Anna Qvarnström, Animal Ecology/Department of Ecology and Evolution, Evolutionary Biology Centre, Uppsala University, Norbyvägen 18D, SE-752 36 Uppsala, Sweden. Tel.: +46 18 471 2662; fax: +46 18 471 6484; e-mail: email@example.com
Given that population divergence in sexual signals is an important prerequisite for reproductive isolation, a key prediction is that cases of signal convergence should lead to hybridization. However, empirical studies that quantitatively demonstrate links between phenotypic characters of individuals and their likelihood to hybridize are rare. Here we show that song convergence between sympatric pied (Ficedula hypoleuca) and collared flycatchers (F. albicollis) influence social and sexual interactions between the two species. In sympatry, the majority of male pied flycatchers (65%) include various parts of collared flycatcher song in their song repertoire (but not vice versa). Playback experiments on male interactions demonstrate that male collared flycatchers respond similarly to this ‘mixed’ song as to conspecific song. Long-term data on pairing patterns show that males singing a converged song attract females of the other species: female collared flycatchers only pair with male pied flycatchers if the males sing the mixed song type. From the perspective of a male pied flycatcher, singing a mixed song type is associated with 30% likelihood of hybridization. This result, combined with our estimates of the frequency of mixed singers, accurately predicts the observed occurrence of hybridization among male pied flycatchers in our study populations (20.45% of 484 pairs; predicted 19.5%). Our results support the suggestion that song functions as the most important prezygotic isolation mechanism in many birds.
Given that population divergence in song plays an important role in speciation in birds, we should expect cases of song convergence to counteract reproductive isolation. There are several observations of song convergence between closely related species in sympatry (e.g. Dobkin, 1979; Helb et al., 1985; Grant & Grant, 1998; Baker & Boylan, 1999; de Kort et al., 2002; Secondi et al., 2003), which may have several different underlying explanations. First, genetic introgression through hybridization may have occurred (i.e. when song is mainly genetically determined; de Kort et al., 2002). Second, the species could have, independently from each other, adjusted to the local sound environment (Wiley, 1991). Third, song convergence may result from heterospecific song learning (e.g. Grant & Grant, 1996,1997b,1998). In the two latter cases, song convergence should be expected to cause hybridization rather than vice versa, but it is generally difficult to quantitatively demonstrate links between phenotypic characters of individuals and their likelihood to hybridize. This is because although many bird species are able to hybridize and produce fertile offspring, they rarely do so (Grant & Grant, 1992; Price & Bouvier, 2002). However, identifying factors that drive hybridization is an efficient way to reveal the necessary prerequisites for prezygotic reproductive isolation under natural situations.
We investigated whether song convergence influences the likelihood of hybridization between two closely related species of old world flycatchers – the pied (Ficedula hypoleuca) and collared flycatcher (F. albicollis). The distributions of the two species overlap in central and eastern Europe and on the Baltic islands of Öland and Gotland. The two species compete over similar food sources and over suitable nest sites (natural tree holes or nest-boxes in deciduous forest) but differ in their robustness to harsh environment (Qvarnström et al., 2005). In sympatric populations, heterospecific pairing occurs regularly (Alatalo et al., 1990; Sætre et al., 1997; Veen et al., 2001) leading to gene flow between the species (Sætre et al., 2003), despite reduced fertility of hybrids (Alatalo et al., 1990). Thus, selection should favour male traits that decrease the probability of attracting heterospecific females. As expected from reinforcement, the differences in male breeding plumage between the species are accentuated in areas where they coexist (Alatalo et al., 1994; Sætre et al., 1997), and experimental evidence shows that this plumage divergence facilitates species recognition (Sætre et al., 1997). By contrast, the song of the pied flycatcher has converged in sympatry to partly resemble the song of the collared flycatcher (Löhrl, 1955; Alatalo et al., 1990; Haavie et al., 2004). The songs of pied and collared flycatchers are complex and may include 100 unique figures (Lundberg & Alatalo, 1992) and are easily distinguishable in the field based on differences in frequency and tempo (Gelter, 1987). In Sweden, the converged song is common in sympatry (approximately 65% of all male pied flycatchers sing a converged song) and appears to be induced by the presence of collared flycatchers (Alatalo et al., 1990), but not through genetic introgression (Haavie et al., 2004). Since the converged song is a result of different degrees of mixing elements of the two typical songs (sometimes referred to as mixed singing Alatalo et al., 1990; Haavie et al., 2004) it is possible to locate converged singers by ear. In pied and collared flycatchers (Eriksson & Wallin, 1986), as in songbirds in general (Searcy & Nowicki, 2000), song is used as a signal both to deter other males and to attract females. The converged song of pied flycatchers may therefore trigger stronger responses from male and female collared flycatchers than the pure pied flycatcher song. Here we investigated how male collared flycatchers react to converged pied flycatcher songs and whether song similarity predicts hybridization between the two species.
