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
O R I G I NA L A RT I C L E doi:10.1111/j.1558-5646.2012.01749.x INTERSPECIFIC COMPETITION ALTERS NONLINEAR SELECTION ON OFFSPRING SIZE IN THE FIELD Dustin J. Marshall1,2,3 and Keyne Monro1,2 1 School of Biological Sciences, The University of Queensland, Brisbane 4072, Australia 2 School of Biological Sciences, Monash University, Clayton 3800, Australia 3 E-mail: [email protected] Received August 18, 2011 Accepted July 16, 2012 Data Archived: Dryad doi:10.5061/dryad.d6v18 Offspring size is one of the most important life-history traits with consequences for both the ecology and evolution of most organisms. Surprisingly, formal estimates of selection on offspring size are rare, and the degree to which selection (particularly nonlinear selection) varies among environments remains poorly explored. We estimate linear and nonlinear selection on offspring size, module size, and senescence rate for a sessile marine invertebrate in the field under three different intensities of interspecific competition. The intensity of competition strongly modified the strength and form of selection acting on offspring size. We found evidence for differences in nonlinear selection across the three environments. Our results suggest that the fitness returns of a given offspring size depend simultaneously on their environmental context, and on the context of other offspring traits. Offspring size effects can be more pervasive with regards to their influence on the fitness returns of other traits than previously recognized, and we suggest that the evolution of offspring size cannot be understood in isolation from other traits. Overall, variability in the form and strength of selection on offspring size in nature may reduce the efficacy of selection on offspring size and maintain variation in this trait. KEY WORDS: Egg size, life-history evolution, maternal effect. Offspring size is a life-history trait of fundamental importance that affects the ecology and evolution of many organisms. Estimating the relationship between offspring size and fitness has been a major goal of evolutionary ecology, and over the past 70 years, we have accumulated a wealth of estimates of the offspring size– fitness relationship (Lack 1947; Bagenal 1969; Stanton 1984; Williams 1994; Bernardo 1996; Marshall and Keough 2008a). Despite this intense research effort, our understanding of the selective forces acting on offspring size remains surprisingly incomplete. For example, while early theory predicted a single, optimal offspring size for any one species (Smith and Fretwell 1974), in nature, offspring size typically varies among populations, mothers, broods, and offspring (Fox and Czesak 2000; Marshall and Keough 2008b; Krist 2011). The maintenance of offspring size 328 variation at all levels of biological organization is thought to be due to environmentally driven variation in selection on offspring size (Einum and Fleming 2002), but this variation in selection (formally defined) has received relatively little attention. The effect of offspring size on offspring performance appears to be highly context dependent. A range of abiotic factors (e.g., temperature) and biotic factors (e.g., competition, predation) can alter the fitness benefits of a particular offspring size (Bernardo 1996; Marshall and Keough 2008a). Yet, despite the increasing number of studies that examine the influence of the local environment on offspring size effects, there are very few formal estimates of selection on offspring size (Einum and Fleming 2000; Dibattista et al. 2007; Johnson et al. 2010), and even fewer that estimate how selection changes across environments. Meanwhile, much of the C 2012 The Society for the Study of Evolution. 2012 The Author(s). Evolution ! Evolution 67-2: 328–337 ! C NONLINEAR SELECTION ON OFFSPRING SIZE observed variation in offspring size is assumed to be adaptive. For example, population-level variation in offspring size is often described as an adaptive response to large-scale environmental variation (Johnston and Leggett 2002). Variation in offspring size among clutches is viewed as an adaptive response driven by mothers encountering different environments (Fox et al. 1997; Allen et al. 2008), and variation in offspring size within clutches is assumed to be an adaptive strategy to cope with environmental undertainty (i.e., bet-hedging Einum and Fleming 2002; Crean and Marshall 2009). For patterns in offspring size variation at any scale to reflect adaptation, variation in selection must exist. Importantly, environmental variation cannot only cause changes in directional (linear) selection in offspring size, it can also cause changes in nonlinear selection. The impact of environmental variation on nonlinear selection on offspring size remains completely unknown. Indeed, while the fitness returns of different offspring sizes are often modeled theoretically as a nonlinear function (Smith and Fretwell 1974), formal estimates of nonlinear selection are rare (Einum and Fleming 2000; Dibattista et al. 2007). Nonlinear selection on offspring size is important because it includes quadratic selection and correlational selection, both of which are likely to be of key importance for an understanding of offspring size evolution. Quadratic selection occurs when either extreme trait values are favored (known as disruptive selection) or intermediate trait values are favored (known as stabilizing selection). In his influential review, Bernardo (1996) predicted that selection on offspring size was likely to be stabilizing, but lamented that tests were lacking. Recent theoretical considerations of maternal bethedging strategies have re-emphasized the importance of estimating quadratic selection on offspring size (Crean and Marshall 2009). Under some circumstances, diversified bet-hedging with regards to offspring size is predicted to be favored, but only when offspring size is subject to stabilizing selection, otherwise, conservative bet-hedging strategies should be favored (Einum and Fleming 2004; Marshall et al. 2008). Those studies that have looked for evidence of stabilizing selection on offspring size have found it (Einum and Fleming 2000; Dibattista et al. 2007), but few studies have examined how environmental variation affects quadratic selection on offspring size. Correlational selection occurs when selection acts on the covariance between traits in combination rather than on individual traits alone (Blows