The divergence of conspecific recognition signals (CRS) among isolated populations facilitates the evolution of behavioral barriers to gene flow. The influence of CRS evolution on signal effectiveness in isolated populations can be assessed by testing the salience of changes in CRS from surviving ancestral populations but founder events are rarely detected. The population history of the North Island (NI) saddleback Philesturnus rufusater is absolutely known following conservation translocations which increased the number of populations from 1 to 15. With one exception there is no gene flow between these populations. The translocations have generated interisland divergence of male rhythmical song (MRS), a culturally transmitted CRS. We conducted an experimental test of behavioral discrimination in NI saddlebacks exposed to familiar and unfamiliar MRS and found that responses were significantly stronger for familiar MRS, consistent with a model of contemporary cultural evolution leading to discrimination between geographic song variants. Significantly, this result demonstrates the rapid tempo with which discrimination of CRS might evolve within isolated populations and supports both bottleneck and cultural mutation hypotheses in CRS evolution. The evolutionary implications of contemporary cultural evolution in the production and perception of CRS merit debate on the time frames over which conservation management is evaluated.

The divergence of conspecific recognition signals (CRS) within isolated or partially isolated populations facilitates the development of behavioral barriers to gene flow (Darwin 1871; West-Eberhard 1983; Foster 1999). This process, in concert with ecological factors, genetics, and geographical isolation (Edwards et al. 2005; Phillimore et al. 2006; Price 2008) likely influences speciation in birds (Grant and Grant 1997; Slabbekoorn and Smith 2002), particularly the spectacular diversification of oscine passerines (Fitzpatrick 1988) but also suboscine passerines (Seddon and Tobias 2007) and seabirds (Mulard et al. 2009). Bird song plays a critical role in species recognition, mate choice, and resource defense, and in many species shows significant geographical variation (Podos and Warren 2007). Consequently, the study of geographical variation in bird song has been critical to our understanding of how diverging CRS can act as reproductive isolating mechanisms (Slabbekoorn and Smith 2002; Podos and Warren 2007; Grant and Grant 2008).

Many studies have shown that birds can distinguish geographical variation in song (Nelson and Soha 2004b; Podos 2007), typically by measuring the response of males to playback of geographical song variants, but also of females (O’Loghlen and Beecher 1999; Hernandez et al. 2009) or pairs (Searcy et al. 1997; Nelson and Soha 2004a; Patten et al. 2004). However, an understanding of how rapidly this perceptual discrimination evolves is lacking (Zuk and Tinghitella 2008). Derryberry (2007) demonstrated that historical white-crowned sparrow Zonotrichia leucophrys song was a less effective CRS than current song within the same population, supporting the evolution of CRS over time. An alternative approach is to test the salience of changes in communication signals from surviving ancestral populations. This approach has seldom been used as founder events are rarely detected (Clegg et al. 2002; Baker et al. 2003; Brunton et al. 2008), and the founder source is often unknown. Subsequently, much of the research investigating geographical variation in bird song has been on continental populations (MacDougall-Shackleton and MacDougall-Shackleton 2001; Searcy et al. 2002; Nelson and Soha 2004b) with varying levels of dispersal and immigration between populations, whereas few studies (Baker and Jenkins 1987; Lynch and Baker 1993; Baker et al. 2001, 2003, 2006) have examined isolated island populations free of these confounding influences. Therefore, isolated island populations are critical for increasing our understanding of evolutionary processes (Grant and Grant 2008; Price 2008), including the tempo of cultural evolution in bird song and the role it might play in population divergence and speciation.

Conservation translocations provide unique scientific opportunities for studying evolutionary processes as detailed records are often kept of the source, size, and age of founder events (Lovegrove 1996; Parker 2008). In addition, quantifying the long-term implications of creating isolated populations through bottleneck events is critical in realizing and evaluating conservation management goals (Lambert et al. 2005). Although the impacts of translocations on genetic diversity have been well studied (Lambert et al. 2005; Taylor et al. 2005, 2007), the effects on behavioral processes are relatively unknown. The North Island (NI) saddleback or tieke Philesturnus rufusater (Holdaway et al. 2001) has been the focus of intense conservation management since 1964 (Lovegrove 1996) and provides an ideal system for studying the time scale of evolution of CRS.

