONS IN RED BISHOPS (EUPLECTES ORIX): DO FEMALES FOLLOW CONDITIONAL EXTRAPAIR STRATEGIES? T W. P. F 1 G M. K Institut für Biologie und Umweltwissenscha en, AG Zoophysiologie und Verhalten, Carl von Ossietzky Universität Oldenburg, Carl von Ossietzky Strasse 9-11, Postfach 2503, D-26111 Oldenburg, Germany A. Although many studies have revealed differences in pa erns of extrapair paternity between species and between populations of the same species, possible differences in female extrapair behavior within populations have received far less a ention. Here, we propose and test the hypothesis that females of the highly polygynous Red Bishop (Euplectes orix) follow two different extrapair strategies, with benefits of extrapair copulations depending on the quality of their social mate ( conditional-extrapair-strategy hypothesis ). In particular, we propose that females mated to low-quality males enhance the genetic quality of their offspring by performing extrapair copulations with males of higher quality than their social mate (in accordance with the good-genes hypothesis), whereas females mated to high-quality males perform extrapair copulations as insurance against temporary infertility of their social mate caused by sperm depletion due to frequent copulations (in accordance with the fertility-insurance hypothesis). Several predictions derived from our hypothesis, regarding differences in frequency and distribution of both unhatched eggs and extrapair young between high-quality and low-quality males, are tested and shown to be supported. Most importantly, hatching failures occurred more o en in territories of high-quality males than in those of low-quality males, and the proportion of unhatched eggs within the territory was positively correlated to breeding synchrony for high-quality males only. Those results suggest that sperm depletion is a risk for females mated to high-quality males with many mating opportunities. The fact that broods of high-quality males without unhatched eggs had a significantly higher proportion of extrapair young than broods of high-quality males with unhatched eggs is consistent with the hypothesis that females mated to high-quality males were successful in performing EPCs as insurance against temporary infertility of their social mates. Furthermore, genetic fathers of extrapair young hatched in territories of low-quality males were found to be of higher quality than the corresponding social fathers, whereas no difference in quality was found between genetic and social fathers of extrapair young hatched in territories of high-quality males. Although we cannot exclude the possibility that the observed pa ern was a result of females mating randomly from within the available pool of extrapair males, it is in accordance with our prediction that females mated to lowquality males perform extrapair copulations with males of higher quality than their social mate to enhance offspring fitness. Overall, the data presented here suggest that females within our study colony adjusted their extrapair behavior to the quality of their social mate, even if we cannot entirely rule out alternative explanations for some of the results obtained. Thus, the study provides support for the conditionalextrapair-strategy hypothesis, which states that within populations, females may follow different extrapair strategies to gain different benefits. Received 14 July 2003, accepted 24 June 2004. 1 E-mail: Thomas.Friedl@uni-oldenburg.de 57

58 F K [Auk, Vol. 122 Key words: conditional-extrapair-strategy hypothesis, Euplectes orix, extrapair paternity, female behavior, fertility insurance, good genes, male quality, Red Bishops, sexual selection, weaverbirds. Fertilizaciones Extra-pareja en Euplectes orix: Siguen las Hembras Estrategias Extra-pareja Condicionales? Resumen. Aunque muchos estudios han indicado que existen diferencias en los patrones de paternidad extra-pareja entre especies y entre poblaciones de la misma especie, las posibles diferencias en el comportamiento extra-pareja de las hembras dentro de una población han recibido mucha menos atención. En este trabajo proponemos y ponemos a prueba la hipótesis de que las hembras de la especie altamente poligínica Euplectes orix siguen dos estrategias extra-pareja diferentes, en las que los beneficios de las cópulas extra-pareja dependen de la calidad de su pareja social ( hipótesis condicional de estrategia extra-pareja ). En particular, proponemos que las hembras apareadas con machos de baja calidad mejoran la calidad genética de sus crías realizando cópulas extrapareja con machos de mayor calidad que su pareja (de acuerdo a la hipótesis de genes buenos), mientras que las hembras apareadas con machos de alta calidad realizan cópulas extra-pareja como un seguro contra la infertilidad temporal de su pareja social causada por agotamiento del esperma debido a cópulas frecuentes (de acuerdo a la hipótesis de fertilidad asegurada). Varias predicciones derivadas de nuestra hipótesis en cuanto a diferencias en la frecuencia y distribución de huevos no eclosionados y pichones extrapareja entre machos de alta y baja calidad, son puestas a prueba y resultan cumplirse. Lo más importante es que los fracasos de eclosión sucedieron más frecuentemente en los territorios de machos de alta calidad que en los de machos de baja calidad, y la proporción de huevos no eclosionados dentro del territorio estuvo correlacionada positivamente con la sincronía de la reproducción sólo para los machos de alta calidad. Estos resultados sugieren que el agotamiento de esperma es un riesgo para las hembras que se aparean con machos de alta calidad que tienen muchas oportunidades para aparearse. El hecho de que las nidadas de los machos de alta calidad sin huevos no eclosionados tuvieron una proporción de pichones extra-pareja significativamente mayor que la de las nidadas de machos del mismo tipo con huevos no eclosionados es consistente con la hipótesis de que las hembras apareadas con machos de alta calidad incrementaron su éxito reproductivo al realizar cópulas extra-pareja como un seguro contra la infertilidad temporal de sus parejas sociales. Más aún, los padres genéticos de los pichones extra-pareja que eclosionaron en los territorios de machos de baja calidad fueron de mayor calidad que los padres sociales correspondientes, mientras que no existió diferencia en la calidad de los padres genéticos y sociales de los pichones extrapareja que eclosionaron en los territorios de machos de alta calidad. Aunque no podemos excluir la posibilidad de que el patrón observado sea el resultado de que las hembras se aparean aleatoriamente con los machos extra-pareja disponibles, este patrón está de acuerdo con nuestra predicción de que las hembras apareadas con machos de baja calidad realizan cópulas extra-pareja con machos de mayor calidad que su pareja social para incrementar la adecuación de sus crías. En general, los datos aquí presentados sugieren que las hembras de nuestra colonia de estudio ajustaron su comportamiento extra-pareja a la calidad de su pareja social, aún si no podemos descartar completamente explicaciones alternativas para algunos de los resultados obtenidos. Así, este estudio respalda la hipótesis condicional de estrategia extrapareja, que propone que al interior de una población las hembras pueden seguir estrategias extra-pareja diferentes para obtener distintos beneficios.

