Song discrimination suggests premating isolation among sympatric indigobird species and host races

Abstract

Indigobirds (Vidua spp.) are host-specific brood parasites that have diversified in a recent radiation apparently driven by host colonization. Behavioral imprinting of both male and female indigobirds on host song is thought to promote rapid speciation because it results in assortative mating between indigobirds associated with a particular host. We conducted a song playback experiment to test whether male indigobirds discriminate among potential competitors based on song. Of particular interest was the behavior of two sympatric host races of the Cameroon indigobird Vidua camerunensis that differ only in host song mimicry and other components of their vocal repertoires. Territorial males of the two V. camerunensis host races and Vidua raricola, a morphologically distinct indigobird species, were tested with playbacks of each other's songs. Males of all three groups responded most aggressively to songs of their own species and/or host race, as evidenced by strong and highly significant differences in a variety of response variables. This differential territorial defense suggests that an intruding male with different songs does not represent a competitive threat and is expected if females mate assortatively with respect to song. Thus, our results provide evidence of premating reproductive isolation among recently evolved indigobird species and host races.

Recent studies of speciation reflect a shift in emphasis from the geographic context of speciation to the processes promoting population differentiation (Kirkpatrick and Ravigné, 2002; Schluter, 1996, 1998). One such process is the divergence of mate recognition signals. While the evolution of postmating reproductive isolation between bird species can be a relatively protracted process (Prager and Wilson, 1975; Price and Bouvier, 2002), premating isolation may result from the relatively rapid divergence of sexually selected traits such as song (Price, 1998). Among birds, song divergence may therefore be of particular importance in the early stages of the speciation process (Edwards et al., 2005; Slabbekoorn and Smith, 2002a).

Ten species of African indigobirds (Vidua spp.) are host-specific brood parasites of various estrildid finches (Payne, 2004). Both behavioral and genetic data suggest that sympatric speciation via host shift has played an important role in the diversification of indigobird species (Klein and Payne, 1998; Payne et al., 1998, 2000; Sorenson et al., 2003). Host colonization is facilitated by behavioral imprinting of indigobirds on the particular host species that raises them. Male indigobirds imprint on the songs of their foster parents and include mimicry of host song in a repertoire that also includes vocalizations unique to indigobirds (Payne et al. 1998). Though they do not sing, female indigobirds also imprint on their host species and mate preferentially with indigobird males that mimic the songs of the host that reared them (Payne et al., 2000). After the colonization of a new host, social imprinting results in simultaneous divergence in male courtship songs and female preferences for male songs, as well as female preferences for host nests (Payne et al., 2000), and therefore leads to reproductive isolation.

The first step in speciation by host shift is the formation of conspecific host races (Bush, 1969). Although most indigobird species are associated with a single estrildid finch host, the Cameroon indigobird Vidua camerunensis is associated with several host species across West Africa (Payne et al., 2005). Indigobirds associated with these hosts have been treated as a single species because they are essentially indistinguishable in both morphology (Payne et al., 2005) and neutral genetic markers (Sefc et al., 2005; Balakrishnan CN and Sorenson MD, in preparation). Near Tibati, Cameroon, and elsewhere in West Africa V. camerunensis males mimic the songs of either African firefinch Lagonosticta rubricata or black-bellied firefinch Lagonosticta rara. Given social imprinting on hosts, these birds may represent two reproductively isolated host races that have not yet evolved morphological differences after recent colonization of a new host (Payne et al., 2005).

As a test of reproductive isolation between indigobird species and host races, we conducted a song playback experiment to determine whether male indigobirds discriminate among potential competitors based on song. This experiment follows the logic of previous avian studies in which song playbacks to males have been used to test for premating reproductive isolation (e.g., Grant BR and Grant PR, 2002a,b; Irwin et al., 2001; Patten et al., 2004; Salomon, 1989). If female indigobirds mate exclusively with males mimicking the appropriate host song, then a male indigobird mimicking the song of a different host does not represent a competitive threat to a territorial male and should not elicit an aggressive response. Alternatively, if females mate with males irrespective of their host association (as advertised by mimicry song), territorial males should be aggressive toward all male indigobirds. Therefore, if males behave adaptively in response to potential competitors, song discrimination by males provides indirect evidence of premating reproductive isolation based on song. We tested male territorial responses in two host races of V. camerunensis and a second species, the Jambandu indigobird Vidua raricola.

