The examination of evolutionary divergence of a morphologically distinct island-dwelling bird population, considered over three microevolutionary timescales has provided a rare empirical insight into the tempo and mode of divergence in a natural system. The Zosterops population on Heron Island has undergone a substantial body size increase since colonization, a maximum of 4000 years ago. Theoretically, sustained directional selection can produce large-scale shifts, even with weak selection pressures (Kingsolver and Pfennig 2004). However, constancy of selection is likely to be an unrealistic assumption in most systems (Kinnison and Hendry 2001; Estes and Arnold 2007). In Heron Island Zosterops, this issue is highlighted when comparing current selection strengths with net selection intensities required to explain the long-term morphological shift. Current selection was higher, frequently an order of magnitude higher, than the net selection intensity required to explain long-term divergence for many traits, and was interpreted as evidence of heterogeneity in form and strength of selection over the divergence period (Kinnison and Hendry 2001).
CHARACTERIZING HETEROGENEITY IN SELECTION
Establishing the presence of heterogeneity in selection operating in a system is a first step to understanding the divergence process. However, reconstructing that heterogeneity across evolutionary time is a difficult task. In the case presented here, three lines of evidence support a scenario of rapid evolution toward a new optimum followed by many generations of minimal change. First, the comparison of mainland and island phenotypes shows that a substantial shift in the optimum phenotype accompanied the colonization of the island environment. Second, tracking the morphology of the island population over a period of four decades demonstrates a lack of directional change that would be consistent with ongoing directional selection. Finally, recent estimates of selection gradients based on survival analysis show that current selection is relatively weak for most traits that have increased significantly in size in the island form.
The rate at which the silvereye population shifted toward the new island optimum after colonization was explored using the displaced optimum model (Estes and Arnold 2007). Under a range of potential adaptive landscapes, the approach to the new optimum occurred in far fewer generations than even the lowest estimate of divergence time (1000 generations). This supports the hypothesis that under novel selective conditions on the island, the bulk of the evolutionary difference observed developed rapidly, possibly in the first couple of hundred generations after colonization. This would represent at least a doubling of the maximum evolutionary rate reported in Table 1, and even if we assume the shift occurred over a conservative 500 generations, all estimates of evolutionary rates would be above the median value (5.3 × 10−3) reported by Kinnison and Hendry (2001). Historically documented colonizations by Zosterops lateralis from Australia to New Zealand and outlying islands indicate that morphological shifts can occur quickly, too quickly to be accounted for by drift alone (Clegg et al. 2002b). We can only speculate on the proximate causes of selection during the period following colonization. However given that the newly established population was experiencing a novel environment in numerous aspects of ecology (e.g., changes in interspecific competition, predator and parasite pressures, and resource availability), physiology (e.g., changes in response to the new abiotic environment), and behavior (e.g., changes in dominance behavior associated with levels of intraspecific competition), it is likely that a complex and multifaceted selection regime was operating.
Following an initial phase of rapid divergence we suggest that the population has since remained relatively stable, with oscillations occurring around the optimum phenotype. In support of this, we find no consistent directional trends for any trait over a 34-year period. Fluctuations in the direction of change also resulted in a lack of net size shifts when comparing the beginning and end of the time period for most juvenile and some adult traits. Additionally, for traits in which a significant net shift was noted across decades, recent significant selection gradients were not directional (e.g., adult tail length increased significantly when comparing 1966 to 1999 measurements, whereas contemporary selection was stabilizing). Conversely, traits that have recently been subject to directional selection show no significant shift across decades (e.g., adult tarsus and culmen length) (Table 5). These results, describing temporal variation in direction and strength and targets of selection, support the view of a population tracking relatively small changes in the environment. A caveat of this interpretation is that we are assuming that the morphological fluctuations are the result of a selective response. Cases in which directional selection has been shown to be operating but the morphological response is in the opposite direction (e.g., Larsson et al. 1998; Garant et al. 2004) or there is no phenotypic response despite the selection pressure (Merilä et al. 2001b) caution that morphological shifts may not be accurately reflecting the underlying selective mechanism. However, detecting fluctuating and episodic selection and shifts in targets of selection is not uncommon in long-term studies (e.g., Grant 1985; Schluter et al. 1991; Grant and Grant 1995, 2002, Badyaev et al. 2000; Coulson et al. 2003). For instance, Grant (1985) showed that directional selection for longer and deeper bills in Geospiza conirostris was counteracted by selection in the opposite direction on correlated traits, on the same trait in the other sex, and at different life stages, thereby constraining evolutionary change. In G. fortis and G. scandens natural selection was also seen to vary between oscillating, directional, episodic, and gradual over a 30-year period due to environmental events that could not be predicted (Grant and Grant 2002).