January 2005] Extrapair Strategies in Red Bishops 59 C, the major aims in behavioral ecology and evolutionary biology is to understand why females of many bird species actively seek extrapair copulations (EPCs) outside their pair bond. Possible benefits of EPCs for female birds include insurance against infertility of the social mate; material benefits provided by the extrapair male in terms of courtship feeding, nest defense, or paternal care; and genetic benefits in terms of higher genetic diversity or higher genetic quality of offspring (e.g. Westneat et al. 1990, Birkhead and Møller 1992, Parker and Burley 1998, Ligon 1999, Griffith et al. 2002). In a related study (Friedl and Klump 2002), we investigated extrapair paternity in the Red Bishop (Euplectes orix), a highly polygynous weaverbird (Passeriformes: Ploceidae) from southern Africa with a high rate of extrapair fertilizations (see Friedl and Klump 1999, 2002). In this species, male reproductive success is mainly determined by the number of nests built on the territory, which explains 85% of variance in reproductive success (Friedl and Klump 1999). Number of nests built is also a good indicator of male quality, given that territorial males that survived and established a territory in the following season (thereby demonstrating good quality) built more nests than males that were not able to establish a territory in the following season (Friedl and Klump 1999). Assuming that male quality has some heritable components, female Red Bishops could perhaps be expected to mate preferentially with males that build many nests, to obtain indirect benefits in terms of enhanced offspring fitness. However, a detailed analysis of female se lement in our Red Bishop colony revealed that the observed mating pa ern can best be explained by random female se lement, with every nest having the same probability of being chosen by a female (Friedl and Klump 2000). So, why did females not mate preferentially with males that built many nests? The apparent lack of a female mating preference for males with many nests does not mean that females had no primary mating strategy. Instead, it probably reflects a wellbalanced trade-off between (1) possible benefits of mating with males that build many nests, in terms of enhanced offspring fitness; and (2) increased costs of sharing a territory with other breeding females, in terms of increased nest predation risk (Friedl and Klump 2000). The mating system of the Red Bishop can thus be characterized by two main features. First, males offering many nests on their territories have a higher probability of a racting females, simply by chance, than males with few nests on their territories. That is because, in a mating system in which every nest has the same probability of being chosen by a female, male mating success is directly proportional to the number of nests built. (2) Second, females adopting a random-se lement rule are more likely to se le in the territory of a high-quality male offering many nests than in the territory of a low-quality male offering few nests. Thus, females have a good chance of ge ing a high-quality mate without incurring any costs of mate-searching behavior. As noted above, the increased risk of nest predation associated with sharing a territory with other breeding females poses a limitation on a female s choice of social mate. Therefore, it seemed possible that females try to escape that limitation by seeking EPCs with males that built many nests (thereby indicating high quality) to obtain the genetic benefit of enhanced offspring quality (in accordance with the good-genes hypothesis). However, detailed analysis of frequency and pa ern of extrapair fertilizations in the Red Bishop on a population-wide scale failed to provide unequivocal evidence in support of the good-genes hypothesis (Friedl and Klump 2002). In particular, it failed to reveal differences between high- and low-quality males in the proportion of extrapair young. As an alternative to hypotheses postulating genetic benefits, the fertility-insurance hypothesis states that females seek EPCs as insurance against infertility of their social mate. Permanent infertility in males should be strongly selected against and thus is expected to be rare in natural populations. However, temporary male infertility caused by sperm depletion seems more likely to occur. For example, in Soay sheep (Ovis aries), frequently mating high-ranking males have been shown to suffer from sperm depletion toward the end of the ru ing season (Preston et al. 2001). Sperm limitation or sperm depletion in males have also been reported for such diverse organisms as moths, sandflies, crabs, and fishes (see review in Wedell et al. 2002). In polygynous bird species, in which male mate-guarding is mostly absent, male

60 F K [Auk, Vol. 122 paternity-insurance behavior usually takes the form of frequent copulations (e.g. Birkhead and Møller 1992). Especially in species with a high degree of simultaneous polygyny, males may be forced to copulate repeatedly with several fertile females to whom they are socially mated, to assure paternity. Given that sperm numbers in ejaculates of birds have been shown to decline a er frequent copulations (Birkhead 1991; Birkhead et al. 1994, 1995; Westneat et al. 1998), males mated to many females simultaneously may suffer from temporary sperm depletion. Red Bishops in our study colony showed a high degree of polygyny: males had 18 social mates within one breeding season (Friedl and Klump 1999) and 3 simultaneously fertile social mates (T. W. P. Friedl unpubl. data). Male Red Bishops that are mated to many females simultaneously are thus likely to suffer from temporary sperm depletion, given that male mate-guarding is absent and males copulate frequently with their social mates to assure paternity (Friedl 1998). In the Red-winged Blackbird (Agelaius phoeniceus), a species with a high degree of polygyny similar to that of the Red Bishop, Gray (1997) found that clutches sired by multiple males had a significantly lower percentage of unhatched eggs than clutches sired by a single male and concluded that the higher hatching success in clutches with mixed paternity was a result of females successfully seeking EPCs to avoid egg infertility caused by sperm depletion in their social mates. Insurance against infertility has also been invoked as an explanation for female extrapair behavior in House Sparrows (Passer domesticus; We on