MATERIALS AND METHODS

Locality and natural history

Fieldwork was conducted in an approximately 10-km² area north and west of Tibati, Cameroon (6°28′ N, 13°34′ E) from October through December of 2001 through 2003. At Tibati, three indigobird species are associated with four hosts. Individual V. camerunensis males representing two putative host races mimic the songs of either L. rubricata or L. rara (Payne et al. 2005). In contrast, V. raricola, which parasitizes zebra waxbill Amandava subflava, is morphologically distinct and significantly differentiated from V. camerunensis in neutral genetic markers (Sefc et al., 2005; Balakrishnan CN and Sorenson MD, in preparation). A third species that is less numerous at Tibati, Wilson's indigobird Vidua wilsoni, was not included in this study. V. camerunensis and V. raricola are distributed syntopically in small agricultural fields and surrounding bush, with individuals of different species or host races sometimes defending adjacent territories.

Male indigobirds are highly territorial during the breeding season, and song playbacks elicit aggressive behavior (e.g., Payne, 1973). Territorial males sing conspicuously from a habitual “call site” at the top of a tree. Female indigobirds visit male call sites in order to choose a mate and typically copulate at the call site. The indigobird social system has been described as a “dispersed lek,” and male reproductive success appears to be highly skewed (Payne RB and Payne K, 1977). Thus, song plays a fundamental role in the indigobird mating system. Based on male singing and territorial behavior, indigobird species show substantial overlap in the timing of their breeding seasons. To simplify description of the results, each of the two V. camerunensis host races and V. raricola will be referred to as a “song population,” and playback songs will be described as either homotypic or heterotypic with respect to host song mimicry.

Playback stimuli

Indigobirds were recorded with a Sony TCDM5 cassette recorder and a Sennheiser ME66 shotgun microphone. During 2001, we recorded more than 100 males on their call sites, generally collecting about 15 min of song data per individual. Recordings were transferred to minidisc for long-term storage after fieldwork each day using a Sony MZ-R700 minidisc recorder. Audiospectrographs were created and analyzed using AviSoft SASLab Pro version 4.3 (AviSoft Bioacoustics, Berlin, Germany). Playback tapes were made from a subset of the recordings made in 2001. Indigobird vocalizations are remarkably complex, incorporating mimicry of host songs, “chatter” calls, and numerous nonmimicry songs, each comprising a unique sequence of notes (Figure 1). Chatter calls are similar across species, whereas both mimicry and nonmimicry songs vary between species (Payne, 1973, 1982) and host races (Balakrishnan CN and Sorenson MD, unpublished data). Young males apparently learn nonmimicry songs from adult male indigobirds that mimic the same host species. Clear recordings (high signal to noise ratio) containing chatter and both mimetic and nonmimetic song components were selected for use in the playback experiment. Each tape consisted of 5 min of continuous, unaltered courtship song from a different male. Nine unique playback tapes were prepared for each of the three song populations.

Figure 1

Representative audiospectrograms of the three general classes of songs (chatter, complex nonmimicry, and mimicry) for song populations of indigobirds. (A) Vidua camerunensis: Lagonosticta rara mimics, (B) V. camerunensis: Lagonosticta rubricata mimics, and (C) Vidua raricola. Note that song repertoires of individual males include many distinct nonmimicry songs (each comprised of a complex series of notes) and several types of mimetic vocalizations. See Payne RB, and Payne LL (1994) and Payne et al. (2005) for side-by-side comparisons of mimetic indigobird songs and host songs.

Playback experiments

Playback experiments were conducted during the breeding season from October to December of 2002 (n = 72 trials) and 2003 (n = 9 trials). We tested male indigobirds because their territorial responses could be readily quantified in the field and because their vocalizations provide unambiguous evidence of host association. In contrast, females of the study species are morphologically indistinguishable and do not vocalize, making it impossible to determine host association or species identity.

