• Open Access

Fishing-induced evolution of growth: concepts, mechanisms and the empirical evidence


  • Katja Enberg,

    1.  Department of Biology, University of Bergen, Bergen, Norway
    2.  Institute of Marine Research, Bergen, Norway
    Search for more papers by this author
  • Christian Jørgensen,

    1.  Department of Biology, University of Bergen, Bergen, Norway
    2.  Uni Research, Bergen, Norway
    Search for more papers by this author
  • Erin S. Dunlop,

    1.  Department of Biology, University of Bergen, Bergen, Norway
    2.  Institute of Marine Research, Bergen, Norway
    3.  Present address: Aquatic Research and Development Section, Ontario Ministry of Natural Resources, Peterborough, Ontario, Canada
    Search for more papers by this author
  • Øystein Varpe,

    1.  Department of Biology, University of Bergen, Bergen, Norway
    2.  The University Centre in Svalbard, Longyearbyen, Norway
    3.  Present address: Norwegian Polar Institute, Tromsø, Norway
    Search for more papers by this author
  • David S. Boukal,

    1.  Department of Biology, University of Bergen, Bergen, Norway
    2.  Institute of Marine Research, Bergen, Norway
    3.  Present address: Laboratory of Theoretical Ecology, Institute of Entomology, Biology Centre AS CR, České Budějovice, Czech Republic
    Search for more papers by this author
  • Loïc Baulier,

    1.  Department of Biology, University of Bergen, Bergen, Norway
    2.  Institute of Marine Research, Bergen, Norway
    3.  Present address: Department of Fisheries Sciences and Technologies, Ifremer, Lorient, France
    Search for more papers by this author
  • Sigrunn Eliassen,

    1.  Department of Biology, University of Bergen, Bergen, Norway
    Search for more papers by this author
  • Mikko Heino

    1.  Department of Biology, University of Bergen, Bergen, Norway
    2.  Institute of Marine Research, Bergen, Norway
    3.  International Institute for Applied Systems Analysis, Laxenburg, Austria
    Search for more papers by this author

Katja Enberg, Institute of Marine Research, Box 1870 Nordnes, N-5817 Bergen, Norway.
E-mail: katja.enberg@imr.no


The interest in fishing-induced life-history evolution has been growing in the last decade, in part because of the increasing number of studies suggesting evolutionary changes in life-history traits, and the potential ecological and economic consequences these changes may have. Among the traits that could evolve in response to fishing, growth has lately received attention. However, critical reading of the literature on growth evolution in fish reveals conceptual confusion about the nature of ‘growth’ itself as an evolving trait, and about the different ways fishing can affect growth and size-at-age of fish, both on ecological and on evolutionary time-scales. It is important to separate the advantages of being big and the costs of growing to a large size, particularly when studying life-history evolution. In this review, we explore the selection pressures on growth and the resultant evolution of growth from a mechanistic viewpoint. We define important concepts and outline the processes that must be accounted for before observed phenotypic changes can be ascribed to growth evolution. When listing traits that could be traded-off with growth rate, we group the mechanisms into those affecting resource acquisition and those governing resource allocation. We summarize potential effects of fishing on traits related to growth and discuss methods for detecting evolution of growth. We also challenge the prevailing expectation that fishing-induced evolution should always lead to slower growth.


In recent years, the potential evolutionary consequences of fishing have received considerable attention (reviewed and discussed in Law 1991; Dieckmann & Heino 2007; Jørgensen et al. 2007; Kuparinen & Merilä 2007; Marshall & Browman 2007; Fenberg & Roy 2008; Hutchings & Fraser 2008; Naish & Hard 2008; Dunlop et al. 2009a). For harvested stocks, the requirements for harvest-induced evolution are typically fulfilled, namely that: (i) fishing is selective on phenotypic traits, either because the gear in use is actively targeting fish with certain traits (see Hamley 1975 for a review of gill net selectivity) or because elevated mortality in general favours traits expressed early rather than late in life; and (ii) there is heritable genetic variability for several of these traits (Gjedrem 1983; Carlson & Seamons 2008). Furthermore, exploitation rates are often several-fold higher than natural mortality (Mertz & Myers 1998) so the overall selection might be strong. Fishing is therefore likely to influence the course of evolution, but what is not immediately obvious is which traits will evolve, in which direction and how quickly will evolution proceed, how important will it be relative to other forces causing phenotypic change, and, finally, how we can best manage any potential consequences.

One life-history trait that is expected to evolve in response to fishing is growth rate. Early literature on fishing-induced evolution of growth was dominated by the hypothesis that fishing would select against fast-growing, large fish because fishing mortality typically increases with body size (eventually exceeding natural mortality in most exploited stocks), and that growth could consequently evolve towards slower rates (Rutter 1902; Miller 1957; Favro et al. 1979; Kristiansen & Svåsand 1998). This hypothesis is intuitive and has been reiterated many times, but serious attempts to test or evaluate it were for a long time restricted to Ricker’s (1981) classic work on the declining size of Pacific salmon. Within the last decade, there has been a resurgence of studies on fishing-induced evolution of growth, both from experiments (Conover & Munch 2002; Biro & Post 2008) and from examinations of wild populations (Edeline et al. 2007; Swain et al. 2007). The work by Conover & Munch (2002) drew considerable media attention and scientific debate (Hilborn 2006; Conover & Munch 2007). The same occurred with the study by Edeline et al. (2007), which was included among the journal Nature’s research highlights of 2007 (Anonymous 2007, see also Conover 2007) and was praised elsewhere (Coltman 2008). However, these studies have revealed that predictions on the direction of evolution of growth are more complex than was thought when the first hypotheses were formulated. In particular, many studies confuse growth rate with size-at-age, or treat growth rate as the evolving trait without considering other traits and processes that influence size-at-age. Here we aim to address the confusion by a concept-oriented review of mechanisms that affect growth (and consequently size-at-age) through their impact on the acquisition and allocation of resources. We further interpret each mechanism in light of fishing-induced evolution, and expose the difficulties associated with the measurement of growth rate from field data. Finally, we consider the achievements of the field as a whole and where future research is needed.

Definitions and Key Concepts

It is often deceptive to think of growth as a single trait, when instead it is the outcome of a complex suite of behavioural, morphological and physiological processes. These processes relate to both incoming resources and subsequent partitioning of those resources, and in particular to how surplus resources are allocated among competing needs. Different components of growth are therefore subject to various selection pressures, making evolutionary change in growth inherently difficult to predict and quantify from data. In this section, we first provide key definitions, and highlight areas where misconceptions typically occur. Our approach and definitions are inspired by foraging behaviour (Lima & Dill 1990), a life-cycle perspective to energy allocation (e.g.Roff 1983; Kozlowski 1992), and energy budgets and flows within individual organisms (e.g.Kooijman 2010).

Acquisition, allocation, and growth

Key processes involved in energy budgeting within an individual are resource acquisition, resource allocation, and growth (Figs 1 and 2). We define resource acquisition as the processes involved in foraging, ingestion and digestion. Resource acquisition controls the total amount of resources available to the organism, and these are allocated to various competing needs.

Figure 1.

