Common garden experiments: measuring extant natural genetic variation in adaptive traits
The first question that must be answered with respect to fishery-induced evolution is ‘what traits are capable of evolving?’ This question can be answered with a comparative approach that determines whether life history traits display adaptive genetic variation among extant stocks of a given species and, if so, which ones and in what manner? To be certain that any genetic variation measured is truly adaptive and not the result of stochastic processes such as drift, this approach works best when comparing variation across multiple locations spanning strong environmental gradients (Endler 1986). The process involves (i) asking to what extent phenotypic traits vary across an environmental gradient, (ii) determining whether trait plasticity or genetic differentiation is the source of this variability, and (iii) identifying the causal agents of selection. Simply put, only if we can demonstrate how and why traits have evolved in response to natural selection can we gain an understanding of how they might evolve in response to an agent of selection imposed by humans.
‘Common garden’ experiments play a crucial role in revealing adaptive genetic variation in the wild because they disentangle environmental from genetic influences across the gradient. In common garden experiments, offspring from different populations are reared under identical environmental conditions. Any among-population differences in phenotypes that persist under common garden conditions must be genetic and would thus prove that wild populations differ genetically. If such genetic variation is strongly correlated with environmental gradients, then it likely represents local adaptation, thus demonstrating that the traits in question are capable of evolving. Within the past decade, common garden approaches have been applied widely to many taxa (Conover et al. 2006). In fishes, such studies initially focused on isolated populations of freshwater species, while more recently they have also been expanded to many marine species. These studies have revealed that genetic adaptation to local environment conditions is common in fish populations and that the patterns of change are highly correlated with environmental gradients, e.g. latitude, temperature, seasonality, ice cover, migration costs, and predator abundance (Table 1).
Table 1. Published common garden experiments on fish (teleosts, chondrichtyes) revealing countergradient adaptations in various traits along given environmental gradients.
|Species||Common name||Trait(s)||Selection gradient||Source|
|Cynoscion nebulosus||Spotted weakfish||Larval growth rate||Temperature, season length||Smith et al. 2008|
|Fundulus heteroclitus||Mummichog||Growth rate, embryo development||Seasonality||Schultz et al. 1996; DiMichele and Westerman 1997|
|Gadus morhua||Cod||Growth rate, food conversion efficiency|| ||Purchase and Brown 2001; Salvanes et al. 2004|
|Body shape||Temperature||Marcil et al. 2006|
|Hippoglossus hippoglossus||Halibut||Growth rate, growth efficiency||Temperature||Jonassen et al. 2000|
|Lepomis gibbosus||Pumpkinseed||Growth rate, cranial ossification||Competition, predation||Arendt and Wilson 1997, 1999, 2000|
|Menidia menidia||Atlantic silverside||Metabolic rate, growth rate, swimming performance, foraging behavior Food consumption rate, growth efficiency, predator vulnerability||Season length||Arnott et al. 2006; Billerbeck et al. 2000, 2001; Chiba et al. 2007; Conover and Present 1990; Lankford et al. 2001; Munch and Conover 2003|
|Menidia peninsulae||Tidewater silverside||Growth rate||Seasonality||Yamahira and Conover 2002|
|Micropterus salmoides||Largemouth bass||Growth rate||Growing season?||Philipp and Whitt 1991|
|Morone saxatilis||Striped bass||Growth rate||Seasonality||Brown et al. 1998; Conover et al. 1997; Secor et al. 2000|
|Notropis atherinoides||Emerald shiner||Growth rate||Temperature||Pegg and Pierce 2001|
|Oncorhynchus keta||Chum salmon||Body shape||Predation, food limitation||Tallman 1986; Tallman and Healey 1991|
|Oncorhynchus nerka||Sockeye salmon||Breeding color||Sexual selection, carotenoid availability||Craig and Foote 2001; Craig et al. 2005|
|Oncorhynchus tshawytscha||Chinook salmon||Ovarian mass||Migration cost||Kinnison et al. 2001|
|Oryzias latipes||Japanese rice fish||Growth rate||Seasonality||Yamahira et al. 2007; Yamahira and Takeshi 2008|
|Poecilia reticulata||Guppy||Sexual body coloration||Carotenoid availability||Grether et al. 2005|
|Pomacentrus coelestis||Neon damselfish||Clutch size, egg size||Temperature?||Kokita 2003|
|Salmo salar||Atlantic salmon||Growth rate, digestion rate||Light/ice cover||Nicieza et al. 1994a,b; Finstad and Forseth 2006|
|Salmo trutta||Sea trout||Standard metabolic rate||River thermal regime||Alvarez et al. 2006|
|Scophthalmus maximus||Turbot||Growth rate, growth efficiency||Temperature||Imsland et al. 2000, 2001|
|Acipenser fulvescens||Lake sturgeon||Growth rate||Unknown||Power and McKinley 1997|