The data was collected from mixed populations of pied and collared flycatchers breeding in nest boxes on the Swedish islands of Öland (57°10′N, 16°58′E) and Gotland (57°10′N, 18°20′E). The study populations on Öland have been monitored during two separate time periods, i.e. 1981–1985 and 2002–2004, and include a total of 947 breeding records of known pairs out of which 195 records include a breeding male pied flycatcher. Male pied flycatchers were breeding with female collared flycatchers in 11 cases. One of these mixed pairs bred together on two consecutive years but all other female collared flycatchers mated with conspecifics at their other breeding attempts. Breeding data on Gotland were collected between 1980 and 2004 and includes a total of 7965 breeding records of known pairs out of which 289 records includes a breeding male pied flycatcher. Male pied flycatchers were breeding with female collared flycatchers in 88 cases. Also this data set includes a mixed pair that bred together on two consecutive years whereas all other female collared flycatchers mated with conspecifics at their other breeding attempts. Thus, the two known cases where a female collared flycatcher hybridized more than once represent cases where the female paired with the same male twice (which per se is rare). However, in general, hybridizing female collared flycatchers do not seem to have a preference for heterospecific males.
Pied flycatchers represent the less common species on both islands and are slowly replaced by collared flycatchers from the most preferred breeding sites. Although the majority of breeding collared flycatchers are local recruits, most breeding pied flycatchers are immigrants from allopatric populations. Less than 3% of the male pied flycatchers breeding at our study sites on Öland are local recruits. As a result, the pattern of genetic introgression is strongly asymmetric and directed at the local collared flycatcher population (Sætre et al., 2003). Thus, although the pedigrees of most pied flycatchers are unknown, birds that are classified as pied flycatchers based on plumage characteristics are highly unlikely to be backcrossed individuals.
The playback experiments were conducted between May and June 2003 on Öland, where pied flycatchers are more common than on Gotland. Song recordings were made on the same island in 2001 and 2003 with microphones mounted on 55 cm parabolic dishes (Classic, Science and PRO II, all products by Telinga, Tobo, Sweden) on DAT and minidisc recorders (Sony TCD-D7, TCD-D10 and Kenwood DMC-J7R). Playbacks were done with the same minidiscs connected to portable loudspeakers (Sony SRS-A47). Playbacks of one of the three types of song (pure pied flycatcher, converged pied flycatcher, collared flycatcher) were broadcasted from a loudspeaker placed underneath the nestbox. Recordings of 18 males (six of each of the three types of song) were used in the experiment, each consisting of eight to 10 strophes. No male was exposed to song recordings of a neighbouring male. The playback recordings were randomized by observers that were unaware of the predictions and a dummy male, matching the species of the song type, was attached to the roof of the nestbox. Eight dummies (four of each species) were used in the experiments. Within each ‘species’ treatment the use of the different dummies was balanced. The experiments were conducted when song activity was high before the nest-building period and each responding male was only used once. In order to reduce variation in male response caused by differences in ownership (i.e. motivation to defend a specific box) or long distance transmission of different song types, trials were only performed at boxes defended by a collard flycatcher male and each experimental trial began when the male was observed in the vicinity of the nest box. One of the 18 recordings was then repeatedly broadcasted during 20 min and the response was measured as whether or not the male attacked the dummy. We have chosen to use his escalated response rather than more subtle responses because males constantly ‘check out’ their boxes and sing aggressively (also when other birds are absent) during this phase of the breeding period.