 Simplified representation of the basic logic of resource flow underlying growth. Acquired resources are allocated to three main components: (i) the ‘running costs’ of an individual including basal metabolism and routine behaviour, (ii) somatic growth including structures and stores, and (iii) reproduction. Before maturation, changes in surplus resources will translate directly into changes in growth, whereas a proportion of the surplus resources is allocated towards reproduction from maturation onwards. A more comprehensive flowchart is shown in Fig. 2.

Figure 2.

 Resource acquisition (left) and allocation (right) are both adaptive processes. The amount of acquired resources is affected by many processes influencing foraging and ingestion, which after digestion translates into available resources. Resources allocated to routine metabolism, reproduction and structural growth are generally not available for later use, whereas resources deposited in stores can be utilized later. Changes in any of the components affecting resource acquisition or in how the acquired resources are allocated may lead to altered growth.

Some of the resources are allocated to running costs of the organism: mainly its basal metabolism (often quantified as resting metabolism), digestion and routine activity (Fig. 2). Basal metabolism represents a set of basic processes needed to remain alive and functional (e.g. maintenance, immune defence and cognition). Note that allocation to, for example, cognition, movement and digestion may increase the organism’s potential for resource acquisition, so allocation and acquisition are not completely separable. Of special importance for this review is the availability of resources after basal metabolism and routine activity have been accounted for, and we term these surplus resources (Fig. 1). Surplus resources are often termed ‘surplus energy’ (Ware 1975; Roff 1983) or ‘growth effort’ (Abrams & Rowe 1996); by referring to ‘resources’ we acknowledge that not only is the total energy important, but the nature of that energy (e.g. its nutrient composition) has consequences as well (e.g.Pulliam 1975; Blount et al. 2000; Grandison et al. 2009). During the immature phase of the life cycle, surplus resources will result in somatic growth, which includes the growth of all tissues in the body except for germ cells; it is therefore distinguished from reproductive investment, which allocates a proportion of the surplus resources to reproduction as the individual becomes sexually mature.

Size is a state, growth is a process

Individual body size is of profound ecological importance. Larger individuals usually have increased survival, are more successful at attracting mates, have higher fecundity, have an enhanced capability to withstand starvation, and are better in competition for resources (Peters 1983). However, food requirements are generally higher for larger animals, and growing big takes time and resources. There are thus several components in a cost–benefit budget that jointly determine the evolutionary advantages and disadvantages of having a particular size. Behavioural and life-history strategies will therefore often be size-dependent and change as an animal grows (McNamara & Houston 1996). From a fisheries perspective, the size distribution of individuals also has a strong influence on population dynamics, including recruitment (Persson et al. 2007) and fisheries yield (Law & Grey 1989; Conover & Munch 2002).

Growth and size are intimately linked and often correlated (for example, fast-growing individuals are often big) but, obviously, they are not the same. For example, large individuals may, when they use their resources for reproduction, grow more slowly than smaller fish. Thus, whereas body size characterizes an individual’s state, growth is the process that leads to that state. For example, it can be risky to grow fast, but once a certain size is attained, the payback in terms of survival or reproduction can be good.

It is tempting to think that selection on growth could be fully understood by studying selection on size, or vice versa, but this is not the case. To be able to grow to a certain size, an individual needs to acquire resources through foraging, usually at a cost of exposure to predators, parasites and infections. From an evolutionary perspective, it is therefore important to separate between the advantages of being big and the costs of growing to a large size. Consequently, there can be selection not only on body size but also on the growth-related processes that allow the individual to attain a certain size. Growth and size are coupled, so selection on one may lead to indirect selection on the other, and growth and size may also be correlated with other traits under selection.

Observing growth directly implies measuring changes in size over time, and growth per unit of time is referred to as growth rate. The most common measures of growth rate in fish are length increment or body mass increment per time (for other definitions see Dmitriew 2011). These measures differ in one important aspect: increase in length is usually irreversible, whereas weight can increase or decrease as stores are deposited and utilized, or as gonad mass is built and spawned.

Growth is evolutionarily optimized, not maximized

In the past, growth was often regarded as a trait that should, under natural selection, evolve towards maximum rates, within the limits of physiological constraints (e.g.Ricklefs 1969). However, there may be several benefits of sub-maximal growth, and it was surprisingly recently that growth became more widely regarded as being evolutionarily optimized rather than maximized (in large part due to the review by Arendt 1997; but see also Case 1978). Two well-documented phenomena clearly support this conclusion in fish. First, when individuals are given excess food after a period of food deprivation, they often grow faster than control groups that are fed ad libitum. This phenomenon, referred to as compensatory growth or catch-up growth (reviewed in Metcalfe & Monaghan 2001; Ali et al. 2003), illustrates how growth rate is normally well below the physiological capacity (Sundström et al. 2007). A wide spectrum of delayed and immediate costs of compensatory growth have been identified (Arendt 1997; Metcalfe & Monaghan 2001; Arendt & Reznick 2005; Mangel & Munch 2005), further suggesting that although faster growth is possible, it implies costs, particularly in terms of survival. The ability for fast growth may, for example, be costly in terms of starvation tolerance: a recent study in European sea bass Dicentrarchus labrax identified a positive correlation between the rate of mass loss during a starvation period and the growth during a subsequent compensatory period (Dupont-Prinet et al. 2010).

Countergradient variation is another phenomenon that shows a clear role for local adaptations in growth (Conover & Present 1990). In the wild, populations experiencing different growth conditions may show comparable growth within a season, even though there might be considerable differences, for example, in the length of the growing season or temperature along a latitudinal gradient. One species where countergradient variation has been studied is the Atlantic silverside, Menidia menidia, a small annual fish found along the east coast of North America: in common garden experiments, individuals from the northern populations grow faster than their southern counterparts, indicating a genetic basis for the difference (Conover & Present 1990). The northern population’s increased growth rate in the lab compensates for shorter growing season in the wild, so that when different populations are sampled in their native environment, the differences in size-at-age are smaller than the local environment would prescribe (Conover & Present 1990). Countergradient variation has also been observed in a number of other fish species and other taxa (Conover & Schultz 1995).

Compensatory growth and countergradient variation demonstrate that growth is evolutionarily ‘optimized’ at levels below the physiological capacity. Which trade-offs cause adaptive differences in growth between individuals, populations and species? Is it possible to predict how the mechanisms underlying growth may evolve in response to fishing? In the remainder of this paper we try to answer these central questions by splitting growth into several processes and analyzing each process separately. We describe selection on growth-related traits in a natural setting, and how that selection may be modified due to fishing. In practice we do this mainly by identifying the underlying trade-offs and evaluating how the costs and benefits change from the natural setting to one in which fishing occurs.

Factors Affecting Growth

Even though growth is usually observed as one variable (the rate of change in size) the process of growth consists of two main components: acquisition of resources and subsequent allocation to different needs. Below we provide a conceptual review of factors affecting observed growth rate. Although we attempt to break growth down into component processes, it is important to keep in mind that there are developmental constraints, shared mechanisms, delayed effects, variable heritabilities, and genetic correlations that may influence phenotypic expression as well as the ability of each trait to evolve. On top of this, growth is also very sensitive to environmental conditions, so for any phenotypic effect one needs to bear in mind that there might be genetic influences as well as phenotypic plasticity.