and Parkin 1991), Razorbills (Alca torda; Wagner 1992), and both Blue Tits (Parus caeruleus) and Great Tits (P. major) (Krokene et al. 1998). On the basis of those considerations and our failure to find differences between proportions of extrapair young in territories of high- and low-quality males (Friedl and Klump 2002), we propose that female Red Bishops in our study colony followed alternative extrapair strategies, depending on the quality of their social mates ( conditional-extrapair-strategy hypothesis ; herea er CESH ). In particular, we propose that females mated to low-quality males may seek EPCs with males of higher quality than their social mate to enhance offspring quality, in accordance with the good-genes hypothesis. On the other hand, females mated to high-quality males may seek EPCs as insurance against temporary infertility in their social mates (caused by sperm depletion resulting from frequent copulations), in accordance with the fertilityinsurance hypothesis. Therefore, all females within the population would be expected to seek EPCs regardless of the quality of their mate, which would explain why we did not find any difference in rate of cuckoldry between high-quality and low-quality males (Friedl and Klump 2002). Several predictions must be fulfilled before the CESH can be accepted as a likely explanation for the pa ern of extrapair paternity found in the Red Bishop colony studied. Some of those predictions do not exclusively support the CESH; a result in accordance with a particular prediction may be explainable by different mechanisms (see below). However, if most or all of the predictions listed below were fulfilled, we would consider such an outcome as being consistent with our hypothesis. On the other hand, if the data were not consistent with several of the predictions, we would conclude that the pa ern of extrapair paternity found in our study colony cannot be explained by the CESH. Assuming that sperm depletion is a common problem among high-quality males and a particular problem during years of high breeding activity, predictions associated with the CESH are as follows: (1) Unhatched eggs should be common. (2) Unhatched eggs should be more common among high-quality males than among low-quality males. (3) Frequency of unhatched eggs among high-quality males should be highest in years of high breeding activity, but unrelated to breeding activity for low-quality males. (4) Frequency of EPCs should be lowest in years of low breeding activity for females mated to high-quality males, but unrelated to breeding activity for females mated to lowquality males. (5) Frequency of unhatched eggs should increase with breeding synchrony for females mated to high-quality males, but not for females mated to low-quality males. (6) In females mated to high-quality males, frequency of extrapair young should be lower in broods containing unhatched eggs than in broods containing no unhatched eggs (following the reasoning by Li eld [1994] and Gray [1997]; see We on and Parkin [1991] for an alternative view). (7) Females mated to low-quality males should seek EPCs with males of higher quality

January 2005] Extrapair Strategies in Red Bishops 61 than their social mate, whereas females mated to high-quality males should seek EPCs with males similar in quality to their social mate. Here, we evaluate each of those predictions separately and discuss data obtained with regard to whether they support the CESH and whether there are alternative explanations for the results observed. M Study site. Extrapair paternity was studied in a colony of Red Bishops in the Addo Elephant National Park, Eastern Cape, South Africa (33 26 S, 25 45 E) during four consecutive breeding seasons (1993 1994 through 1996 1997). The study was conducted under license from the Provincial Administration of the Cape of Good Hope, Chief Directorate, Nature and Environmental Conservation (License nos. 625/93, 653/94, 144/95, 92/96) and with permission from the South African National Parks Board. The breeding site was a circular small dam (~250 m 2 ) completely surrounded by an inner belt of bulrushes (Typha capensis) and an outer belt of common reeds (Phragmites australis). Territories appeared to be similar in quality, with most territories adjacent to the water and consisting mainly of bulrushes. Territories provided no food (foraging always takes place away from territories) or other material benefits for females, other than the nests, which were not a limited resource, given that most territories contained one or more empty nests at any time during the breeding season (Friedl 2004). Natural history of the study species. The Red Bishop is a sexually dimorphic weaverbird species with a wide distribution in sub-saharan Africa. Breeding males have black and brilliant orange scarlet plumage, whereas females and immatures are brownish above and whitish below. In the nonbreeding season, males have an eclipse plumage (i.e. nonbreeding plumage) that resembles that of females. Red Bishops are gregarious at all times. During the nonbreeding season, they are nomadic, moving around in large mixed-species flocks (together with other bishops, weavers, and widows) sometimes numbering thousands of birds and frequenting grasslands and agricultural land. Red Bishops breed in colonies, which are most o en found in reedbeds or bullrush stands around water. At the beginning of the breeding season, males in breeding plumage establish small territories of a few square meters. Males construct several nests within their territory, to which they try to a ract females, which use the nests for breeding. Territories are vigorously defended throughout the breeding season, and male intruders are usually chased away immediately upon entering the territory. When a female enters a territory, the male performs courtship displays until she either flies off or permits copulation. Male mate-guarding is absent, but pair copulations are performed frequently during the fertile period, which is shortly before and during the egg-laying stage, a er which the female is usually ignored by the territory owner. Females are solely responsible for incubation and nestling provisioning, and males do not show any form of parental behavior. Median clutch size of 670 clutches produced in the colony during the study period was 3 eggs (range = 1 5), with 98.8% of clutches containing 2 4 eggs. All copulations observed during the study were initiated by females