Do male barn swallows (Hirundo rustica) experience a trade-off between the expression of multiple sexual signals?
- First Online:
- Cite this article as:
- Saino, N., Romano, M., Sacchi, R. et al. Behav Ecol Sociobiol (2003) 54: 465. doi:10.1007/s00265-003-0642-z
- 95 Views
Some sexual selection models envisage exaggerated male secondary sexual characters to be costly and therefore reliable indicators of the quality of potential mates to choosy females. If male secondary sexual characters have a natural selection cost, they may be linked to each other by reciprocally constraining relationships that would prevent individual males from increasing their level of multiple signaling. Barn swallow (Hirundo rustica) males have at least two costly signals relevant to socio-sexual interactions: tail length and song. Under the hypothesis that a trade-off exists between male signals, we manipulated the maintenance cost of tail ornaments to test whether this reduced the quantity and quality of song, a condition-dependent, phenotypically plastic signal. Contrary to our expectation, tail elongation had no effect on singing activity and song complexity. However, tail-elongated males produced songs with longer terminal parts ('rattles'). Long rattles are associated with highly competitive social contexts and high circulating levels of testosterone, suggesting that tail-elongated males were more frequently involved in either male-male aggressive or inter-sexual interactions. Therefore, this study shows that males are not displaying multiple signals at the maximum possible level, implying that this system is open to unreliable communication. However, long-term trade-offs between signal expression and viability may prevent males from displaying both signals at higher rates.
KeywordsCondition-dependent signalsHirundo rusticaMultiple signalsReliable communicationSexual selection
Sexual selection leads to the evolution of characters that confer an advantage in mate acquisition, but exceed the natural selection optimum (reviewed by Andersson 1994; Møller 1994). One fundamental consequence of larger investment in reproduction by females is that such exaggerated characters are more likely to evolve in males than in females. Honest-advertisement models of sexual selection propose that condition-dependent male secondary sexual characters function as reliable signals of male quality. According to these models, low-quality males are prevented from signaling at high levels because sexual signals are costly to produce and/or maintain, and because such costs are differentially larger for low-quality compared to high-quality males (Zahavi 1975, 1977; Andersson 1986, 1994; Heywood 1989; Grafen 1990; Iwasa et al. 1991).
However, males of many species have more than a single exaggerated trait (e.g., song and feather ornamentation of male birds) (Rowe 1999). The question of the function of multiple signals in inter-sexual communication systems has been repeatedly treated by theoreticians (e.g., Møller and Pomiankowski 1993; Iwasa and Pomiankowski 1994; Johnstone 1995). Relatively few empirical studies have investigated the combined effects of multiple signals on female choice (e.g., Zuk et al. 1990; Omland 1996a, 1996b), and none, to the best of our knowledge, has experimentally analyzed the trade-offs that may constrain the simultaneous expression of signals relevant to sexual selection.
In general, males are expected to display multiple signals at a level that maximizes their mating success. However, if male secondary sexual characters are condition-dependent and entail the bearer with production and/or maintenance costs, the expression of one signal may have to be traded against the expression of other signals. This is the case because any increment in the expression of one signal will negatively affect the ability of individual males to sustain the costs of other signals. This scenario is complicated by the fact that signals may be affected by individual condition over different time scales, and they may also differ in phenotypic plasticity, i.e., in the degree to which their expression can be tuned according to variation in costs and current individual condition. For example, feather signals that are produced during molt, months before pair formation and reproduction, cannot be altered until the subsequent molt. In contrast, behavioral displays such as song may have a dual function of mate attraction and territorial defense, and they may vary in expression over time scales several orders of magnitude shorter than morphological signals. Differences in timing of production of different signals in conjunction with high unpredictability of environmental conditions, and thus costs of signals, suggest that not all males may display at the optimal level relative to their quality. Hence, some males may display at higher rates than suggested by their condition, while others may do the reverse.
Nonetheless, the existence of a trade-off among signals should result in the depression of plastic signals under poor environmental conditions. This prediction may be tested by investigating the effects of manipulation of a single signal on the expression of other signals. Any experimental increase in the expression (and thus the cost) of a single signal is expected to depress the expression of other signal(s). However, the trade-off between secondary sexual characters may not emerge if the cost of secondary sexual characters is compensated by reduced investment in other traits. If that is the case, no trade-off between sexual signals will emerge because the cost of an experimental exaggeration of one signal will be paid in currencies other than sexual ornamentation. In addition, positive feed-back mechanisms between the expression of secondary sexual characters may result in positive rather than negative correlations between the expression of signals. For example, males with experimentally enhanced morphological ornaments may be more frequently involved in intra-sexual aggressive interactions or extra-pair matings. This may affect their hormonal (e.g., androgen) profile resulting from altered social and sexual behavior and cause enhanced expression of behavioral secondary sexual characters (e.g., song) which are controlled by the same hormonal factors.
