- Email: [email protected]
Females of the lekking great snipe do not prefer males with whiter tails
ANIMAL BEHAVIOUR, 2000, 59, 273–280 doi: 10.1006/anbe.1999.1301, available online at http://www.idealibrary.com on
Females of the lekking great snipe do not prefer males with whiter tails STEIN ARE SÆTHER*, PEDER FISKE*†, JOHN ATLE KA r LA r S† & JENS MAGNE GJUL*
*Department of Zoology, Norwegian University of Science and Technology †Norwegian Institute for Nature Research (Received 19 April 1999; initial acceptance 28 May 1999; final acceptance 1 October 1999; MS. number: 6202)
A previous experimental study of great snipe, Gallinago media, has reported an effect on male mating success of the amount of white in their tails. That result is one of a very limited set of existing experimental results supporting a female mate preference for a morphological trait in animals. However, a later observational study did not find any correlation between amount of white and male mating success. If females sample a limited number of males, their preferences need not result in strong relationships between mating success and trait values in males, possibly explaining the failure to find the predicted correlation. Yet, females of lekking species are thought to have ample opportunities for mate sampling. To resolve these contrasting results, we present in this paper (1) a larger correlational study (several leks during 10 years) showing no relationship between male mating success and whiteness of tails (measured in several ways), and most importantly (2) evidence that individual females do not mate predominantly with males with very white tails among those males that each female samples. These results show that females do not prefer males with whiter tails as mates, within the contemporary natural variation in the trait. They also show that there is no sexual selection of the trait at present. This does not necessarily imply that white tails are not a sexually selected adaptation in males, but the mechanisms are likely to have been different from direct mate choice of whiter tails per se.
because they have weak mating preferences. Even when females have strong unanimous preferences and are able to choose according to their taste, low correlations between male mating success and preferred traits would ensue if each female samples only a few males (Benton & Evans 1998). A few studies have found that manipulating traits of males in the field causes increased mating success, both in lek breeders (Ho ¨ glund et al. 1990a; Andersson 1992; Petrie & Halliday 1994) and in some other species (e.g. Andersson 1982; Møller 1988; Grether 1996). One of these studies involved increasing the amount of white on the tail feathers of great snipe, Gallinago media, males (Ho ¨ glund et al. 1990a). The group of males with experimentally increased white had more matings than controls, corroborating earlier correlational indications of a relationship between male mating success and amount of white on tails (Ho ¨ glund & Lundberg 1987). Ho ¨ glund et al. (1990a) further showed that males with increased white did not enjoy an advantage in male–male interactions, suggesting that the increased mating success was due to a direct female preference for white tails. Yet, in another study of the same species, Fiske et al. (1994) failed to find any correlation between tail whiteness and mating success.
Behavioural biologists have long been interested in the evolution of female mate preferences (Cronin 1991). Such preferences could cause sexual selection and rapid evolution of male traits, and thus explain the striking sexual dimorphism of many species. Studies of lekking species are of particular interest in this context because (1) females receive no help from males in rearing offspring which could otherwise explain the evolution of preferences, (2) females can compare several males in a short time, and (3) males are not monopolized for a long time by any one female. The distribution of mating success is often skewed among males on leks, and female preferences may explain this variation. Several studies have indirectly attempted to identify female preferences by searching for correlates of male mating success (reviewed by Fiske et al. 1998). In general, variation in male traits seems able to explain only a little of the variation in male mating success. However, this could be because females sample a restricted number of males, and not necessarily Correspondence: S. A. Sæther, Department of Zoology, Norwegian University of Science and Technology, N-7491 Trondheim, Norway (email: [email protected]). P. Fiske and J. A. Ka˚la˚s are at the Norwegian Institute for Nature Research, Tungasletta 2, N-7485 Trondheim, Norway. 0003–3472/00/020273+08 $35.00/0
2000 The Association for the Study of Animal Behaviour
273
2000 The Association for the Study of Animal Behaviour
274
ANIMAL BEHAVIOUR, 59, 2
How are we to reconcile these contrasting results? To attempt this, we present the following in this paper: (1) a larger correlational study on the relationship between white on tails and mating success in great snipe for several leks over 10 mating seasons; (2) an examination of the crucial assumption that our measure of white reflects the same as in Ho ¨ glund et al. (1990a); and (3) an analysis of the mating decisions of individual females in relation to the amount of white on the tails of those males each female visited. The results strongly suggest that female great snipes do not prefer males with whiter tails. We conclude that there is no evidence of sexual selection currently acting on this trait in this species and that the earlier experimental results can best be explained by chance events. We emphasize the importance of studying individual females for investigating female mate preferences, and of not simply focusing on the effects of the outcome of mate choice on male mating success. METHODS
Field Methods Our study site is in Ga˚va˚lia (6217 N, 936 E), Norway (see Løfaldli et al. 1992). The study population occupies several leks where snipes have been caught in mist nets and individually marked with coloured leg rings, with permission from the Norwegian Directorate for Nature Management. The data presented here were collected from mid-May to mid-June in 1989–1998 mainly on two leks, but in some years also on other leks as resources allowed. One to four observers (depending on the number of males present) made observations at the leks from 2300 to 0230 hours, the time when females visit leks (Ka˚la˚s et al. 1995). The number of males at each site was usually between 10 and 25, but varied somewhat during the mating season and between leks. Each observer surveyed continuously all visits by females to an assigned subset of the males, and noted all copulations and solicitations by females.
Male Mating Success Since individual females often mate several times in a mating season (Fiske & Ka˚la˚s 1995) we did not simply sum the total number of matings a male received as a measure of mating success. Instead, we tried to estimate the number of individual females that mated with the male. We did this by combining the records for identified females with the minimum number of different unmarked females mating with the male. This latter number was calculated by the criterion that an unmarked female could mate on 3 consecutive nights (Fiske & Ka˚la˚s 1995). We could sometimes keep track of several unmarked females and thus distinguish between different unmarked females mating on the same night, but not across nights. We calculated mating success only for males that were observed to be territorial on at least 5 nights. A successful male is defined as a male that
obtained at least one mating in the specific season. Some females were only seen to solicit copulation and not actually to copulate. These females were also included in a male’s mating success since they had probably mated with that male (Fiske & Ka˚la˚s 1995). Deaths during the mating season are rare in the species (Ho ¨ glund et al. 1992), and we thus feel confident that differential survival did not confound mating success in this study (see Grether 1996). Altogether the summed mating success for the males on the leks and years presented here was 484 females (including 334 seen to copulate), based upon over 2000 female solicitations (including over 800 copulations). The number of marked females seen to solicit mating was 168 female-years (122 individuals). The numbers presented of males at each lek in each year does not necessarily include all males visiting the leks since we excluded males observed for less than 5 days, and some males observed for more than 5 days were not captured.
Measures of White on the Tail Males fan out their white tails during their display behaviour, a rather conspicuous flash of white to a human observer’s eye. The standard measure of the amount of white on the tail in great snipe (Ho ¨ glund et al. 1990a, b, 1992; Fiske et al. 1994; Sæther et al. 1994) is the distance from the first dark spot on the outermost feather to the tip of the same feather (for a drawing of a great snipe tail, see e.g. Ho ¨ glund et al. 1990a). We used this measure (to the nearest 0.1 mm) for the left-hand part of the tail in this paper, and called it ‘measured white’. Measured white varied considerably from male to male (1986–1997, XSD=16.773.13 mm, N=913, range 8.1–32.0, CV 18.66%). It also differs between populations (Ka˚la˚s et al. 1997) and years (Ho ¨ glund et al. 1992), is more extensive on males than females (Ho ¨ glund et al. 1990b) and increases somewhat with age (Ho ¨ glund et al. 1990a, 1992; Sæther et al. 1994). We tried to catch all males every year. In the few cases that we did not obtain measured white for a male in a given year (N=11 of those assigned mating success), we used the value for the next year if this was available. If it was not available, we used the value for the previous year, but only if the male was at least 2 years old. Measured white might not truly reflect the amount of white in a tail. Two tails with the same value for measured white might look very different. If so, any relationships between whiteness and mating success might be obscured. Another measurement of tail whiteness is the number of tail feathers with more than 50% white. This measurement is perhaps even more crude as the two outermost feathers are always at least 50% white, and the maximum number is only four (one bird had five). To obtain a better measure of whiteness we took photographs in the field of the birds’ tails when they were caught, so we could rank the tails later according to the amount of white. The majority of males will readily fan out their tails in the hand, and we photographed the tails (Fig. 1) while trying to ensure that the outer webs of the outermost feathers were exposed.
SÆTHER ET AL.: WHITE TAILS AND FEMALE PREFERENCES
Figure 1. Photographs of four male great snipe tails, to indicate the range of variation used to rank the amount of white on the outer tail feathers, where (c) is of the tail of the most successful male in the study.