Playbacks were broadcast at natural volume, approximating that of the focal bird, using a Sony TCDM5 tape deck and a Sony SRS-A27 speaker placed on or near the ground at a distance of 15–20 m from the call site of the territorial male. In order to provide a focal point for aggressive responses, a taxidermic mount of a male indigobird was attached to a branch near the speaker at a height of approximately 1 m. To control for the possibility of differential responses based on morphology of the intruding male, a specimen of Vidua purpurascens was used in all experiments; V. purpurascens does not occur in West Africa and is morphologically distinct from both of the focal species in this study. The speaker and taxidermic mount were placed in a location that allowed focal birds to perch in close proximity.

Nine males of each song population were challenged with recordings of other males from each of the three song populations (total: 27 males and 81 trials). To avoid pseudoreplication (Kroodsma, 1989), each individual within a song population was presented with a unique set of three recordings, the order of which was shuffled systematically to control for possible habituation across trials (Table 1). All playbacks were conducted between 0700 and 1100 h, and successive trials involving a given individual were separated by a minimum of 2 days. All males tested were present at their call sites for at least 2 days prior to the first playback trial. Studies of banded indigobirds show that individual males generally occupy a single call site throughout the breeding season and sometimes return to the same call site the next year (e.g., Payne, 1973).

Table 1

Design of playback experiment for each study population

	Focal bird
Trial	No. 1	No. 2	No. 3	No. 4	No. 5	No. 6	No. 7	No. 8	No. 9
1	A1	B2	C3	A4	B5	C6	A7	B8	C9
2	B1	C2	A3	B4	C5	A6	B7	C8	A9
3	C1	A2	B3	C4	A5	B6	C7	A8	B9

Each of nine focal birds per song population was tested with a unique set of three recordings, with the order of song types systematically shuffled among males. A1–A9 (nine unique recordings of Vidua camerunensis, L. rara mimics), B1–B9 (nine unique recordings of V. camerunensis, Lagonosticta rubricata mimics) and C1–C9 (nine unique recordings of Vidua raricola). The same set of 27 recordings was used for playbacks to each of the three song populations.

Each trial included a 5 min period prior to playback during which the focal male was recorded, 5 min of song playback, and a 5 min post-playback recording period. During playback, behavioral responses of the focal male were noted by dictation onto a minidisc recorder. Response measures included: time of first response (latency), duration of response (duration), number of flights toward the speaker (flights), number of hops toward the speaker (hops), amount of time spent within 1 m of the speaker (close), amount of time spent out of sight (absent), and the number of mimicry songs performed by the focal male (mimicry songs). During the first 10 trials, we also included a 5 min period prior to preplayback recording during which the taxidermic mount was presented without any accompanying song. No birds showed any response to the mount alone, so this procedure was omitted in subsequent trials.

Statistical analysis

Statistical analyses were conducted in SYSTAT, version 5.2 (Wilkinson, 1992). All response measures were log (x + 1) transformed because of high variance and skew; many of the response measures had values of zero for multiple trials. Because response measures were correlated, we used principal components analysis (PCA) to reduce the data to a set of uncorrelated response variables. Principal components were used as response measures in a nested analysis of variance (ANOVA) design in which individual focal birds were nested within song populations. Trial number (1, 2, or 3 for each focal bird) and experiment date were included as covariates in the ANOVA to test for possible effects of habituation to playbacks and seasonal changes in responsiveness. Pairwise, post hoc comparisons were made using Fisher's least significant difference test with correction of significance values for multiple comparisons (Benjamini and Hochberg, 1995).

RESULTS

Responses to homotypic versus heterotypic songs

Responses to playback of homotypic song were clearly stronger than responses to heterotypic song. Males responding to homotypic song typically made repeated flights back and forth over the mount or a slow, gradual approach toward the mount. In contrast, heterotypic song elicited either no response or a rapid, but short-lived approach that was followed by the resumption of normal courtship singing at a distance from the playback speaker and taxidermic mount. Consistent differences were observed in all seven response measures, with birds of each song population responding more strongly to homotypic than heterotypic song (Table 2). The first two axes in the PCA explained 72% of the variation in the data. The first principal component (PC1) explained 54% of the variation and can be interpreted as an index of aggression, with positive loadings for duration (0.87), close (0.75), flights (0.89), and hops (0.83) and negative loadings for latency (−0.71), absent (−0.51), and mimicry song (−0.46).

