Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Current Zoology 59 (1): 8–19, 2013 Ecological speciation in Darwin’s finches: Parsing the effects of magic traits Jeffrey PODOS1*, Rie DYBBOE2, Mads Ole JENSEN3 1 2 3 Department of Biology, University of Massachusetts, Amherst MA 01003, USA Department of Biology, University of Copenhagen, The August Krogh Bldg., Universitetsparken 13, DK-2100, Copenhagen, Denmark Horsens Statsskole, Studentervænget 2, 7000 Horsens, Denmark Abstract Many recent studies of ecological speciation have focused on “magic trait” scenarios, in which divergent selection on viability traits leads inextricably to corresponding divergence in mechanisms, especially mate recognition systems, that facilitate assortative mating. Speciation however may also proceed via other scenarios, such as when populations experience directly selected or random divergence in mate recognition systems. The relative contributions of magic trait versus other scenarios for speciation remain virtually unexplored. The present study aims to test the relative contribution of the magic trait scenario in the divergence of populations of the medium ground finch Geospiza fortis of Santa Cruz Island, Galapagos. First, we assess differences in G. fortis song between a northern population (Borrero Bay) and a southeastern population (El Garrapatero), differences that we propose (along with other within-island geographic song variations) have arisen via scenarios that do not involve a magic trait scenario. Pairwise comparisons of raw and composite (PC) song parameters, as well as discriminant functions analyses, reveal significant patterns of song divergence between sites. Second, we test the ability of territorial males at Borrero Bay to discriminate songs from the two sites. We find that G. fortis males can discriminate within-island song variants, responding more strongly to local than to “foreign” songs, along 3 raw and 1 composite response measures. Third, we compare these findings to prior data sets on song divergence and discrimination in Santa Cruz G. fortis. These comparisons suggest that song divergence and discrimination are shaped less strongly by geographic sources than by morphological (beak-related) sources. We thus argue that interpopulation song divergence and discrimination, fundamental elements of assortative mating in Darwin’s finches, can be fostered in early stages of divergence under magic trait as well as alternative scenarios for speciation, but with more emphasis on the magic trait scenario, at least for this species on this island [Current Zoology 59 (1): 8–19, 2013]. Keywords Galapagos, Bird song, Geographic variation, Acoustic communication Ecological speciation occurs when divergent natural selection among isolated or partly isolated populations drives the evolution of their reproductive isolation (Dobzhansky, 1951; Mayr, 1963; Rice and Hostert, 1993; Schluter, 2001). One scenario for ecological speciation that has garnered particular attention in recent years involves so-called “magic traits” (Gavrilets, 2004). The magic trait scenario proceeds when two functionally distinct types of traits -- (i) traits subject to disruptive natural selection and (ii) traits that govern patterns of mating -- are either the same or share the same genetic loci. When either condition is met, disruptive natural selection should lead inexorably, as if by magic, to non-random, within-type mating and thus enhanced reproductive isolation (Gavrilets 2004). The “classic” scenario for magic trait speciation, on which the present paper focuses, occurs when divergent selection on viReceived June 26, 2012; accepted Oct. 26, 2012. ∗ Corresponding author. E-mail: [email protected] © 2013 Current Zoology ability traits (e.g., on body size, color pattern, feeding morphology) leads to necessary, correlated divergence in cues or signals that mediate mate selection (Schluter, 2001; Servedio et al., 2011). An emerging empirical question about the classic magic-trait scenario concerns its actual contributions to reproductive isolation, relative to other possible speciation mechanisms. Such mechanisms include differential mutation patterns under uniform selection regimes, hybridization, polypoloidy, or drift (Schluter, 2001; Coyne and Orr 2004). As noted by Servedio et al. (2011, p. 392), “the importance of magic traits in speciation depends on the degree to which they cause an increase in reproductive isolation, i.e. their ‘effect size’.” In many cases, speciation surely proceeds independently of the magic trait scenario. Moreover, for systems even in which magic traits can be identified, such traits might, PODOS J et al.: Ecological speciation in Darwin’s finches in certain times or places, be trivial. For instance, magic traits should no longer influence reproductive isolation if and when divergent selection pressures are relaxed (Haller et al., 2012), or if and when preferences in specific populations come to rely on mating cues or signals unlinked to divergent selection. Under such scenarios, speciation will hinge on the evolution of other factors such as divergence in mate recognition systems via drift, or direct selection on mate recognition systems unrelated to viability selection. What then is the relative contribution of the magic trait scenario versus other scenarios for signal divergence and the evolution of reproductive isolation (Haller et al., 2012; Servedio et al., 2012)? Darwin’s finches of the Galápagos Islands are a potentially useful study system within which to address this question. In a recent literature review, Darwin's finches were identified as one of 18 taxa in nature for which available evidence suggests that ecological speciation proceeds via the classic magic trait scenario (Servedio et al., 2011). This scenario seems applicable to Darwin's finches because a principal locus of adaptive divergence, beak form and function, also influences the expression of vocal mating signals. To elaborate, Darwin's finches are known to adapt readily to their feeding environments, i. e., food availability and interspecific competition, via rapid evolution in beak morphology and function (Gibbs and Grant, 1987; Herrel et al., 2005; Grant and Grant, 