and Brooks 2003). Estimating correlational selection on offspring size and other traits is particularly important because the benefits of offspring size are unlikely to be independent of the context of other traits. In other words, the fitness consequences of increased offspring size are likely to depend on other traits that the offspring expresses. Ignoring the potential for correlational selection on combinations of offspring size and other offspring traits could lead to erroneous conclusions about the benefits of increased offspring size, either overestimating or underestimating these benefits (depending on the distribution of values of other traits in the population). Overall, there is a clear need to examine whether selection (both linear and nonlinear) on offspring size varies among environments. Here, we address this important knowledge gap by applying a well-known multiple regression approach to estimate both directional and nonlinear selection on offspring size, and two other traits: a morphological trait (zooid size) and a life-history trait (rate of colony senescence) in a colonial marine invertebrate (see Materials and Methods for a description of these traits). These two traits were selected because they are important determinants of performance in colonial organisms. In marine benthic communities, interspecific competition can be intense—space is a limiting resource and overgrowth by other species can result in mortality and reduced fecundity (Hart and Marshall 2009). We therefore manipulated the level of interspecific competition that our focal colonies experienced with three treatments: the absence of interspecific competition, the presence of moderate interspecific competition, and in the presence of intense interspecific competition. Materials and Methods STUDY SPECIES AND SITE Watersipora subtorquata is an encrusting bryozoan and an abundant member of the fouling community on man-made structures along the coast of Australia. Watersipora filters planktonic food from the water column, and as a colony, it grows outwards with the leading edge of zooids (the feeding units that make up the colony) growing, with a middle band of reproductive zooids, and an inner region of senescing and senesced zooids. It broods its larvae for approximately 1 week, whereupon the nonfeeding larvae are released and spend only minutes to hours in the plankton before settling and metamorphosing (Marshall and Keough 2004). Previous studies on this species have shown that offspring size can strongly affect postmetamorphic performance, but the effects of offspring size are highly variable over large spatial scales (Marshall and Keough 2004, 2008b, 2009). Our experiment was done at Redcliffe Marina, Brisbane, Australia in Austral Spring– Summer, 2008. Reproductively mature Watersipora colonies were collected from Redcliffe, transported to the laboratory, and induced to spawn after a being kept in constant darkness in lightproof aquaria for 72 h. We collected larvae from at least 45 colonies and took care to collect larvae of a range of sizes from each colony though we did not track which larva came from which colony. We collected the spawned larvae, settled them on pieces of acetate sheeting in a drop of seawater, and then allowed the settlers to EVOLUTION FEBRUARY 2013 329 D. J. M A R S H A L L A N D K . M O N RO metamorphose for 48 h in a constant temperature cabinet at 23◦ C. The metamorphosed settlers, known as ancestrulae, were then measured using standard techniques under a dissecting microscope (Marshall and Keough 2008b), and the acetate sheets (with the settlers attached) were then glued to PVC settlement plates (100 × 100 × 8 mm). The range of offspring sizes used in this study reflected the natural variation in offspring size displayed by this species in the field (Marshall and Keough 2008b). TRAIT AND FITNESS MEASUREMENTS We selected two traits along with offspring size that are thought to be important determinants of performance in modular organisms—senescence rate and module size. We chose these traits in particular as we hypothesized that each could influence selection on offspring size. Offspring size We used size of settlers approximately 12 h after settlement as our measure of offspring size. Settler size is highly correlated with larval size in Watersipora, and the correlation is independent of the length of the larval period (Marshall and Keough 2004), such that offspring size can be reliably inferred from measurements of newly settled juveniles. Zooid size Zooid size is an important morphological trait in bryozoans as it can affect a range of life-history parameters in modular organisms (McKinney and Jackson 1991). For example, larger modules may acquire more resources but take more resources to produce in the first place. After 4 weeks in the field, we returned the colonies to the laboratory and measured the size of zooids within the colony by photographing each colony under a dissecting microscope. We measured zooid size at one time only because collecting these data is extremely time consuming (every zooid was measured to generate a mean zooid size for the colony), and we wanted to minimize the time that the colonies spent in the laboratory. Pilot studies suggested that zooid size remains relatively constant over time in Watersipora. Senescence Senescence, where portions of the colony become inactive or undergo apoptosis, is common in many bryozoans, but the adaptive significance of senescence in this group generally remains unclear (McKinney and Jackson 1991; Orive 1995). In Watersipora, as the colony ages, the central (i.e., oldest) portions of the colony senesce, and the senescing region slowly spreads from the center (Hart and Keough 2009). The rate at which senescence occurs varies among colonies and may be related to how quickly resources are transferred from senescing zooids to reproduction, given that the zooids immediately adjacent to the senescing re330 EVOLUTION FEBRUARY 2013 gion tend to be those that are reproductive at that time (Hart and Keough 2009). Thus in Watersipora, senescence has been described as an adaptive strategy by which the least productive zooids in the colony (i.e., those in the center of the colony that access water that has been depleted of planktonic food) senesce and free up resources that are used for reproduction elsewhere (Hart and Keough 2009). Zooids that have senesced are easily distinguished from zooids that remain alive, with senesced regions clearly