Table 2

Untransformed means of response measures for each of the three focal bird song types when presented with each of three different stimulus songs

	Playback type
Focal species Response variable	Lagonosticta rara mimic	Lagonosticta rubricata mimic	Vidua raricola
Vidua camerunensis: L. rara mimics (n = 9)
Latency	15.55 (13.92)	121.11 (134.36)	145.00 (136.18)
Duration	272.67 (51.18)	50.89 (65.79)	69.78 (100.91)
Close	76.00 (88.34)	0.22 (0.67)	0.33 (1.00)
Flights	9.67 (6.34)	2.00 (2.12)	0.88 (1.17)
Hops	4.11 (3.82)	0.88 (1.26)	1.22 (1.64)
Absent	10.89 (21.70)	80.78 (96.21)	98.44 (72.72)
Mimicry song	0.11 (0.33)	3.33 (4.24)	1.89 (3.89)
V. camerunensis: L. rubricata mimics (n = 9)
Latency	158.11 (131.11)	65.66 (97.28)	115.78 (114.67)
Duration	88.78 (94.51)	189.11 (104.92)	104.94 (85.00)
Close	3.11 (9.33)	10.22 (24.53)	1.67 (5.00)
Flights	1.78 (1.86)	5.44 (5.17)	1.44 (2.01)
Hops	1.67 (2.55)	5.44 (5.17)	1.67 (1.66)
Absent	45.89 (58.17)	35.00 (69.73)	28.33 (76.91)
Mimicry song	2.00 (2.55)	0.33 (0.50)	2.11 (4.17)
V. raricola (n = 9)
Latency	72.44 (99.81)	90.89 (122.45)	36.56 (69.07)
Duration	83.89 (99.09)	105.67 (100.56)	240.22 (91.87)
Close	3.56 (6.00)	5.33 (10.59)	42.89 (32.45)
Flights	2.22 (3.23)	3.88 (4.99)	9.11 (3.66)
Hops	1.33 (1.50)	1.33 (1.58)	4.44 (3.88)
Absent	93.67 (108.82)	117.89 (84.81)	7.88 (17.69)
Mimicry song	1.67 (2.60)	0.67 (1.11)	0.67 (1.66)

Responses to homotypic playbacks are shown in bold. Standard deviations are given in parentheses. Latency, duration, close, and absent are measured in seconds; the remaining response measures are measured as simple counts.

A nested ANOVA using PC1 as the response measure indicated a highly significant interaction between focal bird song population and playback song (F_4,45 = 17.29, p < 10⁻⁵). Specifically, male territorial responses were strongest when the simulated intruder was from the same song population as the focal male. Pairwise tests show that each of the three indigobird song populations responded more strongly when presented with homotypic as compared to either of the two heterotypic songs (Figure 2). Thus, the two V. camerunensis host races responded no more aggressively to each other than to heterospecific recordings of V. raricola. Individual birds differed significantly in aggressiveness (F_24,45 = 1.97, p < .05) but there were no significant effects of habituation (trial number) or experiment date on aggressive responses.

Figure 2

Average PC1 scores (±SE) for each of the three focal male song types when presented with each of three different types stimulus songs. (A) Playback of Lagonosticta rara mimics, (B) playback of Lagonosticta rubricata mimics, and (C) playback of Vidua raricola (mimics Amandava subflava). Each mean is based on nine trials. All pairwise comparisons between homotypic and heterotypic playbacks were significant by Fisher's least significant difference test (*p < .05, **p < .01, ***p < .001).

PC2 had a strongly positive loading for mimicry song (0.72) and a strongly negative loading for absent (−0.74), simply reflecting a negative correlation between the number of mimicry songs delivered by focal males and the time they were away from the call site during the playback trial. However, neither this response measure was significantly related to any of the independent variables or covariates in the nested ANOVA analysis nor did the interaction of focal male song population and playback song significantly affect magnitude of PC2. Likewise, no significant results were obtained for PC3, which explained only 9.7% of the variation in the data (eigenvalue = 0.68).