Key processes: resource acquisition and allocation

Organisms have a limited amount of resources they can use, not only for growth, but also for storage, reproduction and maintenance. That resources are finite can, as illustrated by Reznick et al. (2000), be depicted as a pie where allocation decisions are represented as slices, with the sum of all slices necessarily constrained by the size of the whole pie. Consequently, increasing the ‘growth-slice’ will decrease the size of some other slice(s), for instance the amount of resources allocated to reproduction. This paradigm underlies much of life-history theory (e.g.Roff 1992) and energy allocation modelling (e.g.Roff 1983; Kozlowski 1992).

However, the total size of the pie depends on resource acquisition. If an individual is efficient in gaining resources, the pie becomes larger (Reznick et al. 2000). Constraints on acquisition may, however, arise through avoidance of predation or through food limitation. Here, we highlight the many traits which could be traded-off with growth rate. We group basic mechanisms into two categories: those affecting resource acquisition (the size of the whole pie) and those governing resource allocation (the relative size of each slice).

Resource acquisition revolves around the processes of foraging leading up to ingestion and digestion (Fig. 2). Foraging involves behavioural and morphological adaptations related to finding and capturing prey, as well as physiological and anatomical adaptations involved in sensing and cognition. Many trade-offs have been extensively studied within the field of foraging ecology (e.g.Stephens et al. 2007), and most of these relate to behaviour and risk-taking. In Table 1, we list many of the trade-offs relating to resource acquisition, and how they may be perturbed by fishing activities.

Table 1.   Processes involved in resource acquisition and their relation to natural and fisheries selection. Mechanisms are grouped by general category, below which more specific mechanisms are listed. The last two columns contain potential consequences for fishing-induced adaptations, and relevant literature examples, preferably on fish.
Mechanism or trade-off in natureRelation to fishingPotential fishing-induced adaptationRelevant literature
Habitat choiceHabitats differ in food availability and exposure to predators, resulting in trade-offs between food intake and survivalStrong preference of fish for a given area may direct fishing effort to such locationsFish frequenting certain habitats where fishing occurs will be selected against, causing, for example, changes in the distributional patterns of the populationBiro et al. (2006) observed differences in habitat use between domestic and wild rainbow trout – with higher growth rates and reduced survival in the risk-prone domesticated strain
Fish that hide, for example flatfish on the bottom, generally move when approached by active gears such as trawlsLeaving the hiding place may expose fish to predatorsA classical review of risk in relation to foraging behaviour is Lima & Dill (1990)
Temporal light regimeAs most fish are visual predators, both foraging rates and piscivore predation change as light levels vary throughout the day (and season)Many types of fishing gear, e.g. baits, lures, and gillnets, are most efficient at times of the day when fish are foraging or activeFish that are active foragers may be caught more often. Could cause reduced growth or shifting foraging to less optimal times of the dayJuvenile sockeye salmon feed most intensively at dawn and dusk, when the ratio between foraging rates and predation risk are most favourable (Clark & Levy 1988). A model suggested reduced growth if passive gears harvest the most active swimmers (Favro et al. 1979)
Vertical light regimeLight extinction down the vertical column creates a visual habitat that changes with depth as well as time. Deeper habitats are safer, but may restrict foraging rates for visual feedersSome pelagic fishing gear is deployed close to the surface, for example purse seinesFish that avoid surface waters could have a a selective advantage but might suffer suboptimal foraging ratesThe best time and depth windows are used first, for example in crepuscular feeding where small pelagic fish come close to the surface to fill their stomachs before nightfall (Strand et al. 2002)
Diet choiceFood items differ in size, energy content, digestibility, and handling time. Several trade-offs between food intake and survival are involvedFishing may change availability of foodForaging can become safer or riskier as a result of fishingBottom trawling may increase food availability for plaice (Hiddink et al. 2008)
Handling timeIncluding food items larger or smaller than the optimal size may increase energy intake but at a cost of increased handling time, during which the forager may be vulnerable to predatorsLonger handling time may lead to higher exposure to certain types of fishing gearTo avoid gear, fish that forage less or reject items that require extra handling time may be favoured. This might lead to decreased acquisitionJuvenile coho salmon forage less and avoid large food items just after a predator has been sighted (Dill & Fraser 1984)
Alternatively, fish may evolve reduced handling times at a cost of less energy extracted or increased digestion timesSeveral bird species reduce handling time in the open, presumably at a cost of reduced digestion rates (Valone & Lima 1987)
 Search timeThere is often a trade-off between ingesting an encountered food item and continuing searching for a more profitable one. This trade-off is influenced by exposure to predators during search and handlingFish that search more have a higher risk of encountering passive gear such as gillnets or baited hooksFish with reduced food search behaviour may be more likely to survive fishing, leading to reduced acquisition ratesVulnerability to angling is heritable, suggesting that consistent differences in feeding motivation or food search behaviour affect experienced mortality in a fishery (Cooke et al. 2007, Philipp et al. 2009)
Ingesting pathogensFood items differ in the likelihood of containing parasites, pathogens, or toxins. Individuals that discriminate less may increase energy intake, but also suffer reduced survival or other long-term negative effectsIncreased mortality from fishing shortens lifespan, discounting long-term negative effects of disease, toxins, or parasitesFishing may favour fish that ignore long-term consequences and instead forage indiscriminately, thus leading to elevated mortality rates late in lifeOystercatchers Haematopus ostralegus avoid large and energy-rich cockles because these have higher infection intensities of a helminth parasite (Norris 1999). Sheep Ovis aries that are strongly motivated to feed include also patches infested with parasite-infected faeces (Hutchings et al. 1999)
Nutrient compositionDiets that maximize energy intake can be low in essential nutrients (e.g.Pulliam 1975). Increasing growth through a diet that focuses more on energy can thus compromise other crucial processesBecause fishing reduces life-span, short-term energy gains may outweigh long-term consequences of a poor dietFish that maximize short-term benefits such as high growth or elevated reproduction through an energy-rich diet may be favoured, although these adaptations may incur costs in terms of e.g. sexual traits, development, or immune defence due to deficits of essential nutrientsCarotenoids are important for immune defence and the expression of sexually selected traits in many fish (Maan et al. 2006)
Ingestion rateDigestion elevates metabolic rate and therefore occupies part of the aerobic oxygen budget. This, in turn, leads to reduced swimming performance and increased predation ratesFishing reduces life-span and favours fast life histories, which may require higher ingestion ratesHigher ingestion rates may be selected to sustain faster growth and more intense reproductionAtlantic silversides adapted to high overwintering mortality forage more intensely, digest faster, but are also predated upon more often (Billerbeck et al. 2001; Lankford et al. 2001)
Avoidance of active fishing gear may be more difficult for fish that digest at high ratesFish that consume less may be more successful at avoiding fishing gear, which could select for decreased energy acquisitionFish that digest fast are eaten more often in predation experiments because their swimming performance is poorer (Lankford et al. 2001)
Baited fishing hooks or traps may be most attractive to hungry fish or individuals with high feeding motivationFish that forage more cautiously may be favoured. This could lead to lower acquisition ratesIn a whole-lake experiment, a bold and fast-growing genotype of rainbow trout was selectively harvested by angling with lures (Biro & Post 2008; see also Dupont-Prinet et al. 2010)
Social behaviourGroups often confer advantages to all the individuals that belong to itWith echosounders, fishermen have the ability to target groups of fish rather than individual fish dispersed over larger volumes of waterFish that are less often in groups may have an advantage in the presence of fishing, but less gregarious behaviour may lead to loss of group benefitsSocial and group behaviour is reviewed in Giraldeau & Caraco (2000) and Krause & Ruxton (2002)
Group foragingSmall to moderate-sized groups often find food more efficiently than solitary individualsFishing that targets groups of fish may favour solitary feedingEnergy acquisition rates may drop. This may be compensated by riskier foraging that leads to increased predationMinnows and goldfish find food faster as group sizes increase up to 20 individuals (Pitcher et al. 1982)
Predator defencePredation risk is diluted in groups (Inman & Krebs 1987). Several anti-predator behaviours involve co-operation between individualsFishing that targets groups may favour individuals that face predators aloneFish that encounter predators alone may escape less often, leading to more risk-averse foraging strategies and reduced energy acquisition and growthGuppies inspect predators in pairs, thus mapping risk as part of efficient foraging behaviour (Griffiths & Magurran 1998)
SchoolingIndividual foraging rates may drop in dense and large schools as is often seen in marine pelagic fish speciesFishing that targets schools may favour solitary fish or fish at the outskirts of a school if that enabled escape from the fishing gearFish that are more often outside the school may have higher resource acquisition rates, but will presumably suffer higher predation ratesReduced foraging performance in schools takes place because visual fields and search volumes overlap (Eggers 1976)