The NI saddleback is an ancient passerine in the family Callaeatidae (Holdaway et al. 2001). Anthropogenic factors reduced the NI saddleback to a single population on Hen Island, New Zealand, by 1910. Although common on that island, the single population was vulnerable to extinction so a series of successful conservation translocations were initiated in 1964 (Lovegrove 1996; Parker 2008). There are currently 15 populations of NI saddleback, including the natural population on Hen Island, a single population on Coppermine Island established through natural dispersal from nearby (about 150 m) Whatupuke Island (Newman 1980) and 13 populations established through translocation (Fig. 1). NI saddleback are weak fliers incapable of crossing water gaps >250 m (Newman 1980) and with the exception of Coppermine and Whatupuke Islands there is no natural dispersal between any of the populations. Furthermore, DNA analysis indicates that even the population on Coppermine was founded by a small number of founders (Lambert et al. 2005), suggesting that dispersal between the two islands is a rare event.

Figure 1.

The translocation history of the North Island saddleback showing the years of translocations and the sizes and sources of the founding population (reproduced from Parker [2008]). The population on Little Barrier was founded from three different islands (Whatupuke, Lady Alice, and Cuvier) and that on Coppermine is the only example of a natural colonization event. The Kapiti Island population was primarily founded by birds from Stanley Island. However, one bird from Hen, one from Lady Alice or Whatupuke, and four birds from Cuvier survived from previous failed translocations (Lovegrove 1996). We used MRS from Hen and Cuvier Islands in playbacks comparing response to familiar MRS on Motuihe Island. We did not use MRS from Tiritiri Matangi Island in our playback experiments, as at least nine founding individuals from Tiritiri Matangi Island survive on Motuihe Island, and subsequently might be familiar with Tiritiri Matangi Island MRS.

NI saddleback use male rhythmical song (MRS) for territorial advertisement, defense, and mate attraction (Jenkins 1977). MRS is a culturally transmitted CRS that is subject to change through learning errors (Jenkins 1977). Subsequently, the cultural bottlenecks associated with initial translocation and postrelease cultural mutation appear to have generated interisland divergence of the MRS (K. A. Parker, unpubl. data), although genetic effects on song learning and copying accuracy might also have influenced this divergence (Forstmeier et al. 2009). This provides a unique opportunity for studying the evolution of the production and perception of NI saddleback MRS (Parker 2008).

We set out to quantify whether the process of change in MRS between islands has any functional significance as a communication signal for individual NI saddleback populations. The efficacy of CRS have typically been tested using controlled and replicated experimental playbacks measuring the behavioral reactions (e.g., singing behavior, approach distance, latency to approach, and leave the playback area) of resident birds to songs simulating conspecific intruders (Irwin et al. 2001; Patten et al. 2004; Podos 2007). We hypothesized that in playback experiments MRS from different islands will be less effective CRS than local island songs in eliciting territorial behaviors and that the responses of NI saddleback will be greater when exposed to MRS that represent their most recent translocation history. Support for these hypotheses would demonstrate that discrimination of geographical variation in CRS can rapidly (<25 years) establish and might also elucidate the relative roles of translocation-induced cultural bottleneck events, or posttranslocation cultural mutation, in song divergence between isolated island populations. When combined with natural selection pressures discrimination of CRS might contribute to population divergence, and ultimately to allopatric speciation (Fitzpatrick 1988; Slabbekoorn and Smith 2002; Price 2008). Alternatively, a lack of discrimination would suggest that there is little functional significance in the short-term divergence of CRS for NI saddleback, and that it is simply a highly variable behavioral trait under no significant selection pressure for intraspecific function.