Table 5. Summary of directions of phenotypic shift and current selection for each trait over three timescales. Trait labels as in Table 1. J, juvenile; A, adult. Pop. Div. is the direction in change of the island compared to the mainland population. Net change indicates if a significant size shift in a positive (+) or negative direction (−) was observed from 1966 to 1999 or if no net shift (no Δ) occurred (but note that no consistent trends were observed for any trait across decades). Codes for total (univariate linear) and direct (multivariate linear) selection are: −ns, negative direction but nonsignificant;+ns, positive direction but nonsignificant; −, significant negative selection; +, significant positive selection; na=not assessed. Bold indicates significant results.
|Tail|| ||no Δ|| ||+ns||−ns||+ns||−ns|
|Tarsus||+||no Δ||no Δ||+ns||−ns||+ns||−|
|CulL|| || ||no Δ||−ns|| ||−ns|| |
|CulW|| ||na||na|| ||+ns|| ||−ns|
Reviews of estimates of selection gradients and differentials across a range of studies indicate generally weak directional selection for morphological traits (median |β| = 0.17) (Kingsolver et al. 2001), tending to be weaker still when calculated using viability (survival) analysis and conducted over a timescale of years, (median |β| = 0.07) (Hoekstra et al. 2001). The estimates of selection presented here are therefore in line with other studies of this kind (morphological studies based on survival across a number of years), and with few exceptions discussed below, are consistent with a generally weak directional selection regime. The weak strength of recent directional selection on individual traits and cases of stabilizing selection on overall body size and some shape parameters during some years supports the idea of maintenance near an optimum as directional selection wanes and stabilizing selection becomes more prevalent (Hunt et al. 2008).
Significant directional selection was detected for culmen measurements and tarsus length from 1999 to 2003. This finding remains compatible with the maintenance of the population near some optimum according to a number of scenarios that have been invoked to explain the existence of stasis in systems that are subject to directional selection (see Merilä et al. 2001b). First, the current directional selection on culmen traits and tarsus length may be transient, and opposing selection at other times and life-history stages may negate any directional shifts (Schluter et al. 1991; Grant and Grant 2002). Second, the traits may appear to be under direct selection due to selection on other traits that are highly genetically correlated. Shifts in direction of selection on any correlated traits could therefore limit change in the traits of interest. Third, selection can only produce a response if it acts on a heritable trait rather than one that is largely or entirely plastic (Falconer and Mackay 1996). Morphological traits such as those of interest here, culmen length and width, and tarsus length, generally have a significant additive genetic variation component in birds (Merilä and Sheldon 2001). For Heron Island silvereyes, analysis in progress indicates heritability estimates in the order of 0.7 for tarsus and 0.2 for culmen measures (S. M. Clegg, F. D. Frenitu, I. P. F. Owens, and M. Blows, unpubl. results). Therefore, for culmen measures at least, there is the possibility that the directional selection detected will not result in an evolutionary response. Finally, even if a trait is highly heritable, such as tarsus length, selection may act entirely or partly on the remaining environmental component of phenotypic variation rather than directly on the genetic component, limiting an evolutionary response (e.g., Alatalo et al. 1990). Long-term studies of wild bird populations have demonstrated selection acting on environmental deviations, however concurrent selection on genetic components were also evident (Merilä et al. 2001a; Garant et al. 2004).