To illustrate that there are also less intuitive costs related to acquisition, consider the example of digestion. In fish, the entire digestion process typically consumes some 10% of the energy content of the ingested food (Tandler & Beamish 1979). However, digestion also requires oxygen and, as aquatic respiration is costly, this is often in conflict with other purposes such as escaping predators (Arnott et al. 2006). This conflict has been demonstrated in a comprehensive set of experiments on the Atlantic silverside along the axis of countergradient variation of growth described above. Northern silverside populations, which experience short growing seasons, increase their growth rate by voluntarily ingesting larger meals (Lankford et al. 2001) and consequently have higher metabolic rates and consume more oxygen (Arnott et al. 2006). In common garden experiments, northern populations are poorer swimmers (Billerbeck et al. 2001), which gives them poorer escapement responses and they are consequently eaten more often by predators (Lankford et al. 2001; Munch & Conover 2003). Physiological mechanisms related to digestion thus translate increased growth rate into higher predation rates. Similar results have also been obtained from experiments by Suzuki et al. (2010), who hypothesized that individual medaka Oryzias latipes being predated upon by an ambush predator might trade off cautiousness to forage more intensely. When fishing elevates mortality rates, speeding up growth rate and reproduction by digesting faster might be favoured, even if it means being exposed to some extra predation mortality. Through this mechanism, fishing might thus increase acquisition rates and, if allocation does not change, also growth rate. Further trade-offs related to resource acquisition and how these may respond to fishing are listed in Table 1.

The allocation of acquired resources is also a dynamic and multifaceted process. Owing to its direct relevance for fitness, allocation to reproduction has received considerable attention, often contrasted with allocation to growth as the only other recipient of energy (e.g.Roff 1983; Kozlowski 1992). Growth trajectories are split in two broad categories depending on the age-schedule of allocation to growth and reproduction: determinate (a sharp transition from allocation to growth to allocation to reproduction, leading to growth curves resembling a hockey-stick) and indeterminate (more gradual transition from allocation to growth to allocation to reproduction and more von Bertalanffy-like growth curves; reviewed in Heino & Kaitala 1996; see also Von Bertalanffy and Pirozynski 1953). In fish, indeterminate growth is common, giving extra degrees of freedom to adult life histories.

Although growth and reproduction can receive a large proportion of resources, they are only two of several resource-demanding processes in an individual. In a more complete picture, investment in other components such as maintenance, immune defence, digestion, morphology, cognition, behaviour, and storage also need to be considered (Fig. 2). Each of these components receives energy and resources in an amount that is likely adaptive in the environment in which the organism has evolved. By down-regulating any of these components, energy and metabolic capacity can be freed for somatic growth. Examples of components with competing demands for energy and resources are given in Table 2.

