Behavioral Ecology Vol. 15 No. 4: 555–563 DOI: 10.1093/beheco/arh051 Repertoire size, sexual selection, and offspring viability in the great reed warbler: changing patterns in space and time Wolfgang Forstmeier and Bernd Leisler Max Planck Research Centre for Ornithology, Vogelwarte Radolfzell, Schloss Moeggingen, Schlossallee 2, D-78315 Radolfzell, Germany Only a few studies have focussed on the consistency of sexual selection patterns in space and time. One such case is the great reed warbler (Acrocephalus arundinaceus), for which studies in Germany in 1981–1982 and Sweden in 1987–1991 strongly suggested that the size of a male’s song repertoire was the target of mate choice and sexual selection. Studying the same German population once again in 1994–2000, we investigated the consistency of these patterns over time as well as between populations. Our reanalysis of the data from 1981–1982 shows that male repertoire size was positively correlated with male pairing success (harem size) and with clutch size (adjusted for seasonal effects), whereas no such correlations were found during 1994–2000 in the same population. We suggest that the earlier correlations were probably caused indirectly by covariation with territory quality, and that a decline in population size has changed the role of territory quality. In the Swedish population, an earlier study found a striking correlation between the size of a male’s repertoire and the viability of its offspring, suggesting good-gene effects. In contrast, we found no such correlation in the German population, neither in 1981–1982 nor in 1994–2000. We conclude that repertoire size does not seem to be a very reliable indicator of variation in male quality. Interestingly, the analysis of data from 1994–2000 showed that male pairing success was strongly correlated with measures of strophe length and immediate versatility, traits that have been found to reflect male longevity. Future studies will have to show whether these performance-related traits are more powerful indicators of male quality than is repertoire size. Key words: female choice, good genes, male-male competition, song, territory quality. [Behav Ecol 15:555–563 (2004)] he case of the great reed warbler, Acrocephalus arundinaceus, is probably the most frequently cited example showing that song is subject to directional female mating preferences. In the laboratory, females preferentially solicited copulations from males with large syllable repertoires (i.e., the number of different syllable types in their song; Catchpole et al., 1986). Moreover, two field studies found that males with large repertoires attracted more females and tended to mate polygynously, whereas males with small repertoires often remained unmated (Catchpole, 1986; Catchpole et al., 1985; Hasselquist, 1994, 1998). Finally, Hasselquist et al. (1996) showed that females paired to males with small repertoires sought extrapair copulations from neighboring males that had larger repertoires. Females may have profited from this behavior because offspring viability was positively correlated with the repertoire size of the genetic father (Hasselquist et al., 1996). Thus, several studies suggested that repertoire size may be an indicator of male quality and that females preferentially mate with good singers. Only a few studies have investigated the consistency of patterns of mate choice and sexual selection in space and time (see Dale et al., 1999). The aim of the present study is to present new data on sexual selection in the great reed warbler, which allows us to make such comparisons between different periods of time as well as between different study populations. Two great reed warbler populations have so far been studied extensively. First, there is a German breeding population at the Fränkische Weihergebiet in northern Bavaria, where de- T Address correspondence to W. Forstmeier, who is now at Department of Animal and Plant Sciences, University of Sheffield, Western Bank, Sheffield S10 2TN, UK, E-mail: [email protected]. Received 25 November 2003; revised 28 August 2003; accepted 28 August 2003. tailed information on the mating system has been collected from 1977 onward (Beier 1981; Catchpole et al., 1985; Leisler et al., 1995, 2000). This population has been studied in terms of how song relates to sexual selection during 1981 (Catchpole, 1986) and 1982 (partly published in Leisler et al., 1995). Second, there is a breeding population at lake Kvismaren in south-central Sweden, where detailed investigations started in 1983 (Bensch and Hasselquist, 1991; Bensch et al., 1998). Studies focussing on song characteristics of male great reed warblers have been conducted between 1987 and 1991 (Hasselquist, 1994, 1998; Hasselquist et al., 1996). Here we present new data collected in the above-mentioned German population during a 7-year period from 1994–2000, and we examine these data in relation to the earlier findings. To facilitate the comparison with the new data, we combine the data from 1981 (Catchpole, 1986) with those from 1982 (Leisler et al., 1995), and we reanalyze some of the former results with an increased power. We also investigate the consistency of patterns between the German and the Swedish population. We focus on three aspects of male reproductive success: (1) harem size, (2) clutch size, and (3) offspring viability. Studying the German great reed warbler population in 1981, Catchpole (1986) found a strong positive correlation between the size of a male’s repertoire and its success in attracting females (measured as harem size). A similar, but less strong, pattern was also observed in 1982 (Leisler et al., 1995). However, in both years, territory quality was a better predictor of harem size than was male repertoire size, and the analyses published so far do not support any conclusion as to whether there was a female preference for good singers beyond that which was explained by territory quality. Thus, females may have aimed for direct benefits resulting from high territory quality and may not have chosen particular males for their high genetic quality. A similar problem applies Behavioral Ecology vol. 15 no. 4 International Society for Behavioral Ecology 2004; all rights reserved. 