Two common geographical patterns have emerged from these studies. The predominant pattern is countergradient variation (CnGV), which occurs when genetic variation in a phenotypically plastic trait is distributed such that it counteracts environmental influences on that trait, thereby making phenotypes appear to be similar when in fact their genotypes are not. Such genetic divergences, which have also been termed ‘genetic compensation’ (Grether 2005), can be revealed only by common garden experiments. CnGV has so far been detected in 21 fish species, including many from marine or estuarine environments that are extensively harvested (Table 1). It is common, too, in numerous other ectotherms including reptiles, amphibians, insects, and marine invertebrates (Conover et al. 2006). Most of the finfish examples involve temperate species in which growth rate has evolved to compensate for the reduction in temperature and length of the growing season that occurs at higher latitudes. Other traits that display CnGV are those mechanistically linked to growth rate such as metabolic rate, feeding rate, growth efficiency, foraging behavior, and body shape. The agent of selection that drives these differences, at least in some species, is size-selective first winter mortality that favors larger body sizes at higher latitudes (Munch et al. 2003; Hurst and Conover 1998; see review by Hurst 2007). Conversely, when genetic variation is distributed in nature such that it accentuates environmental influences on a plastic trait, the pattern is known as cogradient variation (CoGV; not shown in Table 1). CoGV has been documented in five fish species and primarily involves morphological characters (Day et al. 1994; Robinson and Wilson 1996; Billerbeck et al. 1997; Yamahira et al. 2006; Ghalambor et al. 2007), although at least one case of CoGV in growth rate has been documented (Arendt and Reznick 2005). For further details, an extensive review of the theory, prevalence, and evolutionary significance of CnGV and CoGV across all organisms, with implications for conservation of resource species, is provided by Conover et al. (2009a).
One of the most thoroughly studied cases of CnGV in growth is the Atlantic silverside, Menidia menidia. In this species, common garden experiments have demonstrated that the genetic capacity for growth increases greatly with latitude along the east coast of North America (Conover and Present 1990). Because this countergradient pattern almost exactly counteracts the threefold decrease in length of the growing season at higher latitudes, adult body size (at age one) is nearly the same at all latitudes. The principal agent of selection is size-selective winter mortality: i.e., there is strong directional selection that favors large body sizes in northern populations (Munch et al. 2003). Faster growth is positively correlated with a suite of covarying traits that together maximize energy acquisition including increased standard metabolism (Billerbeck et al. 2000; Arnott et al. 2006), food consumption and conversion efficiency (Present and Conover 1992), and foraging activity (Chiba et al. 2007). Also displaying a positive genetic correlation with growth rate is egg production rate (Conover 1992). However, there is a cost associated with higher rates of tissue synthesis. Fast growth is negatively correlated with swimming speed (Billerbeck et al. 2001; Munch and Conover 2004) and vulnerability to predation (Lankford et al. 2001), Hence, in the north where size selective winter mortality dominates, fast growth is favored despite the trade-offs with swimming performance and predation vulnerability. At southern latitudes, on the other hand, the time constraint on growing season length and the severity of winter is reduced, while predation intensity is increased. Under these conditions, genotypes that acquire energy and grow at lower rates and therefore have higher metabolic scope for swimming and evading predators have higher fitness (Arnott et al. 2006; Chiba et al. 2007).
How does knowledge of the prevalence of CnGV in growth rate and other physiological and behavioral traits help us understand fishery-induced evolution? The answer is three-fold. First, it proves that juvenile growth rate is not generally maximized by natural selection as was previously thought by early life history theorists (see review by Arendt 1997). Instead, growth is optimized by stabilizing selection and thereby is fine-tuned to the adaptive landscape in any given habitat. When an unfished stock is exploited, the imposed mortality on adults shifts the adaptive landscape, causing selection for a new phenotypic optimum, thereby disrupting the fine-tuning between growth rate and the natural environment. Second, it proves that despite the extreme plasticity of growth and metabolism in response to environmental factors such as temperature or food level in the wild, genetic variation remains a very important component of the growth rate expressed by individuals within a population and thereby the productivity among populations. Plasticity and genetic variation are not mutually exclusive and in fact may act antagonistically (as in CnGV) or synergistically (as in CoGV). Third, it demonstrates that size selective processes such as winter mortality are capable of driving growth rate evolution in the wild. These observations set the stage for the possibility of fishery-induced evolution.