To see if such ranking was reliable, we asked 12 university students independently to rank the photographs from three leks in 1994 according to the amount of white. There was strong agreement between the 12 judges (Kendall coefficient of concordance: lek HN: W=0.85, N=24; lek AH: W=0.85, N=12; lek ØS: W=0.90, N=11; all P values <0.01). We thus felt confident that ranking was a reliable measure, and used the rankings of only one judge in subsequent analyses of ranked white. Tails were ranked separately for each year and lek. The ranks generally correlated strongly with the measured amount of white (N=24 lek-years, mean r=0.76, mean sample size=15.6, mean P=0.0003). However, the correlations varied from 0.56 to 0.93 with an outlier at 0.25, supporting our impression that measured white might give less than perfect information about the extent of white in the tail. We therefore analysed male mating success and female mate choice in relation to the ranked photographs, as well as to measured white.
Analyses Male lekwide relationships 1989–1998 We compared successful and unsuccessful males in relation to both measured and ranked white. This was done separately for each lek and year. Similarly, we correlated mating success with both measures of white. Some of the data (for 1989–1992) have been analysed previously (in slightly different ways by Fiske et al. 1994), but are included here for completeness and because relationships with ranked photographs were not included in the previous analyses. In addition, to avoid pseudo-
replication since some males appeared in many years, we calculated average values of measured white over all years for each male. We compared these average values for those males that mated at least once versus those that did not. For this latter analysis we calculated the least detectable effect size for power=0.8, and the power for detecting Cohen’s medium effect size (d=0.5; Cohen 1988). For power analyses, we used GPower 2.1.1 (Erdfelder et al. 1996).
Individual females 1994–1998 To see if females preferred to mate with males with white tails, we compared the amount of white of the males a female was observed to have visited (using both measured and ranked white). In this way we avoided the problem of limited mate sampling in making inferences about mate preferences. We did this by asking (1) if females tended to mate with the male with the most white out of those sampled and (2) if females tended to mate with males with more than the average amount of white out of those sampled. We called the male(s) with which a female mated her chosen male(s), and the others she visited her rejected males. For all analyses we used female-years as the sample size, pooled over all years (here 1994–1998) and leks, and treated data for individual females as independent for each mating season. One of the males in the sample (Fig. 1c) was exceptionally successful. Over the 5 years, this male mated with 37 of the 79 females in the sample of females from which we had observations from both their chosen and at least one rejected male. We therefore carried out additional analyses excluding all observations, in any year, with this male
275
276
ANIMAL BEHAVIOUR, 59, 2
Table 1. Amount of white on left tail (mm) and ranked amount of white from photographs, for males without and with mating success (at least one female copulating or soliciting) on different great snipe leks in different years Measured white Unsuccessful Year
Lek
X±SD
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1997 1994 1995 1998 Sum Mean Mean*
HN HN HN HN HN HN HN HN HN HN AH AH AH AH AH AH AH AH AH AH BJR ØS ØS SAL
16.4±2.1 16.1±2.1 17.0±3.2 16.7±1.2 17.4±3.6 18.3±3.3 20.5±6.4 17.3±3.4 17.3±4.9 16.5±2.2 16.9±2.3 16.0±1.5 16.1±2.3 21.1±1.2 15.1±1.7 16.7±2.4 14.6±2.5 18.2±3.8 17.1±3.0 16.4±1.9 16.7±1.3 17.2±2.8 16.1±1.5 16.0±1.3 16.80±2.80 16.76±2.91
Ranked photographs
Successful N
11 11 13 7 7 11 2 13 14 11 7 6 12 2 8 8 7 6 11 11 5 4 3 10 168 7.0
X±SD 17.3±2.5 16.8±2.2 16.7±2.4 17.2±4.4 14.9±3.7 16.7±3.3 17.1±2.8 16.1±1.8 15.6±1.5 15.3±2.6 18.4±3.2 19.2±6.0 17.1±3.6 16.5±1.8 17.9±2.7 16.5±2.4 16.6±2.0 15.5±2.3 16.5±1.6 16.9±3.3 16.3±2.7 15.6±4.3 20.2±1.8 16.6±2.4 16.98±2.57 16.76±2.62
N
t
P
N
P
8 11 17 12 12 7 7 4 4 7 9 7 6 6 7 6 6 4 8 6 5 4 2 3 200 8.3
−0.82 −0.74 0.30 −0.32 1.47 0.96 1.22 0.67 0.69 1.08 −1.03 −1.27 −0.70 3.22 −2.44 0.18 −1.58 1.28 0.49 −0.41 0.26 0.63 −2.87 −0.64
0.43 0.47 0.77 0.75 0.16 0.35 0.26 0.51 0.50 0.30 0.32 0.23 0.49 0.02 0.03 0.86 0.14 0.24 0.63 0.69 0.80 0.56 0.06 0.54
0.66 0.90 0.48 0.84 0.14 0.90 0.56 0.41 0.96 0.48 0.68 0.18 0.96 0.07 0.28 0.90 0.09 0.91 0.55 0.66 0.22 0.69 0.20 0.29
−0.02 −0.01
0.42 0.46
19 22 30 19 19 14 9 17 18 18 16 13 18 8 15 14 12 10 19 17 10 8 5 13 363 15.1
0.54 0.57
P values for ranked photographs are exact probabilities from Mann–Whitney U tests. *Weighted for sample size.
(six of the 37 females were included since they also mated with other males). This also caused some females not choosing this male to be excluded, since this was their sole rejected male (N=5). We used randomization tests to calculate probability values for obtaining by chance the observed number of females mating with the male with the most white. Since this probability is strongly affected by the number of males visited, which varied between females, we cannot compute this as a simple binomial probability. Instead, we used the following randomization technique to obtain a proper null model. We first counted the number of females, of the k females in the sample, mating with the male with the most white as y. To calculate the one-tailed probability of obtaining y by chance, we randomly drew for each female i one number j from the set {mi(1), ni mi(0)} where mi is the number of males with which the female mated and ni is the number of males visited. We then calculated i=k i=1 ji 10 000 times to obtain an expected distribution under the null-hypothesis. We counted the number of times i=k i=1 ji ≥y as x. The one-tailed probability of y was then calculated as x/10 000. We analysed if females mated with males with more than the average amount of white in two ways. First, we used parametric statistics to see if there was a difference in the amount of white (mean values for each female)
between chosen and rejected males (or Wilcoxon signedranks tests, for ranked white). Second, we calculated the fractional rank of the preferred males (rank divided by number of males visited, 1.0=mating with the one with most white). To obtain the probability of the observed mean fractional rank of chosen males, z, occurring by chance, we used the following randomization approach. A number hi was randomly drawn for each female i from the set {1, 2 . . . ni} and the ratio li =hi/ni computed. If a female had mated with more than one male we calculated l as many times as the number of chosen males she had, and used the average. We then calculated the simulated mean fractional rank over all k females, i=k i=1 li/k, 10 000 times and counted w as the number of times i=k i=1 li/k ≥z. The one-tailed probability of z was obtained as w/10 000. Randomization programmes were compiled using Resampling Stats 4.0.7 (Resampling Stats Inc., Arlington, VA, U.S.A.). RESULTS
Sexual Selection on Amount of White For measured amount of white, the weighted mean of means over all years and all leks was identical for successful and unsuccessful males (16.76 mm; Table 1). We
SÆTHER ET AL.: WHITE TAILS AND FEMALE PREFERENCES
Table 2. Relationships between male mating success (no. of females copulating or soliciting) and amount of white on tail on different great snipe leks in different years Measured white
Ranked photographs
Year
Lek
r
N
P
rS
P
rS
N
P
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1997 1994 1995 1998 Sum Mean Mean*
HN HN HN HN HN HN HN HN HN HN AH AH AH AH AH AH AH AH AH AH BJR ØS ØS SAL
0.27 0.09 −0.14 0.20 −0.08 −0.42 −0.53 −0.26 −0.21 −0.23 0.32 0.06 0.05 −0.75 0.47 −0.05 0.12 −0.31 0.19 0.17 −0.30 0.44 0.65 0.16
19 22 30 19 19 18 9 17 18 18 16 13 19 8 15 14 13 10 19 17 10 8 5 13 369 15.4
0.26 0.68 0.47 0.41 0.75 0.09 0.14 0.31 0.39 0.36 0.22 0.85 0.84 0.03 0.08 0.87 0.70 0.38 0.44 0.52 0.41 0.28 0.23 0.60
0.17 0.14 −0.01 0.24 −0.22 −0.31 −0.34 −0.15 −0.18 −0.26 0.31 0.27 0.06 −0.62 0.40 −0.08 0.29 −0.34 −0.06 0.14 −0.38 −0.18 0.78 0.16
0.47 0.53 0.96 0.31 0.35 0.22 0.37 0.56 0.46 0.30 0.24 0.35 0.80 0.10 0.13 0.78 0.33 0.35 0.80 0.58 0.29 0.67 0.12 0.61
0.09 0.05 −0.08 −0.03 0.21 −0.22 0.35 −0.25 −0.02 −0.16 0.10 0.37 −0.01 −0.54 −0.29 −0.03 0.42 −0.03 0.21 0.12 0.45 −0.05 0.78 0.32
0.69 0.82 0.70 0.90 0.38 0.43 0.36 0.33 0.93 0.52 0.70 0.20 0.97 0.17 0.29 0.92 0.18 0.93 0.38 0.64 0.20 0.90 0.12 0.29
0.43 0.45
−0.01 −0.00
0.45 0.49
0.07 0.05
19 22 30 19 19 14 9 17 18 18 16 13 18 8 15 14 12 10 19 17 10 8 5 13 363 15.1
−0.00 −0.00
0.54 0.58
N is the same for the two correlation coefficients of measured white. *Weighted by sample size.