visible as black regions on the colony. We calculated the proportion of the colony that had senesced at the end of the experimental period as our measure of senescence. Fitness We estimated reproduction weekly and defined our measure of individual fitness as the total number of offspring produced by each colony over the entire experimental period. We recognize that this estimate of fitness probably does not encompass the entire reproductive output of the colonies but we are confident that this estimate is a good and most practical predictor of true fitness. Importantly, estimates of fecundity across the study period were correlated positively (correlation coefficients among weeks ranged between 0.484 and 0.74), suggesting that there was no trade-off between early and late reproduction. Zooids that are brooding larvae are also clearly distinguishable in Watersipora and reproduction can be estimated based on digital photographs of the colony. Watersipora colonies brood their offspring for less than a week in Brisbane, and brooded larvae that are identifiable in 1 week have been released from the colony by the subsequent week. COMPETITION TREATMENTS AND FIELD DEPLOYMENT New settlers were numbered and randomly assigned to one of three treatments: competition-free, intermediate competition, and high competition. Settlers assigned to the competition-free environment were attached to blank settlement plates that were cleaned weekly throughout the experiment to remove any organisms that had naturally colonized the plates in the field. Settlers assigned to the intermediate competition environment were attached to blank settlement plates that were allowed to accumulate settlement from other organisms (including tubeworms, ascidians, bryozoans, and sponges) throughout the experimental period. Settlers assigned to the high competition environment were attached to settlement plates on which a fouling community had been allowed to develop by placing the plates into the field for 3 weeks prior to the initiation of the experiment. Previous studies show that these treatment conditions reflect the range of natural conditions and are known to affect performance in this species (Marshall and Keough 2009). In this study, our treatments resulted in very different growing conditions for our focal colonies: individuals in the NONLINEAR SELECTION ON OFFSPRING SIZE competition-free environment were the largest and most fecund at the end of the experiment, and individuals in the high competition environment were the smallest and the least fecund (see Results and Figures S1 and S2). Note that because the treatments were applied at the scale of settlement plates, we had 60 independent replicates of each treatment. Once the settlers were attached onto the settlement plates, we assigned settlement plates haphazardly to PVC backing plates (400 × 400 × 12 mm), taking care to ensure that there was an equal number of plates from each treatment per backing plate. We then deployed the backing plates into the field by suspending the settlement plates facedown 1 m below the water surface (for a detailed description of the field deployment, see Marshall and Keough 2009). Each week, plates within each backing panel were rearranged haphazardly so as to preclude artefactual covariance between the values of our focal traits, local environmental conditions, and fitness (Rausher 1992). As our focal traits varied at the scale of plate, this was the appropriate scale of rearrangement. We deployed a total of 180 settlers (60 in each treatment) across 30 backing plates. We monitored the survival, growth, senescence rate, and reproduction of our focal colonies weekly for 6 weeks in the field by photographing each colony with a digital camera in the field. On a few occasions (n = 9), we were unable to adequately discern colony size, reproductive status, or senescence rate from our photographs and as such, some measures for some colonies were omitted from the final analyses. None of the traits was correlated with another (correlation coefficients varied between 0.02 and 0.09), precluding the possibility of indirect selection being generated by one trait on another. advantage of generating standardized estimates of selection that can be later used in meta-analyses of selection to great effect (Kingsolver et al. 2001). Canonical analyses essentially extend the traditional multiple regression approach and they have the advantage of providing a more complete picture of selection on trait combinations, and can be more sensitive than the standard multiple regression approach (Blows and Brooks 2003). The outputs of canonical analyses are more difficult to interpret and are less amenable to making comparisons across studies, however, and so we present results from both the multiple regression analyses and the canonical analyses. We used partial F tests to test for differences in the strength and form of selection experienced by individuals in the different environments, following the model reduction approach described in Chenoweth and Blows (2005). ESTIMATING SELECTION GRADIENTS Standardized linear (β) and nonlinear (γ) selection gradients were estimated in each environment by regressing relative fitness (relative to the population mean in each environment) on standardized phenotypic traits. Importantly, we used separate models to estimate linear and nonlinear gradients as appropriate and we calculated the appropriate coefficient values for each type of selection (Stinchcombe et al. 2008). We then used randomization tests of significance of all gradients as such tests avoid the problems of parametric tests on non-normal distributions (Reynolds et al. 2010) when transformation of fitness is inadvisable. The results of these analyses are presented in Figure S3. NONLINEAR SELECTION DATA ANALYSIS We first determined whether there was an effect of the competitive environment that colonies experienced on colony size and colony reproductive output with repeated measures ANOVA. We then explored the effect of including the random factor "Backing plate" in our analyses, but there were no significant interactions between the random factor and the fixed factor of interest and backing plate explained little of the variation so Backing plate was excluded from the final model. There was very little mortality (<10%) across the experimental period, and the little mortality that did occur was unrelated to our traits of interest, so we did not analyze mortality formally. We found