556 to the Swedish study population. Male repertoire size was positively correlated with harem size, but this pattern was confounded by variation in male age (Hasselquist, 1994, 1998). In this population, repertoire size and pairing success both increased with age, and the correlation between repertoire and harem size was nonsignificant when male age was statistically controlled for (Hasselquist, 1994, 1998). Catchpole (1986) found a particularly strong positive correlation between male repertoire size and reproductive success. This was primarily owing to variation in harem size, but also to a positive trend between repertoire size and clutch size (r ¼ .27; p ¼ .092; data from 1981) (Catchpole, 1986). This latter correlation may have arisen from (1) differential allocation (i.e., females paired with attractive males increased their investment by producing larger clutches; see Burley, 1986; Sheldon, 2000), (2) differential attraction (i.e., good singers attracted higher-quality females; see Burley, 1986), or (3) variation in territory quality (i.e., good singers held higher-quality territories which facilitated laying larger clutches). Although the available data do not allow us to distinguish between these alternatives, we want to test whether the observed trend between clutch size and repertoire size would still hold for an increased data set. Studying the Swedish great reed warbler population, Hasselquist et al. (1996) found a positive correlation between male repertoire size and the viability of the offspring, suggesting that repertoire size reflects the genetic quality of males. We use the data collected during the first (1981–1982) and second (1994–2000) study period to test whether this finding also holds for the German population. In addition to repertoire size, we focus on a song parameter termed ‘‘syllable switching,’’ which is the number of syllabletype switches within a strophe. This measure reflects strophe length as well as immediate versatility. As described elsewhere, we found that syllable switching (but not repertoire size) was positively correlated with male longevity, suggesting that this song measure might reveal aspects of male quality (Forstmeier W, Hasselquist D, Bensch S, and Leisler B, unpublished data). Unfortunately, measures of syllable switching are not available for the earlier studies. METHODS Population data The study site is located in the Franken area of southern Germany (49 409 N, 10 519 E, 290 m above sea level). For a detailed description of the area, see Beier (1981) and Leisler et al. (1995). The great reed warbler population breeding at this site has been monitored intensively since 1973. Data have been collected at least weekly each year from the end of April onward during the whole breeding season, and every second day during the peak breeding-season in June (averaging approximately 38 visits per year). We provided almost all adult birds with color rings and nestlings with metal rings. We mapped all singing males and searched all reed fringes for nests from the water side. For each nest we tried to determine the date of deposition of the first egg (assuming that females lay one egg per day) as well as clutch size and the number of young fledged. We defined the size of a male’s harem as the number of females that were nesting together with him (males defend the nests of their females). For the present study, we focus on the breeding seasons from 1994–2000, as song recordings of males were collected during this period (see below). On average, 23.1 males (range ¼ 18–32) were holding a territory in the area each year, resulting in a total of 162 male breeding seasons. This sample comprised 87 individual males, and 86 of them were Behavioral Ecology Vol. 15 No. 4 color-ringed. We tried to collect one song recording from each male in each year, but because of time limitations and restrictions of the permit, we managed to sample only a subset of the males present. No male was recorded twice within the same season. On average, we recorded 15.6 males (67% of those present) per breeding season (range ¼ 12–22), resulting in a total of 109 recordings stemming from 62 individual males (71% of all males). We carefully examined whether taking this subset introduced any systematic bias into our sample. However, males that we managed to record did not differ significantly from those that we failed to record in any trait measured: wing length (see below; t 73 ¼ 1.2; p ¼ .25), body mass (see below; t 73 ¼ 0.3; p ¼ .76), harem size (KruskalWallis test; v2 ¼ 2.0; df ¼ 2; p ¼ .39), male age (first-year versus older; Fisher’s Exact test; v2 ¼ 2.9; p ¼ .12), or whether male age was known from ringing as a nestling or not (see below; Fisher’s Exact test; v2 ¼ 0.6; p ¼ .51). We measured male pairing success in terms of the number of females attracted (harem size). Out of the 109 male breeding seasons, males remained unmated in 28 cases (26%), monogamy occurred in 59 cases (54%), bigamy was recorded 21 times (19%) and the status of one male could not be determined with certainty (leaving us with n ¼ 108). The population-wide mean of harem size varied considerably between the years (from 1.22–0.61). We used two different methods of accounting for this interannual variation. First, when estimating the effect of song characteristics on harem size, we used generalized linear models with a Poisson error distribution (using SAS, procedure: ‘‘genmod’’; options: ‘‘error ¼ poisson’’ and ‘‘dscale’’) for which year could be entered as a separate factor (i.e., a dummy variable accounting for sex-ratio variation). We used male breeding seasons (n ¼ 108) as the independent statistical unit. This widely used approach maximizes the power for detecting patterns but assumes that there is no major effect of the individual male. We selected this approach (rather than deleting repeated measures of the same male), because we were concerned with making a type II error (not finding what earlier studies have found) rather than falsely rejecting the null hypothesis. Second, to avoid the possible effect of pseudoreplication inherent in the first approach, we used lifetime averages of males (n ¼ 62) in the second approach. We first standardized all harem sizes in relation to the population mean of harem size of the respective breeding season (referred to as residual harem size; ranging from 1.22 to 1.24 females). This procedure is to scale male pairing success in relation to the number of females available. Second, we calculated the mean of these residuals for each male (averaged among years). Multiple measures of a male’s song characteristics were also averaged among years (without prior standardization). Neither repertoire size nor syllable switching of individual males increased with age (Forstmeier W, Hasselquist D, Bensch S, and Leisler B, unpublished data), thus taking lifetime averages did not