showing the typical copulation solicitation display of passerine birds, regardless of whether the copulations were within-pair copulations or EPCs. Whenever a male tried to mount a female without being solicited, he was repulsed by vigorous pecks. The few EPCs observed were performed in the territory of the extrapair male, which indicates that females actively visited males for the purpose of seeking EPCs. For more details of Red Bishop breeding behavior, see Friedl (2004). General methods. Information about locations of territories, identities of territory holders, and social parents of nestlings were obtained through detailed daily observations of male and female behavior throughout the breeding seasons. Throughout the study, all nests were checked daily, and we recorded all eggs that were fully incubated but nevertheless did not hatch (i.e. all unhatched eggs found in clutches where at least one young hatched). From those data, we were able to calculate percentage of broods within the territory of a male that contained at least one unhatched egg (percentage of broods with unhatched eggs [PBU]) and percentage of unhatched eggs out of all eggs in those broods (percentage of eggs unhatched, PEU) for every territorial male. However, because all subsequent analyses yielded essentially the same results for PBU and PEU, we present detailed

62 F K [Auk, Vol. 122 results for PEU only. Partial clutch losses (i.e. loss of eggs to predation before hatching date) occurred in only 15 of the 182 fully incubated clutches with fledglings for which paternity was determined and that were subsequently used in analyses. Furthermore, PEU did not differ significantly between fully incubated clutches with and without egg losses before hatching (Fisher exact test, P > 0.7). Thus, predation is unlikely to have influenced our results. A weakness of our study is that we did not inspect unhatched eggs for embryonic development. Thus, although we assume that hatching failures in our study colony were caused by eggs not being fertilized (for reasons given below), we cannot rule out the alternative possibility that hatching failures were caused by embryonic mortality. We calculated a breeding-synchrony index for each clutch in which at least one egg hatched by counting the number of all breeding a empts within the territory that were started within the last five days of clutch initiation (i.e. all breeding a empts from day 5 to day 0, where day 0 is the day of first egg in the clutch). Thus, the breeding-synchrony index represents number of simultaneously fertile females within the territory during the fertile period of the female that produced the clutch in question. For each territorial male, we then calculated an average breeding-synchrony index for all clutches in his territory in which at least one egg hatched. That average breeding-synchrony index could then be related to the proportion of unhatched eggs within the territory. Paternity was determined using nonradioactive, multilocus DNA fingerprinting with the digoxigenin-labeled oligonucleotide probe (GGAT) 4. Genetic analysis and paternity assignment were conducted following the procedures described in detail by Friedl and Klump (1999, 2002). Data analysis. Our hypothesis proposes that females follow conditional extrapair strategies depending on the quality of their social mate. To test whether the predictions of this hypothesis are fulfilled, we had to divide the males into two groups: high-quality and low-quality. We used number of nests as our measure of male quality. In Red Bishops, number of nests built by males within their territories is the most important determinant of male mating success, explaining 85% of the observed variance (Friedl and Klump 1999). In all four study seasons, number of nests built was significantly correlated to time during which a male was able to defend a territory (Spearman correlation coefficients r s ranging from 0.813 to 0.944; all P < 0.001). Given that holding a territory and building many nests is energetically demanding (see Friedl and Klump 1999), one would expect that males building many nests and holding a territory for a long time would have lower probability of establishing a territory in the following breeding season, compared with males that built fewer nests and held their territory for a shorter period. Contrary to that expectation, males that demonstrated their quality by surviving to and establishing a territory in the subsequent breeding season built more nests than males that were not able to establish a territory in the following season (Friedl and Klump 1999), which suggests that those differences indeed reflect variation in male quality. If number of nests built reflects male quality, one would expect that a male s status as below or above average in number of nests built is stable from year to year. In the study period, five of seven territorial males present in three seasons and three of five territorial males present in four seasons were consistently below or above average in number of nests built. Because those data represent a mixture of independent and non-independent data, they cannot be analyzed adequately with standard parametric or nonparametric tests. Therefore, we performed a resampling procedure to test whether the observed consistency among seasons is different from what would be expected from a random distribution. We randomly assigned a status of below- or above-average to seven males, 3 each (for the seven territorial males present in three seasons); and to five males, 4 each (for the five territorial males present in four seasons); calculated number of males that were consistent in status among the three and four seasons, respectively; and repeated that procedure 10,000. The probability of five or more out of seven males being consistently below or above average in three seasons was 1.45%; probability of three or more out of five males being consistently below or above average in four seasons was 1.82%. Thus, males for which we had data for either three or four years exhibited a statistically significant tendency to build either above- or below-average numbers of nests. (Using a more conventional binomial