strong effects of our treatments on the performance of our focal colonies; we also found that some mean trait values were affected by our treatments (see Results and Figures S1 and S2). We estimated selection on our traits of interest with a multiple regression approach (Lande and Arnold 1983), along with an extension of this approach as advocated by Phillips and Arnold (1989). The multiple regression approach is probably more familiar and intuitively appealing to many readers, and it has the It can be difficult to interpret the strength and significance of nonlinear selection from examination of the γ matrix alone, as nonlinear selection is frequently strongest on trait combinations (Phillips and Arnold 1989; Blows and Brooks 2003). Therefore, to assess multivariate selection on traits, we performed a canonical rotation of the γ matrix. Canonical analysis determines the normalized eigenvectors (mi ), which contain the loadings of the original traits on canonical axes, and their associated eigenvalues (λi ), which describe the form of nonlinear selection (Phillips and Arnold 1989; Blows and Brooks 2003). Positive eigenvalues indicate concave selection along eigenvectors (which may indicate disruptive selection), and negative eigenvalues indicate convex selection (which may indicate stabilizing selection). The significance of eigenvalues was tested using the double linear regression method of Bisgaard and Ankenman (1996), with Pvalues calculated from permutation tests (Reynolds et al. 2010). We used thin-plate splines (with a smoothing parameter that minimized the generalized cross validation score) to visualize the major axes of the fitness surface identified by canonical analyses (Blows and Brooks 2003). As only significant eigenvectors EVOLUTION FEBRUARY 2013 331 D. J. M A R S H A L L A N D K . M O N RO were visualized, the fitness surface was plotted on a single axis for the competition-free environment, and on two axes for the intermediate- and high-competition environments. It is important to note that the classic regression approach to measuring selection assumes that all focal traits affect fitness directly. Thus, we also used path analyses of selection (see Scheiner et al. 2000) to explore a different causal structure among our focal traits—namely, that offspring size and senescence affect fitness directly, but that offspring size and zooid size also affect fitness indirectly via their impacts on senescence. We restricted these analyses to linear selection only, given the complexity of modeling indirect nonlinear effects within this framework. Because comparisons of model fit in each competitive environment found little support for path models over regression models (based on differences in model AIC scores, converted to P-values to aid interpretation; Anderson 2008), we present these results in Figure S3 only. Results EFFECTS OF THE ENVIRONMENT ON PERFORMANCE The environment that colonies experienced had a strong effect on both the growth rates and the total reproductive output of the colonies (Effects on growth: Treatment: F 2,160 = 90.2, P < 0.001; Treatment × Time: F 12,960 = 78.15, P < 0.001; Figure S1; Effect on reproduction: F 2,160 = 59.67, P < 0.001; Figure S2A). Colonies that experienced no competition grew and reproduced the most while colonies that experienced competition from an established community grew and reproduced the least. There was a significant difference in the amount of colony senescence among the different treatments: colonies experiencing high competition had a higher proportion of senesced zooids than colonies experiencing intermediate or no competition (F 2,131 = 3.73, P = 0.0265; Figure S2B). There was no significance difference in either the size of zooids or initial offspring size among the treatments (Zooid size: F 2,131 = 0.3024, P = 0.739; Offspring size: F 2,131 = 1.02, P = 0.361). DIFFERENCES IN SELECTION AMONG ENVIRONMENTS There was marginal evidence for linear selection varying among environments (F 3,157 = 2.25, P = 0.084). Explorations with ANCOVA suggested that linear selection on offspring size differed among environments (F 2,155 = 3.23, P = 0.042), but no differences in selection on the other two traits across environments. We found strong evidence that nonlinear selection varied among environments (F 6,139 = 2.52, P = 0.023). Further analyses suggested that quadratic selection differed among environments and 332 EVOLUTION FEBRUARY 2013 Reproductive output PATH ANALYSES 1.5 1.0 0.5 0.0 -2 -1 0 1 2 3 4 M3 Less senescence Larger zooid size More senescence Smaller zooid size Figure 1. Fitness of Watersipora subtorquata colonies in a competition-free environment showing multivariate selection on phenotypic trait combinations. Individuals with intermediate values of offspring size, zooid size, and senescence rate had the highest fecundity. Line represents best quadratic approximations of selection based on multivariate canonical analyses. marginal evidence for differences in correlational selection among environments (quadratic: F 3,145 = 3.06, P = 0.030; correlational: F 3,145 = 2.35, P = 0.074). Exploration with ANCOVA suggested that this difference in quadratic selection among environments was driven entirely by differences in quadratic selection on offspring size among environments (F 2,145 = 3.93, P = 0.022). Inspection of the data and the β and γ matrices suggested that there was very little selection on our traits of interest in the most benign environment, but there was strong selection in the other two environments (Table 1). In the intermediate environment, there was strong stabilizing selection while in the harshest environment, positive directional selection acted on offspring size. Our canonical analyses revealed a complex picture of selection on our traits of interest. We detected significant convex selection acting on one eigenvector (m3) of trait combinations in the competition-free environment (Table 2). In this environment, senescence loads positively, offspring size loads very little, and zooid size loads negatively, and the visualization of the fitness surface along this canonical axis suggests intermediate values of zooid size and senescence yield the highest fitness (Figure 1). In the intermediate-competition environment, strong convex selection acted along two eigenvectors (Figure 2): on one NONLINEAR SELECTION ON OFFSPRING SIZE Table 1. Standardized gradients of directional selection (β) and nonlinear selection (γ) on offspring