Importantly, these alternative scenarios are of interest to both conservation and evolutionary biologists. First, they demonstrate the tempo and mode by which CRS discrimination might evolve, thereby complicating long-term conservation management (Anthony and Blumstein 2000; Buchholz 2007; Caro 2007; Angeloni et al. 2008). Second, they illustrate the potential for conservation management actions, such as translocations, to be potential microevolutionary events, challenging us to look beyond the short time scales (i.e., years or decades) measuring and evaluating conservation actions and to consider the long-term implications of conservation biology in an evolutionary context (Ashley et al. 2003; Stockwell et al. 2003).


Twenty NI saddlebacks were translocated to Motuihe Island from Tiritiri Matangi Island in 2005 (Parker and Laurence 2008). In November 2008, the population was estimated at 45–50 saddlebacks using the methods described in Parker and Laurence (2008), including 13 territorial pairs and 9 of the original translocated birds.

MRS was recorded using a Sony Hi MD MZ-NH700 mini disc (Sony, Park Ridge, NJ) and Sennheiser M66 shotgun microphone (Sennheiser, Old Lyme, CT). All MRS present in the small (45–50 birds) Motuihe Island NI saddleback population were recorded by one researcher over 1 week in November 2008. Hen and Cuvier Islands were each visited for two 1-week periods in 2005 (Hen, April and August; Cuvier, May) and 2006 (Cuvier, April) by 3–5 researchers who systematically sampled MRS across each island. As controls, endemic sympatric grey warbler (Gerygone igata) songs were also recorded on Motuihe Island.

NI saddleback MRS’ are easily distinguished both visually and aurally and all recordings were inspected (FFT = 256, Hann window, 5.8 ms, 50% overlap) using Raven 1.2.1 (Cornell Lab of Ornithology, Ithaca, NY) (Fig. 2). Subsequently, MRSs were randomly selected for playback experiments following comparison and categorization of spectrograms of recordings made on each island.

Figure 2.

Sonograms showing a single exemplar of each of the four classes of stimuli used in the playback experiments on Motuihe Island; (A) control grey warbler; (B) Hen Island MRS; (C) Cuvier Island MRS; (D) Motuihe Island MRS.

Playback experiments were conducted on 10 pairs of Motuihe Island NI saddleback 1–5 December 2008. NI saddleback form stable long-term pair bonds between male and female birds, share and defend a territory year round, forage together, and engage in pair bonding behaviors by giving sexually dimorphic quiet calls (Jenkins 1977). The behavior of test pairs was recorded together for two reasons. First, behavioral responses to territorial intrusions are initiated and maintained by both sexes, although only males give MRS in response to intruders. Second, only 5 of the 10 experimental pairs contained ≥1 color banded known-sex birds (NI saddleback can only be reliably sexed in the hand (Jenkins and Veitch 1991) or when giving sexually dimorphic quiet calls (Jenkins 1977)).

Songs from Motuihe Island were considered familiar but we did not play back self and immediately adjacent neighbors. We chose two types of unfamiliar songs: MRS recorded from the current Cuvier Island population, which had provided the founding population for Tiritiri Matangi Island (24 birds translocated in 1984), and MRS recorded from the current Hen Island population (the only remaining natural population), which provided the founding population for Cuvier Island (29 birds translocated in 1968) (Fig. 1; Parker 2008). Unfamiliar songs from Tiritiri Matangi Island were not used because 5 of the experimental pairs contained at least 1 founder from the 2005 translocation (from Tiritiri Matangi Island to Motuihe Island) that might have been familiar with the MRS.