compared successful and unsuccessful males in 24 lekyears, and found only two significant differences (one in each direction). There was no difference between these two groups of males in white as judged from the ranking of photographs either (Table 1). The weighted mean rank correlation coefficients over all years and all leks were close to zero for both measured white and ranked photographs (Table 2), and not significant in any single lek-year. Including each individual male only once by using average values for measured white gave similar results. For 118 successful males (in at least 1 year) the averaged measured amount of whiteSD was 16.622.95 mm and for 152 unsuccessful males, 16.622.71 mm (t268 =0.023, P=0.98). Least detectable effect size (for power=0.8) was d=0.34, and for medium effect size (d=0.5) power=0.98. Restricting the analysis of averaged measured white to birds at least 2 years old gave similar results (successful males: XSD=16.723.10 mm, N=91; unsuccessful males: 16.702.33 mm, N=80; t169 =0.049, P=0.96, least detectable effect size d=0.43, power for medium effect size=0.90).
Mate Choice for Whitest Males
only 18 out of 79 mated with the male with the most white among those the female visited and 61 mated with another male (Fig. 2). The probability of obtaining 18 or fewer females mating with the male with the most white in the tail out of 79 was 0.029 (given the observed number of males visited for each female, randomization test, 10 000 simulations, mean simulated value=25.9), indicating that females actually avoided mating with the male with most white. However, this result was probably due to one particularly successful male with a low amount of white. When this male was excluded from the sample, 13 of 43 females mated with the male with the most white (Fig. 2). The probability of obtaining 13 or fewer females mating with the male with the most white on the tail out of 43 was 0.33 (randomization procedures as above, mean simulated value=14.9). Forty-seven out of 79 females mated with the male with the least amount of white among those sampled (Fig. 2; randomization procedures as above: P=0.0001, mean simulated value=25.00), but when the most successful male was excluded the figures were 14 out of 43, 14 being the most likely outcome of all by chance (randomization procedures as above: P=0.57, mean simulated value=14.0).
Measured white
Ranked photographs
Among those females observed both at territories of males they mated with and males they did not mate with,
Twenty-two of 78 females mated with the male with the most white of those visited, as judged by ranking of
277
ANIMAL BEHAVIOUR, 59, 2
(a) 80 Most white chosen Most white rejected Expected no. choosing most white
70 60 50 40 30 20 10 No. of females
278
0
MW
RP
MW
RP
(b) 80 Least white chosen Least white rejected Expected no. choosing least
70 60 50 40 30 20
result was confirmed by a two-way ANOVA using amount of white as the dependent variable, and chosen or not and female identity as factors (choice: F1,92.81 =13.79, P<0.001; female: F78,78 =0.605, P=0.99; interaction: F78,159 =0.78, P=0.89; only females observed at both their chosen and at least one rejected male are included). When we excluded the most successful male there was no difference in the mean amount of white of chosen and rejected males (chosen: XSD=16.992.45 mm; rejected: 17.061.85; paired t test: t42 = 0.138, P=0.89). This was also confirmed by a two-way ANOVA using amount of white as dependent variable (choice: F1,47.01 =0.02, P=0.89; female: F42,42 =0.72, P=0.87; interaction: F42,74 =13.79, P=0.71). The mean fractional rank of the male(s) chosen was 0.57 (rank divided by number of males visited, 1=mating with the one with most white). The one-tailed probability of obtaining a mean fractional rank of 0.57 or lower in this sample was 0.001 (randomization test: 10 000 simulations, mean simulated value=0.66). The mean fractional rank of the males rejected was 0.70 (paired t test: t78 = 2.949, P=0.004). Excluding the most successful male yielded a mean fractional rank for chosen males of 0.70, which was close to what was expected by chance (randomization as above: mean simulated value=0.67, probability of getting 0.70 or greater=0.19). For rejected males the mean fractional rank was 0.67 (paired t test: t42 =0.49, P=0.63).
10 0
MW RP All males included
MW RP Most successful male excluded
Figure 2. The number of females that mated with the male with (a) the most white on the tail and (b) the least white on the tail of those visited and the number expected from null model simulations, for measured white (MW) and white ranked from photographs (RP). Also shown is the number of females that did not mate with such a male.
Ranked photographs Using ranked photographs we found that chosen males were ranked to be less white than rejected ones (Wilcoxon signed-ranks test: Z= 2.369, N=78, P=0.018). Again, excluding the most successful male resulted in no significant differences in amount of white between chosen and rejected males (Z= 1.251, N=42, P=0.21). DISCUSSION
photographs (Fig. 2a; randomization probability of 22 or less=0.16, mean simulated value=26.5). Excluding the most successful male gave 17 out of 42 (randomization probability of 17 or greater=0.21, mean simulated value=14.1). Thirty-nine out of 78 mated with the male with the least amount of white (Fig. 2b; randomization probability of 39 or more=0.003, mean=26.5) but when the most successful male was excluded the proportion dropped to 11 out of 42 (randomization probability of 11 or less=0.19, mean=14.1).
Mate Choice for Whiter Males Measured white The mean amount of white of those males with which a female mated was not larger than the mean amount of white of males with which she did not mate. Rather, there was on average a 1.55-mm difference in the opposite direction (chosen: XSD=15.342.54 mm; rejected: 16.892.18; paired t test: t78 =3.653, P<0.001). This
It is of paramount importance to the credibility of any field of empirical inquiry that results be replicated and negative results published. In behavioural ecology, the value placed on novel results is so high that these goals may suffer and be neglected. Publication bias and selective reporting (unconscious as it may be) are serious obstacles to scientific progress, and seems to be responsible for overstating the strength and biological significance of some relationships studied by behavioural ecologists (Palmer 1999; see also Thornhill et al. 1999). This may also be true for the strength of current sexual selection on male traits in animals. Our results may be summarized as follows: neither measure of white was related to male mating success (Tables 1, 2). Both measures of white showed that females did not prefer to mate with the male with most white on the tail of those visited (Fig. 2a). By including the most successful male, there were in fact more females than expected that mated with the male with the least amount of white (Fig. 2b). For neither measure of white, analysed
SÆTHER ET AL.: WHITE TAILS AND FEMALE PREFERENCES
in several different ways, did we find any indications whatsoever that females preferred whiter males as mates. We did, however, find the opposite: the rejected males had significantly more white because many females mated with one particular male that had only a little white in his tail. We think it is safe to say that the lack of any positive relationships was not due to low statistical power, since the sample sizes were high both in the male-centred correlational analyses and in the femalecentred individual mate choice analyses. This was confirmed by a power analysis of successful and unsuccessful males. In the experiment where Ho ¨ glund et al. (1990a) manipulated great snipe tails, 16 of 17 matings went to manipulated males and only one to control males. This is significantly different, with a chi-square test, from the expected frequencies based on number of experimental and control birds. Altogether five males achieved matings, but 10 of these went to one male. Since so few males were successful, they might have been so in the absence of any manipulation, and in fact the frequency of manipulated males achieving mating success was not significantly higher than control birds. The best predictor of male mating success in great snipe seems to be mating success the previous year (Fiske et al. 1994). Thus, a better experimental design would be to block for this effect. Another problem was that mating success was measured as number of matings and not number of different females mating. Although Ho ¨ glund et al. (1990a) tried to count repeated copulations by the same female only as one, they could not know if the same female returned and mated on a new visit, which we now know that females in fact often do (Fiske & Ka˚ la˚ s 1995). Ho ¨ glund & Alatalo (1995) reanalysed the experimental data, using a paired t test of 12 pairs of manipulated and control neighbours. The onetailed probability of the observed difference in number of copulations was 4%. Again, this analysis suffers, albeit less, from the problem of counting copulations instead of females, and the choice of test statistic is questionable because of mating skew. Even if we do accept the results at face value, they could still be due to chance effects. Another problem (pointed out by Bennett et al. 1994) with the experiment was the use of Tipp-Ex fluid. The fluid reflects in the ultraviolet, and may not be perceived by the birds as having the same colour as white feather. The large effect the most successful male had on our results illustrates the problems with such studies on animals that can show large mating skew. Because this very successful male happened to have a tail with little white, a seemingly strong preference for males with little white disappeared when we removed this individual male from the sample. Something similar might have occurred in Ho ¨ glund et al.’s experiment: one very successful male happened to be manipulated to be more white. Without proper controls (both previously successful and unsuccessful males, preferably also manipulations in both directions) it is difficult to conclude much from results involving few very successful males and many unsuccessful males. It seems plausible that great snipes have evolved white tails through sexual selection. This is hinted at because