Focal male vocalizations during and after playback trials

Further examination of the singing behavior of focal males suggests different responses to playback of homotypic versus heterotypic song. When challenged with heterotypic songs, indigobirds generally continued singing their normal, mixed repertoire. In contrast, when challenged with homotypic songs, indigobirds often became silent or primarily gave chatter calls. A reduction in mimicry song in aggressive situations (i.e., homotypic playbacks) was suggested by the negative loading of mimicry song in PC1 (−0.46). An ANOVA using the nested design described above and mimicry song as a univariate response measure indicated a nonsignificant interaction between playback song and focal bird song population (F_4,52 = 2.12, p = .09), but with a trend toward less mimicry song in response to homotypic playbacks. A stronger negative relationship was evident in a partial regression of aggressive response (as measured by PC1 for the six nonvocal response variables) and the number of mimicry songs (F_1,50 = 12.06, p = .001). Thus, those indigobirds responding most aggressively sang less mimicry song during playback trials. Counts of mimicry and nonmimicry vocalizations given by focal males revealed no apparent differences in vocal repertoire between the 5 min periods before and after playback trials, indicating that the playback trials had no lasting effects on the singing behavior of focal males.

DISCUSSION

Male indigobirds showed dramatically different responses to experimental playback of homotypic versus heterotypic song, indicating an ability to recognize and discriminate against potential competitors even in the absence of other phenotypic cues. If these differential territorial responses are adaptive, our results suggest that male indigobirds singing heterotypic song are not significant sexual competitors and, in turn, that females are discriminating in their choice of mates, preferring males mimicking the songs of their particular host. Indiscriminate mating by females would select for generalized territorial defense against any intruding male indigobird. Previous studies have used playbacks to males and similar logic to test for evidence of reproductive isolation (e.g., Grant BR and Grant PR, 2002a,b; Irwin et al., 2001; Patten et al., 2004; Salomon, 1989).

Rapid divergence of sexual signals after the colonization of a new niche has been observed in other taxa (Danley and Kocher, 2001; Patten et al., 2004; Slabbekoorn and Smith, 2002b) and may be of general importance in speciation (Beltman et al., 2004). This process is particularly well understood in indigobirds in which the acquisition of new songs and mating preferences is an essentially automatic consequence of host colonization (Payne et al., 1998, 2000, 2002; see also Beltman and Haccou 2005; Beltman et al., 2004). Our results suggest that two morphologically and genetically indistinguishable populations of V. camerunensis represent sympatric host races isolated by song, perhaps exemplifying the early stages of speciation by host shift. Even if females mate exclusively with males of the same song type, however, imperfect specificity in female host choice (i.e., where they lay eggs) could lead to continuing gene flow between host races (see Payne and Sorenson, 2004); but this would not select for generalized territoriality among males.

While consistent with the overall model of host imprinting determining both male sexual signals (song) and female mate preferences, our results allow only an indirect inference about female behavior. If mate choice is costly to females (e.g., Byers et al., 2005) and there is no particular advantage to mating with a male of the same song type, selection on females might favor indiscriminate mating. In general, however, mate choice based on song is probably strongly favored in indigobirds because it produces offspring that are well adapted to their hosts. Young indigobirds mimic the host-specific mouth markings of host young (Payne, 2004, 2005) and male song mimicry, by advertising success in the nest of a particular host species, serves as a signal of mouth marking genes appropriate to that host (Payne et al., 1998). Just after the colonization of a new host, however, this relationship would not hold, such that selection on female choosiness might be relaxed or reversed. The expression of normal parenting strategies by the new host would likely result in immediate selection on indigobird mouth markings even in the absence of specific antiparasite adaptations (see Schuetz 2005 a,b), and as mouth markings evolved to match those of a new host, selection would again favor mate choice based on song. Throughout this process, however, selection to avoid mating with other sympatric indigobirds with divergent mouth markings (V. raricola and V. wilsoni at our study area) should continue. We do not know if the two V. camerunensis host races in Cameroon differ in mouth markings, but it is unlikely that relaxed selection on female choice during this transient stage would undo behaviors that have been advantageous over the longer term of indigobird evolution. Finally, the dispersed lek mating system of indigobirds, combined with choice based on signals that can be detected at a distance, likely minimizes the costs of female choice in this system.