size, senescence rate, and zooid size across competition environments in Watersipora subtorquata. Quadratic and correlational selection gradients are the diagonal and off-diagonal elements of γ respectively, errors shown in parentheses below each estimate (P < 0.05 shown in bold). γ Competition free Offspring size Senescence rate Zooid size Intermediate competition Offspring size Senescence rate Zooid size High competition Offspring size Senescence rate Zooid size Table 2. β Offspring size Senescence rate Zooid size 0.093 (0.074) −0.021 (0.079) −0.036 (0.080) 0.001 (0.058) −0.050 (0.093) −0.090 (0.016) −0.078 (0.065) 0.086 (0.0880) −0.053 (0.069) 0.038 (0.133) −0.003 (0.131) 0.061 (0.134) −0.205 (0.093) 0.015 (0.148) 0.016 (0.177) −0.118 (0.089) −0.079 (0.142) −0.129 (0.091) 0.422 (0.157) −0.162 (0.151) −0.150 (0.156) 0.140 (0.129) −0.388 (0.209) −0.111 (0.167) 0.0107 (0.101) −0.180 (0.197) −0.064 (0.113) Matrix of eigenvectors (mi ) and their associated eigenvalues (λ) from the canonical analysis of γ for total reproductive output, under different levels of competition (P < 0.05 shown in bold). Selection λ (Std. dev.) Competition free m1 m2 m3 Intermediate competition m1 m2 m3 High competition m1 m2 m3 Original traits Offspring size Senescence rate Zooid size 0.042 (0.097) −0.061 (0.123) −0.111 (0.098) 0.782 0.617 0.084 −0.345 0.542 −0.765 −0.518 0.569 0.637 −0.084 (0.083) −0.161 (0.085) −0.208 (0.066) 0.002 0.234 0.972 0.753 0.639 −0.156 −0.658 0.732 −0.175 0.280 (0.147) 0.017 (0.146) −0.211 (0.111) 0.806 −0.372 0.460 −0.591 −0.466 0.659 0.031 0.802 0.596 eigenvector (m2), offspring size loads lightly while both senescence and zooid size load heavily; on the other eigenvector, offspring size loads very heavily while other traits load very little (Table 2; Figure 2). In the high-competition environment, canonical analysis showed both significant concave and convex selection acting along two eigenvectors of trait combinations (m1 and m3 respectively, Table 2). On the first eigenvector, offspring size loads positively and strongly while senescence rate loads negatively and zooid size loads very little. The fitness surface suggests that larger offspring sizes and low senescence rates yield the highest fitness returns (Figure 3). On the third eigenvector, all three EVOLUTION FEBRUARY 2013 333 D. J. M A R S H A L L A N D K . M O N RO Discussion The strength and form of selection on offspring size varied among the different environments in the field: nonlinear selection on offspring size was strongest in the harsher environments and weak to nonexistent in the most benign environment. Our results also suggest that failing to consider nonlinear selection could lead to an incomplete understanding of the relationship between offspring size and fitness. Larger offspring size Less senescence Smaller zooids Larger zooids More Senescence Smaller offspring size More senescence Larger zooids Smaller zooids Less senescence Figure 2. Fitness of Watersipora subtorquata colonies in an intermediate-competition environments showing multivariate selection on phenotypic trait combinations. Individuals with in- termediate values of offspring size, zooid size, and senescence rate had the highest fecundity. Surface represents best quadratic approximations of selection based on multivariate canonical analyses. Larger offspring size More senescence Larger zooids Larger offspring size Less senescence Smaller offspring size Less senescence Smaller zooids Smaller offspring size More senescence Figure 3. Fitness of Watersipora subtorquata colonies in a highcompetition environments showing multivariate selection on phenotypic trait combinations. Individuals with the greatest values of offspring size, zooid size, and lowest values for senescence rate had the highest fecundity. Surface represents best quadratic approximations of selection based on multivariate canonical analyses. traits load positively and approximately equally and the fitness surface suggests that an intermediate phenotype (with regards to the linear combination of all three traits) is favored (Figure 3). Overall, in the high-competition environment, the phenotype with the highest fitness returns combined the highest values for m1 and intermediate values of m2 (but with the latter of lesser importance). 334 EVOLUTION FEBRUARY 2013 CAUSES AND CONSEQUENCES OF ENVIRONMENTAL VARIABILITY IN SELECTION ON OFFSPRING SIZE We found no nonlinear selection on offspring size in the most benign environment. Our study joins a growing list showing that under benign conditions, selection on offspring size is weaker than under harsher conditions (Mousseau and Fox 1998), but it is important to note that this pattern is not universal (Allen et al. 2008). In nature, it is probably rare for Watersipora to encounter entirely competition-free environments for sustained periods of time, so we suspect that selection in this environment only plays a minor part in shaping offspring size variation. Nevertheless, our finding does illustrate that excluding the effects of interspecific competition (as is typical of most laboratory-based studies of offspring size) can result in the dramatic underestimation of selection on offspring size. In the two harsher environments, nonlinear selection was strong but context dependent. We found evidence for stabilizing selection on offspring size in the intermediate environment and evidence for nonlinear selection for increased offspring size in the harshest environment. In the harshest environment used here, competition for food is intense and so increased size (and presumably, energy reserves) may assist offspring in coping with this competition. The detection of stabilizing selection on offspring size in the intermediate environment is harder to explain; we expected similar benefits of increased offspring size in the intermediate environment if food competition alone drove our effects. It could be that the two surrounding communities (the intermediate and the harshest environment) were composed of different species and this difference led to a difference in selection on offspring size. Regardless, this effect warrants further exploration as it was not anticipated. The detection of stabilizing selection on offspring size in two environments has some interesting implications for the way we view offspring size effects. Traditionally, it has been tempting to view the performance of larger offspring as invariably either greater than, or at least equal to, the performance