Table 2.   Traits associated with resource allocation and their relation to natural and fisheries selection. Traits are grouped by general category, below which more specific mechanisms are listed in some cases. The last two columns contain potential consequences for fishing-induced adaptations, and relevant literature examples, preferably on fish.
Trait group
Specific trait
Mechanism or trade-off in natureRelation to fishingPotential fishing-induced adaptationRelevant literature
MaintenanceRepair of genetic material, replacement of damaged proteins (Hawkins 1991), and renewal of cells and tissue help maintain a well-functioning soma. For example, the theory of aging partly ascribes senescence to build-up of damage inflicted by mutations (Medawar 1952) and reactive oxygen molecules (Harman 1956)Fish investing less in maintenance could be less able to escape active fishing gear such as trawlsOften there is a trade-off between investment in somatic growth and investment into maintenance, such that fishing might favour fish investing less in growthFerox trout is a morph that lives longer and grows larger in co-existence with regular brown trout, presumably due to piscivory allowing a different allocation to growth and maintenance (Mangel & Abrahams 2001). Reduction in allocation to maintenance was hypothesized to enable lower assimilation efficiency but higher conversion efficiency, allowing better predator avoidance, in fast growing damselfly larvae under time and predation stress (Stoks et al. 2005)
Immune defenceBuilding and maintaining general and specific immune defence is energetically costly (Lochmiller & Deerenberg 2000). Down-regulating these mechanisms can increase the costs of having an infection and the probability of injury or death (Møller & Saino 2004)Fishing reduces lifespan and hence the benefits from investment in immune defenceIndividuals that reduce investment in immune defence (for long-term survival) may have more resources for growth and reproductionDe Block et al. (2008) found reduced investment in immune defence in damselflies with compensatory growth
Individuals with reduced immune defence may be more often infected, which could lead to reduced escapement ability, especially from active gearsFish with stronger immune defence would be selected for. Increased investment in immune defence could lead to lower growth ratesSimilarly, the damselflies with higher immune investment had lower rates of compensatory growth (De Block et al. 2008)
Fish with low immune defence may be more prone to fatal damage due to contact with fishing gear  
Neural development and cognitionCognition enhances food-finding, anti-predator behaviours, and within-species social interactions, but sensory organs, neural tissue, and investment in learning are costly. The brain alone usually requires 2–8% of the metabolic rate in vertebrates, but can demand as much as 60% in some species of electric fish (Nilsson 1996)It might be speculated that ‘smarter’ fish would be better at avoiding fishingIf more resources were to be diverted to neural development and cognition, the resources would no longer be available for growth or other resource-demanding processesMiller (1957) suggested that angling would select for intelligence. Haddock have learned to swim through gillnet mesh but the relationship to cognition has not been studied (Özbilgin & Glass 2004).
Learning often involves a short-term cost to acquire a long-term gain. Because fishing reduces lifespan, it also reduces the potential for the long-term gainCostly learning for long-term benefits may evolve to lower levelsIncreasing mortality rate in a model led to reduced levels of learning but faster learning rates (Eliassen et al. 2007)
MorphologyAnti-predator strategies can involve spines, large or sharp fins, and other morphological features. Feeding on specific types of resources can require specialized mouth parts. These structures can be energetically costly to build, use, or carryCatching fish with gillnets is based on individuals getting tangled in the net, thus this kind of gear might select against spines, large or sharp finsReduced investment in costly morphological features may free resources for growth or reproductive investmentGillnet selectivity curves, including their dependence on body profiles, are reviewed in Hamley (1975)
A stout body shape causes swimming to be more costly when compared to more streamlined fish (Petrell & Jones 2000; Boily & Magnan 2002)Good swimming performance improves avoidance of active fishing gears, but is often low for stout fishBy selectively fishing the roundest or stoutest fish, nets could select for more streamlined body shapes. This could cause adaptive changes in growth or reproductive investmentFemale guppies carrying large internal broods have poorer swimming performance than females without broods (Ghalambor et al. 2004)
Fish also differ in their general body plan, from being long and slim to stout and roundGillnets are often selective on body shape, with slim fish better at escaping, but also with very deep-bodied fish such as Tilapia having narrower selectivity curves (Hamley 1975)Slimmer fish could be favoured. That could lead to reduced growth, reduced reproductive investment, or reduced energy storageTrends, presumably evolutionary, towards leaner fish have been observed in Pacific salmonids (Ricker 1981, 1995) and lake whitefish (Thomas & Eckmann 2007; Thomas et al. 2009)
MigrationsMany species move extensively between foraging areas and spawning areas, or follow prey on feeding migrationsFishing concentrated in space or time may cause variation in mortality that can cause selectionMigration patterns may evolve to avoid fishingIncreased marine fishery on anadromous species may delay seawards migration or cause increased freshwater residency (Thériault et al. 2008)
Fishing may cause changes in size distribution, which may have effects on migrations because swimming costs are size-dependentEarlier maturation at smaller size may have correlated consequences for migrationsFishing-induced evolution of smaller adult size was correlated with shorter spawning migrations in a model for Atlantic cod (Jørgensen et al. 2008)
Energy storageFood availability is often unpredictable, or predictably bad in certain periods. Energy storage may function as a buffer against poor conditions, but also as capital allocated to reproduction at a later stage (see Reproduction below)Gillnets select on girth and could remove fish with large energy stores faster than fish with smaller energy storesIndividuals investing less in stores would be favoured. These individuals could be at higher risk of starvation, but could allocate more to growth or income breeding. If energy stores are for capital breeding, selection towards income breeding would be expectedCompensatory growing juvenile halibut decreased energy storage, presumably resulting in increased starvation risk (Hurst et al. 2005)
Somatic growthLarge size is advantageous for many purposes including survival and foraging, but growth requires time and resourcesFishing gear or practice may be directly or indirectly selective on size or on any process related to growthFishing can cause slower or faster growth (Fig. 3A)Pink salmon declined in size over time, and fastest in areas with the most size-selective gear (Ricker 1981). In cod, selection on size fluctuated (Swain et al. 2007)
ReproductionReproduction is among the most dominant processes diverting energy away from growth. In fish, allocation to reproduction is negligible in early life, and it is therefore convenient to characterize reproductive investment through its ontogenetic dimension (timing) and intensityFishing reduces life-span, implying that events later in life matter less for fitnessReduced longevity selects for increased reproductive output early in lifeLaw (1979a) showed theoretically that shorter lifespan favours increased reproductive investment
Fishing activities may also target more specific phenotypes or reproductive behavioursAny adaptation would be case-specificSome fishing-induced evolutionary effects are listed in Dieckmann et al. (2009)
Timing of maturationFish spend energy and time on the maturation process to develop primary and secondary sexual characters. When reproduction begins, it occupies resources that otherwise could have been used for growth or other purposesFishing reduces life spanFish that mature earlier can reproduce more times before they are fished, and could be favoured by selection. Earlier maturation will normally lead to reduced size-at-age in post-mature fish (Fig. 3B)Fishing-induced evolution towards earlier maturation was reviewed by Dieckmann & Heino (2007). The influence of earlier maturation on size-at-age has been quantified in Pardoe et al. (2009)
Some fisheries target spawning aggregationsFish may lower their exposure to fishing gear by delaying maturation or spawning less frequentlyOne of the first papers to suggest delayed maturation if fish were harvested on spawning grounds was Law & Grey (1989)
Fishing gear is often size-selective or targets specific phenotypes in other waysSize-selective gear may create refuges for certain maturation sizes and ages. Gear selective on other traits may cause other refugesGillnets have bell-shaped size-selectivity, and may favour maturation large and late in life (Boukal et al. 2008; Jørgensen et al. 2009)
ReproductiveinvestmentReproductive investment determines the amount of resources channelled towards maturation and reproduction. In the broad sense this includes not only gonads but also spawning migrations, competition for mates, and parental careFishing reduces lifespanWith short lifespan it is more beneficial to increase reproductive investment than to grow, which can be seen as an investment in future reproduction (Law 1979b). Higher investment in reproduction will lead to slower post-maturation growth and a more determinate growth pattern (see Fig. 3C)Increased fecundity or reproductive investment has been demonstrated in several harvested species in the wild, for example plaice (Rijnsdorp 1991; Rijnsdorp et al. 2005), cod (Yoneda & Wright 2004), and haddock (Wright 2005). The reverse trend has been observed in experiments where populations have been translocated to environments with lower predation rates (Reznick et al. 1990)
Skipped spawningIndeterminate growth in fish implies a phase following maturation where there is a gradual transition from both growth and reproduction to only reproduction. For late-maturing fish, skipping spawning seasons may be adaptiveFishing reduces lifespanIndividuals that skip spawning may prioritize growth over reproduction even after maturation. If expected lifespan is reduced, the delayed benefit may never materialize and skipped spawning become less commonTemporal trends in skipped spawning driven by fishing was investigated in an energy allocation model, leading to reduced rates of skipped spawning as an adaptation to modern fishing pressure (Jørgensen & Fiksen 2006; Jørgensen et al. 2006)
Capital-versus income-breedingReproductive tactics are represented along the whole continuum from pure capital breeding to income breeding. Capital breeders rely on stored resources for reproduction whereas income breeders rely on concurrent food intake. Salmon, herring and capelin are examples of capital breedersLarge gonads or large energy stores in preparation for spawning typically make fish rounder, with higher probabilities of being caught with e.g. gillnetsFish that have leaner body shapes may better slip through gillnets, but will reproduce less or depend on income breeding to a greater extent. Trade-offs involving internal resource storage are complex and link to a range of life history traits (e.g.Varpe et al. 2009)Lake whitefish got leaner over time, presumably as a response to a gillnet fishery (Thomas & Eckmann 2007; Thomas et al. 2009)
Parental carePost-fertilization care of offspring requires parental presence and investments of time or energyFish that guard offspring may be easy to find or aggressive and active and therefore easy to fishFishing may select for reduced parental care. Alternatively, the selection pressure induced by parental care may slow fishing-induced evolution on other traits(Dunlop et al. 2007) found that maturation tendency evolved less when paternal care was included in their model. Drake et al. (1997) found reduced parental care and increased cuckoldry among bluegills in heavily harvested lakes, presumably a side-effect of decreasing maturation size and age
  Mating successFemales are often choosy about whom they mate with. In such cases of sexual selection, mating success and thereby reproductive success among males depends on the expression of a display trait or another phenotypic traitIn many species females prefer larger males. Size-selective fishing gear can then remove the most sexually attractive partnersIf larger partners are preferentially chosen as mates, this selection pressure acts in opposition to the most normal fishing-induced In a model, size advantage among males, be it through female choice or male-male competition, slowed evolutionary rates under moderate fishing pressures but not when harvesting was very strong (Hutchings & Rowe 2008)
Male–male competition for mates may favour traits that have no survival value in their environment, for example display traits, a large body size, or weaponryIn many species males fight for territories, for a central position in a lek, or for access to females, an often size is important for the outcome selection towards smaller size. This may slow evolutionary rates 
HermaphroditismSequential hermaphrodites reproduce as one sex early in life and thereafter change sex and reproduce as the other sex. If one sex is rare then it may pay off to change to that sexFishing often truncates age- and size-structure so that the later sex may become underrepresentedBecause each sex contributes equally to the next generation, the rarer sex is selected for. Fishing can therefore cause evolution towards sex change at smaller sizeIn an individual-based model, maturation to the first sex and sex change both evolved to become earlier, and in some cases hermaphroditism was lost and the populations evolved separate sexes (Sattar et al. 2008)
PhenologyFish are adapted to the seasonality in their living environment, for example, timing of spawning run is usually dependent on suitable water flow in the river allowing for upward migration. Likewise, spawning in general is often timed to benefit the survival and growth of eggs and larvaeFisheries sometimes have consistent seasonality caused by fishing within regulated or preferred time-windows, seasonal weather conditions, or closing of fisheries after a quota is reachedSelective harvesting on seasonal timing could cause a shift in the timing of important events such as migration or spawning run, possibly with related changes in resource acquisition and allocation, and thereby change the life history strategy (McNamara & Houston 2008)Late sockeye salmon migrants have been fished the most, and the run has become progressively earlier presumably as an evolutionary adaptation (Quinn et al. 2007)
Phenology  If variation in size is correlated with variation in phenology (e.g. if large individuals arrive earlier at the spawning grounds), then selection on size and potentially growth could arise. For example, large and fast-growing coho salmon smolts migrated earlier than their smaller and slower-growing conspecifics (Sundström et al. 2010)Complex patterns of natural and fishing-induced selection arise on local populations of salmon harvested in different parts of their spawning migration (Kendall & Quinn 2009)