2006). Beak divergence is thought to drive, as an incidental consequence, divergence in the acoustic structure of songs, given the functional role of beak morphology and dynamics in vocal production (Nowicki, 1987; Podos, 2001; Podos and Nowicki, 2004a, b; Podos et al., 2004; Huber and Podos, 2006; Herrel et al., 2009). Divergence in song structure, in turn, likely drives within-type assortative mating, Table 1 9 consistent with two main observations: isolating mechanisms among Darwin's finch populations and species (as with many other island songbird species) are primarily pre-mating (Grant and Grant, 1992); and, in these birds, song is the key mating signal in mate selection (Ratcliffe and Grant, 1985; Grant and Grant, 1997, see also Grant et al., 2000). This magic trait scenario must, however, be weighed against other factors that shape finch song evolution and divergence (Table 1). One such factor for bird song evolution is selection for transmission efficacy, in which signal structure evolves to minimize song degradation and attenuation, or to enhance detectability or distinctiveness against background noise (e. g., Morton, 1975; Ryan and Brenowitz, 1985; Wiley 1991; Seddon, 2005; Tobias et al., 2010). In Darwin’s finches, adaptations for minimizing degradation and attenuation appear to play some role in interspecies song divergence (Bowman, 1979), although that role is relatively minor given that many populations and species inhabit the same or acoustically similar habitats. A case of song evolution for distinctiveness in response to background noise has been recently reported for Darwin's finches of Daphne major island, in which a newly established population of the large ground finch, Geospiza magnirostris, which sings songs with slow trill rates, appears to have driven the evolution of songs of the other resident species, G. fortis and G. scandens, towards higher trill rates (Grant and Grant, 2010). Selection for transmission efficacy in either context (minimizing degradation, maximizing detectability) can be considered as adaptive and ecologically-driven, yet would not fall under the classic magic-trait scenario because evolutionary change in the vocal phenotype results from direct selection rather than emerging as a secondary correlate of divergent selection on viability traits. Mechanisms of song divergence considered in this study Mechanism of song divergence Classic magictrait scenario? Scale for divergence Morphological adaptation, correlated song evolution Yes Between morph Acoustic adaptation for signaling efficacy No Between site Random drift (including cultural evolution) No Citations for study species Podos 2001, Huber and Podos 2006 Bowman 1983, Grant and Grant 2006 Between site (and morph) Grant and Grant 1996, Goodale and Podos 2010 The first two mechanisms are adaptive, resulting in song x ecology correlations, yet only the first corresponds to the classic magic trait scenario (Servedio et al. 2011). Acoustic adaptation does not fit under the classic magic trait scenario because it is driven directly by local acoustic ecology, rather than as an indirect byproduct of selection on a correlated viability trait. The present study considers two scales of song divergence -- between morph and between site -- and also asks how these variations correspond to the strength of birds' perceptual discrimination. Our between-morph comparison removes the effects of acoustic adaptation, given that the morphs occupy the same acoustic environments. Between morph differences can thus be viewed as driven largely by the classic magic-trait scenario, with perhaps some contributions from drift. Our between-site comparisons remove the effects of the classic magic-trait scenario, by being limited to a single morph class. 10 Current Zoology Another factor that can shape song evolution is drift. Song divergence by drift occurs as fully or partly isolated populations undergo distinct trajectories of random evolution, either in genes that underlie the song phenotype, or in vocal memes in taxa such as Darwin's finches that learn to sing through imitation (genetic or cultural evolution, e. g., Jenkins, 1978; Slater, 1986; Payne, 1996; Irwin et al., 2008). Drift likely accounts for moderate to strong degrees of structural variation within and across Darwin's finch populations, as suggested by analyses of song divergence within sites over time (Grant and Grant, 1996; Goodale and Podos, 2010) and across sites within islands (Snodgrass and Heller, 1904; Bowman, 1979, 1983; Ratcliffe, 1981; Podos, 2007; see also Podos and Warren, 2007). Song evolution by drift also does not fit the magic-trait scenario, because it can occur independently of patterns of adaptive evolution. How then do song variations driven indirectly by the divergence of beaks (magic trait scenario) compare to song variations driven by direct selection for signaling efficacy in divergent habitats, and random song variations generated by drift (alternative scenarios) in their influence on reproductive isolation? We here address this question for Geospiza fortis, the medium ground finch, of Santa Cruz Island, Galapagos. G. fortis shows unusually extensive variation in beak morphology across the archipelago (Grant et al., 1985), and different populations across Santa Cruz exhibit distinct beak size distributions (Hendry et al., 2006). Moreover, two Santa Cruz populations, one currently (El Garrapatero) and one historically (Academy Bay), show beak size distributions that are distinctly bimodal, Fig. 1 Vol. 59 No. 1 with small morphs and large morphs separated by a size gap (Hendry et al. 2006, see Fig. 1 for site locations). At El Garrapatero, the songs of the two morphs differ in a manner consistent with a vocal constraint hypothesis, which predicts that large-beaked morphs should sing with lower vocal performance (Huber and Podos, 2006). Furthermore, the two morphs are favored by disruptive selection (Hendry et al., 2009), and mate assortatively by morph (Huber et al., 2007). It thus seems reasonable to ascribe song differences among morphs, and its consequences, to ecological divergence and the magic trait scenario (as in Podos, 2001; Servedio et al., 2011) as well as, to some extent, the