introduce any bias caused by variation in male longevity. We adopted the same procedures when reanalyzing the data from 1981 (mean harem size ¼ 1.00; song recordings available for n ¼ 37 males; Catchpole, 1986) and 1982 (mean harem size ¼ 1.03; song recordings available for n ¼ 27 males, 15 of which were new to the study area; Leisler et al., 1995). This time, however, we did not enter a year effect into the generalized linear models, owing to the very small interannual variation in mean harem size. To examine variation in clutch size in relation to song characteristics, we adopted the same two methodological approaches as above: (1) we used generalized linear models with Poisson error structure, considering each clutch as statistically independent (1994–2000: n ¼ 105; 1981–1982: n ¼ 66) and Forstmeier and Leisler • Song and sexual selection entering date of clutch initiation as a covariate to account for seasonal changes in clutch size, and (2) we first controlled for seasonal effects by taking the residuals from a regression of clutch size over the date of deposition of the first egg measured in days after the first of May (y ¼ 5.665 0.032x; r ¼ .50; p , .0001; n ¼ 966; data from 1973–2000). Second, we calculated mean residual-clutch size for each male (averaged among all nests in a male’s lifetime; 1994–2000: n ¼ 47 males; 1981–1982: n ¼ 39 males). When capturing males in mist-nets, we measured wing length and body weight. Nineteen out of the 62 males were trapped in more than one season, so multiple measures were sometimes available. Male age was known exactly from ringing as a nestling for 25 of the 62 males (40%). For the others, which we observed for the first time as adults, we can only provide an estimate of minimum age (at least 1 year when first recorded as adult), but it is likely that many of these birds were actually 1 year old (data not shown). We measured the survival of the offspring produced by the 62 males included in this study (data from 1994–2000). These males altogether fledged 418 young (in the years from 1989– 2000), and 44 of them (9.5%) were seen again as recruits in the following years (up until 2001). As 98 of 165 adult breeders (59%) in the study area (all males and females from 1994– 2001) had not been ringed as nestlings in the area, we estimate that only about two of five recruits will return to their natal area. This sets some limitations on the estimation of offspring survival rates in relation to song characteristics of their father. A second problem is that we cannot be sure whether all of these young were actually sired by their social fathers. A paternity study conducted in this population (in 1992–1997) showed that 19 out of 194 nestlings (9.8%) were sired by an extrapair male (Leisler et al., 2000). This paternity analysis included 140 offspring, which are part of the 441 fledglings (33%) considered here. Fourteen of these (10%) were sired by an unknown extrapair male. By excluding these 14 young from our analysis, we can expect to reduce the confounding effect of extrapair paternity by approximately one third. We also looked at offspring survival of those males that had been recorded from 1981–1982 (n ¼ 48 color-ringed males). These males altogether fledged 291 young (in the years from 1977–1986), and 18 of them (6.2%) were seen again as recruits in later years. Song characteristics Male great reed warblers have two different kinds of song that differ in length. Unmated males sing long strophes (long song) consisting of a wide variety of high-amplitude elements, whereas mated males that are guarding their female produce short strophes (short song) that consist of relatively few, stereotyped, low-amplitude syllables (Catchpole, 1983). Apparently in an effort to attract a second female, some mated males start producing long song again once their first female is incubating (Hasselquist and Bensch, 1991). Playback experiments in the laboratory as well as in the field have shown that long song is attractive to females, whereas short song functions as a territory defense (Catchpole, 1983; Catchpole et al., 1986). We recorded males that were singing a minimum of 30 consecutive strophes of long song. Thirty-three males were recorded in only 1 year, 17 males in 2 years, seven males in 3 years, four males in 4 years, and one male in 5 years (109 recordings in total). Recordings were made by using a Sony TC-D5 PROII recorder and a Sennheiser K6M67 microphone and were analyzed with Avisoft SASLab Pro 3.4 using the following settings: sampling frequency, 22,050 Hz; 16 bits; 557 time resolution, 5.8 ms; and bandwidth, 111 Hz. From each of the 109 recordings, we extracted the following two song parameters. Repertoire size We categorized syllable types (all done by the same person to ensure consistency of classifications) by visual inspection of printed spectrograms. We defined the number of different syllable types found in 30 consecutive strophes as repertoire size. As shown in Catchpole (1986), new syllable types only rarely turned up after the first 20 strophes had been analyzed. Our measure of a male’s repertoire size showed significant repeatability among years (R ¼ .34; p ¼ .005; for details, see Forstmeier W, Hasselquist D, Bensch S, and Leisler B, unpublished data). Differences in mean repertoire size that seem to exist between various studies (see Catchpole, 1986; Fischer et al., 1996; Hasselquist, 1994; this study) simply reflect methodological differences, that is, splitting versus lumping of syllable types by different observers (Forstmeier W, Hasselquist D, Bensch S, and Leisler B, unpublished data). Note that the estimates of repertoire size of birds from the first period of study (1981–1982) were made by a different person (Catchpole, 1986) and are therefore not directly comparable to our estimates of males from 1994–2000. In cases in which such direct comparison was needed, we ztransformed repertoire estimates within studies. Syllable switching For each strophe, we counted the number of syllable-type switches within the strophe. This parameter is highly correlated with strophe length measured in seconds (r ¼ .78), as long strophes contain more syllable switches than do short strophes. Besides that, syllable switching can be interpreted as a measure of immediate variety, as it typically equals repertoire size of the strophe minus one. Nevertheless, there was only a very weak positive correlation between syllable switching (measured per strophe and averaged among 30 strophes) and total repertoire size (measured per 30 strophes; r ¼ .22, p ¼ .023, n ¼ 109 recordings; and r ¼ .14, p ¼ .27, n ¼ 62 male lifetime means). Our measure of a male’s syllable switching showed significant repeatability among years (R ¼ .36; p ¼ .003) (see Forstmeier W, Hasselquist D, Bensch S, and Leisler B, unpublished data). Strophe length as measured in seconds was slightly less repeatable (R ¼ .28; p ¼ .017) and therefore did not seem worth a separate analysis. However, as strophe length is a widely used measure in the birdsong literature, we present some of the main results for strophe length as well. RESULTS Throughout this section, generalized linear models (GLMs) based on male breeding seasons as the statistical unit are presented in the main text, whereas the corresponding regression analyses based on lifetime averages of males are used for graphical illustration and are presented in the figure legends. Repertoire size and male pairing success During 1981–1982, male repertoire size had a strong positive effect on harem size (GLM, v21;62 ¼ 9.6; p ¼ .002) (Figure 1a). This correlation was not confounded by variation in male age (GLM, age: v21;61 ¼ 0.9; p ¼ .34; repertoire size: v21;61 ¼ 0.9; p ¼ .004). However, as outlined in Catchpole (1986), variation in territory quality seemed to be a major confounding factor. Territory quality was estimated in terms of the amount of reedwater interface (‘‘edge length’’ measured in meters), which 558 Figure 1 Relation between a male’s repertoire size and his success in attracting females in 1981–1982 (r ¼ .42; n ¼ 52; p ¼ .002) (a) and in 1994–2000 (r ¼ .04; n ¼ 62; p ¼ .78) (b). The difference between the two effect sizes is significant (p ¼ .042). Residual harem size is a measure of male pairing success relative to that of other males present in the same year. Average values are shown for individuals present in more than one season. The fact that repertoire size estimates in panel b are considerably larger than in panel a reflects methodological differences (splitting versus lumping of syllable types by different observers). appears to be important in both feeding and nesting ecology (Bensch et al., 2001; Catchpole et al., 1985; Leisler et al., 1995). Harem size was strongly positively correlated with edge length (GLM, v21;60 ¼ 16.9; p , .0001; data for two males missing), and repertoire size was also strongly correlated with edge length (r ¼ .44; n ¼ 62; p , .001). When both parameters were entered together, edge length, but not repertoire size, had a significant effect on harem size (GLM, edge length: v21;59 ¼ 7.9; p ¼ .005; repertoire size: v21;59 ¼ 2.4; p ¼ .125). The same result was obtained with the second method of analyzing lifetime averages. By using partial regression (as done by Catchpole, 1986), we found no significant effect of repertoire size on residual harem size when edge length was statistically controlled for (redge¼const ¼ .20; df ¼ 47; p ¼ .18). Thus, we could not detect a female preference for males with large repertoires beyond that which was already explained by variation in territory quality. In contrast to these patterns, we found no positive effect of repertoire size on harem size during the second period of Behavioral Ecology Vol. 15 No. 4 study, 1994–2000 (GLM, year: v26;100 ¼ 13.2; p ¼ .041; repertoire size: v21;100 ¼ 0.8; p ¼ .376) (Figure 1b). When the full data set from both study periods within the same model was used, and separate slopes for the repertoire-size effect for the two periods were estimated, the difference in slopes was not large enough to reach statistical significance (p ¼ .092). However, with the second approach of analyzing lifetime averages, the effect sizes of repertoire size differed significantly between the two study periods (p ¼ .042) (Figure 1). To understand this discrepancy between the two study periods, it would be important to know how territory quality was related to repertoire size and harem size during this second study period. Unfortunately, no territory characteristics were measured after 1982. Thus, we have no direct way of estimating these correlations, but we can test whether specific territories were associated with these parameters in some way. During the 7 years of the second study period, 20 territories had been occupied consecutively by two or more different males (53 territory owners). An ANOVA shows that there was not a significant association between those territories and the repertoire size of the males occupying them (F19,33 ¼ 1.1; p ¼ .37), nor was there an association between territories and residual harem size (F19,33 ¼ 1.3; p ¼ .23). In contrast, we found a highly significant association between territories and the body weight of the territory owners (F17,22 ¼ 3.4; p ¼ .004), which means that owners of the same territory were more similar in body weight to each other than expected by chance. This indicates that the above analysis had the power to detect such associations. However, it might be argued that such an analysis comprises only the territories of highest quality (those that were occupied repeatedly), and that the main variation in territory quality might be found between multiply occupied territories and those that had been occupied only once. Yet again, the owners of these two types of territories did not differ in repertoire size (t107 ¼ 0.7; p ¼ .49; using male breeding seasons as independent samples), and these territories also did not differ in residual harem size (t106 ¼ 0.7; p ¼ .48). Clutch size in relation to repertoire size For the first period of study, we found a significant positive effect of a male’s repertoire size on the clutch size produced by its females (GLM, accounting for seasonal effects: date: v21;63 ¼ 17.6; p , .0001; repertoire size: v21;63 ¼ 4.6; p ¼ .032) (Figure 2a). Variation in territory quality did not seem to be a confounding factor, as the correlation between edge length and clutch size was very weak (GLM, date: v21;63 ¼ 18.9; p , .0001; edge length: v21;63 ¼ 0.4; p ¼ .536). When edge length as a covariate was entered, there was still a positive effect of repertoire size on clutch size (GLM, date: v21;62 ¼ 17.4; p , .0001; edge length: v21;62 ¼ 0.1; p ¼ .745; repertoire size: v21;62 ¼ 4.2; p ¼ .040). In contrast, repertoire size was not positively correlated with clutch size during the later years (GLM, date: v21;102 ¼ 69.2; p , .0001; repertoire size: v21;102 ¼ 0.4; p ¼ .529) (Figure 2b). When the full data set from both study periods within the same model was used, and separate slopes for the repertoiresize effect for the two periods was estimated, these