The main aim of this experimental study of the barn swallow (Hirundo rustica) was to investigate the trade-off between male secondary sexual characters. This was done by manipulating a morphological secondary sexual character (tail length) and assessing the effect of this treatment on singing behavior, both in terms of singing activity and song quality. We predicted that an experimental increase in male tail length should result in depression of singing behavior, because male tails impose maintenance costs which increase with the size of the character (see below), and because the cost of singing increases with time devoted to singing activity.
Secondary sexual characters of male barn swallows
The barn swallow is a semi-colonial, socially monogamous passerine bird. Sexual size dimorphism is small, with the exception of the two outermost tail feathers, which are approximately 17% longer in males than in females in our Italian population (Møller et al. 1995a). Correlational and experimental studies have repeatedly shown that females prefer males with long tails as social mates and extra-pair fathers of their offspring (Møller 1988, 1994; Saino et al. 1997d). Naturally long-tailed males have higher survival prospects, stronger immune responses and smaller parasite loads than short-tailed males (Møller 1991a, 1991b, 1994; Saino et al. 1995, 1999, 2002c; Saino and Møller 1994, 1996). This suggests that tail length may serve as a condition-dependent, reliable signal of male quality. Male tail ornaments carry maintenance costs. In fact, tail manipulation experiments using the same procedure we used in the present study have shown that artificial tail elongation results in reduced long-term survival, expression of tail ornaments the following year, foraging efficiency, size of insect prey, humoral immune response, and increased physiological costs of flying, whereas tail shortening generally results in the opposite effects (see Møller 1989, 1994; Møller and de Lope 1994; Møller et al. 1995b; Saino et al. 1997a, 1997b). We suggest that these costs are mediated through flight costs of long tails, reduced foraging efficiency, and hence ultimately locomotion.
Song also plays a role in inter-sexual selection. Male barn swallows produce long, complex songs, which consist of largely non-stereotyped sequences of 19.8 syllables on average (range 9–36), arranged in singing bouts of 2.9 songs (range 1–17) (Galeotti et al. 1997; unpublished data). Males vary considerably in their singing activity and quality of their song (Møller 1994; Galeotti et al. 1997; Saino et al. 1997c). Singing rate is a condition-dependent character, since it is negatively affected by ectoparasite infestation (Møller 1991c) and negatively correlates with immunoglobulin levels (Saino et al. 1997c). Males with a high singing rate are more likely to sire offspring in their own brood compared to males that sing infrequently, independent of the length of their tails (Møller et al. 1998). However, song also appears to be relevant in male intra-sexual behavior. For example, several spectrographic features of song vary with social context, in terms of distance to nearest neighbors in a colony and operational sex ratio (Galeotti et al. 1997). The terminal, harsh part of the song (the 'rattle') is particularly emphasized in highly intra-sexually competitive contexts, and its duration positively correlates with plasma testosterone levels (Galeotti et al. 1997).
Thus, male barn swallows possess at least two different secondary sexual signals which are relevant in reproductive contexts. These traits differ in plasticity and timing of production. Tail feathers are produced during the annual molt in the African winter quarters in December to February. Song is produced during the entire breeding season (late March to August in our study area), but also in the winter quarters, and it may be affected by individual condition during breeding (Saino et al. 1997c) and socio-sexual environment in the breeding colonies (Galeotti et al. 1997).
The correlation between song and tail length has consistently been shown to be weak in different European breeding populations (e.g., Møller 1994; Saino and Møller 1995; Møller et al. 1998; N. Saino, unpublished data). No significant correlation has been demonstrated between singing rate and song quality (Galeotti et al. 1997). Hence, associations among different secondary sexual characters in the barn swallow are generally weak. However, correlations may not reveal trade-offs among secondary sexual characters if individuals of different quality have different optimum trait combinations.