effects of drift. By contrast, given that the morphs at El Garrapatero live and nest in the same habitat, often side by side in the same trees and feeding flocks (S. K. Huber, L. DeLeon, A. Hendry, and J. Podos unpublished data), song differences between morphs could not be attributed to divergence in acoustic ecology. Overall, variations that occurs within-island (our focus here) likely represent early stages of divergence, with ready opportunity for gene flow among populations. By contrast, between-island comparisons are more likely to illuminate later stages of divergence, given that gene flow between islands is restricted, and ecological separation typically more extensive. We present tests of song variation between Borrero Bay and El Garrapatero, and behavioral responses of birds at Borrero Bay to playback of song from both localities. This latter component of our study is the reciprocal of an analysis in Podos 2010, in which birds at El Garrapatero were played these same songs. We also revisit published data on variations in song structure, Map of Galápagos and Santa Cruz Island, showing the three locations featured in this paper New data introduced here compares song structure of male G. fortis between Borrero Bay and El Garrapatero, and also includes playback trials in which territorial males at Borrero Bay were presented with song stimuli from their local site and from males recorded at El Garrapatero. These data are evaluated, in the discussion section, together with prior analysis of (i) song variation between El Garrapatero and Academy Bay (Podos, 2007) and between beak-size morphs at El Garrapatero (Huber and Podos 2006), (ii) previously reported playback data with birds at El Garrapatero on song discrimination by site (Podos 2007, Podos 2010) and beak-size morph (Podos 2010), and (iii) reported genetic differences by site and morph (Huber et al., 2007, de Leon et al., 2011). PODOS J et al.: Ecological speciation in Darwin’s finches song playback, and genetic data within and across Santa Cruz sites (Fig. 1, Huber and Podos, 2006; Hendry et al., 2006; Podos, 2007, 2010; de Leon et al., 2010). These approaches together allow a first assessment of the relative contributions of magic trait versus alternative scenarios in finch speciation. More specifically, we propose that magnitudes of song variation between morph and between site will roughly parallel the relative contributions of magic trait versus alternative scenarios, respectively, in song divergence (Table 1). If, for instance, we find that songs vary much more between morphs than between sites, we could infer a greater role for the magictrait scenario in song divergence. Moreover, comparing birds’ abilities to discriminate between-morph versus between-site song variants provides context for interpreting the functional importance of those signal variations. If, for instance, birds show greater facility in discriminating spatial versus morph-associated song variants, we would infer a greater effective role for the speciation scenarios alternative to that of magic traits. 1 Materials and Methods 1.1 Geographic variation in song structure Songs of male G. fortis were recorded at Borrero Bay (Fig. 1) during 4–5 day visits in February and March of 2004, 2005 and 2007, using Sony TCD-5M cassette recorders or Marantz PMD660 digital recorders, and Sennheiser omnidirectional microphones (K6/ME62 or MKH20) mounted in 55.9 cm Telinga Pro Universal parabolas. None of the birds recorded were banded but, as in all Darwin’s finches, males sing a single, acoustically distinct and highly stereotyped song type, which allowed us to confirm later that no individual bird was included twice in our analyses. Songs from our second site, El Garrapatero, were the same used in a prior analysis of geographical variation in song structure (Podos, 2007). Our sampling at El Garrapatero has been long-term and comprehensive, and all birds recorded at that site had been banded, measured, and categorized to morph class (small or large, see Hendry et al., 2006, Huber et al., 2007). In the present analysis we only included songs from El Garrapatero of small-beaked morphs. This is because large morphed birds are very rare at Borrero Bay (Hendry et al., 2006), and characterizing the contributions to song variations of non-beak related factors requires that we limit our comparisons to within the same beak-size classes, i. e., to eliminate from consideration contributions to song variation of the magic-trait scenario. Our sample size for statistical analysis was 17 males from Borrero Bay 11 and 19 males from El Garrapatero. For the El Garrapatero sample, we analyzed up to 10 songs from each singer, and calculated mean values for four acoustic parameters: song duration, trill rate, minimum frequency, and maximum frequency (as described in Podos, 2007). A fifth parameter, frequency bandwidth, was calculated as the difference between maximum and minimum frequency. The same parameters were calculated from a single high quality song from each of our Borrero Bay sample. As in Podos (2007), we compared values of acoustic parameters between sites using Mann-Whitney U tests, to test for spatial variation by site. Also as in Podos (2007), we calculated, across the pooled sample, Principal Component Analysis (PCA) on the correlation matrix for these five raw song parameters (see Table 2 for factor loadings), and compared PC factors by site using Mann-Whitney U tests. These statistical analyses were conducted using the R statistics platform (http://www.r-project.org/). Finally, we performed Discriminant Functions Analysis (DFA), first on the raw song parameters and then on PC factors 1−3, to assess how well these parameters or factors predicted the geographical locality (Borrero Bay or El Garrapatero) at which songs were recorded. The DFA was conducted in Systat 11 (SPSS, Chicago, IL, USA). Table 2 site PC loadings for the analysis of song variation by Song parameter PC1 (40.3%) PC2 (35.0%) PC3 (20.8%) Trill rate 0.308 0.634 0 0.694 Duration 0.644 -0.286 Minimum frequency 0.028 0 0.176 Maximum frequency 0.648 -0.284 -0.698 Frequency bandwidth -0.263 -0.659 0 Proportion of variance explained by each PC factor is indicated