of smaller offspring (Marshall et al. 2006; Jorgensen et al. 2011). However, our detection of stabilizing selection suggests that, in this species, very large offspring will have lower performance than offspring of intermediate size under some conditions. This finding is in NONLINEAR SELECTION ON OFFSPRING SIZE keeping with some studies that have similarly found that larger offspring do not always perform better and sometimes even do worse than smaller offspring (Kaplan 1992; Einum and Fleming 2000; Dibattista et al. 2007; Jacobs and Sherrard 2010). As noted in the Introduction, the form of selection on offspring size will have a critical impact on the type of bet-hedging that mothers may employ to cope with unpredictable environments (Crean and Marshall 2009). Our finding suggests that, in Watersipora, selection may favor mothers that employ a diversified bet-hedging strategy when the offspring environment is unpredictable. Importantly, stabilizing selection on offspring size does not mean that mothers should produce the particular offspring size that selection most favors (Marshall and Uller 2007). Instead, the optimal offspring size that mothers should produce will probably be product of both the offspring size–performance relationship and the trade-off between offspring size and number (Smith and Fretwell 1974; Wolf and Wade 2001; Johnson et al. 2010). To estimate selection on mothers, rather than offspring, our estimates of selection on offspring size would have to be combined with estimates of a size-number trade-off that mothers face as recently done by Johnson et al. (2010), but such data are lacking for our species. Our finding that selection strongly favored larger offspring in the harshest environment is in keeping with offspring provisioning patterns in this species. Previous work has shown that Watersipora colonies grown in the presence of a competitive community produce larger offspring than colonies grown in the absence of competitors (Marshall and Keough 2009). Our results here would suggest that this plasticity in offspring provisioning represents an "anticipatory maternal effect" (sensu Marshall and Uller 2007), whereby colonies respond to environmental cues in their own habitat and provision their offspring accordingly. It should be noted however that offspring from a single colony probably encounter both habitat types in nature because of the scale of dispersal; hence mothers face the daunting challenge of provisioning their offspring on the chance that they will encounter either environment. A number of recent reviews have highlighted the importance of variability in selection for determining the tempo and trajectory of evolution (Orr 2009; Bell 2010; Calsbeek 2012). Our results show that selection is highly variable over small spatial scales. Furthermore, our environments span the range of successional stages in the surrounding community that Watersipora will experience, sometimes within a single lifetime (recall that to create our harshest environment, we allowed a community to develop for only 3 weeks prior to starting the experiment and W. subtorquata colonies can live for twice this time). As such, our results suggest that environmentally driven variation in selection will reduce its efficacy and may maintain variance in offspring size. CORRELATION SELECTION ON OFFSPRING SIZE Offspring size effects have long been recognized as highly context dependent: the local environmental conditions strongly affect the offspring size–performance relationship (Mousseau and Dingle 1991; Bernardo 1996). Importantly, our findings suggest that offspring size is also context dependent with regards to other trait values—the fitness return of a particular offspring size depends on values of other traits. For example, canonical analysis suggested that the fitness benefits of increased offspring size are increased in individuals with lower senescence rates when they experience intense competition. Such correlational selection should, over time, result in increased negative covariance between offspring size and senescence rates (Blows and Brooks 2003). More generally, there is some evidence for covariance between offspring size and other traits at both the population level (Johnston and Leggett 2002; Bashey 2008; Walsh and Reznick 2010), and within populations (Martin and Pfennig 2010). Coevolution between offspring size and other traits has also been implicated in a range of other systems (Krug and Zimmer 2000; Walsh et al. 2006; Benton et al. 2008; Miles and Wayne 2009) suggesting offspring size often interacts with other traits to affect fitness. It would therefore be interesting to examine whether correlational selection operates on offspring size and other traits in organisms where covariance between these traits has already been detected. For example, Pfennig and Martin (2009) did not formally analyze correlational selection on offspring size and "carnivory traits" in spadefooted toads (Pfennig and Martin 2009), but we suspect that correlational selection is acting in this species, given increases in offspring size alter selection on other carnivory traits. We suggest that the prevalence of correlational selection on offspring size has been underestimated more generally. Biologists have long recognized that offspring size has pervasive longitudinal influences throughout the life history (Lack 1947). For example, some studies show that initial offspring size can affect subsequent size throughout the life history (Dias and Marshall 2010). The presence of correlational selection on offspring size in combination with other traits, however, suggests that variation in maternal investment will also generate selection on traits other than offspring size where there would otherwise be none (Blows and Brooks 2003). Thus, the influence of initial maternal investment on the phenotype of offspring may be more pervasive than has been recognized previously—affecting selection on traits that are not directly correlated with offspring size. For example, variation in initial offspring provisioning could constrain the range of values of other traits that yield a fitness benefit for offspring. This maternal effect is more insinuating than those typically discussed, and warrants further exploration. On the other hand, the presence of correlational selection also raises the possibility that the range of offspring sizes that yield high fitness EVOLUTION FEBRUARY 2013 335 D. J. M A R S H A L L A N D K . M O N RO returns