We recorded all MRS present on Motuihe Island as both the NI saddleback population (45–50 birds, including approximately 13 territorial males) and potential habitat (about 25 ha) on the island are small. Seven MRS (mean ± SD; 7.6 ± 7.8; range 1–23) were recorded, 5 of which were repeatedly recorded (10.2 ± 7.9; range 3–23) and heard during sampling and playback experiments. NI saddleback MRS are sung throughout the day, year round, particularly in response to conspecific intruders or other territorial disturbances (such as people). Singing behavior peaks at dawn year round but is particularly intense during the breeding season (Murphy 1989). However, two of the seven MRS recorded were only heard on the single occasion on which they were recorded, despite a constant researcher presence over two 5-day periods. Therefore, these two MRSs were excluded from the playback experiments, due to doubts as to their representativeness as familiar MRS; they might represent adherent or rare MRS that are not in use as territorial signals. Subsequently, 2 exemplars of each of the repeatedly sampled 5 MRS from Motuihe Island, 10 unfamiliar randomly selected MRS from both Hen and Cuvier Islands, and 10 different Motuihe Island grey warbler songs were selected to be included in playback sequences for each of the 10 test pairs (3 sequences/pair, 30 sequences total). Sequences consisted of 4 min of control songs (1 song/min, 5 songs total), followed by 8 min of silent observation, then 4 min of saddleback MRS (1 song/min, 5 songs total), followed by a further 8 min of silent observation. Playback order was randomized, pairs were not tested more than twice in a single day (≥4 h between tests), and immediate neighbors were not consecutively tested. Playbacks were conducted 0630–1930 h local time using a portable CD player and a Remote Audio Speak Easy Speaker. Playback stimuli were 82–87 dB at 1 m from the speaker, as measured by a sound pressure level meter (Radio Shack 33-2055; Fort Worth, TX).

Prior to playback test territories were checked for subject presence (NI saddleback pair within 80 m of the speaker). Five minutes of preplayback observations confirmed the behavioral status of test pairs; playbacks were delayed if territorial confrontations were occurring. During preplayback, playbacks and observation periods the location of the test pairs relative to the speaker, MRS/min and chatter calls/min were recorded by a single observer 12 m from the speaker. Distance was estimated at 0–4 m from the speaker, 4–8 m, 8–12 m, 12–20 m, and >20 m with flagging tape on vegetation to 12 m to assist estimation.

The responses of test pairs during experiments were grouped into a single 12-min period. Responses compared between control and test stimuli included: MRS/min (male birds only), chatter calls/min, the closest approach to the speaker (the median measure for each distance category, that is, an approach to 0–4 m was analyzed as an approach to 2 m. Approaches >20 m were analyzed as 21 m), the latency to approach to ≤12 m of the speaker (minutes from the start), and the latency to leave the playback area (proportion of the 12 min experimental period within 12 m of the speaker). The latter two responses were restricted to ≤12 m of the speaker due to difficulties in detecting birds in dense vegetation at 12–20 m. A more intense response to playback by test pairs was considered to be increased calling behavior (MRS from males only, chatter calls from both sexes), a closer approach to the speaker, a short latency to approach and a longer latency to leave the playback area.

Kroodsma et al. (2001) suggested that to avoid pseudoreplication the number of test stimuli used should be treated as the sample size rather than the number of test subjects. In our experiments, we used a random sample of 10 different MRS from each of the 2 unfamiliar islands. Ideally, we would have also used 10 different familiar MRS from Motuihe but as described above we are confident that a maximum of 7 MRS types exist on Motuihe, 2 of which were only recorded and heard on one occasion and might not actually be representative MRS. In any case, the five familiar MRS used in the experiments represent at least 70% of the familiar song diversity on Motuihe. We also controlled for pseudoreplication at the level of the recording or rendition of an MRS type by any particular individual by using two different exemplars of each of the five Motuihe MRS.

Each pair's responses to control songs were compared to familiar MRS, and to unfamiliar MRS (average pair responses to Cuvier and Hen MRS), using Wilcoxon's matched pair tests. We also used the same test to check for order effects by comparing behavior 1 min immediately before and after the control songs. We also checked if responses to playback were from a single bird or both members of the pair using Friedman's analysis of variance (ANOVA).

Each test pair's behavioral responses to Hen, Cuvier, and Motuihe MRSs were then compared using Friedman's ANOVA and Kendall's coefficient of concordance. NI saddlebacks do not give a singular response to behavioral intrusions and strategies vary between individuals. Therefore, as a final test we conducted a principal components analysis (PCA) using all the behavioral responses to construct the principal components and repeated the above statistical comparisons. Alpha was set at 0.05 for all statistical tests.