in parentheses. The two most strongly weighted values for each factor are indicated in bold. 1.2 Playback study Subjects Playback trials were conducted with 9 male G. fortis at Borrero Bay. We ran trials during four days, 26 February through 01 March 2007. Each morning we located males that were actively defending a territory, recorded songs from each, and ran playback trials. All males were defending nests, some of which contained eggs or chicks. Playback stimuli We here used the same playback stimuli from the 12 Current Zoology geographic song discrimination component of Podos (2010). In brief, we identified high-quality versions of songs recorded during our 2004 and 2005 visits to Borrero Bay (see above for recording methods). We assumed that all of our Borrero Bay singers were smallmorphed, based on annotations taken during recording and also given that large-morphed birds at Borrero Bay are rare (Hendry et al., 2006). Following Podos (2010), and as with our analysis of geographical variation in song structure (see above), we accounted for the presumed small-morph bias in our Borrero Bay playback stimuli by restricting our selection of playback songs from El Garrapatero to small-morphed singers from that site. For each of 18 playback stimuli (2 sequences for each of the 9 subjects), we prepared a playback sequence consisting of 18 repetitions of each song, played every 10 s for 3 min. Playback trials One local song sequence and one foreign song sequence were played to each subject during separate trials. Each playback sequence was presented only once during the entire experiment, so as to avoid pseudoreplication (Kroodsma et al., 2001). For each trial, a portable playback speaker (Mineroff SME-AFS) was placed on the ground within the subject’s territory, facing a nest the male was seen to be defending. The same speaker location was used for each bird for playback of both local and foreign song stimuli. Playback stimuli were played from an Apple iPod, saved in uncompressed ‘.wav’ format at a 44.1 kHz sample rate. The maximum amplitude of all sequences was standardized to a constant value, 90 dB at 1 m, using test tones and a Radio Shack sound level meter. Before a given trial, four observers (all authors plus an assistant) positioned themselves around the subject’s territory, and annotated information about subject location and behavior, using portable digital recorders. Each trial consisted of the 3 minutes of song playback, and was only initiated when the focal bird was in sight and within ~ 20 m of the speaker. The vocal behavior of subjects during trials was recorded using a Marantz PMD660 digital recorder (44.1 kHz sampling rate, uncompressed “.wav” files) and a Sennheiser omnidirectional microphone mounted in a Telinga parabola. Behavioral patterns noted were flights, songs, and horizontal distance to the speaker, in terms of position when perched and when flying over the playback speaker. After each trial distance estimates were confirmed or refined using a tape measure. The second trial for each Vol. 59 No. 1 bird was conducted on the same morning as the first trial, after an interval of at least one hour so as to minimize habituation. As in many prior studies of responses to geographic song variants by territorial male songbirds, we predicted that birds would respond more strongly to local, familiar song variants. Analysis of playback responses Upon return to the laboratory, audio recordings of each trial were visualized using spectrograms, generated via the “StripChart” function of Signal 4.0 (Beeman, 2002). This enabled us to annotate all relevant response behaviors and note their timing to the nearest second. Quantification of response behaviors followed the methods in Podos (2007) and Podos (2010). To account for possible correlations among raw response parameters and to generate composite response indexes, we conducted Principal Component Analysis on the raw response parameters (see Table 3 for factor loadings), using the R statistical platform. Raw response parameters and PC factors 1–3 were compared among local and foreign sites, using Wilcoxon signed-rank tests. Table 3 PC loadings of response parameters from playback trials Response parameter PC1 (41.8%) PC2 (28.9%) PC3 (12.3%) minimum distance (perched) -0.558 0.011 -0.309 minimum distance w/overflights) -0.535 0 -0.407 latency to first flight -0.203 -0.595 0.494 flights/min 0.453 0.37 -0.211 latency to first song 0.314 -0.527 -0.107 songs/min -0.242 0.482 0.662 2 Results 2.1 Geographic variation in song structure Representative songs from Borrero Bay are illustrated in Fig. 2. As can be seen, song structure across birds in this population is highly variable, an observation also reflected in the wide scatter among points in PCA summary plots (Fig. 3). Three lines of evidence indicate that Borrero Bay songs are structurally distinct from their El Garrapatero counterparts. First, visual inspection of spectrograms (representative samples of G. fortis songs from El Garrapatero can be found in Huber and Podos, 2006; Podos, 2007) reveals differences by site in song type structure. That is, most song types at each site have no obvious counterparts at the other site. Second, songs from the two sites are seen to diverge statistically in three parameters: trill rate (faster at Bor- PODOS J et al.: Ecological speciation in Darwin’s finches Fig. 2 13 Representative songs from the Borrero Bay population of Geospiza fortis Broad variation in song structure within this population is evident, consistent with observations of song variation within other G. fortis populations. Several of these songs are similar in structure to songs recorded at El Garrapatero (e.g., Fig.1 in Podos, 2007), although most do not appear to have a specific El Garrapatero counterpart. A statistical analysis of raw song parameters revealed two differences in song by population, in trill rate (with Borrero Bay birds singing faster songs) and in minimum frequency (with Borrero Bay birds singing lower minimum frequencies, see text, Table 4). rero Bay), minimum frequency (lower at Borrero Bay, Table 4), and PC3, which mainly represents maximum frequency and duration (Table 2, Fig. 3). All three of these results retain significance after sequential Bonferroni corrections. Third, our DFA revealed significant separation of songs according to site (for raw acoustic parameters, Wilk’s lambda = 0.404, F1, 34 = 11.44, P < 0.001; for PC factors, Wilk’s lambda = 0.448, F1, 34 = 13.13, P < 0.001). DFA was also able to classify the majority of songs to their correct locations (for raw parameters, 16 of 17 Borrero Bay songs and 15 of 19 El Garrapatero songs; for PC factors, 17 of 17 Borrero Bay songs and 16 of 19 El Garrapatero songs). 