the trait (1) is sexually dimorphic (Ho ¨ glund et al. 1990b), (2) increases somewhat with age (Ho ¨ glund et al. 1990a, 1992; Sæther et al. 1994), (3) is different from those of all other species in the genus Gallinago (Cramp & Simmons 1983), (4) is more pronounced in more southern populations that experience darker nights (Ka˚ la˚ s et al. 1997), and (5) is frequently displayed at leks (Lemnell 1978). There are many ways in which sexual selection could have favoured whiter tails. Our data do not support the claim that females have evolved any preferences for white tails, but it cannot be ruled out that the shape of preference functions of females (Widemo & Sæther 1999) is such that males with less white than the natural variation would have been discriminated against. At least we can say that females do not discriminate between males based on contemporary variation in this trait. The benefit for males of having white tails could be through interactions with other males (e.g. by using them as a threat signal). It could be that the trait is not costly to produce or costly in itself to carry, so that all males have more or less white tails, and the information depends instead on how the tail is used and not on how white it is. Another possibility is that it is not the amount of white per se but the patterns created by the black and white patches that contain information. To display such patterns, the tail must be more or less white. This could be used, for example, in individual recognition. Tail marks may also have an amplifier function of feather quality (Hasson 1991), since absence of melanin in white patches may increase damage and abrasion (Fitzpatrick 1998). The present study does not provide support for that theory. In addition, we only rarely see much abrasion and damage of great snipe tail feathers, as opposed to wing feathers (Sæther et al. 1994), probably because these short tails are not exposed to much physical contact with the surroundings. Although we did not find evidence for any female preference, our methods and our approach of comparing the sampled males of individual females should be of general interest and more widely employed (see also Petrie et al. 1991). This will be more appropriate for the study of female mate choice than the traditional emphasis on sexual selection. The latter approach is of more interest for the evolutionary consequences of female mate choice on male traits, but not for understanding mate choice itself. Studies of mate preferences should focus on individual females (Widemo & Sæther 1999) and the males available to each female (Benton & Evans 1998), and not simply on the effects of preferences on the distribution of matings among males. It was for this reason we compared males that each female rejected with those that she chose. In lekking species however, it is not obvious that females only sample males whose territories they visit. This will depend on the nature of the trait of interest: is it a long-range or short-range signal? In the present case it is likely that if females prefer white tails, they cannot examine this from outside the territory. But despite ample opportunity to do so within territories, they did not prefer males with whiter tails. Given the reasonably large sample sizes obtained, from several lek sites and over many years, using both a sexual selection
279
280
ANIMAL BEHAVIOUR, 59, 2
male approach and an individual-level female mate choice approach, we feel confident to conclude that female great snipes do not prefer males with whiter tails. We have assumed that if females preferred white tails, they would all do so. What we have rejected therefore is the hypothesis that females unanimously prefer whiter tails. We cannot reject a hypothesis that females somehow use information contained in the tail patterns of males in their mate choice. It is in this respect intriguing that males often display their tails clearly directed towards visiting females (personal observations), as if a male’s tail is something that a female is paying attention to after all. Acknowledgments We thank our field assistants (especially Sten L. Svartaas and Tord Bretten) for making it all possible, and Jacob Ho ¨ glund and the anonymous referees for comments. References Andersson, M. 1982. Female choice selects for extreme tail length in a widowbird. Nature, 299, 818–820. Andersson, S. 1992. Female preference for long tails in lekking Jackson’s widowbirds: experimental evidence. Animal Behaviour, 43, 379–388. Bennett, A. T. D., Cuthill, I. C. & Norris, K. J. 1994. Sexual selection and the mismeasure of color. American Naturalist, 144, 848–860. Benton, T. G. & Evans, M. R. 1998. Measuring mate choice using correlation: the effect of female sampling behaviour. Behavioral Ecology and Sociobiology, 44, 91–98. Cohen, J. 1988. Statistical Power Analysis for the Behavioral Sciences. 2nd edn. Hillsdale, New Jersey: L. Erlbaum. Cramp, S. & Simmons, K. E. L. (Eds) 1983. The Birds of the Western Palearctic, Vol. 3. Oxford: Oxford University Press. Cronin, H. 1991. The Ant and the Peacock. Cambridge: Cambridge University Press. Erdfelder, E., Faul, F. & Buchner, A. 1996. GPOWER: a general power analysis program. Behavior Research Methods, Instruments, and Computers, 28, 1–11. Fiske, P. & Ka˚la˚s, J. A. 1995. Mate sampling and copulation behaviour of great snipe females. Animal Behaviour, 49, 209–219. Fiske, P., Ka˚la˚s, J. A. & Sæther, S. A. 1994. Correlates of male mating success in the lekking great snipe (Gallinago media): results from a four year study. Behavioral Ecology, 5, 210–218. Fiske, P., Rintama¨ki, P. T. & Karvonen, E. 1998. Mating success in lekking males: a meta-analysis. Behavioral Ecology, 9, 328–338. Fitzpatrick, S. 1998. Birds’ tails as signaling devices: markings, shape, length, and feather quality. American Naturalist, 151, 157–173.
Grether, G. F. 1996. Sexual selection and survival selection on wing coloration and body size in the rubyspot damselfly Hetaerina americana. Evolution, 50, 1939–1948. Hasson, O. 1991. Sexual displays as amplifiers: practical examples with an emphasis on feather decorations. Behavioral Ecology, 2, 189–197. Ho ¨ glund, J. & Alatalo, R. V. 1995. Leks. Princeton, New Jersey: Princeton University Press. Ho ¨ glund, J. & Lundberg, A. 1987. Sexual selection in a monomorphic lek-breeding bird: correlates of male mating success in the great snipe Gallinago media. Behavioral Ecology and Sociobiology, 21, 211–216. Ho ¨ glund, J., Eriksson, M. & Lindell, L. E. 1990a. Females of the lek breeding great snipe, Gallinago media, prefer males with white tails. Animal Behaviour, 40, 23–32. Ho ¨ glund, J., Ka˚la˚s, J. A. & Løfaldli, L. 1990b. Sexual dimorphism in the lekking great snipe. Ornis Scandinavica, 21, 1–6. Ho ¨ glund, J., Ka˚la˚s, J. A. & Fiske, P. 1992. The costs of secondary sexual characters in the lekking great snipe (Gallinago media). Behavioral Ecology and Sociobiology, 30, 309–315. Ka˚la˚s, J. A., Fiske, P. & Sæther, S. A. 1995. The effect of mating probability on risk taking: an experimental study in lekking great snipe. American Naturalist, 146, 59–71. Ka˚la˚s, J. A., Kuresoo, A., Luigujo ˜ e, L. & Svartaas, S. L. 1997. Morphometrical comparison between Estonian and Norwegian great snipes (Gallinago media). Proceedings of the Estonian Academy of Sciences Biology Ecology, 46, 115–122. Lemnell, P. A. 1978. Social behaviour of the great snipe Capella media at the arena display. Ornis Scandinavica, 9, 146–163. Løfaldli, L., Ka˚la˚s, J. A. & Fiske, P. 1992. Habitat selection and diet of great snipe Gallinago media during breeding. Ibis, 134, 35–43. Møller, A. P. 1988. Female choice selects for male sexual tail ornaments in the monogamous swallow. Nature, 332, 640–642. Palmer, A. R. 1999. Detecting publication bias in meta-analyses: a case study of fluctuating asymmetry and sexual selection. American Naturalist, 154, 220–233. Petrie, M. & Halliday, T. 1994. Experimental and natural changes in the peacock’s (Pavo cristatus) train can affect mating success. Behavioral Ecology and Sociobiology, 35, 213–217. Petrie, M., Halliday, T. & Sanders, C. 1991. Peahens prefer peacocks with elaborate trains. Animal Behaviour, 41, 323–332. Sæther, S. A., Ka˚la˚s, J. A. & Fiske, P. 1994. Age determination of breeding shorebirds: quantification of feather wear in the lekking great snipe. Condor, 96, 959–972. Thornhill, R., Møller, A. P. & Gangestad, S. W. 1999. The biological significance of fluctuating asymmetry and sexual selection: a reply to Palmer. American Naturalist, 154, 234–241. Widemo, F. & Sæther, S. A. 1999. Beauty is in the eye of the beholder: causes and consequences of variation in mating preferences. Trends in Ecology and Evolution, 14, 26–31.