slopes differed significantly (p ¼ .034). Correlation between repertoire size and offspring viability Thirty-five of the 48 males from the first period of study produced at least one fledgling in any of the years from 1977– 1986 (mean ¼ 8.3, range ¼ 2–23 fledglings). Eighteen of these 291 fledglings (6.2%) were resighted as recruits. Forstmeier and Leisler • Song and sexual selection Figure 2 Relation between a male’s repertoire size and the size of the clutches produced by his females in 1981–1982 (r ¼ .30; n ¼ 39; p ¼ .065) (a) and in 1994–2000 (r ¼ .21; n ¼ 47; p ¼ .16) (b). The difference between the two effect sizes is significant (p ¼ .024). Residual clutch size is adjusted for variation in laying date. Following the methods used by Hasselquist et al. (1996: Figure 2a), we plotted the number of a male’s lifetime recruits over the number of lifetime fledglings, excluding the males that produced no fledglings. The equation of this regression was y ¼ 0.072x 0.084 (r ¼ .428; n ¼ 35; p ¼ .010). As males without fledglings cannot produce any recruits, we forced this regression through the origin (y ¼ 0.065 [60.014 SE]x; t34 ¼ 4.6; p , .0001) and took the residuals from this regression. In contrast to the findings of Hasselquist et al. (1996), the residual number of lifetime recruits was not significantly correlated with repertoire size (r ¼ .10; n ¼ 35; p ¼ .55) (Figure 3a). However, this correlation coefficient did not differ significantly from that obtained by Hasselquist et al. (1996) (r ¼ .32; n ¼ 48 versus r ¼ .10; n ¼ 35; p ¼ .32), and the power of our test, assuming an effect size as found by Hasselquist et al. (1996), was only 62%. During the second period of study, 44 of the 62 males produced at least one fledgling in any of the years from 1989– 2000 (mean ¼ 9.5, range ¼ 3–26 fledglings). Up until 2001, 44 of these 418 fledglings (9.5%) were resighted in the study area 559 Figure 3 Relation between a male’s repertoire size and the relative postfledging survival of his offspring in 1981–1982 (r ¼ .10; n ¼ 35; p ¼ .55) (a) and in 1994 –2000 (r ¼ .11; n ¼ 44; p ¼ .50) (b). The difference between the two effect sizes is not significant (p ¼ .374). The relative survival of a male’s offspring is quantified by taking the residuals from a regression of the number of lifetime recruits over the number of lifetime fledglings that were produced by a male. as recruits. We took residuals from a regression of the number of lifetime recruits over the number of lifetime fledglings (y ¼ 0.107 [60.017 SE]x; t43 ¼ 6.4; p , .0001). Also in this sample, the residual number of lifetime recruits was not positively correlated with repertoire size (r ¼ .11; n ¼ 44; p ¼ .50) (Figure 3b). This correlation coefficient differed significantly from that obtained by Hasselquist et al. (1996) (p ¼ .046), and the power of our test was 71%. When the data from 1981–1982 were pooled with those from 1994 –2000 (adjusting for the differences in repertoire size estimates), the overall correlation was close to zero (r ¼ .03; n ¼ 79; p ¼ .77) and the power of this test was 91%. A problem with this result is that the analysis does not control for the occurrence of extrapair paternity, which should concern approximately 10% of the offspring (Leisler et al., 2000). Paternity data are available for 140 out of 418 fledglings (33%) from the second period of study, and these include 14 extrapair young (10%). By deleting these 14 offspring with unknown genetic fathers from the above analysis (three out of 44 males were concerned), we can hope to reduce the confounding effect of extrapair paternity by approximately one third. However, the correlation between Behavioral Ecology Vol. 15 No. 4 560 residual lifetime recruits and repertoire size did not change toward a positive direction (r ¼ .12; n ¼ 44; p ¼ .45). Finally, we repeated the analysis, basing it on the survival of the 126 fledglings for which we could confirm that the social father had also been the genetic father. Note that this leaves us with a greatly reduced sample of only 21 males. However, again, we could not find a positive correlation between offspring survival and repertoire size of the father (r ¼ .10; n ¼ 21; p ¼ .60). Syllable switching and sexual selection Syllable switching was measured only during the second period of study. This song trait showed a significant positive effect on harem size (GLM, year: v26;100 ¼ 12.9; p ¼ .045; syllable switching: v21;100 ¼ 6.6; p ¼ .010) (Figure 4a). The effect changed only marginally when the outlier (Figure 4a) was removed (GLM, year: v26;99 ¼ 11.6; p ¼ .071; syllable switching: v21;99 ¼ 6.6; p ¼ .010). When the more widely used parameter strophe length as a substitute of syllable switching was used, a significant effect was found as well (GLM, year: v26;98 ¼ 15.4; p ¼ .017; strophe length: v21;98 ¼ 5.9; p ¼ .015; data for two males lacking). Unlike in the first study period, male age was a potentially confounding factor in this analysis, as it significantly affected harem size (GLM, year: v26;100 ¼ 15.9; p ¼ 0.015; age: v21;100 ¼ 6.0; p ¼ .014). However, when age was entered as a covariate, the effect of syllable switching was still significant (GLM: year: v26;99 ¼ 16.3; p ¼ .012; age: v21;99 ¼ 4.9; p ¼ .027; syllable switching: v21;99 ¼ 5.5; p ¼ .019). This was also true when strophe length was used as a substitute of syllable switching (data not shown). Territory quality did not seem to be a confounding factor, as there was no association between multiply occupied territories and the syllable switching of the males occupying them (F19,33 ¼ 1.1; p ¼ .40). Moreover, multiply occupied territories did not differ from singly occupied territories in terms of syllable switching of their owners (t107 ¼ 1.2; p ¼ 0.24; using male breeding seasons as independent samples). A male’s syllable switching had no positive effect on clutch size (GLM, date: v21;102 ¼ 66.4; p , .0001; syllable switching: v21;102 ¼ 1.2; p ¼ .276) (Figure 4b) or offspring viability (r ¼ .09; n ¼ 44; p ¼ .57) (Figure 4c). Excluding the 14 extrapair young from the latter analysis did not affect the outcome (r ¼ .07; n ¼ 44; p ¼ .65) ,and restriction to the cases in which paternity was known did not yield a significant result either (r ¼ .04; n ¼ 21; p ¼ .87). DISCUSSION Repertoire size and male pairing success Figure 4 The number of syllable-type switches per strophe in the songs of males in relation to a male’s pairing success (r ¼ .36; n ¼ 62; p ¼ .004) (a), clutch size produced by a male’s females (r ¼ .14; n ¼ 47; p ¼ .36) (b), and