We cannot fully assess which, if any, of these alternatives may limit evolutionary change despite directional selection on tarsus and culmen measures without a more complete understanding of the underlying quantitative genetic parameters in the population. However, morphological trends over the previous decades indicate that changes in the direction of selection at different times is a plausible mechanism to explain why traits under current directional selection have not exhibited an evolutionary shift over decades. Short series of trends in the data suggest that at times in previous years, larger bills in both juveniles and adults may have conferred a selective advantage, but reversals of direction of selection have resulted in no net change. Our selection study may have therefore coincided with a time period when having a larger bill directly improved chances of survival. There is also some indication that opposing selection directions at different life-history stages may operate. The multivariate selection coefficient for culmen width in juveniles is matched by a coefficient of similar magnitude, but opposite sign in adults, although the adult estimate is not significant due to small sample sizes. Selection on correlated traits is also possible in this system. Significant or near-significant correlational selection was noted for a limited set of traits, and depending on the genetic correlation between them, the response to selection may be limited.
ENVIRONMENTAL INFLUENCES ON MORPHOLOGY
The proximate environmental causes of morphological fluctuations across decades and years are not easily identified. Two types of large-scale recurring environmental disturbances that possibly influence the pattern of morphological fluctuations are tropical cyclones and El Niño events. Tropical cyclones may act as a selective agent, with a particular phenotype gaining a selective advantage during and immediately after a cyclone, similar to what has previously been shown to occur in birds that experience extreme and abrupt environmental events (e.g., Bumpus 1899; Brown and Brown 1998). High mortality rates from cyclones (Brook and Kikkawa 1998; McCallum et al. 2000) could also provide conditions conducive to rapid, drift-mediated phenotypic shifts. El Niño events are less abrupt, lasting from months to years, and lead to lower than expected average rainfall with severe events characterized by extreme and extended drought conditions. However, despite the potential for either cyclones or El Niño events to affect morphological fluctuations in the population, we found little evidence that either type of event produced a consistent effect in the Heron Island silvereye population. Surviving a cyclone may simply be due to chance but drift may not have an opportunity to affect the phenotypic distribution due to rapid recovery from postcyclone population decreases (McCallum et al. 2000). El Niño events vary in their severity and therefore, it may be unreasonable to expect consistent impacts across separate events. A more detailed analysis of survival at these times would be required to fully assess the impact of these complex environmental events, as has been done for Darwin's finches on the Galápagos Islands where selective episodes were driven by El Niño cycles (Grant and Grant 1993).
Environmental fluctuations that result in an interplay of factors affecting food abundance and population density (Brook and Kikkawa 1998; McCallum et al. 2000) are a likely explanation for the variation in survival probability and selection for different ages, seasons and years seen across the survival study. The age-related differences in survival may reflect successful passage through earlier selection filters, or be due to increases in competence gained through experience throughout the lifetime of an individual (Forslund and Pärt 1995). The evidence that adult birds are not released from selection, despite having passed through previous selection filters, along with changes in aggressive behavior across the lifetime of individuals (Kikkawa 1987), and shifts in juvenile and adult foraging strategies (Catterall et al. 1989; Jansen 1989) implies an important role for experience gained with age in determining survival.
Previous studies of the population have noted that when population densities were very high, larger individuals had higher overwinter survival (Kikkawa 1980; Robinson-Wolrath and Owens 2003). Under high-density conditions, links between bill size and foraging ability or efficiency, such as that found in other avian species (e.g., Boag and Grant 1981; Grant 1986; Smith 1987; Benkman 1993) may be more pronounced. Alternatively, a less direct mechanism may be at play, where larger sized individuals have an advantage in aggressive interactions and thereby control of food resources (Robinson-Wolrath and Owens 2003). In the Heron Island population, larger bill size and overall body size has been shown to be positively associated with the proportion of fights won (Kikkawa 1980), making this mechanism a contender for explaining the advantages of large bill over short timescales. However, whether fluctuations in these traits seen over decades and years represent a plastic or genetic response to selection requires additional quantitative treatment.