Song and species identity
Data on male song and species identity of male and female flycatchers were collected during 1982–1990 and in 2003 from populations breeding on Öland, Gotland (4% pied flycatchers), and on the Swedish mainland (59°50′N, 17°40′E, 100% pied flycatchers). The proportion of pied flycatchers has declined from 29% to 18% on Öland between the two study periods while the relative frequency of the two species has remained fairly constant at the other locations. We recorded the song of 16 collared flycatchers and 75 pied flycatchers on a Sony TC-D5M or a Nagra SN using a Dan Gibson Electronic Parabolic Microphone (EPM) model P-650. Recordings were analysed using a UNISCAN II Spectrograph, model 4600 (Multigon Industries, Mt. Vernon, NY, USA). Prior to the analysis, the songs of all pied flycatchers were classified by ear (by DE and TP) as either pure (i.e. typical allopatric pied flycatcher song) or converged (i.e. the male switches between the two typical pied and collared flycatcher songs). Additionally, the song of 18 males was classified as either converged or pure song (done by TP) without being recorded. We performed a principal component analysis based on the correlation matrix of nine different song variables: minimum frequency, maximum frequency, median value of the mean frequencies of song elements included in the song, mean value of the frequency range of the figures included in the song, frequency modulation expressed as the mean number of inflection points in the modulation of the figures included in the song, length of the entire song, mean length of the figures included in the song, total number of figures included in the song and number of different figures in the song. The choice of measurements in this study was based on earlier studies that found them useful in characterizing the species (Wallin, 1985; Gelter, 1987). To compare quantitatively between the song of male collared flycatchers and male pied flycatchers classified as singing either a pure or converged song, we performed an analysis of variance based on the PC1 scores. All males were caught after the song recording, identified to species and individually marked. All breeding males and females were caught again when feeding their offspring as part of the long-term studies. Species identity of previously unmarked individuals was based on plumage characters. The majority of the females are local recruits with known pedigrees and lifetime breeding records. Male F1 hybrids are relatively easily recognized by their plumage traits (intermediate between the two parental species). Individuals belonging to later generations of back-crosses cannot be identified based on plumage traits, but such back-crossed individuals are expected to be of collared flycatcher phenotype (Sætre et al., 2003). In the analysis we used one pairing for each male and that was the first female he attracted after being recorded.
The first principal component (PC1) and second principal component (PC2) explained 34.9% and 29.18%, respectively, of the total variation in our data on song variables extracted from 16 recordings of male collared flycatchers and 75 recordings of pied flycatchers (Table 1). For a detailed comparison between the song of different sympatric and allopatric populations of pied and collared flycatchers (see Haavie et al., 2004). The three classified types of song (i.e. collared flycatcher song, converged pied flycatcher song and pure pied flycatcher song) differed significantly in the PC1 (anova, F2,88 = 82.84, P < 0.0001). The converged song was on average intermediate between pure pied and collared flycatcher (and is the result of different degrees of mixing elements of the two typical songs, Fig. 1) but some male pied flycatchers sang a song that is virtually indistinguishable from the song of collared flycatchers (Fig. 2). Thus, some male pied flycatchers appear to have learned their entire song repertoire from heterospecific individuals and do use elements of the typical pied flycatcher song.
Table 1. Correlation matrix of the song variables and the two first components (PC1 and PC2) obtained in the principal component analysis.