2.2 Playback study Playback trials for two subjects were interrupted by neighboring birds, and data from these subjects were thus excluded from further analysis. In both cases, interruptions occurred during playback of local song, and consisted of a neighbor either responding strongly to the playback stimulus itself, or engaging with the focal male in a chase. Responses of the remaining birds, sorted by category, are illustrated for raw response parameters in Fig. 4. Three of these parameters, all related to flight, showed significant differences by stimulus class, in terms of uncorrected p-values (minimum approach distance, perched, Z = -2.152, P = 0.031; latency to first flight, Z = -2.032, P = 0.042; flight rate, Z = 1.960, P = 0.050). PC1, which corresponded to flight behavior (closer approaches, quicker latency to fly, higher flight rate) also showed significant differences by condition, with stronger responses to local song stimuli than to foreign song stimuli (PC1, Borrero Bay stimuli = 0.830 ± 0.512 SE, El Garrapatero stimuli = -0.830 ± 0.587 SE, Z = 2.366, P = 0.018). Statistical values for non-significant response parameters were as follows: minimum approach distance w/ overflights, Z = -1.638, P = 0.101; latency to first song, Z = -1.270, P = 0.204; song rate, Z = 0.255, P = 0.799; PC2, Borrero Bay stimuli = 0.360 ± 0.695 SE, El Garrapatero stimuli = -0.360 ± 1.804 SE, Z = 1.183, P = 0.237. 14 Current Zoology Fig. 3 Multivariate analysis of song variation between Borrero Bay (closed circles) and El Garrapatero (open squares) Shown are PC2 and PC3 plotted as a function of PC1. PC loadings are shown in Table 2: PC1 represents, roughly, duration and maximum frequency; PC2 varies positively with trill rate and negatively with frequency bandwidth; and PC3 varied positively with duration and negatively with maximum frequency. Differences by population in PC3 were statistically significant (see Table 4, text). 3 Discussion Our first main finding is that songs of Geospiza fortis Vol. 59 No. 1 have diverged in structure across two Santa Cruz populations ~25 km distant, Borrero Bay and El Garrapatero. Other instances of within-island song divergence in Darwin's finches have been reported (Snodgrass and Heller, 1904; Bowman, 1979, 1983; Ratcliffe, 1981; Podos, 2007), but the present case seems especially pronounced. Statistically significant differences were identified in two song parameters, trill rate and minimum frequency, as well as in one composite measure of song structure, PC3. These differences are more extensive than those reported in a similar comparison of songs between El Garrapatero and Academy Bay, sites separated by ~11 km (Podos, 2007; Fig. 1). Songs between those sites were found to differ in only one raw song parameter (minimum frequency), not two, although they did also differ in a PC factor (Podos, 2007). Also, discriminant functions analysis was able to categorize songs to site with greater accuracy in the Borrero Bay x El Garrapatero comparison (86.1% and 91.7% for raw and PC factor data, respectively) than for the El Garrapatero x Academy Bay comparison (78.9% and 70.2%, Podos, 2007). This pattern of greater differences in song between the more distant sites might be best explained by drift, by clinal variation in acoustical ecology, or by some combination of the two. Random variations in song should accumulate over increasing distances (e.g., Slater, 1986), as in ‘isolation by distance’ models of genetic divergence. Ecologically speaking, the two study sites in the present comparison present roughly similar vegetation profiles (both populations are low-elevation arid zone with landscapes featuring Opuntia echios, Acacia macracantha, Cordia lutea, and Bursera graveolens), although they still might exert slightly differential pressures for song adaptations that minimize song degradation. The sites probably also differ in available acoustic Table 4 Summary statistics (mean ± SD) for five song parameters and three PC factors from two sites, Borrero Bay and El Garrapatero. Significant uncorrected P-values are shown in bold Locality Song parameter Between-site comparison Borrero Bay El Garrapatero Mann-Whitney U scores Z approximation Uncorrected P value Trill rate (Hz) 2.52±0.56 2.08±0.35 Duration (s) 0.78±0.21 0.85±0.14 240.5 2.5 0.012 186.5 0.79 0.428 Minimum frequency (kHz) 1.98±0.23 2.31±0.20 278 3.69 <0.001 Maximum frequency (kHz) 4.93±1.21 4.73±0.95 174 0.39 0.692 Frequency bandwidth (kHz) 2.95±1.13 2.42±0.97 209 1.51 0.132 PC1 0.42±1.50 -0.37±1.27 208.5 1.49 0.136 PC2 0.33±1.60 -0.30±0.96 189.5 0.89 0.375 PC3 -0.68±0.84 0.61±0.75 285 3.91 <0.001 PODOS J et al.: Ecological speciation in Darwin’s finches Fig. 4 15 Results from playback trials, means ± SE for the 6 raw response parameters shown Significant differences were found in 3 of the 4 flight response parameters, with birds responding more strongly to local songs (Borrero Bay) than foreign songs (EI Garrapatero, * uncorrected P < 0.05). A parallel stronger response to local song was also documented for PC1, which loaded most strongly with flight response (see Table 3, text). space (Luther, 2009; Grant and Grant, 2010), especially given the relative paucity of large-morphed G. fortis at Borrero Bay. As such, Borrero Bay birds might be free to occupy a broader acoustic space with their songs than they would be able to in the presence of large-morphed acoustic competitors. Indeed, birds from Borrero Bay sing songs with relatively low minimum frequencies and high trill rates, patterns which at El Garrapatero appear characteristic of large-morphed birds (Huber and Podos, 2006). Turning then to the central question raised in this paper, how do drift and/or