Females of the lekking great snipe do not prefer males with whiter tails STEIN ARE SÆTHER*, PEDER FISKE*†, JOHN ATLE KA r LA r S† & JENS MAGNE GJUL*
*Department of Zoology, Norwegian University of Science and Technology †Norwegian Institute for Nature Research (Received 19 April 1999; initial acceptance 28 May 1999; final acceptance 1 October 1999; MS. number: 6202)
A previous experimental study of great snipe, Gallinago media, has reported an effect on male mating success of the amount of white in their tails. That result is one of a very limited set of existing experimental results supporting a female mate preference for a morphological trait in animals. However, a later observational study did not find any correlation between amount of white and male mating success. If females sample a limited number of males, their preferences need not result in strong relationships between mating success and trait values in males, possibly explaining the failure to find the predicted correlation. Yet, females of lekking species are thought to have ample opportunities for mate sampling. To resolve these contrasting results, we present in this paper (1) a larger correlational study (several leks during 10 years) showing no relationship between male mating success and whiteness of tails (measured in several ways), and most importantly (2) evidence that individual females do not mate predominantly with males with very white tails among those males that each female samples. These results show that females do not prefer males with whiter tails as mates, within the contemporary natural variation in the trait. They also show that there is no sexual selection of the trait at present. This does not necessarily imply that white tails are not a sexually selected adaptation in males, but the mechanisms are likely to have been different from direct mate choice of whiter tails per se.
because they have weak mating preferences. Even when females have strong unanimous preferences and are able to choose according to their taste, low correlations between male mating success and preferred traits would ensue if each female samples only a few males (Benton & Evans 1998). A few studies have found that manipulating traits of males in the field causes increased mating success, both in lek breeders (Ho ¨ glund et al. 1990a; Andersson 1992; Petrie & Halliday 1994) and in some other species (e.g. Andersson 1982; Møller 1988; Grether 1996). One of these studies involved increasing the amount of white on the tail feathers of great snipe, Gallinago media, males (Ho ¨ glund et al. 1990a). The group of males with experimentally increased white had more matings than controls, corroborating earlier correlational indications of a relationship between male mating success and amount of white on tails (Ho ¨ glund & Lundberg 1987). Ho ¨ glund et al. (1990a) further showed that males with increased white did not enjoy an advantage in male–male interactions, suggesting that the increased mating success was due to a direct female preference for white tails. Yet, in another study of the same species, Fiske et al. (1994) failed to find any correlation between tail whiteness and mating success.
Behavioural biologists have long been interested in the evolution of female mate preferences (Cronin 1991). Such preferences could cause sexual selection and rapid evolution of male traits, and thus explain the striking sexual dimorphism of many species. Studies of lekking species are of particular interest in this context because (1) females receive no help from males in rearing offspring which could otherwise explain the evolution of preferences, (2) females can compare several males in a short time, and (3) males are not monopolized for a long time by any one female. The distribution of mating success is often skewed among males on leks, and female preferences may explain this variation. Several studies have indirectly attempted to identify female preferences by searching for correlates of male mating success (reviewed by Fiske et al. 1998). In general, variation in male traits seems able to explain only a little of the variation in male mating success. However, this could be because females sample a restricted number of males, and not necessarily Correspondence: S. A. Sæther, Department of Zoology, Norwegian University of Science and Technology, N-7491 Trondheim, Norway (email: [email protected]). P. Fiske and J. A. Ka˚la˚s are at the Norwegian Institute for Nature Research, Tungasletta 2, N-7485 Trondheim, Norway. 0003–3472/00/020273+08 $35.00/0
2000 The Association for the Study of Animal Behaviour
273
2000 The Association for the Study of Animal Behaviour
274
ANIMAL BEHAVIOUR, 59, 2
How are we to reconcile these contrasting results? To attempt this, we present the following in this paper: (1) a larger correlational study on the relationship between white on tails and mating success in great snipe for several leks over 10 mating seasons; (2) an examination of the crucial assumption that our measure of white reflects the same as in Ho ¨ glund et al. (1990a); and (3) an analysis of the mating decisions of individual females in relation to the amount of white on the tails of those males each female visited. The results strongly suggest that female great snipes do not prefer males with whiter tails. We conclude that there is no evidence of sexual selection currently acting on this trait in this species and that the earlier experimental results can best be explained by chance events. We emphasize the importance of studying individual females for investigating female mate preferences, and of not simply focusing on the effects of the outcome of mate choice on male mating success. METHODS
Field Methods Our study site is in Ga˚va˚lia (6217 N, 936 E), Norway (see Løfaldli et al. 1992). The study population occupies several leks where snipes have been caught in mist nets and individually marked with coloured leg rings, with permission from the Norwegian Directorate for Nature Management. The data presented here were collected from mid-May to mid-June in 1989–1998 mainly on two leks, but in some years also on other leks as resources allowed. One to four observers (depending on the number of males present) made observations at the leks from 2300 to 0230 hours, the time when females visit leks (Ka˚la˚s et al. 1995). The number of males at each site was usually between 10 and 25, but varied somewhat during the mating season and between leks. Each observer surveyed continuously all visits by females to an assigned subset of the males, and noted all copulations and solicitations by females.
Male Mating Success Since individual females often mate several times in a mating season (Fiske & Ka˚la˚s 1995) we did not simply sum the total number of matings a male received as a measure of mating success. Instead, we tried to estimate the number of individual females that mated with the male. We did this by combining the records for identified females with the minimum number of different unmarked females mating with the male. This latter number was calculated by the criterion that an unmarked female could mate on 3 consecutive nights (Fiske & Ka˚la˚s 1995). We could sometimes keep track of several unmarked females and thus distinguish between different unmarked females mating on the same night, but not across nights. We calculated mating success only for males that were observed to be territorial on at least 5 nights. A successful male is defined as a male that
obtained at least one mating in the specific season. Some females were only seen to solicit copulation and not actually to copulate. These females were also included in a male’s mating success since they had probably mated with that male (Fiske & Ka˚la˚s 1995). Deaths during the mating season are rare in the species (Ho ¨ glund et al. 1992), and we thus feel confident that differential survival did not confound mating success in this study (see Grether 1996). Altogether the summed mating success for the males on the leks and years presented here was 484 females (including 334 seen to copulate), based upon over 2000 female solicitations (including over 800 copulations). The number of marked females seen to solicit mating was 168 female-years (122 individuals). The numbers presented of males at each lek in each year does not necessarily include all males visiting the leks since we excluded males observed for less than 5 days, and some males observed for more than 5 days were not captured.
Measures of White on the Tail Males fan out their white tails during their display behaviour, a rather conspicuous flash of white to a human observer’s eye. The standard measure of the amount of white on the tail in great snipe (Ho ¨ glund et al. 1990a, b, 1992; Fiske et al. 1994; Sæther et al. 1994) is the distance from the first dark spot on the outermost feather to the tip of the same feather (for a drawing of a great snipe tail, see e.g. Ho ¨ glund et al. 1990a). We used this measure (to the nearest 0.1 mm) for the left-hand part of the tail in this paper, and called it ‘measured white’. Measured white varied considerably from male to male (1986–1997, XSD=16.773.13 mm, N=913, range 8.1–32.0, CV 18.66%). It also differs between populations (Ka˚la˚s et al. 1997) and years (Ho ¨ glund et al. 1992), is more extensive on males than females (Ho ¨ glund et al. 1990b) and increases somewhat with age (Ho ¨ glund et al. 1990a, 1992; Sæther et al. 1994). We tried to catch all males every year. In the few cases that we did not obtain measured white for a male in a given year (N=11 of those assigned mating success), we used the value for the next year if this was available. If it was not available, we used the value for the previous year, but only if the male was at least 2 years old. Measured white might not truly reflect the amount of white in a tail. Two tails with the same value for measured white might look very different. If so, any relationships between whiteness and mating success might be obscured. Another measurement of tail whiteness is the number of tail feathers with more than 50% white. This measurement is perhaps even more crude as the two outermost feathers are always at least 50% white, and the maximum number is only four (one bird had five). To obtain a better measure of whiteness we took photographs in the field of the birds’ tails when they were caught, so we could rank the tails later according to the amount of white. The majority of males will readily fan out their tails in the hand, and we photographed the tails (Fig. 1) while trying to ensure that the outer webs of the outermost feathers were exposed.
SÆTHER ET AL.: WHITE TAILS AND FEMALE PREFERENCES
Figure 1. Photographs of four male great snipe tails, to indicate the range of variation used to rank the amount of white on the outer tail feathers, where (c) is of the tail of the most successful male in the study.