We conducted the experiment in four colonies located in cattle farms in an area east of Milano (Northern Italy) during spring/summer 2001 (see also Saino et al. 2002a, 2002b). Adults were captured soon after arrival at the breeding colonies, starting in early April, and were individually marked with color bands and with markings on their breast and belly feathers to allow recognition during behavioral observations. Sex was determined by inspection of the cloacal protuberance and was confirmed at later recaptures by inspection for presence (female) or absence (male) of an incubation patch, and by observation of sexual and incubation behavior (only females incubate). Males were sequentially assigned to one of four tail manipulation treatments (sample sizes see Fig. 2). Males of the first group were subjected to shortening of the two outermost tail feathers by removing a 20 mm-long piece approximately 10 mm from the feather base. Males of the second group were just handled with no tail manipulation. Males of the third group had their tail cut and re-glued without altering the original length, thus serving as a second control group. Males of the fourth group had their outermost tail feathers elongated by 20 mm by inserting a piece of feather at approximately 20 mm from the feather base. In each tail manipulation session, males were arranged in random order before starting sequential assignment, so that tail manipulation treatment did not reflect capture order. These procedures of tail manipulation were similar to those adopted in several previous experiments (e.g., Møller 1994; Saino et al. 1997a, 1997d). All males were manipulated at least 10 days before laying of the first egg by their social mate. No significant variation in pre-manipulation tail length or any other morphological variable, including three measures of the bill, tarsus length and keel length, wing chord, and central rectrix length existed among the four experimental groups of males (ANOVA, F-values associated with P>0.05 in all cases), suggesting that the groups were homogeneous in composition.
The day following tail manipulation we started daily observations of behavior of focal males. Each morning (except Sundays), between 0700 and 1000 hours we performed a 1.0–1.5 h observation session where every second minute we recorded whether focal males were in their territory and whether they were singing. Up to five males could be observed simultaneously. Singing rate of an individual was subsequently expressed as the proportion of observations (taken every second minute) in which the male was singing, averaged across all observation days spanning from 9 days before laying of the first egg by their social mate to the day of laying of the penultimate egg. Singing rates were computed over this period because it presumably corresponds to the female fertile period (Møller 1994). The methods of recording and expressing singing rate are similar to those adopted in several previous studies, and singing rates have been shown to be highly repeatable within days (Møller 1990, 1991c; Galeotti et al. 1997; Møller et al. 1998). High repeatability (Falconer 1981) was also found in this study (F=8.03, df=5, 6, P=0.012, R=0.78).
Spectrographic analysis of songs
Song recordings were made by a Sony TCD-D7 DAT tape recorder connected to an AKG C451 shotgun microphone with a CK 9 hypercardioid capsule placed 1–5 m from the focal individual. Recordings were analyzed by Avisoft SAS Lab Pro 3.93 (Specht 2001) computer program. The best resolution was achieved by analyzing songs in the 0–10 kHz frequency range, with a sampling rate of 20,000 samples/s, band width 200 Hz, frequency resolution 80 Hz, and time resolution 32 ms.
For each song we measured the following variables: (1) song duration; (2) number of syllables; (3) number of different types of syllables (see Galeotti et al. (1997) for a description of repertoire); (4) mean peak amplitude frequency (the frequency where most of the energy was concentrated) computed across syllables; (5) number of pulses in the rattle; and (6) duration of the rattle. In the analyses we used mean values computed across all songs available for each individual. The number of songs available per individual ranged between 5 and 36. We tested for consistency of spectrographic song features within individuals in one-way analyses of variance of individual songs where male was entered as a factor. All song features showed highly significant variation among individuals (F>4.77, df=57, 573 and P<0.001 for all variables). R2-values from these analyses of variance were larger than 0.48 for variables 1–3 and 5–6 whereas R2 was 0.32 for mean peak amplitude frequency. Within-individual coefficients of variation of spectrographic variable values were not significantly correlated with the number of songs recorded for each individual (r-values ranging from −0.24 to 0.10, n=58, P>0.05 in all cases). Hence, song features were highly repeatable among songs of the same individuals, and the observed variability within individuals did not depend on the size of the sample of songs. These results suggest that the songs we recorded were representative of an individual's song features and variation in sample size of songs among individuals did not markedly affect our analyses.
We adopted a generalized linear model (GLM) approach in all analyses of the effect of tail manipulation on male singing behavior. In these analyses, colony and tail manipulation were included as factors and pre-manipulation tail length as covariate, while allowing for two-way interactions among predictor variables. Singing rates were expressed as proportions, and we therefore assumed a binomial error distribution. A backward selection procedure of non-significant main effects and interactions was adopted to obtain parsimonious statistical models including only significant terms. At each step of the backward selection procedure, the independent predictor that provided the smallest partial contribution to the model was excluded. Interactions were excluded from the models before main effects. Breeding date (i.e., laying date of the first egg) did not significantly affect song variables in any GLM analysis and was therefore excluded from final models by backward selection of non-significant predictors in all cases. GLM analyses assuming a gaussian error distribution were used to investigate the effect of tail manipulation, colony and original tail length on song spectrographic quality. In the correlation analyses among song variables, sequential Bonferroni correction was applied to control for type I statistical errors due to 15 simultaneous tests.