acoustic adaptation (alternative scenarios for speciation) compare to morphology (magic-trait scenario) as drivers of song divergence? The relationship between song divergence and morphology has been assessed for beak-size morphs at El Garrapatero (Huber and Podos, 2006), a comparison that minimized the potentially confounding effects of acoustic environment, given that these birds nest and sing in the same habitats. In that study, morph songs showed divergence in all parameters assessed: minimum and maximum frequencies, frequency bandwidth, trill rate, and "vocal deviation", the latter being a composite measure of vocal performance that considers trill rate 16 Current Zoology and frequency bandwidth together relative to the species' putative performance boundary (Podos, 2001; Huber and Podos, 2006). Direct comparison of the Huber the and Podos (2006) morphological divergence data set to the geographic divergence data sets (here and in Podos, 2007) is hindered by the fact that not all of the same song parameters were measured across studies, and that Huber and Podos (2006) did not include a discriminant functions analysis. Nevertheless, different patterns of trait variation are apparent. Minimum frequency is seen to vary by both morph and site, being higher in small-morphed birds and at El Garrapatero as compared to Borrero Bay and Academy Bay. Similarly, trill rate also varies by morph and one of the site comparisons, El Garrapatero and Borrero Bay (Table 4), being higher at Borrero Bay and in large- beaked morphs. Yet divergence in other song parameters assessed is predicted only by morph designation; these parameters include maximum frequency, frequency bandwidth, and vocal deviation. The greater breadth of parameters that vary with morph designation suggests to us that morphology has the more pronounced influence on song divergence. This finding in turn tentatively supports a greater role for the magic trait scenario than the alternative scenarios in song divergence. Ultimately, however, sorting magic trait from alternative scenarios for speciation requires attention not just to signal structure but also to signal discrimination. To illustrate, certain signal parameters might diverge as a correlated effect of morphological divergence, but have no effect on mating patterns (and thus the evolution of reproductive isolation) if those signal parameters are not recognized or used in mate assessment. In many bird groups including Darwin's finches, reproductive isolation is principally pre-mating in nature, which means that females’ perception, discrimination, and preferences for certain signal features are the key determinants of speciation. The ideal test for magic trait versus alternative scenario effect sizes in Darwin's finches would thus compare female preferences for song traits linked to morphology versus song traits explained by acoustic adaptation or drift. However, studies of song preferences in female Darwin's finches have so far proven intractable (S. K. Huber and J. Podos, unpublished data), and the majority of published studies on song function in Darwin's finches have focused instead on responses to song playback in territorial males (e. g., Ratcliffe and Grant, 1985; Grant and Grant, 2002a, b). Typically we expect males to be less discriminating than females in Vol. 59 No. 1 their reactions to song, because of lower costs associated with false positives (Searcy and Brenowitz, 1988; Ratcliffe and Otter, 1996). Indeed, in two studies of sparrows in the same family as Darwin’s finches (Emberizidae), females were seen to respond positively to a narrower range of song stimuli than did males (Searcy et al., 2002; Danner et al., 2011). We thus expect any discriminations males make among stimulus classes to also occur in females, whereas a lack of discrimination among song categories by males is less likely to tell us anything about how females would respond. Thus, we presume that any perceptual discriminations (by morph or site) we can document for males would translate into patterns of assortative mating (by morph or site) in females, and as such inform the likelihood of magic-trait versus alternative scenarios for speciation. This leads to the second main finding from our study, which is that males responded more strongly to local songs (Borrero Bay) than to foreign songs (El Garrapatero, see Fig. 4). This finding parallels those in two prior studies of geographic song discrimination: for El Garrapatero males played local songs and songs from Academy Bay (Podos, 2007); and less strongly, in a reciprocal study to the present one, for birds at El Garrapatero played local songs and songs from Borrero Bay (Podos, 2010). Returning the central question of this paper, how do these geographical discriminations compare to the discrimination of morphologicallylinked song variations? Does a greater degree of linked divergence in song parameters translate into a higher effect size for a magic-trait scenario versus alternative scenarios? An informal ranking of the four available sets of playback data is illuminating. As illustrated in Table 5, we find that the overall discrimination of morphological variants was stronger than in the three tests of geographic variant discrimination. This result suggests a comparatively large role for the magic-trait scenario in the song divergence and discrimination. As a side note, within the geographic discrimination data sets, the strongest discrimination was in the El Garrapatero x Academy Bay discrimination, and the two weakest discriminations were in the Borrero Bay x El Garrapatero comparisons. This latter result is at odds with our song divergence data, which documented greater (not less) divergence in the Borrero Bay x El Garrapatero comparison versus the El Garrapatero x Academy Bay comparison. This outcome is puzzling given that song sets that have diverged more in structure should be easier to discriminate. PODOS J et al.: Ecological speciation in Darwin’s finches Table 5 17 Ranking of song discrimination scores across 4 playback data sets Playback study closest approach (perched) closest approach (w/overflights) latency to first flight flight rate PC1 PC2 Average Rank Geographic 1: Males at El