postfledging survival of a male’s offspring (r ¼ .09; n ¼ 44; p ¼ .57) (c). With the outlier removed these correlations were as follows: (a) r ¼ .36, n ¼ 61, p ¼ .005; (b) r ¼ .19, n ¼ 46, p ¼ .21; and (c) r ¼ .06, n ¼ 43, p ¼ .69. For further details see legends of Figures 1 through 3. Two previous studies have found a positive correlation between the repertoire size of male great reed warblers and harem size. First, studying the German great reed warbler population in 1981, Catchpole (1986) found a very strong correlation between repertoire size and harem size (r ¼ .49), and this correlation remained to some extent when territory quality was statistically controlled for by using partial regression (redge¼const ¼ .28; n ¼ 37; p ¼ .093; Catchpole, 1986). After combining these data with those collected in the same population in the following year, our analysis shows that this partial correlation is very weak at best (redge¼const ¼ .20; n ¼ 50; p ¼ .18). We conclude that the strong correlation between repertoire size and harem size (r ¼ .42; based on the full data set) is probably caused indirectly. Males with large repertoires occupied the territories of highest quality, and females preferentially settled in the best territories. Forstmeier and Leisler • Song and sexual selection Second, studies on the Swedish breeding population also found a significant positive correlation between repertoire size and harem size (r ¼ .36; Hasselquist, 1994, 1998). In this case, it appeared that the correlation was indirectly caused by covariation with male age. After the age-related increase in repertoire size was controlled for (which exists in the Swedish but not the German population; Forstmeier W, Hasselquist D, Bensch S, and Leisler B, unpublished data), there was not even a positive trend for males with larger age-adjusted repertoire size to attract more females than others (Hasselquist, 1994, 1998). In conclusion, both studies could not detect a significant female preference for pairing with males having large repertoires in addition to what was attributable to variation in age or territory quality. The new results presented in this article (from 1994–2000) are in agreement with this interpretation, as no female preference was apparent. The strong correlation observed between repertoire size and territory quality during the first period of study suggests that repertoire size might be related to male-male competition (as found in several other species: Hiebert et al., 1989; Krebs et al., 1978; Yasukawa, 1981) rather than to female choice. For the remaining question, namely, why the relationship between repertoire size and territory quality has changed over the past two decades, we suggest the following explanation. When the first study was conducted, there were on average 38 males holding territories in our study area. In the later years the numbers of territory owners ranged from 18 to 32 males only (mean ¼ 23). If one assumes that this population decline was not caused by deterioration of the breeding habitat (which is suggested by an increasing annual reproductive success; data not shown), but by other factors (data not shown), then the intensity of male-male competition may have been relaxed during the later period. It might well be that all males had the opportunity to settle in territories of high quality, and with the lack of variation in territory quality, the correlation with repertoire size may have disappeared. This idea is supported by the fact that we could not detect any female preferences for certain territories (ANOVA using repeatedly occupied territories, and comparison between singly and repeatedly occupied territories). Clutch size in relation to repertoire size During the first period of study, we found that females paired to males with large repertoires laid larger clutches, which, assuming that repertoire size reflects male quality, would be consistent with idea of differential allocation (Burley, 1986; Sheldon, 2000). However, this finding should be interpreted with caution. First, we cannot distinguish between differential allocation and differential attraction, that is, the possibility that males with large repertoires attracted higher-quality females. Second, we cannot exclude the possibility that females adjusted their clutch size in response to some aspect of territory quality other than edge length, and that repertoire size was also correlated with this unknown aspect of territory quality. In other words, controlling statistically for variation in edge length does not mean that all aspects of territory quality are held constant (see also Sheldon, 2000). Thus, given that we could not detect a female preference for males with large repertoires (partial regression), it appears somewhat unlikely that females differentially allocated resources in response to male repertoire size. Rather it might be that repertoire size was strongly correlated with territory quality (which is only partly described by edge length) and that females adjusted their clutch size in relation to territory quality. Interestingly, the correlation between clutch size and repertoire size has changed significantly between the two periods of study 561 (Figure 2), just as the correlation between harem size and repertoire size has disappeared (Figure 1). We consider it likely that these changes over time were caused by the same factor, namely, by a loss of the covariation between repertoire size and territory quality. Repertoire size and offspring viability Despite high statistical power, we found no positive correlation between a male’s repertoire size and the probability of survival of his offspring. Although we were not able to check the paternity of most of the offspring included in the analysis, there were no indications that partially controlling for this confounding factor would shift the correlation in a positive direction. Moreover, if it is true that males with low repertoire size are typically cuckolded by extrapair males with larger repertoires (Hasselquist et al., 1996), we should expect that males with large repertoires do not lose paternity in their own nests. This means that our estimates of offspring survival (the y-axis in Figure 3b) would be most reliable for males with large repertoires. However, for the top 25% of males with largest repertoires offspring survival was still 0.35 SD below the overall mean. This indicates that the lack of a positive correlation between repertoire size and offspring survival cannot solely be attributable to the failure to fully determine