We only recorded singing rates (in the four colonies) and song (two colonies) from a sub-sample of treated males for practical reasons, i.e., limitation of time available to obtain records of singing rates and song and the possibility to regularly observe and approach individual males given the presence of bulls in the stables. The individuals with recorded songs represented a random sample, and we are therefore confident that sub-sampling did not bias the results.
Correlations among the six song features measured in the 58 barn swallow (Hirundo rustica) males of the four experimental groups, and between each feature and the scores for two principal components. Uncorrected significance values are shown. Asterisks indicate correlations that were significant after sequential Bonferroni correction for 15 simultaneous tests. Variance explained by the first two principal components was 50.3% for PC1 (eigenvalue 3.018) and 31.4% for PC2 (eigenvalue 1.881)
Duration of song
Number of syllables
Types of syllables
Pulses in the rattle
Peak amplitude frequency
Number of syllables
Types of syllables
Pulses in the rattle
Peak amplitude frequency
Individual scores for the second principal component (PC2) showed no significant variation in relation to colony or tail length (F-values associated with P>0.05 in all cases). However, a significant effect of tail manipulation was observed, with tail-elongated males having the largest scores and tail-shortened males the smallest scores on PC2 (Fig. 3; F=4.69, df=3, 54, P=0.001). Post-hoc Bonferroni tests of differences between pairs of tail-treatment groups showed that tail-elongated males had larger PC2 scores than tail-shortened (P=0.002) or unmanipulated (P=0.01) males (Fig. 3). These results imply that males with an experimentally elongated tail produced longer 'rattles', with more pulses, than those with a shortened tail, and the two control groups produced 'rattles' with intermediate features.
In this study, we manipulated the size of a morphological secondary sexual character, length of the two outermost tail feathers, of male barn swallows and tested the effect on the expression of song. We found no significant effect of tail manipulation on singing activity, expressed as the frequency of song production during the female fertile period. Song duration, peak fundamental frequency and complexity, in terms of number and diversity of syllables were also unaffected by tail manipulation. However, the terminal part of the song, the rattle, was exaggerated in males that had their tail ornaments experimentally enlarged.
These findings contradict the prediction of a negative effect of tail elongation on singing rate that was based on the hypothesis that trade-offs are constraining the simultaneous expression of male secondary sexual characters because these are costly. The energetic costs of vocal displays have been measured only for a few bird species. Most studies suggest that singing is costly (Lambrechts and Dhondt 1988; Vehrencamp et al. 1989; Eberhardt 1994), although other studies have reached different conclusions (e.g., Horn et al. 1995). No direct estimates of the cost of singing are available for the barn swallow. However, an experiment in which parasite load was manipulated showed a negative effect of parasites on singing rate (Møller 1991c). In addition, humoral immune response negatively correlates with singing rate in this species (Saino et al. 1997c). Thus, findings from previous studies suggest that singing rate is a condition-dependent trait, and that males in relatively poor condition may be unable to sustain the costs of high singing activity. In addition, singing in the barn swallow has an obvious time cost because it competes with other activities such as foraging, since males sing from their nesting territory where no foraging activity normally occurs. Several experimental studies have shown that a long tail has larger maintenance costs than a short tail (e.g., Møller 1994; Møller and de Lope 1994; Saino et al. 1997a, 1997b). These costs are large in terms of fitness since tail-elongation reduces annual survival rate considerably, whereas tail-shortening has the opposite effect (Møller and de Lope 1994). These costs seem to arise from effects of manipulation on foraging ability (Møller et al. 1995b), and they must ultimately arise from the effects of experimentally manipulated phenotype on locomotion. Hence, both long tails and high singing rates seem to be costly to male barn swallows, although these costs have not been quantified in energetic terms nor on commensurable scales.