Garrapatero, local vs Academy Bay songs (Podos 2007) 2 1 3 3 1.5 4 2.42 Geographic 2: Males at El Garrapatero, local vs. Borrero Bay songs (Podos 2010) 4 4 4 1 4 1 3.00 Geographic 3: Males at Borrero Bay, local vs. El Garrapatero songs (this study, Fig. 4) 3 3 2 4 3 2 2.83 Morphological: Males at El Garrapatero, same vs. other morph songs (Podos 2010) 1 2 1 2 1.5 3 1.75 For each data set, responses to treatment (geographic or morph) were evaluated by 4 raw and 6 PC response variables, and the strength of the differences by treatment indicated by Wilcoxon signed-rank test values. Here we ranked those values by strength, via comparison of uncorrected p-values from Wilcoxon values across the studies. Scores were ranked from 1−4, for strongest to weakest discrimination. Two additional response parameters, song rate and latency to first song, did not differ by condition in any of the four experiments, and are thus not considered here. Average rank is shown in the last column. The overall strongest discrimination was to the morph treatment. Needless to say, much remains to be learned about if and how spatial and morphological song variations translate into actual mechanisms and patterns of population divergence. The present study and analysis is best considered as only preliminary, given the limited number and arbitrary selection of sites used in our study. However, we can view our results to date within the context of the work of de Leon et al. (2010), who documented patterns of genetic divergence both across sites and between morphs for Santa Cruz Geospiza fortis. A main finding from de Leon et al. (2010) is that genetic divergence among morphs substantially outweighs genetic divergence by site (the same sites studied here), typically by at least an order of magnitude (e.g., FST value for El Garrapatero large vs. small morph = 0.016, FST values for El Garrapatero small morph vs. Borrero small morph = <0.001). Once again the direction is consistent with a magic-trait divergence scenario, but to a much greater degree for the song data. How do we account for the greater disparity (between morph and site) in genetic data than in song data? Perhaps females show disproportionately stronger realworld discriminations than we detected in territorial males, in a way that accentuates the differences in male discriminations that we report. Exploring this possibility in more quantitative terms will require additional research, ideally incorporating data on female preferences for morphological versus geographical axes of song divergence. To summarize, our focal species, Geospiza fortis, exhibits extensive song divergence across Santa Cruz island sites, as indicated by statistical comparisons of song structure across three sites and within one site with a bimodal distribution. Variation in the relative strength of song differences suggest that song divergence is more pronounced when it is linked to morphological divergence than when it occurs via alternative mechanisms of divergence, including acoustic adaptation or drift. Likewise, available evidence from playback data, admittedly limited, suggests that discrimination is stronger for morphologically- mediated than geographicallymediated song divergence. Together, these data offer a first suggestion that the magic trait scenario outweighs alternative scenarios in the evolution of inter-population reproductive isolation, at least for this species on this island. Acknowledgements We thank the Galápagos National Park Service and Charles Darwin Research Station for logistical support, and Morten Harboe Kirkegaard, Andrew Hendry, and Luis Fernando de Leon for assistance in the field. Jenny Boughman, Sarah Goodwin, Jofre Carnicer, and an anonymous referee offered insightful comments on prior drafts. Supported by NSF grants IBN-0347291 and IOS-1028964 to JP, and awards from the Hede Nielsen Foundation and the 3rd Galathea expedition to MOJ. References Beeman K, 2002. SIGNAL 4.0. Belmont, MA: Engineering De- 18 Current Zoology sign. Bowman RI, 1979. Adaptive morphology of song dialects in Darwin's finches. Journal für Ornithologie 120: 353–389. Bowman RI, 1983. The evolution of song in Darwin's finches. In: Bowman RI, Berson M, Leviton AE ed. Patterns of Evolution in Galápagos Organisms. San Francisco, CA: American Association for the Advancement of Science, Pacific Division, 237–537. Coyne JA, Orr HA, 2004. Speciation. Sunderland MA: Sinauer Associates, Inc. Danner JE, Danner RM, Bonier F, Martin PR, Small TW et al., 2011. Female, but not male, tropical sparrows respond more strongly to the local song dialect: Implications for population divergence. American Naturalist 178: 53–63. de Leon LF, Bermingham E, Podos J, Hendry AP, 2010. Divergence with gene flow as facilitated by ecological differences: Within-island variation in Darwin's finches. Philosophical Transactions of the Royal Society B-Biological Sciences 365: 1041–1052. Dobzhansky T, 1951. Genetics and the Origin of Species. New York: Columbia University Press. Gavrilets S, 2004. Fitness Landscapes and the Origin of Species. Princeton, NJ: Princeton University Press. Gibbs HL, Grant PR, 1987. Oscillating selection on Darwin's finches. Nature 327: 511–513. Goodale E, Podos J, 2010. Persistence of song types in Darwin's finches Geospiza fortis over four decades. Biology Letters 6: 589–592. Grant BR, Grant PR, 1996. Cultural inheritance of song and its role in the evolution of Darwin's finches. Evolution 50: 2471–2487. Grant BR, Grant PR, 2002a. Lack of premating isolation at the base of a phylogenetic tree. American Naturalist 160: 1–19. Grant BR, Grant PR, 2002b. Simulating secondary contact in allopatric speciation: An empirical test of premating isolation. Biological Journal of the Linnean Society 76: 545–556. Grant BR, Grant PR, 2010. Songs of Darwin's finches diverge when a new species enters the community. Proceedings of the National Academy of Sciences of the United States of America 107: 20156–20163. Grant BR, Grant PR, 1998. Hybridization and speciation in Darwin's finches: The role of sexual imprinting on a culturally transmitted trait. In: Howard DJ, Berlocher SH ed. Endless Forms: Species and Speciation. Oxford: Oxford University Press, 404–422. Grant