Common garden experiments have several limitations. First, in order to confidently rule out confounding environmental effects, common garden experiments need to start with the earliest ontogenetic stage of a given species (usually fertilized eggs, preferably from parents that have been maintained in a common garden). This minimizes the possibility that irreversible effects of environment on phenotypes prior to the beginning of the experiment do not persist. While this may prove extremely difficult in species with high early life mortalities and special larval food requirements (e.g., many tropical reef fishes), it has also led to a bias in literature towards traits that are expressed relatively early during ontogeny (larval and juvenile stages) and thus require relatively short rearing protocols (weeks to months). For most fish species, attempts to investigate adult traits are not feasible because of very long rearing times (years). Relatively few common garden experiments have compared size and/or age at maturity among populations, even though this trait is strongly suspected to have evolved in many exploited fish stocks (Dieckmann and Heino 2007; Jørgensen et al. 2007; Heino and Dieckmann 2008). The argument for fisheries-induced evolution of age at maturity in cod, for example, would be greatly enhanced if there was a common garden study demonstrating a genetic basis for natural variation in this trait among populations. Experimental studies of age at maturity that do exist involve very short-lived species (swordtails: Kallman and Borkoski 1978; guppies: Reznick and Ghalambor 2005; platyfish: Sohn 1977). Second and most importantly, studies of the existing level of adaptive genetic variation among wild stocks are not informative about the rate at which such traits may have evolved or will evolve in the future. We know only that the divergence occurred sometime after at least partial separation from a common ancestor which in many cases may have been thousands of generations ago. To predict the potential for evolution in the future requires knowledge of the level of additive genetic variation currently existing within populations. An important exception involves recent introductions of species to novel environments (e.g., Haugen and Vøllestad 2000, 2001; Hendry et al. 2000) or as part of a planned field translocation as further discussed below. In these cases, rates of contemporary evolution are measurable. Finally, while common garden experiments on wild populations work well within the spatial domain, they are rarely feasible in the temporal domain. It is not possible to compare the genetic basis of extant trait variation in a given population today to what it was a century ago. However, common garden experiments at different points in time have been used to measure rates of evolution in field experiments that were planned in advance (see guppy studies described below).
Fishes display an enormous diversity of life history patterns. We suspect most ecologists and evolutionists would agree that such divergent life histories likely evolved as a function of selection and adaptation operating throughout the long evolutionary history of fishes. There is valid scientific uncertainty, however, about the time frame required for such evolutionary divergences to transpire. Originally, the perception was that ecological dynamics operate on immediate to decadal time scales whereas evolutionary dynamics involve millennia, but there are now many examples of rapid evolution in nature occurring after only a few generations of selection (e.g., Hendry et al. 2000; Reznick and Ghalambor 2005). There is also uncertainty about the multivariate nature of selection. Not only must we contend with a tangled web of genetic and environmental covariance and interaction terms, but there is also a tangled web of positive and negative genetic covariances among traits that can constrain or accelerate rates of evolution. Selection experiments are the principal tool for measuring the rate of evolution of any given trait and, more importantly, the correlated evolution of trait clusters that are genetically linked to the target of selection (Fuller et al. 2005).
The main goals of a selection experiment are to (i) demonstrate that phenotypic selection on a given trait translates into genotypic selection, (ii) identify concomitant changes in correlated traits, and (iii) measure the rate of such evolutionary changes. The rate of evolutionary change is a product of the trait heritability (the extent to which phenotypes are determined by genes transmitted from parents) and the selection differential (the change in mean phenotype of parents caused by selection). The heritability is an intrinsic biological parameter while the selection differential varies with the intensity of selection which, in the case of fishing, is imposed by the harvest regime. Hence, many selection experiments purposely impose very severe selection differentials because this will require the fewest number of generations to provide an accurate measure of heritability. The rate of evolution imposed by some lesser selection differential is then easily calculated (e.g., Brown et al. 2008). Hence, the idea of fishery-induced selection experiments is not to directly mimic ‘real-world fisheries’ (at least initially) but to develop and refine theory from which testable predictions can be derived (see Benton et al. 2007).