Diverse interpretations of the lack of effect of tail manipulation on singing rate are possible. First, tail manipulation has no immediate effect on male condition, and maintenance costs of an elongated tail become apparent only later in the breeding season or during migration and wintering (Rubolini and Schiavi 2002). This possibility is unlikely because we have shown that tail manipulation has physiological effects on hematocrit and immune response as well as foraging efficiency within a few weeks after manipulation (Møller 1994; Møller and de Lope 1994; Møller et al. 1995b; Saino et al. 1997a, 1997b), in addition to long-term effects on annual survival (Møller 1989, 1994; Saino et al. 1997a). Second, physiological trade-offs may exist between signal expression and other fitness-related traits, so that an experimental increase of tail length is detrimental to traits affecting viability but not to singing. Physiological pathways mediating trade-offs among signals may not have evolved because intense selection for high mating success prevents depression of individual signals such as song below a certain threshold. This should particularly be the case in short-lived species such as the barn swallow where risk of mortality due to stochastic factors may be high and most individuals only have a single breeding season. Hence, lack of a negative effect of tail elongation on singing rate does not necessarily imply that either or both signals are produced or maintained at no cost, but rather that the cost of exaggerated signals is paid in another currency. Third, the mechanisms that control the expression of one secondary sexual character (tail length in our study species) may also directly control the expression of other traits, such as singing activity. This interpretation also seems unlikely because this mechanism would lead to consistent (either positive or negative) correlations between the two traits, while we have shown repeatedly that no significant association exists between tail length and singing rate (Møller 1994; Saino and Møller 1995; Møller et al. 1998). Finally, tail elongation could have resulted in reduced singing activity, but this effect was compensated by larger rate of aggressive interactions in which tail-elongated males could have been involved (see also below).
Tail elongation significantly enhanced the expression of the terminal part of the song, the rattle. Two mechanisms, possibly mediated by androgens, could have caused this effect on song quality. In a previous study, we have shown that males sing longer rattles when they are in highly competitive social contexts (Galeotti et al. 1997). In addition, we found a positive correlation between male testosterone plasma levels and rattle duration (Galeotti et al. 1997). Hence, androgens, which mediate the expression of male aggressive behavior in birds (Wingfield et al. 1987), may also enhance rattle duration. Males with long tails may be more frequently involved in aggressive interactions with neighboring individuals, and this may have a positive feed-back effect on their androgen levels and thus enhance the aggressive component of their song. In the present study, despite 2-week sampling effort involving 1–1.5 h of daily observations in the morning, when social and sexual interactions peak (Møller 1994), we recorded too few male-male fights to allow statistical treatment. However, no effect of tail manipulation on the rate of male-male fights was found in a previous study (Møller 1993). Hence, it seems unlikely that longer rattles are the result of increased fighting rates. Nonetheless, most territorial interactions may be mediated by song and not result in direct fights between males. Tail-elongated males could have been more frequently involved in territorial displays mediated by song because they are perceived by neighboring individuals as relatively strong competitors for paternity (Møller et al. 1998).
Alternatively, increased sexual attractiveness of tail-elongated males may positively feed back on male androgen profiles because such males are more frequently accepted as extra-pair mates by females (e.g., Møller 1994; Saino et al. 1997d). Future tests may clarify the consequences of expression of male ornaments on androgen profile mediated by socio-sexual interactions.
Our findings that tail elongation did not significantly affect singing rate and resulted in enhanced expression of a component of song are thus novel and have general implications for the study of inter-sexual communication mediated by multiple signals. In particular, the lack of negative effect of tail elongation on song performance raises the question of what is limiting the expression of different male signals. In addition, it suggests that the secondary sexual characters we studied may not function as a system of condition-dependent reliable indicators of quality because males apparently could enhance the expression of one signal without depressing other signals.
Studies of the consequences of experimentally altered signals have shown that individuals displaying at artificially high levels are faced with negative feed-back mediated through aggressive social interactions. In a study of Harris sparrows (Zonotrichia querula) (Rohwer 1977; Rohwer and Rohwer 1978), deceptive individuals were experimentally created by an increase in badge size. However, such males also received more aggressive attacks than predicted by chance, suggesting that cheating was socially controlled. In a study of house sparrows (Passer domesticus), males were prevented from displaying unreliable signals of social dominance by frequent social assessment of ability to dominate conspecifics (Møller 1987). The present results suggest that male barn swallows do not experience an immediate cost by displaying experimentally elongated tails in terms of reduced ability to sing. Therefore, direct physiological trade-offs may not exist between male tail ornamentation and determinants of singing rate.
In conclusion, this study suggests that no short-term trade-offs exist between two male secondary sexual characters in the barn swallow. However, reliability of sexual signals may be enforced by trade-offs between song and viability operating on a longer time-scale.
This study was supported by MURST COFIN2000 and COFIN2001 grant to N.S. and P.G. The experiments complied with current Italian law.