Acoustic signal evolution is influenced by features of the physical habitat, community composition, ambient noise, phylogenetic history, morphological and physical constraints inherent to the species . Our data suggest that several of these factors may be involved in the evolution of acoustic signalling in Zosterops. The two Zosterops species we studied have clearly distinct contact calls, mainly differing in the frequency range and the starting frequency of calls. Further, most populations within the two species are differentiated from each other in features of their contact calls. Differentiation was stronger in the isolated mountain populations of Z. poliogaster than in the panmictic lowland Z. abyssinicus. Call patterns in both species show a geographic cline suggesting similar effects of environmental selection. When considering the absence or presence of the congener, parapatric populations have relatively similar contact calls. In contrast, the sympatric populations of Z. poliogaster in the highland of Nairobi shows the strongest differentiation in contact calls from the lowland Z. abyssinicus suggesting reproductive character displacement. The sympatric Z. abyssinicus population also differed from all conspecific populations and shows a distinct shift away from Z. poliogaster with respect to total call length (Figure 1). In conclusion, it appears that both differences in local environmental conditions as well as selection due to co-existence with congenerics play an important role in the evolution of acoustic signalling in the genus. In the following we discuss these findings in detail.
The two investigated species of Zosterops are clearly differentiated in their contact calls (Figure 1). Contact calls are considered an important, species-specific, social trait, as shown in previous studies for the genus Zosterops, suggesting a function for the maintenance of flock structure as well as for mate recognition [33, 34]. Interestingly, the call patterns of some populations of Z. poliogaster strongly overlapped with Z. abyssinicus, but never in sympatry, indicating that divergent selection may be relaxed in non-overlapping populations.
In addition to species differentiation, local populations of both species significantly differed in their calls, but the degree to which populations have diverged was different between the two species. In the lowland Z. abyssinicus, most populations showed little or no divergence in acoustic parameters (except for the population from Hunters Lodge, which was represented by a limited number of recordings, N = 78). In contras, local populations of the highland Z. poliogaster were much stronger differentiated (Figure 1). These different intraspecific patterns could be explained by the contrasting distributions of the two species: Zosterops abyssinicus is widely distributed in the lowland savannahs of East Africa with strong gene flow between local populations and very limited genetic differentiation  whereas Z. poliogaster populations are geographically isolated with little gene flow leading to strong genetic and morphologic differentiation [35, 36]. The strong acoustic differentiation is also in line with the findings of Baker , who studied the bioacoustics in two isolated Zosterops populations and concluded that song patterns can evolve rapidly in this genus even across limited geographical space. However, whether the observed strong differentiation is the result of slight differences in local selective regimes or simply the product of drift cannot be inferred at this point.
In both species acoustic divergence follows geographic clines (Figure 2A,B), but not geographical distance between sites (Figure 2C,D). A correlation of longitude or latitude and acoustic parameters is a common finding in a variety of organisms (e.g. katydids, frogs and birds, [38–41]). In katydids such clinal variation in song patterns has been interpreted to be the result of hybridization and selection along the geographical gradient . While we found clinal variation in both species, the differences among local populations in relation to longitude and latitude are more pronounced in the mountain species. This is likely due to lower connectivity and therefore less individual exchange among mountain populations as compared with the interconnected populations of the lowland taxon. Interestingly the clinal change is directed similarly in both species suggesting that similar environmental forces might be responsible. Several studies have focused on the traits influencing the rates of the evolution of syllable frequencies in different latitudes, altitudes and environments [42, 43]. Analyses show that song frequency differences have evolved more rapidly at high latitudes, which may indicate an increased intensity of sexual selection [42, 43].
Differentiation in geographic proximity and sympatry
Besides the geographically isolated (allopatric) populations of both species, we recorded and analysed contact calls from three locations where the two species occur in parapatry (Mt. Kasigau, Taita Hills and Chyulu Hills). Our analyses indicate that calls of both species (highland and lowland taxa) are less different from each other than the average divergence found between both species. The call similarity in these parapatric populations might be the result of several processes: 1) if contact calls are a cultural trait and the species specific call patterns are learned rather than inherited, close proximity to a congener (but not complete overlap) might lead to the accidental admixture of acoustic signals among taxa during the imprinting phase, which has been shown for brood parasites . Mixed singing due to heterospecific copying rather than introgression has also been shown in flycatchers . 2) Alternatively, intermediate call patterns can be the result of occasional hybridization of two closely related species, if calls are genetically determined . 3) A more recent study has suggested that interactions may occur more frequently between evolutionarily ‘old’ species; in such cases species interactions can drive phenotypic convergence across entire radiations . In our case it is unknown whether contact calls are genetically fixed or learned. Furthermore, the evolutionary age of the species cannot be determined with certainty, therefore, it is difficult to distinguish between these alternatives. Phylogenetic analyses on the group are currently being undertaken and will allow testing of the latter hypothesis.
Our analyses suggest that when both species occur in sympatry (Nairobi population), the calls of each species show different deviations from the average call parameters. While for Z. abyssinicus, coexistence has relatively little effect on its contact call, Z. poliogaster exhibits a strong shift in call characteristics when compared to allo- and parapatric populations. The song pattern of the sympatric population does not overlap with the congener, and, represents the acoustically most distinct (conspecific) population. In addition, the acoustic space of the sympatric population is smaller when compared to all other Z. poliogaster highland populations or any population of Z. abyssinicus. The picture changes when populations are found in sympatry. Generally, sympatric species are rare in the genus Zosterops, which has led to the suggestion that species diversity is limited by competition with congeners . In systems with multiple closely related species, interspecific competition for signal space  and selection against incorrect mate choice and hybridization [17, 18] can lead to reproductive character displacement. Hence, if contact calls play a role in reproductive isolation in the system, it is expected that these species will have distinct calls when they occur sympatrically. This suggests that the coexistence with the congener leads to reproductive character displacement.
Character displacement is a common mechanism to reduce the risk of hybridization when closely related species occur in sympatry with examples from crickets , frogs [18, 17] and birds [20, 21]. Reproductive character displacement can either be symmetrical (meaning that both species diverge from their patterns exhibited in allopatry) or it can be asymmetrical (where only one species displaces) . A study of sympatric populations of tinkerbirds for example showed a displacement of song patterns in both sympatric species . In contrast, reproductive character displacement in sympatric populations of chorus frog species was asymmetrical . In this example the rarer species displaced their calls, whereas the more common species at the location remained stable . In our study, only Z. poliogaster, which likely has the smaller effective population size, showed character displacement, congruent with the previously described pattern of asymmetrical character displacement . Our observation of reproductive character displacement suggests that contact calls might be under sexual selection in the genus Zosterops and could represent a prezygotic isolation barrier [48, 14]. In addition, natural selection might further act to prevent hybridization because hybrids might be at a disadvantage as both species have different habitat preferences [31, 32, 49]. Therefore, there might also be ecological selection against hybrids.
Our study indicates distinct calls for the two White-eye species, with different intraspecific variances among local populations. These differences in their call variability reflect their contrasting spatial distribution, with strong disjunction in the highland Z. poliogaster populations and large and interconnected populations in Z. abyssinicus. Contact calls of heterospecific populations in parapatry are more similar to each other if compared with both, allopatric and sympatric populations. In contrast, when the species occur in sympatry, reproductive character displacement appears to occur leading to strongly divergent call patterns compared to divergence in calls of allopatric or parapatric populations.