Selection experiments on captive populations
There are two ways of carrying out selection experiments on captive populations (Fuller et al. 2005). The difference lies in the way the investigator intends to bring about phenotypic change in parental generations. In ‘artificial selection experiments,’ individual parents are directly chosen by the investigator for breeding and mated based on specific trait values. The offspring of such matings are reared and then sorted again by the investigator prior to the next breeding cycle. In such cases, selection is highly artificial so as to provide precise control over the parental phenotypes chosen for breeding. This approach is valuable primarily to test the heritability of a specific trait (e.g., coloration, number of gill rakers, etc.) and to measure correlated characters that are dragged along by selection on the targeted trait. In ‘natural selection experiments,’ on the other hand, genetically homogeneous populations are subdivided and assigned to two or more environmental treatments that differ only in the parameter of interest (e.g., temperature, density, predators, etc.). Selection is exerted by the differential effects of environmental parameters and the populations self reproduce. The investigator merely monitors the rate of divergence, if any, among populations over multiple generations. The advantage of this approach is that the responsible environmental agents are precisely known, yet the researcher does not directly control reproductive success or other components of selection except those imposed by the environment. Each treatment can be replicated with multiple populations so as to correct for other sources of variation that can cause divergence such as genetic drift. The drawback of this approach, as compared with artificial selection, is that it may prove difficult to distinguish between traits responding to selection and those that change indirectly because of genetic covariances (Fuller et al. 2005).
One of the first to conduct experimental simulations of fishery dynamics was Ralph Silliman. In a series of overlooked studies on captive guppy and tilapia populations, Silliman measured the ecological relationship between fishing mortality rate and yield, thereby providing an empirical basis for the basic stock production models that were emerging at that time (e.g., Silliman 1968, 1971, 1972). Then he turned his attention to evolutionary change. Silliman (1975) established two brood stocks of about 200 Tilapia mossambica derived from mixed source populations of unspecified origin. Once these captive populations were established, he subjected them to either a random or a large size-selective harvest scheme, removing 10–20% of the population every two months. The target of selection was body thickness, which was tightly correlated with length, and it was only large body thicknesses that were removed by fishing. After a period of only 3 years (roughly six generations), an evolutionary response was apparent. The selectively-fished population displayed diminished yield and male (but not female) fish grew slower and attained smaller body sizes on average than those from the control group (Fig. 1). Unfortunately, there are limitations in interpreting the outcome of Silliman’s experiment. First, the mixed origin of the brood stocks may have introduced higher levels of genetic variation than would normally occur within a single population and this may have increased the likelihood of a rapid evolutionary response. Furthermore, because he did not interbreed the mixed origin brood stocks for several generations prior to the start of the exploitation trials, effects of genetic linkage disequilibrium may have influenced the results. Second, there was no replication of treatments (only one control and one harvested population) so it is not clear that the divergence was actually caused by fishing as opposed to genetic drift. Finally, the divergence occurred in only one gender, suggesting the possibility that sexual selection may have contributed to the outcome. Despite these issues, Silliman’s work was ground-breaking in attempting to provide the experimental evidence of fishery-induced evolution. Sadly, his work has been virtually ignored (e.g., Silliman 1975 has been cited only nine times).
Figure 1. Outcome of a selection experiment on Tilapia mossambica (redrawn after Silliman 1975). A control population was harvested randomly while another was harvested selectively with respect to size (all individuals >25 mm body thickness) every 2 months over a period of 3 years. At the end, 46 size-matched fish each were reared for 150 days. Males from the selectively-fished population grew much slower than the control, while no such response was apparent in females.
Download figure to PowerPoint
Edley and Law (1988) used experimental captive populations of Daphnia magna (Crustacea, Branchiopoda) to track the evolutionary response to size-selective harvest. Their experiment involved replicated treatments of mixed clonal populations with culling based on the selective removal of either large or small body sizes. They observed rapid evolutionary responses to culling. Selective removal of large individuals resulted in lower population yields, decreased size and age at maturation, and lower individual growth rates. Selective removal of small individuals produced the opposite response. While this work has received a moderate amount of attention in the life history evolution literature (39 citations), it did not stimulate widespread concern about fishery-induced evolution probably because the taxon and its life history were viewed as being too far removed from that of harvested fishes.