January 2005] Extrapair Strategies in Red Bishops 63 test would yield similar results, with P = 0.013 for five of seven males being consistent over three years and P = 0.016 for three of five males being consistent over four years.) Given those results, we are confident that number of nests built is a good indicator of male quality. We therefore categorized all territorial males within a breeding season as having built more or less than the average number of nests within that season and assigned them to the groups of highquality and low-quality males, respectively. We were able to identify genetic fathers of 19 out of 61 (31%) extrapair young from territories of high-quality males (paternity determined for a total of 373 nestlings) and for 5 out of 15 (33%) extrapair young from territories of low-quality males (paternity determined for a total of 56 nestlings). The difference in total number of nestlings with known paternity between low-quality and high-quality males is simply a consequence of skewed mating success among territorial Red Bishop males, with half of the territorial males (i.e. high-quality males) accounting for >85% of all fledglings (see above). Three broods from territories of high-quality males contained more than one extrapair young sired by the same extrapair male; those cases entered the analysis only once, decreasing the sample size of genetic fathers of extrapair young from territories of high-quality males from 19 to 14. Four males were found to be the genetic father of extrapair young in different broods. However, we regarded extrapair young from different broods with the same genetic father as independent data points, because they represented independent mating decisions of different females. To test whether social and genetic fathers of extrapair young differed with regard to male quality, we compared the males in terms of number of nests built and whether they had a territory in the previous season (which reflects age or prior residency as well as quality) or in the subsequent breeding season (which reflects survival and quality). All data analyses were conducted with the so ware package SPSS for Windows, version 11.0, using nonparametric statistical procedures (Bortz et al. 1990, Sokal and Rohlf 1995). When analyzing pa erns of unhatched eggs and extrapair paternity for all four breeding seasons combined, we used every male only once (selected at random) to assure statistical independence. All P-values given are two-tailed. R Overall frequency of unhatched eggs. During the study, a total of 403 clutches produced hatchlings (i.e. those clutches were fully incubated). Those 403 clutches contained 1,190 eggs that were fully incubated, of which 102 (8.57%) failed to hatch. In 86 (21.3%) of those clutches, at least one egg did not hatch. Frequency of unhatched eggs in relation to male quality. Overall, high-quality males had a higher PEU than low-quality males (high-quality males: n = 40, median = 7.2%, range = 0.0 28.6%; low-quality males: n = 28, median = 0.0%, range = 0.0 33.3%; Mann-Whitney-U test: U = 409, P = 0.041; every male used only once and selected at random). Thus, hatching failures occurred more o en in territories of high-quality males than in those of low-quality males, in accordance with the CESH. Occurrence and frequency of PEU for both low-quality and high-quality males for each season separately is shown in Figure 1. In the 1993 1994 and 1996 1997 seasons, high-quality males had a significantly higher PEU than lowquality males; in the 1994 1995 season, there was a strong tendency in the same direction (see Fig. 1). Only the 1995 1996 season differed in that there was even a weak tendency for low-quality males to have a higher PEU (see Fig. 1). Frequency of unhatched eggs in different seasons. The 1995 1996 season was characterized by an unusually low amount of breeding activity, with only 226 eggs laid in 85 clutches, as compared with >550 eggs laid in >170 clutches in each of the other three breeding seasons investigated (Friedl 2002). The reduced breeding activity was probably caused by an unusually low amount of rainfall during the 1995 1996 season, which in turn led to reduced availability of both grass seeds and insects, the main food resources of adults and nestlings of the Red Bishop (Friedl 2002). Furthermore, in the 1995 1996 season, average number of females per territorial male was significantly lower than in any of the other three seasons (Kruskal-Wallis test: H = 9.1, df = 3, P = 0.028; 1993 1994: median = 5.5, range = 0 13; 1994 1995: median = 5.0, range = 0 12; 1995 1996: median = 3.0, range = 0 7; 1996 1997: median = 5.0, range = 0 18). Differences between seasons with respect to number of females per territorial male were even greater when only high-quality males are considered (Kruskal-Wallis test: H = 29.3, df = 3, P < 0.001).

64 F K [Auk, Vol. 122 F. 1. Frequency distribution of percentage of unhatched eggs within the territory (PEU) for lowquality and high-quality males for each study season. Size of dots represents number of data points. Two-tailed P-values refer to comparisons using the Mann-Whitney U-test. Thus, sperm depletion in males caused by frequent copulations with their social mates to assure paternity was less likely to occur in the 1995 1996 season. Consequently, the CESH predicts that the proportion of unhatched eggs within territories of high-quality males should be lower in the 1995 1996 season than in the other three seasons. As predicted, PEU of highquality males differed significantly between 1995 1996 and the other seasons (Mann- Whitney-U test: U = 175, P = 0.002), being lower in the 1995 1996 season (median PEU in season 1995 1996: 0.0%; median PEU in the other three seasons: 7.7 12.15%). There was no significant difference between years with respect to PEU for low-quality males; in fact, there was a tendency toward a higher PEU in the 1995 1996 season than in the other three seasons (Mann- Whitney U-test: U = 19.5, P = 0.12). Frequency of cuckoldry in different seasons. The CESH also proposes that females mated to highquality males should perform EPCs for fertility insurance less o en in seasons with reduced breeding activity, given that sperm depletion in their social mate is less likely to occur. In contrast, cuckoldry level is not expected to differ between seasons of different breeding activity for lowquality males. Our results are in accordance with those predictions. There was a significant difference between 1995 1996 (the season with reduced breeding activity) and the other three seasons (with similar high levels of breeding activity) in the proportion of high-quality males that suffered from being cuckolded (Chi-square test: χ 2 = 4.25, df = 1, P = 0.039). Although only 33.3% of all high-quality males were affected by cuckoldry in 1995 1996, 70% of all high-quality males, on average, had extrapair young within their territories in each of the other three seasons (range = 57 79%). In contrast, proportion of low-quality males that suffered from being cuckolded did not differ between 1995 1996 (66.7%) and the other three seasons (average = 44.4%; χ 2 = 0.509, df = 1, P = 0.48). Proportion of unhatched eggs within the territory in relation to breeding synchrony. For high-quality males, there was a significant positive correlation between their average breeding-synchrony index (see above) and the proportion of unhatched eggs within their territories (r s = 0.33, P = 0.04, n = 39; every male used only once and selected at random; see Fig. 2A). No such relationship was