Mean frequency of elements
Length of entire song
Length of figures
No. of figures
No. of different figures
We used playback experiments to investigate whether the convergence in song influences heterospecific communication. Male collared flycatchers responded significantly differently depending on the type of song they were exposed to (Fig. 3). The song of collared flycatchers triggered an aggressive response more often than the pure song of pied flycatchers (Fisher's exact test P = 0.018, n = 19), and the converged song triggered an aggressive response more often compared to the pure song of pied flycatchers (Fisher's exact test P = 0.045, n = 19). In fact, no male collared flycatcher responded aggressively to the pure song of pied flycatchers. There was no significant difference in the number of males that responded to the converged song of pied flycatchers relative to the song of collared flycatchers (Fisher's exact test P = 1.0, n = 22). Thus, male collared flycatchers responded to converged song as if it was a conspecific signal.
If females perceive song in a similar way to males (e.g. Riebel et al., 2002), they may be attracted to converged song. Whether or not a male pied flycatcher sings a converged song may therefore influence his risk of pairing with a heterospecific female. Indeed, we found that male pied flycatchers singing a converged song paired more often with a female of the other species than did males singing a pure song (Fisher's exact test, P = 0.014, n = 66; Table 2). In fact, all female collared flycatchers that paired with a male pied flycatcher had chosen a male singing a converged song (Fig. 4). Furthermore, the risk of hybridization increased as the song increased in similarity to collared flycatcher song (Table 2; Fig. 2). From a male pied flycatcher's perspective, a converged song type was associated with 30% likelihood to hybridize. None of the female collared flycatchers that paired with male pied flycatchers with known song type had paired with heterospecific males before. In the analyses on the effect of song on the observed pairing patterns only sympatric pied flycatcher were included and the results remain the same when the few males recorded during the later study period (2003) are excluded.
Table 2. Logistic regression analyses of the effect of song type of pied flycatchers in sympatry with collared flycatchers on the probability of heterospecific pairing.
−2 Log LR
LR, likelihood ratio.
*Song scored in the field as being of converged or pure type.
†Öland or Gotland.
‡First principal component extracted from nine song variables.
Based on the facts that; (i) 65% of all pied flycatchers breeding at our study sites are estimated to be mixed singers, (ii) the total number of breeding male pied flycatchers on Öland and Gotland during the study periods is 484 and (iii) that 30% of the mixed singers are expected to hybridize (see above), the predicted total number of mixed pairs including a male pied flycatcher and female collared flycatcher is 94.4. The observed number of such pairs was 99.
Our results show that male pied flycatchers that sing songs similar to collared flycatchers experience a higher likelihood to be involved in heterospecific pairing, and thus hybridize. Playback experiments suggest that the underlying reason for heterospecific pairing is that male pied flycatchers singing a converged song type are sometimes perceived as conspecifics by individuals of the other species.
The fact that only male pied flycatchers singing a converged song paired with heterospecific females is consistent with the suggestion that song functions as the most important prezygotic isolation mechanism in many birds (Martens, 1996). However, we do not interpret our results as song being the sole species recognition cue used by female flycatchers. When selecting their mates, female flycatchers approach several singing males, inspect each male and the nest-site he defends, and then make a choice among the sampled males probably based on a number of different criteria (Lundberg & Alatalo, 1992 and references therein). There are several reasons to assume that female collared flycatchers initially perceive male pied flycatchers as potential mates when they sing a converged song. First, it has been experimentally demonstrated that female collared flycatchers are attracted to collared flycatcher song (Eriksson & Wallin, 1986) and we found that a male pied flycatcher's risk of heterospecific pairing increased as the song increased in similarity to collared flycatcher song. Second, male collared flycatchers responded to converged song as if it was a conspecific signal, and females may perceive song in a similar way (e.g. Riebel et al., 2002). However, in order to fully explain why a specific female collared flycatcher decided to settle with a male pied flycatcher we would need to take a number of additional factors into account. There may be constraints involved, such as availability of conspecific mates, limited experience, or errors in imprinting on other species-specific traits (e.g. Grant & Grant, 1997b). Since most hybridizing female collared flycatchers paired with conspecific males at their other known breeding attempts we consider errors in imprinting to be an unlikely explanation. It has also been suggested that female choice of heterospecific males could be adaptive from the female's point of view through direct benefits of mate choice combined with multiple mating, conspecific gamete precedence and possibly sex-ratio adjustments of young towards the most fertile sex (Veen et al., 2001; Marshall et al., 2002). In any case, singing a converged song appears to be a critical condition for males to attract heterospecific females in the first place. Thus, male pied flycatchers singing a species-specific pied flycatcher song probably avoid hybridization simply by not attracting prospective females of the other species. That song convergence plays an important role in explaining mixed pairing between male pied flycatchers and female collared flycatchers is further supported by the fact that the observed number of such pairs in our database is close to the predicted number based on the frequency of converged song and the estimated hybridization risk associated with this behaviour.