Conover and Munch (2002) conducted the first fishery-induced selection experiment that involved a harvested marine fish, the Atlantic silverside. This experiment was motivated by the knowledge that this species displays CnGV in growth in the wild (Conover and Present 1990): i.e., growth rate is optimized at any given latitude by the tradeoff between the benefits of growing fast to ensure winter survival and the costs of rapid growth which entails diminished swimming speed and increased vulnerability to predation (as described above). A cluster of additional physiological and behavioral traits covary with growth rate in the wild (Table 1). The Conover and Munch (2002) harvest experiment was designed to measure the rate of evolution of these traits in response to severe size selection imposed by removing the largest 90% or the smallest 90% of the population each generation after the fish had grown to adult size. The results were striking. After only four generations the population yields and mean weights of fish diverged dramatically. Populations subjected to large size harvest quickly evolved lower growth rates, yields and mean fish weights while the small-size harvested populations did the reverse. Moreover, the same cluster of traits that vary with growth across latitudes in the wild (Table 1) also coevolved with growth rate in the fishing experiment (Walsh et al. 2006). Evolved differences in juvenile growth rate were positively correlated with changes in food consumption, growth efficiency, behavioral willingness to forage, fecundity, egg volume, larval size at hatch, larval viability, larval growth, and vertebral number (Walsh et al. 2006). Hence, the populations that evolved slower growth experienced correlated declines in a broad array of traits that collectively determine the per capita rates of energy flow and reproductive output, leading to an overall reduction in fitness. Because the experimental design included replicate populations for each fishing regime, as well as nonfished control populations, there is no question that size-selective harvest caused the genetic changes observed. The results of this experiment, combined with the knowledge of CnGV and its adaptive significance in the wild, provides irrefutable evidence that size selective mortality can cause evolutionary changes that influence the physiology, growth, behavior, and productivity of marine fish populations.
Once selected lines have been developed or identified, they provide the opportunity to test for the reversibility of fishery-induced evolution after fishing pressure is relaxed. If fishery managers are to be precautionary in their approach to conservation, the issue of reversibility is crucial because it determines the rate at which changes wrought by fishing selection might be undone by natural selection acting alone after fishing ceases. Many have speculated that the reversals will be very slow (Law and Grey 1989; Law 2000; de Roos et al. 2006; Dieckmann and Heino 2007; Swain et al. 2007) but, until recently, little hard data existed to address this question. Conover et al. (2009b) filled this empirical gap by extending the Menidia experiment for an additional five generations during which size selection at harvest was relaxed across all lines. They found that those populations evolving smaller size and lower productivity under fishing pressure displayed a very gradual but significant rebound in size after selective fishing ceased. This shows that harvested populations have an intrinsic capacity for reversal of the detrimental evolutionary effects of fishing but the rebound rate may be much slower (2.5 times longer in the Menidia experiment) than that caused initially by directional selection on fish size (Conover et al. 2009b).
Selection experiments in the field
Selection experiments can also be carried out in the field. Such experiments involve the introduction of a known composition of genotypes to a natural environment and tracking their relative fitness over time. This approach has the advantage of greater realism because environmental factors are not controlled by the investigator but has the disadvantage that agents of selection may be of multiple sources and therefore not always easy to identify. Such trials may involve tracing the fate of genotypes over a segment of the life history as they experience episodes of selection. Better still, self replenishing populations may be planted in contrasting environments or subjected to alternative forms of selection and then be allowed to diverge over multiple generations.
Biro and Post (2008) marked and released two genotypes of juvenile trout (Oncorhynchus mykiss) – slow growing/shy versus fast growing/bold – into replicate small, natural lakes, and four months later subjected them to intensive gillnet fishing that removed approximately 70% of the population. Independent of size, gillnets removed twice as many specimens of the fast/active genotype as compared with the slow/shy one. This experiment provided the first direct confirmation that fishing gear selectively alters the genotypic composition of a population. It also shows how behavioral differences associated with growth, rather than body size, may be the target of selection. However, because these populations were not self-reproducing the overall effect of such fishery-induced selection on life history evolution and stock productivity are unknown.