PC1 explained 65% of the variation in raw response parameters and loadings were consistent with a closer approach to playback of familiar Motuihe Island MRS and unfamiliar Cuvier Island MRS, approaching the playback area more rapidly, and a greater latency to leave the playback area when compared to unfamiliar Hen Island MRS. PC2 only explained 17% of the variation in raw response parameters (Table 1) and is not included in further analyses.

Table 1.  Principal component factor loadings for the analysis of behavioral responses of Motuihe Island saddleback to playback of grey warbler songs (control) and Motuihe, Cuvier, and Hen Island MRS. The highest values for each PC factor are shown in bold and the percentage of variation across all responses explained by each PC factor is shown in parentheses.
Behavioral response PC1 (65%) PC2 (17%)
MRS 0.56−0.66
Chatter calls 0.560.65
Approach distance−0.93−0.07
Latency to approach−0.95 0.00
Latency to leave0.93−0.06

The behavioral responses and PC1 scores of test pairs to familiar and unfamiliar MRSs were all significantly greater than those given to control songs (Table 2; Fig. 3). There were no approaches to the playback area and no significant order effects when calling behavior (mean ± SD; MRS before = 0.13 ± 0.17; MRS after = 0.03 ± 0.11; Wilcoxon Z = 1.60, N= 10, P= 0.11; chatter calls before = 0.07 ± 0.14; chatter calls after = 0.03 ± 0.11; Z = 0.53, N= 10, P= 0.60) and closest approach to the speaker (approach before = 12.30 ± 8.96 m; approach after = 4.20 ± 8.85 m; Z = 1.86, N= 10, P= 0.06) were compared 1 min before and after control playbacks. With few exceptions the response to playback was from the territorial pair rather than a single bird (Motuihe = 1.9 ± 0.32 birds; Cuvier = 1.8 ± 0.42; Hen = 1.8 ± 0.42) and there was no significant difference in the number of birds (the territorial pair vs. a single bird) responding to calls from Motuihe, Cuvier, or Hen Islands (Friedman χ2= 1.00, N= 10, P= 0.61).

Table 2.  Descriptive statistics (means±SD) and test statistics (Wilcoxon's matched pairs test) for response variables and PC1 comparing playbacks of control song (grey warbler) with playbacks of familiar MRS (Motuihe Island) and unfamiliar MRS (Cuvier and Hen Island) on Motuihe Island.
Response StimulusComparison
 Control (N=10)Familiar (N=10)Unfamiliar (N=10)Control/familiarControl/unfamiliar
Wilcoxon Z-scoreP-valueWilcoxon Z-scoreP-value
MRS  2.00±1.7910.7±7.51 2.1±4.552.670.0121.50<0.01
Chatter calls  2.43±1.21 7.9±8.08 6.3±5.902.110.04 2.19 0.02
Approach distance 16.90±3.17 3.4±4.43 8.7±8.522.800.01 2.41 0.01
Latency approach 11.10±1.69 2.9±3.24 6.4±4.622.700.01 2.24 0.01
Latency leave  0.04±0.060.69±0.400.38±0.382.700.01 2.52 0.01
PC1 −0.93±0.240.92±0.700.01±0.872.700.01 2.41 0.02
Figure 3.

Behavioral responses and PC1 of Motuihe NI saddleback to playback of control grey warbler song, familiar Motuihe Island MRS and unfamiliar Cuvier, and Hen Island MRS. Means ± SE are shown. N= 10 pairs. The bars labeled a and b on the MRS and PC1 graphs are significantly different (P < 0.05) but there were no significant differences in the other behavioral responses measured following playback.