January 2005] Extrapair Strategies in Red Bishops 65 F. 2. Relationship between breeding synchrony index (see text for calculation) and percentage of unhatched eggs within the territory. Size of dots represents number of data points. (A) Highquality males (n = 39). (B) Low-quality males (n = 28). found for low-quality males (r s = 0.08, P > 0.60, n = 28; every male used only once and selected at random; see Fig. 2B). Those results suggest that for a clutch in the territory of a high-quality male, probability of containing unhatched eggs increased with number of fertile females within the territory at time of clutch initiation. Proportion of extrapair young in broods with and without unhatched eggs. We first calculated, for each male, the average proportion of extrapair young in all his broods of a season, with and without unhatched eggs. We then compared average proportions of extrapair young in broods of high-quality males with unhatched eggs to average proportions of extrapair young in broods of high-quality males without unhatched eggs. When data for all four study years were combined (every male used only once and selected at random), broods of highquality males without unhatched eggs had a significantly higher proportion of extrapair young than broods of high-quality males with unhatched eggs (see Fig. 3). Pairwise comparison of social and genetic fathers of extrapair young. Results of pairwise comparisons of social and genetic fathers for extrapair young hatched in territories of high- and lowquality males are shown in Figure 4. Although the sample size of extrapair young from territories of low-quality males for which the extrapair father could be identified was small (n = 5), we found a significant difference in number of nests built (our measure of male quality). Sires of extrapair young from territories F. 3. Comparisons of observed proportions of extrapair young within the territory between broods of high-quality males with and without unhatched eggs. Box plots show medians, interquartiles, 10% and 90% percentiles, and means (black circles). Two-tailed P-value refers to a comparison using the Mann-Whitney U-test. of low-quality males built significantly more nests than the males they cuckolded (see Fig. 4A), which suggests that females mated to lowquality males obtained EPCs with males who were of higher quality than their social mate. In contrast, there was no significant difference

66 F K [Auk, Vol. 122 F. 4. Results of a pairwise comparison of number of nests built between social and genetic fathers of extrapair young. Box plots show medians, interquartiles, 10% and 90% percentiles, and means (black circles). Two-tailed P-values refer to pairwise comparisons using the Wilcoxon test. (A) Extrapair young hatched in territories of low-quality males. (B) Extrapair young hatched in territories of high-quality males. between numbers of nests built by extrapair and within-pair males when comparisons were made for extrapair young found on territories of high-quality males (see Fig. 4B). Genetic fathers of extrapair young from territories of low-quality males were more o en territory owners in the previous season than were social fathers (Chi-square test: χ 2 = 6.67, df = 1, P = 0.01), which indicates that females mated to low-quality males preferred extrapair males older than their social mates. However, probability of being a territory owner in the following season did not differ between genetic and social fathers of extrapair young from territories of low-quality males (χ 2 = 0.4, df = 1, P = 0.53). For extrapair young from territories of high-quality males, there were no significant differences between genetic and social fathers in territory ownership in either the previous or the next breeding season (Chi-square tests; both P > 0.10). That sires of extrapair young that hatched in territories of low-quality males were of higher quality than the cuckolded social fathers does not necessarily imply that females mated to low-quality males actively chose extrapair males of higher quality. It is also possible that females obtained extrapair fertilizations from high-quality males simply by random choice. That is because, for females mated to low-quality males, even a random choice of an extrapair male would, on average, result in higher quality of the extrapair male as compared with the low-quality social mate. One means of distinguishing between those two possibilities is to determine whether extrapair males were of above-average quality. The hypothesis of active female choice of high-quality extrapair males would be supported if extrapair partners built an above-average number of nests. If they did not, it would suggest that females more likely selected extrapair males at random. We tested that hypothesis a er eliminating the focal male s number of nests from the calculation of the population mean. There was a slight bias for extrapair sires to build an above-average number of nests (four of five), but the trend was not significant (Binomial-test, P = 0.38). For extrapair young that hatched on territories of high-quality males, sires and presumptive fathers (i.e. territory holders) did not differ significantly in quality (see above). However, because the power of that test was low, we conducted an additional test to determine whether sires of extrapair young were of above-average quality: they were not (8 of 14 above average: Binomial-test, P = 0.79). Our analyses, therefore, do not allow us to reject the hypothesis that females chose their extrapair partners randomly. However, we note that the power of our tests were severely compromised by small sample sizes.