Clearly, playback experiments with females would allow a more direct test of reproductive isolation, but are logistically difficult because female indigobirds do not respond well to playbacks in the field (e.g., Payne, 1973). In white-crowned sparrows, females appear to be less discriminating than males in their responses to local and foreign dialects (Nelson and Soha, 2004), whereas the consequences of song recognition learning are similar in male and female zebra finches (Riebel et al., 2002). Other studies comparing male and female responses are difficult to evaluate because experimental procedures for males and females often are necessarily different (Ratcliffe and Otter, 1996; Reibel et al., 2002). In addition, males and females may respond to different components of the song repertoire.

Consistent with this possibility, the changes we observed in song repertoire by focal birds during playback of homotypic song suggest different functions for different components of indigobird song. Payne (1979) found correlations between song type (chatter, complex nonmimicry, and mimicry) and behavioral context. In particular, chatter calls were delivered more often in aggressive contexts, whereas mimicry song was used more often in sexual contexts. The results presented here provide corroborating experimental evidence that mimicry songs function primarily in courtship and are rarely delivered in aggressive contexts. Chatter vocalizations, which are similar in tonal quality and note structure across song populations (Figure 1), may elicit initially aggressive responses even in heterotypic playbacks and may explain observations of occasional interspecific aggression in indigobirds (e.g., Payne and Gropschupf, 1984). After an initial response, males may modify their behavior once an intruding male has been identified by its mimicry and/or complex nonmimicry songs.

We used recordings of natural song sequences including both mimetic and nonmimetic songs to simulate territorial intrusions by males of different song populations and therefore cannot assess the relative contributions of these song components in eliciting male territorial responses. In a previous study, male Vidua chalybeata and V. purpurascens in southern Africa responded more strongly to conspecific than heterospecific nonmimicry songs, but were generally unresponsive to mimicry songs when each was played back separately (Payne, 1973). In contrast, playbacks to a sample of four captive females indicated a preference for conspecific mimicry song, but no preference for nonmimicry song (Payne, 1973), suggesting that male and female indigobirds might respond to different attributes of song. A lack of response to individual song components, however, does not necessarily mean that they are not important when delivered in the context of a natural mixed repertoire. For example, the mimetic songs of intruding males, when delivered along with chatter and nonmimicry vocalizations, may contribute to discrimination by territorial males.

While the available evidence indicates that both male and female indigobirds respond more strongly to conspecific song (Payne, 1973; Payne et al., 2000; this study), further experiments using artificial combinations of mimicry and nonmimicry songs from different song populations are needed to assess the relative roles of these vocalizations in eliciting both male territorial responses and female mate preferences. Responses to the full courtship repertoire, however, are perhaps most relevant to tests of reproductive isolation. Our study supports the principal role of song in indigobird speciation by providing clear experimental evidence that male indigobirds discriminate among potential competitors based on song and extending this result to the case of morphologically and genetically indistinguishable host races in the wild. Only if females respond to song differently than males and the lack of male response to heterotypic intruders is maladaptive would our results be consistent with an alternative conclusion of interbreeding between individuals of different song populations.

We thank P. Campbell, T. Kunz, T. Lenz, I. Lovette, R. Payne, C. Schneider, K. Sefc, A. Shedlock, F. Wasserman, M. Hauber and three anonymous reviewers for comments on this manuscript. R. Payne provided advice on fieldwork and help with identification of host song mimicry. K. Njabo, J. Dongmo, and J. Cooper provided assistance in the field. We thank R. Fotso for helping us initiate our research in Cameroon. Permits for work in Cameroon were kindly provided by the Institute of Agricultural Research for Development and the Ministry of the Environment and Forests. This paper is based on work supported by the National Science Foundation under Grant Nos. 0089757 and 0309249.