There were significant differences in the number of MRS given by test males in response to playback of familiar and unfamiliar MRS (Friedman χ2= 11.73, N= 10, P < 0.01) (Fig. 3). A post hoc nonparametric multiple comparison (Zar 1999) showed that males gave significantly more MRS in response to familiar Motuihe Island MRS than to unfamiliar Cuvier Island MRS (q= 3.95, P < 0.05) or Hen Island MRS (q= 4.11, P < 0.05) but there was no significant difference between the responses to unfamiliar Cuvier and Hen MRS (q= 0.16, P > 0.05). There were no significant differences in latency to approach the speaker (χ2= 5.89, N= 10, P= 0.05), chatter call rates (χ2= 0.65, N= 10, P= 0.72), approach distance (χ2= 3.26, N= 10, P= 0.20), or latency to leave the playback area (χ2= 3.30, N= 10, P= 0.31).

Responses by test pairs to familiar and unfamiliar MRS differed significantly according to PC1 (χ2= 7.40, N= 10, P= 0.03). A post hoc nonparametric multiple comparison (Zar 1999) showed significantly more responses to familiar Motuihe Island MRS when compared to unfamiliar Hen Island MRS (q= 3.48, P < 0.05). There were no significant differences between familiar Motuihe Island MRS and unfamiliar Cuvier Island MRS (q= 3.16, P > 0.05), or between unfamiliar Cuvier Island MRS and unfamiliar Hen Island MRS (q= 0.32, P > 0.05) (Fig. 3).


Our playback experiments of conspecific song types are consistent with the rapid establishment of behavioral discrimination between geographic song variants by NI saddleback pairs following serial translocations. Critically, these results support the prediction of the contemporary evolution hypothesis that familiar MRS is a more effective CSR than unfamiliar MRS. First, territorial NI saddleback pairs showed consistent behavioral discrimination in all responses separately and the combined first principal coordinate when control heterospecific grey warbler songs was compared to that given to either unfamiliar or familiar conspecific MRS (Table 1). Second, the first principal coordinate showed significant discrimination of familiar Motuihe MRS over unfamiliar Hen to be consistent with a shorter latency to approach the playback area, a closer approach to the playback speaker, and a greater latency to leave the playback area (Table 2; Fig. 3).

When the different behavioral responses to playbacks are examined separately the number of MRS given by males in response to unfamiliar Hen and Cuvier MRS was significantly less when compared to familiar Motuihe MRS, and post hoc tests showed the greatest response to familiar Motuihe MRS. There were no statistically significant differences in the other behaviors or with increasing distance from the ancestral population in response to playback of MRS.

Although many studies demonstrate discriminatory behaviors in playbacks of familiar versus unfamiliar song (Searcy et al. 1997; Nelson and Soha 2004b; Patten et al. 2004; Podos 2007), we know of few wild population studies (Derryberry 2007) with such detailed information on the pace (<25 years) at which this discrimination has evolved. Distinguishing the causal mechanisms that have led to the divergence of MRS between islands is more difficult. We have demonstrated that MRS from different islands is less recognized as a conspecific territorial signal. This might be explained by two hypotheses. Under a cultural bottleneck hypothesis when NI saddleback are translocated to a new island there would be a loss of song diversity in translocated populations relative to the severity of the bottleneck, and a subsequent loss of response to MRSs absent within translocated populations. Alternatively, a cultural mutation hypothesis would predict that the social transmission of MRS is subject to a high rate of learning errors, with a subsequent increase in cultural mutations (Jenkins 1977) and rapid changes in song diversity both within and between populations. Future work focusing on reciprocal experiments in ancestral populations might fully resolve these two hypotheses, but our results provide an initial insight into these alternatives. If cultural mutation has led to divergence Motuihe Island NI saddleback pairs should treat both Hen Island and Cuvier Island MRS as unfamiliar when responses are compared to familiar Motuihe MRS. In contrast, if a cultural bottleneck has led to divergence then the response to unfamiliar Cuvier and familiar Motuihe should have been more similar than to the unfamiliar ancestral MRS from Hen Island, particularly given that the initial bottleneck was severe (Fig. 1) relative to subsequent bottlenecks, where cultural diversity would have already been reduced. These hypotheses are not mutually exclusive and our findings offer support to both. The behavior of test pairs, as represented by PC1 support the bottleneck hypothesis. However, when examined individually the behavior of males, as measured by MRS call rates in response to playback, support the cultural mutation hypothesis. A bioacoustic analysis of MRS from all islands is beyond the scope of the analyses presented here. However, initial results (K. A. Parker, unpubl. data) also support both hypotheses and we suspect the contribution of each will be relative to individual island history, that is, how far any island is removed from the ancestral population on Hen Island (Fig. 1).