January 2005] Extrapair Strategies in Red Bishops 67 D Frequency of hatching failures. Unhatched eggs were common in our study population, with 8.6% of fully incubated eggs failing to hatch. Here are some comparative data on frequency of hatching failures reported in other species: 3 4% for Great Tits and 4 8% for Blue Tits (Kempenaers et al. 1996, Krokene et al. 1998), 6.5% for House Sparrows (Cordero et al. 1999), 6.8% for Collared Flycatchers (F. albicollis; Török et al. 2003), and quite exceptional 15.8% for Tree Swallows (Tachycineta bicolor; Kempenaers et al. 1999). Thus, frequency of unhatched eggs in our Red Bishop study colony was high, but not exceptionally so. However, given that the median clutch size in Red Bishops is only three eggs, one unhatched egg constitutes a loss of one-third of the current breeding a empt for a female. Therefore, we think that the risk of hatching failure can be considered an important selective force in Red Bishops and that females adjusting their extrapair behavior to that risk should be at a selective advantage. Do female Red Bishops follow conditional extrapair strategies? Our data are consistent with nearly all the predictions of the CESH (see above) and suggest that female Red Bishops within our study colony follow two different extrapair strategies, depending on the quality of their social mate. As predicted by our hypothesis, high-quality males had higher proportions of unhatched eggs within their territories than low-quality males. Furthermore, for high-quality males, proportions of unhatched eggs and of males that lost paternity within their territories were significantly lower in the 1995 1996 season (when sperm depletion was less likely than in the other three seasons). In comparisons among years, low-quality males showed no significant difference in proportions of unhatched eggs and of males that lost paternity within their territories. High-quality males exhibited a positive correlation between proportion of unhatched eggs and number of simultaneously fertile females within the territory (see Fig. 2), in accordance with our hypothesis. Number of fertile females within the territory might even be used by females to assess the risk that their mate is temporarily sperm-depleted. In addition, broods of high-quality males with unhatched eggs had a significantly lower percentage of extrapair young than broods of high-quality males without unhatched eggs (see Fig. 3), which suggests that the EPC-seeking behavior of females was effective as insurance against temporary infertility of their social mates. Finally, we found that extrapair young hatched in territories of low-quality males were sired by males of higher quality than their social father, whereas there were no differences in quality between sires and social fathers of extrapair young hatched in territories of high-quality males. Both pa erns were predicted by the CESH. However, we cannot exclude the possibility that females simply mated randomly from within the available pool of extrapair males. Further research must a empt to more critically evaluate that final prediction of the CESH. Are there alternative explanations for our results? The seven predictions listed in the introduction are not exclusively supportive of the CESH; some results that are consistent with the CESH might also be explained by other mechanisms. A weakness of the present study is the fact that we did not inspect unhatched eggs for signs of embryonic development (see above). Therefore, we cannot rule out the possibility that hatching failures were caused by embryonic mortality, rather than egg infertility. In a review of egg hatchability in birds, Koenig (1982) found that hatching failures occurred more o en in species with higher frequency of sexual interactions with conspecifics during the female s fertile period. It might thus be argued that the results reported here are a ributable to breeding-density effects on both egg hatchability and EPC rates. Red Bishop females in territories of high-quality males may be more likely to suffer from hatching failures (caused by embryonic mortality rather than egg infertility) than females in territories of low-quality males, because of higher levels of female female aggression, male harassment, or both. Hatching failures caused by high levels of sexual interactions could also explain both the lower incidence of unhatched eggs found in the 1995 1996 season (with reduced overall breeding activity) and the positive correlation between proportion of unhatched eggs and breeding synchrony in high-quality males. Although female female aggressions have been reported for numerous polygynous bird species (Slagsvold and Li eld 1994), behavioral observations revealed that in Red Bishops,

68 F K [Auk, Vol. 122 females usually tolerate other nesting females on the territory and do not show aggressive behavior toward them (Friedl 2004). There is li le competition among females nesting on the same territory, because male Red Bishops provide no parental care or food resources within the territory (Friedl and Klump 1999). Furthermore, female Red Bishops se led more o en than expected by chance on territories already containing other nesting females, which suggests that resident females did not discourage further se lement (Friedl and Klump 2000). Thus, female female aggression is unlikely to cause hatching failures in Red Bishops. What about male harassment? Given that, in general, males have more to gain through EPCs than females, it might be argued that EPCs should be mainly male-driven. The pa ern of hatching failures observed here could then be a ributed to embryonic mortality caused by male harassment of females as has been found, for example, in Dunnocks (Prunella modularis; Davies 1985). However, our behavioral observations suggest that EPCs are driven by females, in that all observed copulations were solicited by females, males cannot coerce females, and observed EPCs were performed on the territory of the extrapair mate. Thus, male harassment seems an unlikely explanation for the observed pa ern of hatching failures in our study colony. Other facts that argue against female female aggression and male harassment as causes of the observed pa ern of hatching failures include our findings (1) of a lower proportion of unhatched eggs in the 1995 1996 season for high-quality males but not for low-quality males, (2) of a positive correlation between breeding synchrony (as assessed for the territory) and proportion of unhatched eggs for high-quality males but not for low-quality males, and (3) that the incidence of extrapair young in the 1995 1996 season (with reduced breeding activity) was lower than in the