is more constrained than has been recognized. If the fitness benefits of increased offspring size are highly dependent on the values of other traits (as they are when correlational selection occurs), mothers may evolve to provision offspring according to the values of other offspring traits. We found evidence for correlational selection on senescence in conjunction with other traits, but this selection depended on the local environment. In environments with less competition, colonies that senesced more had higher reproductive output than colonies that senesced very little, whereas in the harshest environment, this pattern was reversed. While there has been much debate regarding the evolution of senescence in clonal organisms and marine invertebrates in particular (McKinney and Jackson 1991; Chadwickfurman and Weissman 1995; Orive 1995; Bayer and Todd 1997; Gardner and Mangel 1997), there are few empirical studies of senescence, and as far as we are aware, no formal estimates of the selective value of senescence in a marine invertebrate. It appears that some senescence can benefit colony fitness, but that the benefits of senescence are highly context dependent— both the context of other life-history traits (i.e., offspring size) and local environmental conditions (intensity of competition) can alter the fitness returns of senescence. An interesting next step will be to determine whether genetic trade-offs between senescence and other components of fitness exist in this species. Overall, we found a complex web of selection acting on offspring size that varied among environments. Our results suggest that further examinations of variability in selection in the field are required as our results suggest that selection can vary over remarkably small spatial and temporal scales. We suspect that nonlinear selection on offspring size and other offspring traits is common but without more studies, it remains impossible to know its prevalence. ACKNOWLEDGMENTS The authors wish to thank the K. Donohue, C. Fox, and two anonymous reviewers for comments that greatly improved the manuscript. We thank S. Chenoweth for statistical advice. We thank Tane Sinclair-Taylor for assistance in the field and laboratory. DJM and KM were supported by grants and fellowships from the Australian Research Council. LITERATURE CITED Allen, R. M., Y. M. Buckley, and D. J. Marshall. 2008. Offspring size plasticity in response to intraspecific competition: an adaptive maternal effect across life-history stages. Am. Nat. 171:225–237. Anderson, D. R. 2008. Model based inference in the life sciences: a primer on evidence. Springer, New York. Bagenal, T. B. 1969. Relationship between egg size and fry survival in brown trout Salmo trutta L. J. Fish. Biol. 1:349–353. Bashey, F. 2008. Competition as a selective mechanism for larger offspring size in guppies. Oikos 117:104–113. Bayer, M. M. and C. D. Todd. 1997. Evidence for zooid senescence in the marine bryozoan Electra pilosa. Invertebr. Biol. 116:331–340. 336 EVOLUTION FEBRUARY 2013 Bell, G. 2010. Fluctuating selection: the perpetual renewal of adaptation in variable environments. Philos. Trans. R. Soc. Lond. B 365:87–97. Benton, T. G., J. J. H. St Clair, and S. J. Plaistow. 2008. Maternal effects mediated by maternal age: from life histories to population dynamics. J. Anim. Ecol. 77:1038–1046. Bernardo, J. 1996. The particular maternal effect of propagule size, especially egg size: patterns models, quality of evidence and interpretations. Amer. Zool. 36:216–236. Bisgaard, S. and B. Ankenman. 1996. Standard errors for the eigenvalues in second-order response surface models. Technometrics 38:238–246. Blows, M. W. and R. Brooks. 2003. Measuring nonlinear selection. Am. Nat. 162:815–820. Calsbeek, B. 2012. Exploring variation in fitness surfaces over time or space. Evolution 66:1126–1137. Chadwickfurman, N. E. and I. L. Weissman. 1995. Life histories and senescence of Botryllus schlosseri (Chordata, Ascidiacea) in Monterey Bay. Biol. Bull. 189:36–41. Chenoweth, S. F. and M. W. Blows. 2005. Contrasting mutual sexual selection on homologous signal traits in Drosophila serrata. Am. Nat. 165: 281–289. Crean, A. J. and D. J. Marshall. 2009. Coping with environmental uncertainty: dynamic bet-hedging as a maternal effect. Philos. Trans. R. Soc. Lond. B 364:1087–1096. Dias, G. M. and D. J. Marshall. 2010. Does the relationship between offspring size and performance change across the life-history? Oikos 119:154– 162. Dibattista, J. D., K. A. Feldheim, S. H. Gruber, and A. P. Hendry. 2007. When bigger is not better: selection against large size, high condition and fast growth in juvenile lemon sharks. J. Evol. Biol. 20:201–212. Einum, S. and I. A. Fleming. 2000. Highly fecund mothers sacrifice offspring survival to maximize fitness. Nature 405:565–567. ———. 2002. Does within-population variation in fish egg size reflect maternal influences on optimal values. Am. Nat. 160:756–765. ———. 2004. Environmental unpredictability and offspring size: conservative versus diversified bet-hedging. Evol. Ecol. Res. 6:443–455. Fox, C. W. and M. E. Czesak. 2000. Evolutionary ecology of progeny size in arthropods. Annu. Rev. Entomol. 45:341–369. Fox, C. W., M. S. Thakar, and T. A. Mosseau. 1997. Egg size plasticity in a seed beetle: an adaptive maternal effect. Am. Nat. 149:149–163. Gardner, S. N. and M. Mangel. 1997. Can a clonal organism escape senescence? Am. Nat. 150:462–490. Hart, S. P. and M. J. Keough. 2009. Does size predict demographic fate? Modular demography and constraints on growth determine response to decreases in size. Ecology 90:1670–1678. Hart, S. P. and D. J. Marshall. 2009. Spatial arrangement affects population dynamics and competition independent of community composition. Ecology 90:1485–1491. Jacobs, M. W. and K. M. Sherrard. 2010. Bigger is not always better: offspring size does not predict growth or survival for seven ascidian species. Ecology 91:3598–3608. Johnson, D. W., M. R. Christie, and J. Moye. 2010. Quantifying evolutionary potential of marine fish larvae: heritability, selection, and evolutionary constraints. Evolution 64:2614–2628. Johnston, T. A. and W. C. Leggett. 2002. Maternal and environmental gradients in the egg size of an iteroparous fish. Ecology 83:1777– 1791. Jorgensen, C., S. K. Auer, and D. N. Reznick. 