To see if such ranking was reliable, we asked 12 university students independently to rank the photographs from three leks in 1994 according to the amount of white. There was strong agreement between the 12 judges (Kendall coefficient of concordance: lek HN: W=0.85, N=24; lek AH: W=0.85, N=12; lek ØS: W=0.90, N=11; all P values <0.01). We thus felt confident that ranking was a reliable measure, and used the rankings of only one judge in subsequent analyses of ranked white. Tails were ranked separately for each year and lek. The ranks generally correlated strongly with the measured amount of white (N=24 lek-years, mean r=0.76, mean sample size=15.6, mean P=0.0003). However, the correlations varied from 0.56 to 0.93 with an outlier at 0.25, supporting our impression that measured white might give less than perfect information about the extent of white in the tail. We therefore analysed male mating success and female mate choice in relation to the ranked photographs, as well as to measured white.
Analyses Male lekwide relationships 1989–1998 We compared successful and unsuccessful males in relation to both measured and ranked white. This was done separately for each lek and year. Similarly, we correlated mating success with both measures of white. Some of the data (for 1989–1992) have been analysed previously (in slightly different ways by Fiske et al. 1994), but are included here for completeness and because relationships with ranked photographs were not included in the previous analyses. In addition, to avoid pseudo-
replication since some males appeared in many years, we calculated average values of measured white over all years for each male. We compared these average values for those males that mated at least once versus those that did not. For this latter analysis we calculated the least detectable effect size for power=0.8, and the power for detecting Cohen’s medium effect size (d=0.5; Cohen 1988). For power analyses, we used GPower 2.1.1 (Erdfelder et al. 1996).
Individual females 1994–1998 To see if females preferred to mate with males with white tails, we compared the amount of white of the males a female was observed to have visited (using both measured and ranked white). In this way we avoided the problem of limited mate sampling in making inferences about mate preferences. We did this by asking (1) if females tended to mate with the male with the most white out of those sampled and (2) if females tended to mate with males with more than the average amount of white out of those sampled. We called the male(s) with which a female mated her chosen male(s), and the others she visited her rejected males. For all analyses we used female-years as the sample size, pooled over all years (here 1994–1998) and leks, and treated data for individual females as independent for each mating season. One of the males in the sample (Fig. 1c) was exceptionally successful. Over the 5 years, this male mated with 37 of the 79 females in the sample of females from which we had observations from both their chosen and at least one rejected male. We therefore carried out additional analyses excluding all observations, in any year, with this male
275
276
ANIMAL BEHAVIOUR, 59, 2
Table 1. Amount of white on left tail (mm) and ranked amount of white from photographs, for males without and with mating success (at least one female copulating or soliciting) on different great snipe leks in different years Measured white Unsuccessful Year
Lek
X±SD
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1997 1994 1995 1998 Sum Mean Mean*
HN HN HN HN HN HN HN HN HN HN AH AH AH AH AH AH AH AH AH AH BJR ØS ØS SAL
16.4±2.1 16.1±2.1 17.0±3.2 16.7±1.2 17.4±3.6 18.3±3.3 20.5±6.4 17.3±3.4 17.3±4.9 16.5±2.2 16.9±2.3 16.0±1.5 16.1±2.3 21.1±1.2 15.1±1.7 16.7±2.4 14.6±2.5 18.2±3.8 17.1±3.0 16.4±1.9 16.7±1.3 17.2±2.8 16.1±1.5 16.0±1.3 16.80±2.80 16.76±2.91
Ranked photographs
Successful N
11 11 13 7 7 11 2 13 14 11 7 6 12 2 8 8 7 6 11 11 5 4 3 10 168 7.0
X±SD 17.3±2.5 16.8±2.2 16.7±2.4 17.2±4.4 14.9±3.7 16.7±3.3 17.1±2.8 16.1±1.8 15.6±1.5 15.3±2.6 18.4±3.2 19.2±6.0 17.1±3.6 16.5±1.8 17.9±2.7 16.5±2.4 16.6±2.0 15.5±2.3 16.5±1.6 16.9±3.3 16.3±2.7 15.6±4.3 20.2±1.8 16.6±2.4 16.98±2.57 16.76±2.62
N
t
P
N
P
8 11 17 12 12 7 7 4 4 7 9 7 6 6 7 6 6 4 8 6 5 4 2 3 200 8.3
−0.82 −0.74 0.30 −0.32 1.47 0.96 1.22 0.67 0.69 1.08 −1.03 −1.27 −0.70 3.22 −2.44 0.18 −1.58 1.28 0.49 −0.41 0.26 0.63 −2.87 −0.64
0.43 0.47 0.77 0.75 0.16 0.35 0.26 0.51 0.50 0.30 0.32 0.23 0.49 0.02 0.03 0.86 0.14 0.24 0.63 0.69 0.80 0.56 0.06 0.54
0.66 0.90 0.48 0.84 0.14 0.90 0.56 0.41 0.96 0.48 0.68 0.18 0.96 0.07 0.28 0.90 0.09 0.91 0.55 0.66 0.22 0.69 0.20 0.29
−0.02 −0.01
0.42 0.46
19 22 30 19 19 14 9 17 18 18 16 13 18 8 15 14 12 10 19 17 10 8 5 13 363 15.1
0.54 0.57
P values for ranked photographs are exact probabilities from Mann–Whitney U tests. *Weighted for sample size.
(six of the 37 females were included since they also mated with other males). This also caused some females not choosing this male to be excluded, since this was their sole rejected male (N=5). We used randomization tests to calculate probability values for obtaining by chance the observed number of females mating with the male with the most white. Since this probability is strongly affected by the number of males visited, which varied between females, we cannot compute this as a simple binomial probability. Instead, we used the following randomization technique to obtain a proper null model. We first counted the number of females, of the k females in the sample, mating with the male with the most white as y. To calculate the one-tailed probability of obtaining y by chance, we randomly drew for each female i one number j from the set {mi(1), ni mi(0)} where mi is the number of males with which the female mated and ni is the number of males visited. We then calculated i=k i=1 ji 10 000 times to obtain an expected distribution under the null-hypothesis. We counted the number of times i=k i=1 ji ≥y as x. The one-tailed probability of y was then calculated as x/10 000. We analysed if females mated with males with more than the average amount of white in two ways. First, we used parametric statistics to see if there was a difference in the amount of white (mean values for each female)
between chosen and rejected males (or Wilcoxon signedranks tests, for ranked white). Second, we calculated the fractional rank of the preferred males (rank divided by number of males visited, 1.0=mating with the one with most white). To obtain the probability of the observed mean fractional rank of chosen males, z, occurring by chance, we used the following randomization approach. A number hi was randomly drawn for each female i from the set {1, 2 . . . ni} and the ratio li =hi/ni computed. If a female had mated with more than one male we calculated l as many times as the number of chosen males she had, and used the average. We then calculated the simulated mean fractional rank over all k females, i=k i=1 li/k, 10 000 times and counted w as the number of times i=k i=1 li/k ≥z. The one-tailed probability of z was obtained as w/10 000. Randomization programmes were compiled using Resampling Stats 4.0.7 (Resampling Stats Inc., Arlington, VA, U.S.A.). RESULTS
Sexual Selection on Amount of White For measured amount of white, the weighted mean of means over all years and all leks was identical for successful and unsuccessful males (16.76 mm; Table 1). We
SÆTHER ET AL.: WHITE TAILS AND FEMALE PREFERENCES
Table 2. Relationships between male mating success (no. of females copulating or soliciting) and amount of white on tail on different great snipe leks in different years Measured white
Ranked photographs
Year
Lek
r
N
P
rS
P
rS
N
P
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1997 1994 1995 1998 Sum Mean Mean*
HN HN HN HN HN HN HN HN HN HN AH AH AH AH AH AH AH AH AH AH BJR ØS ØS SAL
0.27 0.09 −0.14 0.20 −0.08 −0.42 −0.53 −0.26 −0.21 −0.23 0.32 0.06 0.05 −0.75 0.47 −0.05 0.12 −0.31 0.19 0.17 −0.30 0.44 0.65 0.16
19 22 30 19 19 18 9 17 18 18 16 13 19 8 15 14 13 10 19 17 10 8 5 13 369 15.4
0.26 0.68 0.47 0.41 0.75 0.09 0.14 0.31 0.39 0.36 0.22 0.85 0.84 0.03 0.08 0.87 0.70 0.38 0.44 0.52 0.41 0.28 0.23 0.60
0.17 0.14 −0.01 0.24 −0.22 −0.31 −0.34 −0.15 −0.18 −0.26 0.31 0.27 0.06 −0.62 0.40 −0.08 0.29 −0.34 −0.06 0.14 −0.38 −0.18 0.78 0.16
0.47 0.53 0.96 0.31 0.35 0.22 0.37 0.56 0.46 0.30 0.24 0.35 0.80 0.10 0.13 0.78 0.33 0.35 0.80 0.58 0.29 0.67 0.12 0.61
0.09 0.05 −0.08 −0.03 0.21 −0.22 0.35 −0.25 −0.02 −0.16 0.10 0.37 −0.01 −0.54 −0.29 −0.03 0.42 −0.03 0.21 0.12 0.45 −0.05 0.78 0.32
0.69 0.82 0.70 0.90 0.38 0.43 0.36 0.33 0.93 0.52 0.70 0.20 0.97 0.17 0.29 0.92 0.18 0.93 0.38 0.64 0.20 0.90 0.12 0.29
0.43 0.45
−0.01 −0.00
0.45 0.49
0.07 0.05
19 22 30 19 19 14 9 17 18 18 16 13 18 8 15 14 12 10 19 17 10 8 5 13 363 15.1
−0.00 −0.00
0.54 0.58
N is the same for the two correlation coefficients of measured white. *Weighted by sample size.