other seasons in broods of high-quality males. None of those results can be explained by breedingdensity effects. To sum up, the data presented here are in accordance with predictions derived from the CESH, which proposes that female Red Bishops mated to low-quality males may seek EPCs with males of higher quality than their social mate to enhance offspring quality, whereas females mated to high-quality males may seek EPCs as insurance against temporary infertility of their social mate. We are aware that there are other possible explanations for some of our results (as discussed above) and that an experimental approach is needed to provide unequivocal evidence for our hypothesis (see also Sheldon 1994). However, in our opinion, the results presented here are both consistent with and can best be explained by the CESH. Variation in female extrapair behavior between and within populations. Recently, a ention has been focused on differences in levels of extrapair paternity among populations of the same species (Petrie and Kempenaers 1998). That some populations of the same species differ in their rates of extrapair paternity, whereas there is no such difference between populations of other species, can be explained only if costs and benefits of EPCs to males and females vary among populations (Petrie and Kempenaers 1998). However, it is likely that costs and benefits of EPCs for females vary even within a population; individual females might, therefore, be expected to adjust their extrapair behavior to potential costs and benefits, as has been suggested by Møller (1992), Wagner (1992), Gowaty (1996), and Wagner et al. (1996). Although the models presented by those authors differ in their circumstances and in factors proposed to affect the probability of females seeking EPCs, the main idea is that, within a population, all females that engage in EPCs gain the same type of benefit. That may be an increase of genetic quality of offspring (Møller 1992, Gowaty 1996, Wagner et al. 1996) or insurance against male infertility (Wagner 1992). The CESH differs from those models in proposing that EPCs can benefit females within a population in more than one way (i.e. increase of genetic quality of offspring or insurance against temporary infertility of the social mate), depending on the quality of the social mate. The possibility that, even within populations, females follow conditional extrapair strategies has important implications for the study of extrapair paternity and sexual selection in general. It suggests that future studies investigating costs and benefits of EPCs for females should pay a ention to the possibility that costs and benefits are not necessarily the same for all females within a population. Moreover, it implies that extrapair behavior of individual

January 2005] Extrapair Strategies in Red Bishops 69 females is not fixed but can be adjusted to variations of potential costs and benefits. The possibility of different female extrapair strategies within populations may also encourage re-analysis of data from studies that failed to provide evidence for any of the proposed benefits of EPCs at the population level. Such a reanalysis may reveal new and interesting results if males and females that are likely to differ with regard to potential costs and benefits of EPCs are analyzed separately. A We are grateful to the National Parks Board of South Africa and the Provincial Administration of the Cape of Good Hope, Chief Directorate, Nature and Environmental Conservation for permission to conduct this study in the Addo Elephant National Park. We thank the park warden, L. Moolman, and the whole park staff for continuous support. A very special thank-you to J. and M. Adendorff for all their help and friendship. We thank B. Kempenaers, J. Li eld, M. T. Murphy, R. Wagner, and an anonymous reviewer for helpful comments that greatly improved the manuscript. Genetic analysis were facilitated by grants from the Leonard Lorenz Sti ung and the Sti ung der Freunde der TU München. T.W.P.F. was supported by grants from the DAAD (German Academic Exchange Service), the Friedrich-Schiedel- Sti ung, the Sti er-verband, and by a scholarship from the Technical University Munich. This study was supported by grants from the Deutsche Forschungsgemeinscha (Kl 608/11-1 and Kl 608/13-1). L C B, T. R. 1991. Sperm depletion in the Bengalese Finch, Lonchura striata. Behavioral Ecology 2:267 275. B, T. R., F. F, E. J. P, A. S. 1995. Ejaculate quality and the success of extrapair copulations in the Zebra Finch. Nature 377:422 423. B, T. R., A. P. M. 1992. Sperm Competition in Birds: Evolutionary Causes and Consequences. Academic Press, London. B, T. R., J. P. V, A. P. M. 1994. Male sperm reserves and copulation behaviour in the House Sparrow, Passer domesticus. Proceedings of the Royal Society of London, Series B 256:247 251. B, J., G. A. L, K. B. 1990. Verteilungsfreie Methoden in der Biostatistik. Springer-Verlag, Berlin. C, P. J., J. H. W, D. T. P. 1999. Within-clutch pa erns of egg viability and paternity in the House Sparrow. Journal of Avian Biology 30:103 107. D, N. B. 1985. Cooperation and conflict among Dunnocks Prunella modularis in a variable mating system. Animal Behaviour 33:628 648. F, T. W. P. 1998. Sexual selection in the Red Bishop (Euplectes orix). Ph.D. dissertation, Technische Universität München, Munich, Germany. F, T. W. P. 2002. The effect of rainfall on the breeding behaviour in the Red Bishop, Euplectes orix. Ostrich 73:181 184. F, T. W. P. 2004. Breeding behaviour of the Red Bishop (Euplectes orix): A synthesis and new observations. Vogelwarte 42:178 190. F, T. W. P., G. M. K. 1999. Determinants of male mating success in the Red Bishop (Euplectes orix). Behavioral Ecology and Sociobiology 46:387 399. F, T. W. P., G. M. K. 2000. Nest and mate choice in the Red Bishop (Euplectes orix): Female se lement rules. Behavioral Ecology 11:378 386. F, T. W. P., G. M. K. 2002. Extrapair paternity in the Red Bishop (Euplectes orix): Is there evidence for the good-genes hypothesis? Behaviour 139:777 800. G, P. A. 1996. Ba les of the sexes and origins of monogamy. Pages 21 52 in Partnership in Birds (J. M. Black, Ed.). Oxford University Press, Oxford. G, E. M. 1997. Do female Red-winged Blackbirds benefit genetically from seeking extrapair copulations? Animal Behaviour 53:605 623. G, S. C., I. P. F. O, K. A. T. 2002. Extra pair paternity in birds: A review of interspecific variation and adaptive function. Molecular Ecology 11:2195 2212. K, B., F. A, A. J. N, A. A. D. 1996. Genetic similarity, inbreeding and hatching failure in Blue Tits: Are unhatched eggs infertile? Proceedings of the Royal Society of London, Series B 263:179 185.