Developmental constraints and delayed effects of growth

Having outlined how resource acquisition (Table 1) and resource allocation (Table 2) are malleable processes that can respond to selection pressures (including those from fishing), it is necessary to understand that constraints may act directly on growth rate. In particular, the development rate of certain body structures may constrain the growth rates of other structures or induce costs related to rapid growth. For instance, rapidly growing snails have thinner shells because the rate of calcium deposition does not keep up with increased growth rates, and individuals with thin shells are more vulnerable to predators (Palmer 1981). Similarly, rapid growth may result in compromised morphology, such as suboptimal body proportions, increased fluctuating asymmetry, and skeletal deformities (see Arendt 1997). For example, a positive correlation between rapid growth rates and the degree of coronary lesions was reported in Atlantic salmon (Saunders et al. 1992). Such effects may be exacerbated by energy-maximizing diets, which may lead to deficiency of essential nutrients (Pulliam 1975). Another study, on zebra finches Taeniopygia guttata, showed that individual variation in resting metabolic rate may be a consequence of the growth trajectory during early ontogeny, with high metabolic rates in adulthood being caused by accelerated growth during earlier life stages (Criscuolo et al. 2008). These examples emphasize that not all costs of growth are immediate, which can make it harder to identify the trade-off in nature.

Effects of the environment

Growth rate is also influenced by environmental conditions such as temperature or the type and abundance of food items. The latter may, in turn, be influenced by density dependence, for example if a large population depletes its food resource. There is also a potential for social effects, for example, reduced numbers of old and large individuals in a population may release an inhibition of maturation on smaller and younger individuals (Kolluru & Reznick 1996). Fishing affects many of these relationships directly, for example by decreasing the number of fish, which can induce an increase in prey abundance and lead to more resources becoming available. Furthermore, phenotypic expression may depend on the environmental conditions through developmental channelling or phenotypic plasticity. It is therefore difficult to conclude whether detectable changes in size-at-age during a period of fishing are due to fishing-induced evolution or indirect effects of fishing on environmental characteristics. The method of Swain et al. (2007) offers one good example of how one may work around this problem. They included proxies for the biotic and abiotic environment, and estimated their effects on growth trajectories.

In this paper we do not aim to review the environmental influences of growth, as there is a general awareness of the phenomenon and it is often corrected for in studies. We merely want to emphasize that whenever a phenotypic trait is quantified, one needs to correct for environmental influences before one can make inferences about underlying genetic differences, be it between populations or trends in time-series analyses. Whenever genetic variation underlies some of the phenotypic variation in a trait, the trait is heritable, and natural or artificial selection may mould that trait over time. However, modelling studies suggest that phenotypic plasticity in growth and maturation dominate the picture, especially over short time scales, and that these may mask underlying evolutionary change that takes place at a comparatively slower pace (see Dunlop et al. 2009b; Enberg et al. 2009). Probably due to such plastic effects, the analysis by Hilborn & Minte-Vera (2008) did not reveal any clear pattern between change in size and fishing mortality in a large number of marine fish stocks.

Fishing-Induced Evolution of Growth

Up until this point, we have been discussing the processes and mechanisms involved in growth with the hope of providing the necessary context for studying how fishing might affect these processes. We argue that an understanding of the fundamentals of resource acquisition and allocation is crucial to being able to predict the direction and extent of fishing-induced evolution of growth and to interpret trends in size-at-age or growth from phenotypic data. For the remainder of this paper, we direct our attention to fishing-induced evolution of growth more specifically, and in particular describe what previous research on the topic has taught us.


Although several studies have suggested that fishing will result in reduced growth rates (e.g.Miller 1957; Conover & Munch 2002; Edeline et al. 2007), these predictions have often been based on the influence of size on survival. If one considers a larger set of mechanisms, as listed in Tables 1 and 2, it becomes more difficult to devise general expectations for how fishing-induced selection might change growth rates (see also Heino & Godø 2002). As is evident from Tables 1 and 2, there are fishing-induced effects that could lead to higher or lower acquisition rates, or to allocation of a greater or lesser share to somatic growth, and there might be selection on several of these mechanisms simultaneously.

The most fundamental consequence of increased mortality from fishing is reduced expected life-span. Fish that accelerate their life history through earlier maturation, increased reproductive investment, or increased resource acquisition and faster growth are thus likely to be favoured. For example, Biro et al. (2005) found that juvenile trout raised in low-food lakes took high risks to achieve rapid growth and thereby shortened their duration of exposure to high predation at small size.

In general, a faster life history (higher acquisition rate and earlier reproduction) will be favoured if the mortality risk associated with it is lower than the extra mortality accumulated with a slower life history (lower acquisition rate and later reproduction) (Williams 1966). Increased mortality from fishing may thus lead to the evolution of faster resource acquisition and altered allocation patterns to accelerate growth, as suggested by Case (1978). This prediction is in contrast to the frequently cited expectation that fishing will cause evolution towards slower growth. Most likely there will be selection pressures acting in both directions simultaneously, and depending on relative strengths, evolution of growth may go in either direction. Models have indeed supported our hypothesis, with some predicting evolution of slower growth (Favro et al. 1979), some predicting evolution of faster growth (Enberg et al. 2009; Jørgensen & Fiksen 2010), and others predicting evolution of either faster or slower growth depending on the size-selectivity of the fishery (Boukal et al. 2008; Andersen & Brander 2009; Dunlop et al. 2009b). These model findings challenge the prevailing notion that fishing will always lead to evolution of slower growth.