compared successful and unsuccessful males in 24 lekyears, and found only two significant differences (one in each direction). There was no difference between these two groups of males in white as judged from the ranking of photographs either (Table 1). The weighted mean rank correlation coefficients over all years and all leks were close to zero for both measured white and ranked photographs (Table 2), and not significant in any single lek-year. Including each individual male only once by using average values for measured white gave similar results. For 118 successful males (in at least 1 year) the averaged measured amount of whiteSD was 16.622.95 mm and for 152 unsuccessful males, 16.622.71 mm (t268 =0.023, P=0.98). Least detectable effect size (for power=0.8) was d=0.34, and for medium effect size (d=0.5) power=0.98. Restricting the analysis of averaged measured white to birds at least 2 years old gave similar results (successful males: XSD=16.723.10 mm, N=91; unsuccessful males: 16.702.33 mm, N=80; t169 =0.049, P=0.96, least detectable effect size d=0.43, power for medium effect size=0.90).
Mate Choice for Whitest Males
only 18 out of 79 mated with the male with the most white among those the female visited and 61 mated with another male (Fig. 2). The probability of obtaining 18 or fewer females mating with the male with the most white in the tail out of 79 was 0.029 (given the observed number of males visited for each female, randomization test, 10 000 simulations, mean simulated value=25.9), indicating that females actually avoided mating with the male with most white. However, this result was probably due to one particularly successful male with a low amount of white. When this male was excluded from the sample, 13 of 43 females mated with the male with the most white (Fig. 2). The probability of obtaining 13 or fewer females mating with the male with the most white on the tail out of 43 was 0.33 (randomization procedures as above, mean simulated value=14.9). Forty-seven out of 79 females mated with the male with the least amount of white among those sampled (Fig. 2; randomization procedures as above: P=0.0001, mean simulated value=25.00), but when the most successful male was excluded the figures were 14 out of 43, 14 being the most likely outcome of all by chance (randomization procedures as above: P=0.57, mean simulated value=14.0).
Measured white
Ranked photographs
Among those females observed both at territories of males they mated with and males they did not mate with,
Twenty-two of 78 females mated with the male with the most white of those visited, as judged by ranking of
277
ANIMAL BEHAVIOUR, 59, 2
(a) 80 Most white chosen Most white rejected Expected no. choosing most white
70 60 50 40 30 20 10 No. of females
278
0
MW
RP
MW
RP
(b) 80 Least white chosen Least white rejected Expected no. choosing least
70 60 50 40 30 20
result was confirmed by a two-way ANOVA using amount of white as the dependent variable, and chosen or not and female identity as factors (choice: F1,92.81 =13.79, P<0.001; female: F78,78 =0.605, P=0.99; interaction: F78,159 =0.78, P=0.89; only females observed at both their chosen and at least one rejected male are included). When we excluded the most successful male there was no difference in the mean amount of white of chosen and rejected males (chosen: XSD=16.992.45 mm; rejected: 17.061.85; paired t test: t42 = 0.138, P=0.89). This was also confirmed by a two-way ANOVA using amount of white as dependent variable (choice: F1,47.01 =0.02, P=0.89; female: F42,42 =0.72, P=0.87; interaction: F42,74 =13.79, P=0.71). The mean fractional rank of the male(s) chosen was 0.57 (rank divided by number of males visited, 1=mating with the one with most white). The one-tailed probability of obtaining a mean fractional rank of 0.57 or lower in this sample was 0.001 (randomization test: 10 000 simulations, mean simulated value=0.66). The mean fractional rank of the males rejected was 0.70 (paired t test: t78 = 2.949, P=0.004). Excluding the most successful male yielded a mean fractional rank for chosen males of 0.70, which was close to what was expected by chance (randomization as above: mean simulated value=0.67, probability of getting 0.70 or greater=0.19). For rejected males the mean fractional rank was 0.67 (paired t test: t42 =0.49, P=0.63).
10 0
MW RP All males included
MW RP Most successful male excluded
Figure 2. The number of females that mated with the male with (a) the most white on the tail and (b) the least white on the tail of those visited and the number expected from null model simulations, for measured white (MW) and white ranked from photographs (RP). Also shown is the number of females that did not mate with such a male.
Ranked photographs Using ranked photographs we found that chosen males were ranked to be less white than rejected ones (Wilcoxon signed-ranks test: Z= 2.369, N=78, P=0.018). Again, excluding the most successful male resulted in no significant differences in amount of white between chosen and rejected males (Z= 1.251, N=42, P=0.21). DISCUSSION
photographs (Fig. 2a; randomization probability of 22 or less=0.16, mean simulated value=26.5). Excluding the most successful male gave 17 out of 42 (randomization probability of 17 or greater=0.21, mean simulated value=14.1). Thirty-nine out of 78 mated with the male with the least amount of white (Fig. 2b; randomization probability of 39 or more=0.003, mean=26.5) but when the most successful male was excluded the proportion dropped to 11 out of 42 (randomization probability of 11 or less=0.19, mean=14.1).
Mate Choice for Whiter Males Measured white The mean amount of white of those males with which a female mated was not larger than the mean amount of white of males with which she did not mate. Rather, there was on average a 1.55-mm difference in the opposite direction (chosen: XSD=15.342.54 mm; rejected: 16.892.18; paired t test: t78 =3.653, P<0.001). This
It is of paramount importance to the credibility of any field of empirical inquiry that results be replicated and negative results published. In behavioural ecology, the value placed on novel results is so high that these goals may suffer and be neglected. Publication bias and selective reporting (unconscious as it may be) are serious obstacles to scientific progress, and seems to be responsible for overstating the strength and biological significance of some relationships studied by behavioural ecologists (Palmer 1999; see also Thornhill et al. 1999). This may also be true for the strength of current sexual selection on male traits in animals. Our results may be summarized as follows: neither measure of white was related to male mating success (Tables 1, 2). Both measures of white showed that females did not prefer to mate with the male with most white on the tail of those visited (Fig. 2a). By including the most successful male, there were in fact more females than expected that mated with the male with the least amount of white (Fig. 2b). For neither measure of white, analysed
SÆTHER ET AL.: WHITE TAILS AND FEMALE PREFERENCES
in several different ways, did we find any indications whatsoever that females preferred whiter males as mates. We did, however, find the opposite: the rejected males had significantly more white because many females mated with one particular male that had only a little white in his tail. We think it is safe to say that the lack of any positive relationships was not due to low statistical power, since the sample sizes were high both in the male-centred correlational analyses and in the femalecentred individual mate choice analyses. This was confirmed by a power analysis of successful and unsuccessful males. In the experiment where Ho ¨ glund et al. (1990a) manipulated great snipe tails, 16 of 17 matings went to manipulated males and only one to control males. This is significantly different, with a chi-square test, from the expected frequencies based on number of experimental and control birds. Altogether five males achieved matings, but 10 of these went to one male. Since so few males were successful, they might have been so in the absence of any manipulation, and in fact the frequency of manipulated males achieving mating success was not significantly higher than control birds. The best predictor of male mating success in great snipe seems to be mating success the previous year (Fiske et al. 1994). Thus, a better experimental design would be to block for this effect. Another problem was that mating success was measured as number of matings and not number of different females mating. Although Ho ¨ glund et al. (1990a) tried to count repeated copulations by the same female only as one, they could not know if the same female returned and mated on a new visit, which we now know that females in fact often do (Fiske & Ka˚ la˚ s 1995). Ho ¨ glund & Alatalo (1995) reanalysed the experimental data, using a paired t test of 12 pairs of manipulated and control neighbours. The onetailed probability of the observed difference in number of copulations was 4%. Again, this analysis suffers, albeit less, from the problem of counting copulations instead of females, and the choice of test statistic is questionable because of mating skew. Even if we do accept the results at face value, they could still be due to chance effects. Another problem (pointed out by Bennett et al. 1994) with the experiment was the use of Tipp-Ex fluid. The fluid reflects in the ultraviolet, and may not be perceived by the birds as having the same colour as white feather. The large effect the most successful male had on our results illustrates the problems with such studies on animals that can show large mating skew. Because this very successful male happened to have a tail with little white, a seemingly strong preference for males with little white disappeared when we removed this individual male from the sample. Something similar might have occurred in Ho ¨ glund et al.’s experiment: one very successful male happened to be manipulated to be more white. Without proper controls (both previously successful and unsuccessful males, preferably also manipulations in both directions) it is difficult to conclude much from results involving few very successful males and many unsuccessful males. It seems plausible that great snipes have evolved white tails through sexual selection. This is hinted at because