We are not aware of any previous studies that have quantitatively demonstrated a link between song convergence and natural occurrence of hybridization, but there is circumstantial evidence pointing in the same direction as our results. In Darwin's finches, song appears to be a relatively more reliable species-specific signal because learning mainly occurs from father to son (Grant & Grant, 1996). Experiments have demonstrated species-specific responses to song playbacks (Ratcliffe & Grant, 1985). Males singing a heterospecific song were rare, but among these few ‘miss-imprinted’ males, hybridization was common: three of nine breeding Geospiza fortis and G. scandens males that sang the other species song hybridized (Grant & Grant, 1998). By contrast, extensive copying of heterospecific song has been reported from a study on a mixed population of indigo (Passerina cyanea), lazuli (P. amoena) and hybrid buntings where males of any bunting morph can sing phases of both species (Baker & Boylan, 1999). When taking plumage morph into account, song type of males did not explain any more of the variance in natural mating associations (Baker & Boylan, 1999). However, since females were shown to use both plumage and song in mate choice in a laboratory setting (Baker & Baker, 1990) the authors argue that it is premature to exclude a role for cross-species song learning in hybridization between the two species of buntings.
In this study we have investigated the consequences of song convergence, whereas the proximate reason for this behaviour remains to be investigated. As mentioned in the introduction, there are three main explanations for why song may converge between two species in sympatry (i.e. due to parallel adjustment to the local environment, genetic exchange, or cultural exchange). The converged song of male pied flycatchers is induced by the presence of collared flycatchers and does not occur in a specific type of habitat (Alatalo et al., 1990). The crucial question is then whether the convergence is mediated through genetic introgression or through learning. In contrast to collared flycatchers, pied flycatchers display an extensive dispersal in this part of the breeding range (Lundberg & Alatalo, 1992). Less than 3% of the male pied flycatchers breeding at our study sites on Öland are local recruits and even if all of them belonged to a backcrossed generation that would not explain the fact that the majority (approximately 65%) of all male pied flycatchers sing a converged song. Based on species-specific genetic markers there is no introgression of genes from collared flycatchers found among locally sampled pied flycatchers (Sætre et al., 2003) including five males known to sing a converged song (Haavie et al., 2004). In addition, anecdotal data on pied flycatchers suggest that learning is involved in the development of song structure and content. A hand reared male pied flycatcher sang a song completely unrelated to any flycatcher song we had heard before and this bird also mimicked the calls of great tits held in nearby aviaries (D. Eriksson and T. Pärt, unpublished data). Löhrl (1955) observed a male pied flycatcher switching song type from pure pied song on arrival from the wintering grounds to a converged song later on in the breeding season. Thus, present evidence suggests that learning song from hetersopecific males is the most likely explanation for song convergence.