genetic paternity in our study. A recent meta-analysis of good-genes effects in sexual selection (Møller and Alatalo, 1999) suggests that the effect of male genetic quality on offspring survival may be quite small (r ¼ .122). Thus, unfeasibly large sample sizes would be required to nullify the existence of such a small effect in a particular case study (a one-tailed power of 95% would be reached with a sample size of 718 males). The distinction between preference and choice In the bird-song literature, there seems to be overwhelming support for the idea that females prefer to mate with males having large repertoires (Catchpole and Slater, 1995). There are two lines of evidence: demonstrations of female mating preferences in the laboratory (see Catchpole, et al., 1986), and correlative findings of field studies in which males with large repertoires were more successful in attracting females (see Catchpole, 1986). Although these two approaches seem to perfectly complement each other, they often do not measure the same thing. It is important to make the distinction between female mating preferences and female choice. In the laboratory, females may readily display their mating preferences because all other factors are being controlled for. In contrast, these preferences may be of little importance under natural circumstances (Searcy, 1992). The outcome of female choice in the field may depend much more strongly on territory characteristics than on the song characteristics of the male (Searcy and Yasukawa, 1996). Because good-genes effects are typically small (see above), it has often been argued that direct benefits (e.g., resulting from high territory quality) should be more important to females than are indirect benefits. This view is supported by the present study (the data from 1981–1982), in which female choice was strongly related to measurable aspects of territory quality, but not to song characteristics, when territory quality was controlled for, which also seems to be the case in many other studies (Searcy and Yasukawa, 1996). Under such circumstances, the outcome of female choice will largely depend on the outcome of malemale competition for the best territories. Assuming that male-male competition and territory quality are the major components contributing to the patterns under consideration (the relation between repertoire size and Behavioral Ecology Vol. 15 No. 4 562 harem size), the low temporal and spatial stability of these patterns is not so astonishing. The intensity of competition may change as the size of the population changes, and it will also depend on the magnitude of differences in habitat quality. In other words, intrasexual selection is highly dependent on ecological circumstances and will therefore be as changeable as these circumstances are. Are repertoires acoustic peacock tails? If large repertoires were very costly to maintain (as costly as a peacock’s tail), we would expect that the ornament would reflect the overall quality of males (Zahavi, 1975). Individuals well adapted to their environment (including resistance to parasites and adaptation to autecological factors) would display larger repertoires than do less well adapted individuals or individuals carrying deleterious mutations. Such differences in the overall quality of individuals would very likely also affect a male’s pairing success and the viability of his offspring. Thus, we would at least have expected to find some positive trend between repertoire size and these fitness-related parameters. As this was not the case, we have to conclude that repertoire size is probably not a very reliable or not a very powerful indicator of male quality. The discrepancy between studies of how repertoire size relates to offspring viability may again result from environmental variation. Offspring viability is a trait that depends on the selective environment, and this selective pressure may vary between populations (as well as over time). This means that certain pleiotropic genes that have positive effects on both repertoire size and offspring survival (i.e., good genes) may be favored in one population, but need not be favored in another population in which offspring survival may depend on a different set of factors. Likewise, we might expect natural selection to favor one type of female mating preferences in one population at one time, but other preferences elsewhere or at other times. This scenario of changing selective environments leading to different patterns of intra- and intersexual selection in different populations seems most likely to apply to situations in which sexually selected male characters are not highly exaggerated. The literature seems full of such examples, in which repeated studies of sexual selection patterns in the same species have yielded different or even contradictory results (see Collins and ten Cate, 1996; Dale et al., 1999; Griffith et al., 1999; Jennions, 1998; Krokene et al., 1998; Mateos and Carranza, 1996; Saetre et al., 1997). Thus, it appears to be quite common that the forces of sexual selection vary greatly in space and time. Syllable switching as an indicator of male quality In the present study, we found a positive correlation between the number of syllable-type switches within a strophe and a male’s pairing success. As shown elsewhere, syllable switching of individual males did not increase with age but was a strong predictor of male longevity (Forstmeier W, Hasselquist D, Bensch S, and Leisler B, unpublished data). This latter correlation causes an apparent increase in syllable switching with age. As older males were more successful in attracting females, male age should be controlled for when analyzing the effect of syllable switching on male pairing success. When doing this, the remaining effect was still significant, which supports the idea that syllable switching reflects some aspects of male quality. Note that we do not claim that there is a female preference for such males, as we cannot disentangle the effects of female choice from those of male-male competition. Syllable switching might reflect variation in male quality for two reasons. First, this song trait is tightly connected to strophe length, and it may be energetically demanding or exhausting to sing long strophes (Suthers and Goller, 1997). Second, long series of syllable-type switches may also reflect neural abilities, as great reed warblers sing with great immediate variety. Apart from the less variable start of a strophe, males avoid introducing the same syllable type twice within a