the trait (1) is sexually dimorphic (Ho ¨ glund et al. 1990b), (2) increases somewhat with age (Ho ¨ glund et al. 1990a, 1992; Sæther et al. 1994), (3) is different from those of all other species in the genus Gallinago (Cramp & Simmons 1983), (4) is more pronounced in more southern populations that experience darker nights (Ka˚ la˚ s et al. 1997), and (5) is frequently displayed at leks (Lemnell 1978). There are many ways in which sexual selection could have favoured whiter tails. Our data do not support the claim that females have evolved any preferences for white tails, but it cannot be ruled out that the shape of preference functions of females (Widemo & Sæther 1999) is such that males with less white than the natural variation would have been discriminated against. At least we can say that females do not discriminate between males based on contemporary variation in this trait. The benefit for males of having white tails could be through interactions with other males (e.g. by using them as a threat signal). It could be that the trait is not costly to produce or costly in itself to carry, so that all males have more or less white tails, and the information depends instead on how the tail is used and not on how white it is. Another possibility is that it is not the amount of white per se but the patterns created by the black and white patches that contain information. To display such patterns, the tail must be more or less white. This could be used, for example, in individual recognition. Tail marks may also have an amplifier function of feather quality (Hasson 1991), since absence of melanin in white patches may increase damage and abrasion (Fitzpatrick 1998). The present study does not provide support for that theory. In addition, we only rarely see much abrasion and damage of great snipe tail feathers, as opposed to wing feathers (Sæther et al. 1994), probably because these short tails are not exposed to much physical contact with the surroundings. Although we did not find evidence for any female preference, our methods and our approach of comparing the sampled males of individual females should be of general interest and more widely employed (see also Petrie et al. 1991). This will be more appropriate for the study of female mate choice than the traditional emphasis on sexual selection. The latter approach is of more interest for the evolutionary consequences of female mate choice on male traits, but not for understanding mate choice itself. Studies of mate preferences should focus on individual females (Widemo & Sæther 1999) and the males available to each female (Benton & Evans 1998), and not simply on the effects of preferences on the distribution of matings among males. It was for this reason we compared males that each female rejected with those that she chose. In lekking species however, it is not obvious that females only sample males whose territories they visit. This will depend on the nature of the trait of interest: is it a long-range or short-range signal? In the present case it is likely that if females prefer white tails, they cannot examine this from outside the territory. But despite ample opportunity to do so within territories, they did not prefer males with whiter tails. Given the reasonably large sample sizes obtained, from several lek sites and over many years, using both a sexual selection
279
280
ANIMAL BEHAVIOUR, 59, 2
male approach and an individual-level female mate choice approach, we feel confident to conclude that female great snipes do not prefer males with whiter tails. We have assumed that if females preferred white tails, they would all do so. What we have rejected therefore is the hypothesis that females unanimously prefer whiter tails. We cannot reject a hypothesis that females somehow use information contained in the tail patterns of males in their mate choice. It is in this respect intriguing that males often display their tails clearly directed towards visiting females (personal observations), as if a male’s tail is something that a female is paying attention to after all. Acknowledgments We thank our field assistants (especially Sten L. Svartaas and Tord Bretten) for making it all possible, and Jacob Ho ¨ glund and the anonymous referees for comments. References Andersson, M. 1982. Female choice selects for extreme tail length in a widowbird. Nature, 299, 818–820. Andersson, S. 1992. Female preference for long tails in lekking Jackson’s widowbirds: experimental evidence. Animal Behaviour, 43, 379–388. Bennett, A. T. D., Cuthill, I. C. & Norris, K. J. 1994. Sexual selection and the mismeasure of color. American Naturalist, 144, 848–860. Benton, T. G. & Evans, M. R. 1998. Measuring mate choice using correlation: the effect of female sampling behaviour. Behavioral Ecology and Sociobiology, 44, 91–98. Cohen, J. 1988. Statistical Power Analysis for the Behavioral Sciences. 2nd edn. Hillsdale, New Jersey: L. Erlbaum. Cramp, S. & Simmons, K. E. L. (Eds) 1983. The Birds of the Western Palearctic, Vol. 3. Oxford: Oxford University Press. Cronin, H. 1991. The Ant and the Peacock. Cambridge: Cambridge University Press. Erdfelder, E., Faul, F. & Buchner, A. 1996. GPOWER: a general power analysis program. Behavior Research Methods, Instruments, and Computers, 28, 1–11. Fiske, P. & Ka˚la˚s, J. A. 1995. Mate sampling and copulation behaviour of great snipe females. Animal Behaviour, 49, 209–219. Fiske, P., Ka˚la˚s, J. A. & Sæther, S. A. 1994. Correlates of male mating success in the lekking great snipe (Gallinago media): results from a four year study. Behavioral Ecology, 5, 210–218. Fiske, P., Rintama¨ki, P. T. & Karvonen, E. 1998. Mating success in lekking males: a meta-analysis. Behavioral Ecology, 9, 328–338. Fitzpatrick, S. 1998. Birds’ tails as signaling devices: markings, shape, length, and feather quality. American Naturalist, 151, 157–173.
Grether, G. F. 1996. Sexual selection and survival selection on wing coloration and body size in the rubyspot damselfly Hetaerina americana. Evolution, 50, 1939–1948. Hasson, O. 1991. Sexual displays as amplifiers: practical examples with an emphasis on feather decorations. Behavioral Ecology, 2, 189–197. Ho ¨ glund, J. & Alatalo, R. V. 1995. Leks. Princeton, New Jersey: Princeton University Press. Ho ¨ glund, J. & Lundberg, A. 1987. Sexual selection in a monomorphic lek-breeding bird: correlates of male mating success in the great snipe Gallinago media. Behavioral Ecology and Sociobiology, 21, 211–216. Ho ¨ glund, J., Eriksson, M. & Lindell, L. E. 1990a. Females of the lek breeding great snipe, Gallinago media, prefer males with white tails. Animal Behaviour, 40, 23–32. Ho ¨ glund, J., Ka˚la˚s, J. A. & Løfaldli, L. 1990b. Sexual dimorphism in the lekking great snipe. Ornis Scandinavica, 21, 1–6. Ho ¨ glund, J., Ka˚la˚s, J. A. & Fiske, P. 1992. The costs of secondary sexual characters in the lekking great snipe (Gallinago media). Behavioral Ecology and Sociobiology, 30, 309–315. Ka˚la˚s, J. A., Fiske, P. & Sæther, S. A. 1995. The effect of mating probability on risk taking: an experimental study in lekking great snipe. American Naturalist, 146, 59–71. Ka˚la˚s, J. A., Kuresoo, A., Luigujo ˜ e, L. & Svartaas, S. L. 1997. Morphometrical comparison between Estonian and Norwegian great snipes (Gallinago media). Proceedings of the Estonian Academy of Sciences Biology Ecology, 46, 115–122. Lemnell, P. A. 1978. Social behaviour of the great snipe Capella media at the arena display. Ornis Scandinavica, 9, 146–163. Løfaldli, L., Ka˚la˚s, J. A. & Fiske, P. 1992. Habitat selection and diet of great snipe Gallinago media during breeding. Ibis, 134, 35–43. Møller, A. P. 1988. Female choice selects for male sexual tail ornaments in the monogamous swallow. Nature, 332, 640–642. Palmer, A. R. 1999. Detecting publication bias in meta-analyses: a case study of fluctuating asymmetry and sexual selection. American Naturalist, 154, 220–233. Petrie, M. & Halliday, T. 1994. Experimental and natural changes in the peacock’s (Pavo cristatus) train can affect mating success. Behavioral Ecology and Sociobiology, 35, 213–217. Petrie, M., Halliday, T. & Sanders, C. 1991. Peahens prefer peacocks with elaborate trains. Animal Behaviour, 41, 323–332. Sæther, S. A., Ka˚la˚s, J. A. & Fiske, P. 1994. Age determination of breeding shorebirds: quantification of feather wear in the lekking great snipe. Condor, 96, 959–972. Thornhill, R., Møller, A. P. & Gangestad, S. W. 1999. The biological significance of fluctuating asymmetry and sexual selection: a reply to Palmer. American Naturalist, 154, 234–241. Widemo, F. & Sæther, S. A. 1999. Beauty is in the eye of the beholder: causes and consequences of variation in mating preferences. Trends in Ecology and Evolution, 14, 26–31.