Challenges in quantifying growth evolution from field data

Fishing-induced evolution of growth has been well documented in selection experiments (e.g.Conover & Munch 2002). In observational field studies, the task is more challenging. In an ideal setting, allelic frequencies at the loci that determine growth rate would be monitored to detect whether growth is evolving (as suggested by Kuparinen & Merilä 2007). Unfortunately, we are still years away from identifying all the genes that contribute to the complex suite of processes and mechanisms affecting growth of wild fish. Until then, much of our analysis of growth evolution in wild populations will have, at best, to focus on a restricted number or genes (or their products) of which various alleles have been associated with differential growth rates (e.g.Case et al., 2006), or more frequently rely on observational phenotypic data, which can provide only indirect evidence for evolution. However, rather than giving up in the absence of genetic data, we believe much can still be learned about fishing-induced evolution of growth from the analysis of phenotypic data. The reality is that phenotypic data are most readily available, notably time-series data for commercial species, and we can still obtain valuable insight as long as the various caveats and challenges are kept in mind.

The first challenge to consider when analyzing phenotypic field data is that size at a given age is influenced by how much resources are diverted to reproduction, if any. In Fig. 3 we give a simplified schematic of how it is possible to attain the same distribution of size-at-age through three different mechanisms: variable growth rate (Fig. 3A), variable timing of maturation (Fig. 3B), and variable investment into reproduction (Fig. 3C). The energetic demands of reproduction are ubiquitous and have strong effects on growth rate that only rarely can be ignored. The challenge of understanding the processes behind changes in size-at-age can be partly overcome by studying juvenile growth because it is not affected by reproductive investment (Heino et al. 2008; Swain et al. 2008).

Figure 3.

 Size at a given age (indicated by the black vertical bars) is influenced not only by growth rate but also by the timing of maturation and the subsequent investment into reproduction. Black lines show growth trajectories for individuals with different trait values, and open circles indicate timing of first reproduction. (A) Different juvenile growth rates, due to adaptive differences in either resource acquisition or allocation to routine metabolism, lead to different growth trajectories and sizes at a given age even before maturation. (B) When the maturation schedule can evolve, individuals with the same juvenile growth rate may have different post-maturation size-at-age depending on their maturation schedule. (C) Variation in reproductive investment can also cause variation in size-at-age later in life, even among individuals with the same growth rate and maturation age. In particular, increased reproductive investment will cause a more determinate growth pattern with a sharp transition between juvenile growth and a more or less fixed adult size (lower growth curve), in contrast to indeterminate growth (top growth curve). In reality, these three different processes are likely to interact and need to be accounted for before changes in size-at-age are equated to evolution of growth. Worth highlighting is that all of these different processes could lead to similar patterns in size-at-age (where the growth trajectories cross the black vertical bar).

Secondly, obtaining representative samples of a fish population to estimate growth rate is difficult. Virtually all sampling methods are size-selective, which is an obvious problem for estimating growth rate, although some biases can be corrected for. The origin of the data makes an important difference: the fishing fleet intentionally targets a certain size range of fishes (usually the larger end of the size spectrum), determined by landing size and gear regulations, geographical distribution, and market valuation of differently sized fish (see Pardoe et al. 2009). Studies based on catch data inherit these biases, and the nature of those biases may change over time as the population or fishing practices change. Scientific surveys are intended to be less selective and more consistent over time, as they follow deliberately designed sampling schemes, aiming also for segments of the population that are not specifically targeted by fisheries (for example smaller and younger individuals). Nevertheless, unbiased sampling over the entire population is virtually impossible.

Thirdly, purely demographic effects confound the estimation of population parameters. Even unselective fishing leads to higher mortality, leaving on average younger and therefore smaller individuals. On top of this, selective harvesting may lead to differential mortalities within a cohort, for example when the largest members of a given age class are harvested and the smaller ones survive, leading to demographic change in mean trait values over time even within one cohort (Sinclair et al. 2002; Swain et al. 2007).

Fourthly, as we have emphasized above, growth rate is generally not a directly observable trait but needs to be inferred from observations of size-at-age. Repeated individual measurements of size-at-age are preferable, for example from recaptures of tagged individuals or through back-calculations of growth patterns from scales or otoliths (e.g.Edeline et al. 2007; Swain et al. 2007), but such samples are more laborious and costly to obtain and are therefore relatively rare.

Last but not least, growth is notoriously plastic, being influenced by both abiotic factors (e.g. temperature) and a range of biotic factors such as predator and prey distributions and density dependence. Conceptually, phenotypic plasticity can for example be accounted for by estimating reaction norms describing how growth varies with environmental factors, similar to what has been proposed to help disentangle phenotypic plasticity and evolutionary change in maturation (Heino et al. 2002; Dieckmann & Heino 2007). In practice, using a similar methodology for growth may turn out to be more difficult because whereas observations of maturation often carry along with them individual-level information on growth (a major source of plasticity in maturation), sources of growth plasticity are many and mostly difficult to measure even at the level of a population, not to mention that of an individual.

Empirical Evidence of Fishing-Induced Evolution of Growth: What Have We Learned?

The body of literature relating to fishing-induced evolution of growth is diverse in both the species studied and the methodology used (Table 3). Most investigations rely on the analysis of time-series data from field studies, but also include experiments performed in laboratory settings (Conover & Munch 2002) and in semi-natural ponds (Biro & Post 2008). In our view, growth evolution has been most rigorously documented in pink salmon (Ricker 1981, 1995) and in the experiments on silversides (Conover & Munch 2002, Conover et al., 2009). Part of the reason these two studies were able to clearly document fishing-induced evolution of growth was because both species examined had a constant maturation age; thus, the confounding factors of maturation and reproductive investment (Fig. 3) could be excluded. However, at the same time that a constant maturation age makes the results easily interpretable, it also limits the generality of the conclusions because most fished species have more flexible reproductive schedules. In the case of silversides, declining food consumption rates were indicative of evolving resource acquisition, and changes in fecundity and egg size indicated that allocation to reproduction had evolved as well (Walsh et al. 2006). In addition, food conversion efficiency in the silversides declined, suggesting changed allocation also to some of the more subtle costs such as immune defence or maintenance.