It appears contradictory that the majority of all the male pied flycatchers sing a converged song at our study site despite the fact that this behaviour is associated with a 30% likelihood of heterospecific pairing. Male pied flycatchers mated with collared flycatcher females lose on average 59% paternity in their broods (Veen et al., 2001) and the remaining offspring (i.e. hybrids) have very low fitness (Alatalo et al., 1990). Thus, heterospecific pairing is associated with a considerable cost, especially from the male's perspective. Since male pied flycatchers singing a converged song are more easily recognized as competitors by the socially and numerically dominant male collared flycatchers, this behaviour may confer an advantage when establishing and maintaining a territory within prime habitats. Thus, convergence in song between species may facilitate interspecific territoriality (e.g. Irwin & Price, 1999) and therefore be associated with a benefit, given that the males manage to attract conspecific females. Another possibility is that the ability to learn the song of neighbouring males confers a selective advantage in allopatric populations (i.e. where all neighbouring flycatchers belong to the same species), and that immigration of pied flycatchers into the hybrid zones slows down an evolutionary response to hybridization in sympatry. We consider the latter explanation most likely since pied flycatchers largely occur in allopatry and have only recently (approximately 150 years ago) come into contact with the collared flycatcher in Sweden (Lundberg & Alatalo, 1992). By contrast, collared flycatchers breed largely in sympatry with pied flycatchers throughout their breeding range and apparently have a much less flexible song learning template as they do not copy the song of pied flycatchers (Gelter, 1987; Haavie et al., 2004). Selection at the hybrid zone may, in time, reduce the occurrence of the converged song, since the converged song occurs more rarely in much older hybrid zones of pied and collared flycatchers in central Europe (approximately 10% compared to 65% on Öland, Haavie et al., 2004). Furthermore, the dispersal of pied flycatchers is more restricted in central Europe (Lundberg & Alatalo, 1992). Heterospecific pairs of female collared flycatchers and male pied flycatchers are also less frequently observed at this older zone (Sætre et al., 2003). Natural selection may alter the probability of song convergence either through reducing the flexibility of the song template (i.e. pure song may become genetically assimilated such that individuals develop species-specific song without external models) or through changes in the song-learning strategy (e.g. through a genetically determined perceptual bias in favour of conspecific song, Nelson & Marler, 1993; Mundinger, 1995; Whaling et al., 1997; Nelson, 2000). Genetic changes in the song-learning program that reduce the likelihood of heterospecific song learning in response to selection against hybridization can be viewed as ‘cryptic’ reinforcement of mating barriers (Haavie et al., 2004). Beecher & Brenowitz (2005) recently reviewed song-learning strategies in songbirds and referred to social factors as the ‘wildcard’ in song learning. Possibly, the high diversity of song-learning strategies, at least partly, results from variation in selection patterns driven by social- and sexual interactions between heterospecific individuals.
Given that both plumage and song have important functions in species isolation – why does plumage diverge in sympatry while song converges? The answer to this question may lie in differences in either the function or the proximate determination of the two types of signals. We consider it unlikely that plumage should be relatively more important for avoidance of hybridization (and therefore diverge at faster rate in sympatry). This is because female flycatchers (like most female birds) are selected to choose their mates as quickly as possible, making it important for them not to waste time and energy on sampling heterospecific males. If anything, song should be relatively more important in this context and the reason for sympatric convergence in song probably lay in the proximate determination of the trait. Signals that are partly learned have the potential to rapidly change, and birdsong has been shown to quickly diverge between allopatric populations (Irwin et al., 2001). In zones of secondary contact, individuals from different populations together constitute the social environment and may therefore learn from each other. Hence, learned traits have the potential to also rapidly converge between populations and our study shows that convergence may lead to hybridization. This pattern may, in turn lead to ‘cryptic’ reinforcement of mating barriers through genetic changes in the proximate determination of song. A challenge for future studies is to make comparisons of genetic divergence in song learning strategies between allopatric and sympatric bird populations.
We thank E.H. Morrow T. Price, K. Räsänen, G.-P. Sætre, N. Svedin and C. Wiley for comments that greatly improved the manuscript and M. Halvarsson, M. Olsson and J. Älvgren for excellent help in the field. We also thank Lars Gustafsson for permission to use long-term data from Gotland. Financial support was obtained from the Swedish Research Council, the Research Council of Norway, the Royal Swedish Academy of Science and Knut and Alice Wallenberg's Foundation.