2011. A model for optimal offspring size in fish, including live-bearing and parental effects. Am. Nat. 177:E119–E135. Kaplan, R. H. 1992. Greater maternal investment can decrease offspring surivival in the frog Bombina orientalis. Ecology 73:280–288. NONLINEAR SELECTION ON OFFSPRING SIZE Kingsolver, J. G., H. E. Hoekstra, J. M. Hoekstra, D. Berrigan, S. N. Vignieri, C. E. Hill, A. Hoang, P. Gibert, and P. Beerli. 2001. The strength of phenotypic selection in natural populations. Am. Nat. 157:245–261. Krist, M. 2011. Egg size and offspring quality: a meta-analysis in birds. Biol. Rev. 86:692–716. Krug, P. J. and R. K. Zimmer. 2000. Developmental dimorphism and expression of chemosensory-mediated behavior: habitat selection by a specialist marine herbivore. J. Exp. Biol. 203:1741–1754. Lack, D. 1947. The significance of clutch size. Ibis 89:302–352. Lande, R. and S. J. Arnold. 1983. The measurement of selection on correlated characters. Evolution 37:1210–1226. Marshall, D. J., R. Bonduriansky, and L. F. Bussiere. 2008. Offspring size variation within broods as a bet-hedging strategy in unpredictable environments. Ecology 89:2506–2517. Marshall, D. J. and M. J. Keough. 2004. Variable effects of larval size on postmetamorphic performance in the field. Mar. Ecol. Prog. Ser. 279:73–80. ———. 2008a. The evolutionary ecology of offspring size in marine invertebrates. Adv. Mar. Biol. 53:1–60. ———. 2008b. The relationship between offspring size and performance in the sea. Am. Nat. 171:214–224. ———. 2009. Does interspecific competition affect offspring provisioning? Ecology 90:487–495. Marshall, D. J. and T. Uller. 2007. When is a maternal effect adaptive? Oikos 116:1957–1963. Marshall, D. J., C. N. Cook, and R. B. Emlet. 2006. Offspring size effects mediate competitive interactions in a colonial marine invertebrate. Ecology 87:214–225. Martin, R. A. and D. W. Pfennig. 2010. Maternal investment influences expression of resource polymorphism in amphibians: implications for the evolution of novel resource-use phenotypes. PLoS One 5:7. McKinney, F. K. and J. B. C. Jackson. 1991. Bryozoan evolution. University of Chicago Press, Chicago. Miles, C. M. and M. L. Wayne. 2009. Life history trade-offs and response to selection on egg size in the polychaete worm Hydroides elegans. Genetica 135:289–298. Mousseau, T. A. and H. Dingle. 1991. Maternal effects in insect life histories. Annu. Rev. Entomol. 36:511–534. Mousseau, T. A. and C. W. Fox. 1998. The adaptive significance of maternal effects. Trends Ecol. Evol. 13:403–407. Orive, M. E. 1995. Senescence in organisms with clonal reproduction and complex life histories. Am. Nat. 145:90–108. Orr, H. A. 2009. Fitness and its role in evolutionary genetics. Nat. Rev. Genet. 10:531–539. Pfennig, D. W. and R. A. Martin. 2009. A maternal effect mediates rapid population divergence and character displacement in spadefoot toads. Evolution 63:898–909. Phillips, P. C. and S. J. Arnold. 1989. Visualizing multivariate selection. Evolution 43:1209–1222. Rausher, M. D. 1992. The measurement of selection on quantitative traits: biases due to environmental covariances between traits and fitness. Evolution 46:616–626. Reynolds, R. J., D. K. Childers, and N. M. Pajewski. 2010. The distribution and hypothesis testing of eigenvalues from the canonical analysis of the gamma matrix of quadratic and correlational selection gradients. Evolution 64:1076–1085. Scheiner, S., R. J. Mitchell, and H. S. Callahan. 2000. Using path analysis to measure natural selection. J. Evol. Biol. 13:423–433. Smith, C. C. and S. D. Fretwell. 1974. The optimal balance between size and number of offspring. Am. Nat. 108:499–506. Stanton, M. L. 1984. Seed variation in wild radish: effect of seed size on components of seedling and adult fitness. Ecology 65:1105–1112. Stinchcombe, J. R., A. F. Agrawal, P. A. Hohenlohe, S. J. Arnold, and M. W. Blows. 2008. Estimating nonlinear selection gradients using quadratic regression coefficients: double or nothing? Evolution 62:2435– 2440. Walsh, M. R. and D. N. Reznick. 2010. Influence of the indirect effects of guppies on life history evolution in Rivulus hartii. Evolution 64:1583– 1593. Walsh, M. R., S. B. Munch, S. Chiba, and D. O. Conover. 2006. Maladaptive changes in multiple traits caused by fishing: impediments to population recovery. Ecol. Lett. 9:142–148. Williams, T. D. 1994. Intraspecific variation in egg size and egg composition in birds: effects on offspring fitness. Biol. Rev. 68:38–59. Wolf, J. B. and M. J. Wade. 2001. On the assignment of fitness to parents and offspring: whose fitness is it and when does it matter? J. Evol. Biol. 14:347–356. Associate Editor: K. Donohue Supporting Information The following supporting information is available for this article: APPENDIX Figure S1. Effect of competition on colony size in Watersipora subtorquata in the field. Figure S2. Effect of competition on (A) total reproductive output; and (B) proportion of colony that had senesced in Watersipora subtorquata. Figure S3. Comparisons of regression models versus path models of selection in different competitive environments. Supporting Information may be found in the online version of this article. Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article. EVOLUTION FEBRUARY 2013 337 Colony"area" (mm2)" 1500" Low" 1000" Intermediate" 500" High" 0" 1" 2" 3" 4" 5" 6" 7" Week" Figure"2)"Effect"of"compeGGon"on"colony"size"in"Watersipora*subtorquata* in"the"field." 500" Total"ReproducGon" a)" 300" 100" b)" ProporGon"senesced" 0.15" 0.10" 0.05" Low" Intermediate" High" Figure"2)"Effect"of"compeGGon"on"a)"total"reproducGve"output;"and""b)" proporGon"of"colony"that"had"senesced"in"Watersipora*subtorquata" ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! Appendix!2!Figure!1.!!Comparisons!of!regression!models!vs#path!models!of!selection!in! different!competitive!environments.!In!regression!models,!paths!(single>arrow!lines)!show! the!direct!effects!(black!lines!and!coefficients)!of!offspring!size,!colony!senescence!and!colony! zooid!size!on!fitness,!given!trait!variances!(V)!and!covariances!(CV,!double>arrow!lines).!In! path!models,!paths!show!the!direct!effects!of!offspring!size!and!senescence!on!fitness,!plus!the! indirect!effects!(grey!lines!and!coefficients)!of!offspring!size!and!zooid!size!on!fitness,!acting! via!their!direct!effects!on!senescence!(with!paths!now!replacing!covariances!between!these! traits).!!Significant!coefficients!are!in!bold.!For!all!environments,!differences!in!the!fit!of! alternative!models!(measured!by!changes!in!their!AIC!scores,!converted!to!P>values!to!aid! interpretation)!are!minimal,!indicating!similar!support!for!each!model!type.!