We have demonstrated that unfamiliar MRS is a less recognized CRS than familiar MRS. But despite rapid and sometimes aggressive responses in our experiments, with one bird perching on top of and pecking a speaker for >1 min during experimental playbacks, our data do not inform us as to the relative strengths of individual MRS in holding a territory. However, given experimental evidence from other species, including song sparrows (Melospiza melodia) (Nowicki et al. 1998), and the primary role of MRS for territorial defense and advertisement in NI saddleback (Jenkins 1977) it is a reasonable assumption that playback responses might correlate with territory holding behavior. Male removal and replacement with a speaker might elucidate this relationship. However, Falls (1988) suggests that speaker replacement experiments might be most effective with visually inconspicuous species, and in NI saddleback delivery of MRS is loud, conspicuous, close to (<10 m), and within sight of any intruder and accompanied by stereotyped body posturing. Therefore, speaker replacement experiments might also require the use of a complex “robotic” model NI saddleback after Searcy et al. (2008).

The heterospecific controls used in these experiments anchor the directionality of behavioral responses, but it is also possible that responses to variation in conspecific and heterospecific social signals are not linear, but rather curvilinear. Under this scenario, NI saddleback might possess a template for CRS that prompts a strong response to intermediate signals but poor responses to both best and poor matches, in a similar manner to Bateson's (1982) work on Japanese quail (Coturnix coturnix japonica) which showed a preference for first cousins (intermediate phenotype match) over siblings (best phenotype match) and unrelated birds (poor phenotype match) in choice tests. Alternatively, there might be a curvilinear response to low, medium, or high performance level of the signal, as hypothesized by de Kort et al. (2008).

The tempo of cultural evolution of NI saddleback song can clearly occur over time spans (<25 years) consistent with a model of contemporary cultural evolution. Given this short time frame, we consider that the NI saddleback currently fits within the early stages of Slabbekoorn and Smith's (2002) model of acoustic divergence with song learning. Under this model, NI saddleback share a common acoustic space, in that all populations remain physiologically capable of producing all the MRS patterns present within other populations and the realized diversity currently observed within individual populations represents phenotypic variation in MRS.

The relationship between gene flow and geographic variation in song in NI saddlebacks in particular, and in songbirds in general, remains unclear. Although some studies have found a correlation between song dialects and restricted gene flow (Irwin et al. 2001; MacDougall-Shackleton and MacDougall-Shackleton 2001) others have not (Soha et al. 2004; Nicholls et al. 2006; Leader et al. 2008). NI saddlebacks are an important model because with the exception of two of the 13 islands there is no natural dispersal between any of the populations. Dispersal between populations could only be emulated via translocations and might be desirable for genetic management to increase heterozygosity (Lambert et al. 2005) but it is not clear what the effect of divergence in the behavioral recognition of and responses to CRS will be if birds are translocated between established populations. In a population of the closely related NI kokako (Callaeas cinerea wilsoni) (Ewen et al. 2006) established on Kapiti Island through translocations from several distant geographic origins, birds mated assortively with individuals from the same source population (Rowe and Bell 2007) but this behavioral mating pattern disappeared with the first island hatched progeny. In parallel, one practical implication of our work is the recommendation that future translocations of NI saddleback should proceed simultaneously from multiple source populations.