strophe. Thus, singing long strophes requires a large repertoire of immediately accessible syllable types. In addition, such measures of immediate variety seem easier to take by the choosing female than are estimates of total repertoire size. Note that the correlation between these two song traits is very low (see Methods). Future field studies will have to show whether syllable switching is a more consistent correlate of male quality than repertoire size seems to be. Laboratory experiments would offer an alternative approach to discover which aspects of song quality are likely to be perceived by females. We thank Lucia Meyer, who analyzed the song recordings for us. We are indebted to Josef Beier for monitoring the great reed warbler population with great accuracy for nearly three decades. The following persons helped carry out some of the field work: Meinrad Kneer, Kirstin Kuczius, Karl-Heinz Siebenrock, and Edith Sonnenschein. Gösta Metzler helped with the analyses. We are grateful to Thorsten Balsby, Clive Catchpole, Dennis Hasselquist, and Jonathan Ekstrom for their valuable comments on earlier versions of the manuscript, and to Hector Castillo, Andrew MacColl, and Markus Neuhäuser for their help with the statistics. We thank the nature conservation authorities (Erlangen-Höchstadt and Ansbach) for permitting access to the various reserves. This project was supported by the Deutsche Forschungsgemeinschaft (LE512/7–2). REFERENCES Beier J, 1981. Untersuchungen an Drossel- und Teichrohrsänger (Acrocephalus arundinaceus, A. scirpaceus): Bestandsentwicklung, Brutbiologie, Ökologie. J Ornithol 122:209–230. Bensch S, Hasselquist D, 1991. Territory infidelity in the polygynous great reed warbler Acrocephalus arundinaceus: the effect of variation in territory attractiveness. J Anim Ecol 60:857–871. Bensch S, Hasselquist D, Nielsen B, Hansson B, 1998. Higher fitness for philopatric than for immigrant males in a semi-isolated population of great reed warblers. Evolution 52:877–883. Bensch S, Hasselquist D, Nielsen B, Nihlén C, Frodin P, 2001. Food resources and territory quality in the polygynous great reed warbler. In: The ecology of reed birds: biosystematics and ecology series, no. 18 (Hoi H, ed). Vienna: Austrian Academy of Sciences Press; 49–71. Burley N, 1986. Sexual selection for aesthetic traits in species with biparental care. Am Nat 127:415–445. Catchpole CK, 1983. Variation in the song of the great reed warbler Acrocephalus arundinaceus in relation to mate attraction and territorial defense. Anim Behav 31:1217–1225. Catchpole CK, 1986. Song repertoires and reproductive success in the great reed warbler Acrocephalus arundinaceus. Behav Ecol Sociobiol 19:439–445. Catchpole CK, Leisler B, Dittami J, 1986. Sexual differences in the responses of captive great reed warblers (Acrocephalus arundinaceus) to variation in song structure and repertoire size. Ethology 73: 69–77. Catchpole CK, Leisler B, Winkler H, 1985. Polygyny in the great reed warbler, Acrocephalus arundinaceus: a possible case of deception. Behav Ecol Sociobiol 16:285–291. Catchpole CK, Slater PJB, 1995. Bird song: biological themes and variations. Cambridge: Cambridge University Press. Collins SA, ten Cate C, 1996. Does beak color affect female preference in zebra finches? Anim Behav 52:105–112. Dale S, Slagsvold T, Lampe HM, Saetre GP, 1999. Population divergence in sexual ornaments: the white forehead patch of Norwegian pied flycatchers is small and unsexy. Evolution 53:1235–1246. Fischer S, Frommolt K-H, Tembrock G, 1996. Gesangsvariabilität beim Drosselrohrsänger Acrocephalus arundinaceus. J Ornithol 137:503– 513. Forstmeier and Leisler • Song and sexual selection Griffith SC, Owens IPF, Burke T, 1999. Female choice and annual reproductive success favor less-ornamented male house sparrows. Proc R Soc Lond B 266:765–770. Hasselquist D, 1994. Male attractiveness, mating tactics and realized fitness in the polygynous great reed warbler (PhD dissertation). Lund University, Sweden. Hasselquist D, 1998. Polygyny in great reed warblers: a long-term study of factors contributing to male fitness. Ecology 79:2376–2390. Hasselquist D, Bensch S, 1991. Trade-off between mate guarding and mate attraction in the polygynous great reed warbler. Behav Ecol Sociobiol 28:187–193. Hasselquist D, Bensch S, von Schantz T, 1996. Correlation between male song repertoire, extra-pair paternity and offspring survival in the great reed warbler. Nature 381:229–232. Hiebert SM, Stoddard PK, Arcese P, 1989. Repertoire size, territory acquisition and reproductive success in the song sparrow. Anim Behav 37:266–273. Jennions MD 1998. The effect of leg band symmetry on female-male association in zebra finches. Anim Behav 55:61–67. Krebs JR, Ashcroft R, Webber MI, 1978. Song repertoires and territory defense in the great tit. Nature 271:539–542. Krokene C, Rigstad K, Dale M, Lifjeld JT 1998. The function of extrapair paternity in blue tits and great tits: good genes or fertility insurance? Behav Ecol 9:649–656. Leisler B, Beier J, Heine G, Siebenrock K-H, 1995. Age and other factors influencing mating status in German great reed warblers (Acrocephalus arundinaceus). Jap J Ornithol 44:169–180. 563 Leisler B, Beier J, Staudter H, Wink M, 2000. Variation in extra-pair paternity in the polygynous great reed warbler (Acrocephalus arundinaceus). J Ornithol 141:77–84. Mateos C, Carranza J, 1996. On the intersexual selection for spurs in the ring-necked pheasant. Behav Ecol 7:362–369. Møller AP, Alatalo RV, 1999. Good-genes effects in sexual selection. Proc R Soc Lond B 266:85–91. Saetre G-P, Moum T, Stanislav B, Miroslav K, Adamjan M, Moreno J, 1997. A sexually selected character displacement in flycatchers reinforces premating isolation. Nature 387:589–592. Searcy WA 1992. Song repertoire and mate choice in birds. Am Zool 32:71–80. Searcy WA, Yasukawa K, 1996. Song and female choice. In: Ecology and evolution of acoustic communication in birds (Kroodsma DE, Miller EH, eds). Ithaca, New York: Cornell University Press; 454– 473. Sheldon B, 2000. Differential allocation: tests, mechanisms and implications. Trends Ecol Evol 15:397–402. Suthers RA, Goller F, 1997. Motor correlates of vocal diversity in songbirds. In: Current ornithology, vol. 14. (Nolan VJr et al., eds).New York: Plenum Press; 235–288. Yasukawa K, 1981. Song repertoires in the red-winged blackbird (Agelaius phoeniceus): a test of the Beau Geste hypothesis. Anim Behav 29:114–125. Zahavi A, 1975. Mate selection: selection for a handicap. J Theor Biol 53:205–214.
© Copyright 2026 Paperzz