PR, Abbott I, Schluter D, Curry RL, Abbott LK, 1985. Variation in the size and shape of Darwin's finches. Biological Journal of the Linnean Society 25: 1–39. Grant PR, Grant BR, 1992. Hybridization of bird species. Science 256: 193–197. Grant PR, Grant BR, 1997. Hybridization, sexual imprinting, and mate choice. American Naturalist 149: 1–28. Grant PR, Grant BR, 2006. Evolution of character displacement in Darwin's finches. Science 313: 224–226. Grant PR, Grant BR, Petren K, 2000. The allopatric phase of speciation: The sharp-beaked ground finch Geospiza difficilis on the Galapagos Islands. Biological Journal of the Linnean Society 69: 287–317. Haller BC, De Leon LF, Rolshausen G, Gotanda KM, Hendry AP, Vol. 59 No. 1 2012. Magic traits: Distinguishing the important from the trivial. Trends in Ecology and Evolution 27: 4–5. Hendry AP, Grant PR, Grant BR, Ford HA, Brewer MJ et al., 2006. Possible human impacts on adaptive radiation: Beak size bimodality in Darwin's finches. Proceedings of the Royal Society B-Biological Sciences 273: 1887–1894. Hendry AP, Huber SK, De Leon LF, Herrel A, Podos J, 2009. Disruptive selection in a bimodal population of Darwin's finches. Proceedings of the Royal Society B-Biological Sciences 276: 753–759. Herrel A, Podos J, Huber SK, Hendry AP, 2005. Evolution of bite force in Darwin's finches: A key role for head width. Journal of Evolutionary Biology 18: 669–675. Herrel A, Podos J, Vanhooydonck B, Hendry AP, 2009. Force-velocity trade-off in Darwin's finch jaw function: A biomechanical basis for ecological speciation? Functional Ecology 23: 119–125. Huber SK, De Leon LF, Hendry AP, Bermingham E, Podos J, 2007. Reproductive isolation of sympatric morphs in a population of Darwin's finches. Proceedings of the Royal Society B-Biological Sciences 274: 1709–1714. Huber SK, Podos J, 2006. Beak morphology and song features covary in a population of Darwin's finches Geospiza fortis. Biological Journal of the Linnean Society 88: 489–498. Irwin DE, Thimgan MP, Irwin JH, 2008. Call divergence is correlated with geographic and genetic distance in greenish warblers Phylloscopus trochiloides: A strong role for stochasticity in signal evolution? Journal of Evolutionary Biology 21: 435–448. Jenkins PF, 1978. Cultural transmission of song patterns and dialect development in a free-living bird population. Animal Behaviour 26: 50–78. Kroodsma DE, Byers BE, Goodale E, Johnson S, Liu WC, 2001. Pseudoreplication in playback experiments, revisited a decade later. Animal Behaviour 61: 1029–1033. Mayr E, 1963. Animal Species and Evolution. Cambridge MA: Harvard University Press. Morton ES, 1975. Ecological sources of selection on avian sounds. American Naturalist 109: 17–34. Payne RB, 1996. Song traditions in indigo buntings: Origin, improvisation, dispersal, and extinction in cultural evolution. In: Kroodsma DE, Miller EH ed. Ecology and Evolution of Acoustic Communication in Birds. Ithaca, NY: Cornell University Press, 198–220. Podos J, 2001. Correlated evolution of morphology and vocal signal structure in Darwin's finches. Nature 409: 185–188. Podos J, 2007. Discrimination of geographical song variants by Darwin's finches. Animal Behaviour 73: 833–844. Podos J, 2010. Acoustic discrimination of sympatric morphs in Darwin's finches: A behavioural mechanism for assortative mating? Philosophical Transactions of the Royal Society B-Biological Sciences 365: 1031–1039. Podos J, Huber SK, Taft B, 2004. Bird song: The interface of evolution and mechanism. Annual Review of Ecology Evolution and Systematics 35: 55–87. Podos J, Nowicki S, 2004a. Beaks, adaptation, and vocal evolution in Darwin's finches. Bioscience 54: 501–510. Podos J, Nowicki S, 2004b. Performance limits on birdsong. In: Marler P, H Slabbekoorn ed. Nature's Music: The Science of PODOS J et al.: Ecological speciation in Darwin’s finches Birdsong. New York: Academic Press, 318–342. Podos J, Warren PS, 2007. The evolution of geographic variation in birdsong. Advances in the Study of Behavior Vol. 37: 403–458. Ratcliffe LM, 1981. Species Recognition in Darwin Finches (Geospiza, Gould). PhD Thesis, McGill University, Canada. Ratcliffe LM, Grant PR, 1985. Species recognition in Darwin finches (Geospiza, Gould). 3. Male-responses to playback of different song types, dialects and heterospecific songs. Animal Behaviour 33: 290–307. Ratcliffe LM, Otter, K. 1996. Sex differences in song recognition. In: Kroodsma DE, Miller EH ed. Ecology and Evolution of Acoustic Communication in Birds. Ithaca, NY: Cornell University Press, 290–307. Rice WR, Hostert EE, 1993. Laboratory experiments on speciation: What have we learned in 40 years? Evolution 47: 1637–1653. Ryan MJ, Brenowitz EA, 1985. The role of body size, phylogeny, and ambient noise in the evolution of bird song. American Naturalist 126: 87–100. Schluter D, 2001. Ecology and the origin of species. Trends in Ecology and Evolution 16: 372–380. Searcy WA, Brenowitz EA, 1988. Sexual differences in species recognition of avian song. Nature 332: 152–154. 19 Searcy WA, Nowicki S, Hughes M, Peters S, 2002. Geographic song discrimination in relation to dispersal distances in song sparrows. American Naturalist 159: 221–230. Seddon N, 2005. Ecological adaptation and species recognition drives vocal evolution in neotropical suboscine birds. Evolution 59: 200–215. Servedio MR, Van Doorn GS, Kopp M, Frame AM, Nosil P, 2011. Magic traits in speciation: 'Magic' but not rare? Trends in Ecology and Evolution 26: 389–397. Servedio MR, Van Doorn GS, Kopp M, Frame AM, Nosil P, 2012. Magic traits, pleiotropy and effect sizes: A response to Haller et al.. Trends in Ecology and Evolution 27: 5–6. Slater PJB, 1986. The cultural transmission of birdsong. Trends in Ecology and Evolution 1: 94–97. Snodgrass RE, Heller E, 1904. Papers from the Hopkins-Standford Galápagos expedition, 1898–1899, XVI: Birds. Proceedings of the Washington Academy of Science 5: 231–372. Tobias JA, Aben J, Brumfield RT, Derryberry EP, Halfwerk W et al., 2010. Song divergence by sensory drive in amazonian birds. Evolution 64: 2820–2839. Wiley RH, 1991. Associations of song properties with habitats for territorial oscine birds of eastern North America. American Naturalist 138: 973–993.