We speculate that translocating between established populations will be challenging but only well-monitored experimental translocations will elucidate the role of familiar MRS and territory holding ability. However, although adult NI saddleback are capable of learning a new MRS Jenkins (1977) suggested that this is infrequent. Young birds moving into a territory typically perfect existing MRS in weeks or months, possibly learnt whereas nonterritorial floaters, but adults might take up to 1 year. It is not clear if there would be a cost to a male NI saddleback singing an unfamiliar MRS or learning a new MRS; however, there is limited evidence from species such as song sparrows that singing a familiar song might be advantageous to territory holding ability (Beecher et al. 2000a,b; Wilson et al. 2000). The NI saddleback social system is based on territorial neighbors sharing the same MRS, and the rare examples of nonconforming birds seem to be temporary phenomenon, with birds either learning the predominant MRS or shifting territories (Jenkins 1977). Density-dependent survival is evident in established NI saddleback populations, territorial vacancies are rare and rapidly filled by local recruits (Armstrong et al. 2005) and birds that can quickly conform might have an advantage. Therefore, reciprocal translocations of adult birds (possibly ≥ 100) might be required to create territorial openings at the respective sites. Several studies in other songbirds show that female birds also prefer familiar song over unfamiliar song (Searcy et al. 1997; O’Loghlen and Beecher 1999; Patten et al. 2004; Hernandez et al. 2009) but such studies are rarer, and exceptions have been reported (Nelson and Soha 2004a). Translocation of eggs or chicks for local song tutoring might also be viable management options but will be expensive logistically and ethically challenging.

The evolutionary impact of geographical variation in song ultimately depends on patterns of signal perception and discrimination (Podos 2007) and existing populations of NI saddleback are likely to continue to diverge acoustically with the existing discriminatory behaviors further reinforced. Over time, NI saddleback might proceed to the latter stages of Slabbekoorn and Smith's (2002) model of acoustic divergence with learning, where song will become a reliable population marker following divergence of potential acoustic variation between island populations. This process might proceed relatively quickly given the small founder populations (generally 20–40 birds) and small offshore and mainland islands in which they can exist (135–3000 ha) (Grant and Grant 1996; Lovegrove 1996). Ultimately, the evolution of behavior, along with genetic drift and ecological factors prompting natural selection, might lead to an allopatric model of speciation within translocated populations of NI saddleback. An oft-stated goal of conservation biology is to preserve evolutionary potential but conservation practice can also direct evolutionary trajectories (Stockwell et al. 2003). Subsequently, there is increasing recognition of the impact of management actions as potential micro-evolutionary events leading to a call for evolutionary enlightened management (Ashley et al. 2003), particularly where conservation mediated changes in behavior might lead to an eventual reduction in the effective population size of managed species (Anthony and Blumstein 2000). This is clearly a long-term phenomenon in a world of pressing short-term conservation problems. Nonetheless, it warrants debate as to the management significance of long-term divergence and also provides a unique research opportunity; that is, although the benefits of long-term studies of ecology are acknowledged, long-term studies of evolution are less recognized (Grant and Grant 2002).

Our study demonstrates how behavioral research can contribute to a short-term understanding of the consequences of conservation management on perceptual selectivity of song variants and a long-term understanding of the evolutionary potential of conservation management of individuals from diverse source populations (Anthony and Blumstein 2000; Buchholz 2007; Caro 2007; Angeloni et al. 2008). Speciation in birds proceeds with the evolution of behavioral barriers to gene flow (Grant and Grant 1997) and evolution is a process, not an event (Grant and Grant 2008). However, with current conservation management programs, evolution might prove to be a process that proceeds at the scale of each translocation event.

Associate Editor: C. Peichel


Research permits for this study were provided by the Massey University Animal Ethics Committee and the New Zealand Department of Conservation. The Motuihe Project funded the translocation of NI saddleback to Motuihe, with particular thanks to J. Lawrence, J. Thompson, and L. Badham. Assistance with field recordings on Hen Island was provided by G. Parker, S. Stoddard, and M. Booth, and on Cuvier Island from W. Ji and B. Evans. The manuscript benefited from thorough constructive reviews from C.L. Peichel, C.A. Stockwell, and an anonymous reviewer. This research is part of KAP's Ph.D. research and was funded by a Marsden Fund through the Royal Society of New Zealand to DHB.