The pioneering experiments by David Reznick and colleagues on guppies (e.g., Endler 1980; Reznick and Bryga 1987; Reznick et al. 1990, 1996) were among the first to explicitly test predictions of life-history theory on multiple generations of natural populations. The authors took advantage of a stream system structured by waterfalls, in which isolated guppy (Poecilia reticulata) populations are adapted to high-predation (downstream) or low-predation (upstream) environments, depending on the abundance and species of co-occuring fish predators. After showing that guppies from high-predation environments consistently mature at smaller sizes and produce more and smaller offspring than their conspecifics in low predation environments, the authors experimentally transplanted guppies from high to low predation-environments (previously guppy free). Many generations later, common garden experiments on the transplanted versus founding populations demonstrated rapid evolution of life history traits in the directions predicted by theory, with significant 5–15% increases in male and female size-at-maturity after 4 and 7.5 years (7–12 generations), respectively. Offspring size, fecundity, and reproduction effort had significantly changed after 11 years (Reznick and Ghalambor 2005). While not specifically designed to measure fishery-induced evolution, these experiments are nonetheless highly relevant because they demonstrate the rapidity of life history evolution in a completely natural setting.
That the implications of these findings on guppy populations for fishery-induced evolution have been largely ignored by fishery scientists is puzzling. Reznick and Ghalambor (2005) argue that guppy populations are a realistic model for fisheries because, like harvested species, they have overlapping generations, the intensity of selection by predators is in the mid- to low range of that imposed by commercial fisheries, and the rate of phenotypic change in guppies is also comparable with that found in exploited fisheries. Moreover, the evolutionary response to predation in guppies involves changes in an array of behavioral and morphological traits similar to those in harvested species (Reznick et al. 2008) and also like those in the Menidia experiments cited above. Ignoring these lessons from guppies apparently reflects (i) a belief that natural predators can be agents of selection but not human predators, or (ii) a taxonomic bias, namely that guppies may evolve rapidly but not the taxons comprising commercially harvested species. Neither of these beliefs seem justifiable to us.
Recent insights from field and laboratory experiments on the Trinidadian killifish, Rivulus hartii, point out the importance of evolutionary responses to the indirect effects of predation-mediated mortality. Walsh and Reznick (2008) examined natural killifish populations adapted to high and low predation intensities. They observed that killifish from high predation localities displayed reduced size and age at maturity, increased reproductive investment and smaller hatchlings. These life history responses are similar to those of the guppies described above and might be attributable to the direct effect of predation. However when rearing second-generation-born killifish at two realistic food levels, the authors found significant interactions between (former) predator environment and food level for most life history traits (i.e., age and size at maturity, fecundity, egg size). These statistical interactions suggest that killifish evolution has not only been directly influenced by predation, but also indirectly by effects of elevated food availability in high predation environments.
Another experimental illustration of fishery-induced selection acting upon behavioral variation involves vulnerability to angling. Philipp et al. (2009) and Cooke et al. (2007) evaluated data from an experimental catch-and-release program involving largemouth bass (Micropterus salmoides) in a large reservoir in Illinois. Between 1977 and 1980, individually angled bass were creeled, tagged, and released. The lake was then drained and those fish angled and released four or more times in 1980 were stocked in ponds and allowed to breed to create a ‘high-vulnerability’ strain (HVF), while those never caught by angling were stocked and bred to produce a ‘low-vulnerability’ strain (LVF). These selected strains were maintained in experimental ponds and were further selected and bred for high-vulnerability or low vulnerability to fishing for three additional generations. The experiment enabled direct estimation of realized heritability for vulnerability, since both the selection differential and the response to selection could be quantified over multiple generations. The analysis of this experiment by Philipp et al. (2009) provided clear evidence that angling vulnerability is indeed a heritable trait (h = 0.15). Redpath et al. (2009) investigated growth and energy characteristics of the two strains, reporting that over a 6 month period LVF fish grew between 9–17% faster than HVF individuals. In addition, Cooke et al. (2007) found that the LVF had lower resting cardiac activities and lower metabolic requirements than the HVF, leading to an estimated 40% reduction in food requirements. Male LVF were also observed to invest less energy in parental care including decreased vigilance against predators. Hence, this selection experiment demonstrated not only that angling vulnerability has a genetic basis in largemouth bass but also that a suite of physiological and behavioral traits are genetically correlated with vulnerability. More importantly, some of these trait correlations were unexpected and would have been difficult to foresee without doing the experiment. For example, it is not obvious why angling vulnerability and parental care behavior should be genetically correlated but the fact that they are has important consequences. As Cooke et al. (2007) pointed out ‘if HVF are selectively harvested from a population, the remaining fish in that population may be less effective in providing parental care, potentially reducing reproductive output.’ They concluded that ‘strong angling pressure in many freshwater systems, and therefore the potential for this to occur in the wild, necessitate management approaches that recognize the potential evolutionary consequences of angling.’