Table 3.   Studies examining evolutionary changes in growth caused by fishing. We have included studies where the authors themselves acknowledge fishing-induced evolution as a possible cause for the observed changes, and some studies which in our opinion suggest evolutionary changes in growth due to fishing.
SpeciesType of studyProxy for growthAgent of selectionMain findingReference
Pink salmon Oncorhynchus gorbuschaCommercial catch statistics and data from sampling programSize at maturation (fixed maturation age and semelparous life history indicate that size at maturation reflects juvenile growth rate)Gillnetting, trollingDecreased size at maturation over timeRicker 1981, 1995
Rainbow trout Oncorhynchus mykissExperimental (semi-natural ponds)Specific growth rateGillnettingBold and fast-growing individuals with active foraging behaviour were first to be fished out from the experimental pondsBiro & Post 2008
Atlantic salmon Salmo salarTime series data from trapping fish on their spawning runWeight and length at ageAnglingDeclines in body weight and length due to selection caused by anglingSaura et al. 2010
Lake whitefish Coregonus clupeaformisMostly commercial catch data, some from sampling programWeight at ageGillnettingSomatic growth rate and condition factor had decreased as a result of girth-selective gillnet fisheryHandford et al. 1977
Common whitefish Coregonus lavaretusCommercial catch data and surveysBack-calculated length at ageGillnettingDeclining somatic growth and increased investment into reproduction due to gillnet harvestingThomas & Eckmann 2007; Thomas et al. 2009
Alpine whitefish Coregonus palaeaData from monitoring program, sampled with gillnetsBack-calculated length at age 1 (reflecting juvenile growth rate), back calculated growth rate from age 1 to capture (reflecting adult growth rate)GillnettingSignificantly negative selection differentials were detected for a coefficient that described growth in the adult phase. No selection on juvenile growthNussléet al. (2009)
Atlantic silverside Menidia menidiaLaboratory experiment over four generationsLength at age 190 days (before maturation)Artificial selectionSelective harvesting on large individuals led to decreased size-at-age, whereas selective harvesting on small individuals led to increased size-at-ageConover & Munch (2002); see also Walsh et al. (2006); Conover et al. (2009)
Atlantic cod Gadus morhuaBottom-trawl survey dataBack-calculated length at ages 3 and 4 yearsTrawlingFisheries-induced evolution towards slower growth rate had likely occurredSwain et al. (2007, 2008)
Pike Esox luciusTime series of gillnet fishing (culling and scientific sampling)Asymptotic length from von Bertalanffy growth curve fitted to individual back-calculated length-at-age dataGillnettingA pattern of selection towards reduced asymptotic length during some periods was observedEdeline et al. (2007, 2009)
Plaice Pleuronectes platessaOtolith data from surveys and market samplesEnergy allocation model fitted to back-calculated individual growth trajectoriesTrawlingThe method specifies a growth model based on allometric scaling functions for acquisition, maintenance, and reproduction, with growth emerging as the resultant outcome of these processes. It enables quantifying temporal changes in each of these processes as well as maturation ageMollet et al. (2010)
Haddock Melanogrammus aeglefinusSurvey data, 1965–2001 year-classesYear-class-specific length-at-day of ages 5–10TrawlingDeclined length at age and age at maturity not explained by temperature or population densityNeuheimer & Taggart (2010)

Despite the difficulty of generalization to many commercial species, the case studies on pink salmon and silversides nonetheless make a strong case that fishing has the capability to induce evolutionary change in resource acquisition and allocation. In the silverside experiment where most confounding factors were controlled, the strength of the evidence approaches proof that fishing can cause evolution of growth, both in terms of changes in resource acquisition and allocation.

The influence of reproduction

As we hope is now appreciated, observing change in size-at-age is not sufficient evidence that growth rate has evolved, as there are many potential factors that could cause those changes. In particular, one needs to exhibit caution when interpreting changes in size-at-age or growth in a species where age at maturation or reproductive investment might change in response to fishing. Such caution seems warranted in the study of Lake Windermere pike by Edeline et al. (2007, 2009) where changes in von Bertalanffy growth parameters (including asymptotic length) were seen as evidence for growth evolution. The von Bertalanffy parameters aggregate size-at-age information and do not allow the separation of resource acquisition from resource allocation. This is a problem because, as Edeline et al. (2007) point out themselves, gonad weight decreased over time, especially in smaller young female pike. This leaves open the possibility that changes in reproductive investment could be underlying the reported trend in asymptotic length. Therefore, although the Windermere pike study has been lauded as providing evidence for fishing-induced evolution of growth (Conover 2007; Coltman 2008), the conclusions that can be drawn are not so straightforward.

Increased reproductive investment is a commonly predicted response to fishing mortality (Law & Grey 1989; Rijnsdorp et al. 2005). The most common explanation for this type of pattern is that those individuals that invest more in current reproduction (even at the cost of slower growth) produce higher numbers of offspring over their lifetime than those that invest less in reproduction. However, it is important to keep in mind that selection could act on reproductive investment in other ways. For example, in pink salmon (Ricker 1981, 1995) and whitefish (Handford et al. 1977; see also Hamley 1975) the high condition factor and high ratio of maximum girth over body length, respectively, imply that gillnets could select on reproductive investment either directly, by removing fish with large gonads, or indirectly, by fishing out individuals with large energy stores that later would be used for reproduction. It is thus interesting to observe that in a fished population of lake whitefish, reproductive investment increased over the same period as sizes decreased (Thomas et al. 2009). Thus, simultaneous change towards lower condition and larger gonads suggests that fecundity selection acts simultaneously with viability selection imposed by the fishing gear.

Correlations between traits

Size-selective fishing lead to changes in life-history, behavioural and morphological traits in the experiments on silversides (Conover & Munch 2002; Walsh et al. 2006; Chiba et al. 2007), and it will be interesting to see whether further experiments can reveal to what degree these concurrent responses are due to genetic correlations (Naish & Hard 2008) as opposed to being independent processes of adaptation. In the pond experiments on rainbow trout by Biro & Post (2008), growth was correlated with behavioural traits; because gillnetting harvested the most active fish, there was selection on feeding behaviour that also led to selection on other physiological traits related to growth. These studies suggest that considering different classes of traits together (e.g. behavioural, morphological, physiological) is necessary to paint the full picture of how fishing can lead to evolutionary growth changes in harvested species.

Selective gear, but on which trait?

A surprisingly diverse range of fishing gear has been shown to exert selection on growth and size-at-age (Table 3).These gears are both active (trawls, angling with lures) and passive (gillnets) and can select directly on girth (trawls, gillnets), feeding motivation (angling with lures), activity (gillnets), and potentially also on swimming and escape ability (trawls). Indirectly, selection on morphological traits such as girth may affect the mechanisms of building stores, growing large gonads and expressing prominent secondary sexual characters. Thus, the ultimate goal in understanding how gear selection may lead to changes in growth rate or size-at-age is piecing together a mosaic of the many mechanisms mentioned in Tables 1 and 2. This perspective may at first glance seem discouraging, but in our opinion it can also inspire new experiments and statistical analyses of available data, which will move the entire field of fishing-induced evolution forward.

Accounting for the environment

As phenotypic plasticity can have strong effects on growth rate, it is important to account for environmental factors. Studies of fishing-induced evolution often rely on time-series analysis or at least on consistent comparisons between separate periods. Unfortunately, appropriate environmental data are often unavailable or only available for part of the time-series. For example, one of the earliest studies of fishing-induced evolution showed that somatic growth rate and condition factor of lake whitefish had decreased, just as one would expect as an evolutionary response to the highly girth-selective gillnet fishery with a large mesh size (Handford et al. 1977). However, the data did not permit Handford to draw strong conclusions about whether the observed changes were evolutionary or caused by phenotypic plasticity, as only limited information about the environment was available. The whitefish population had concurrently undergone a collapse with delayed recovery, indicating large changes in density dependence and potentially also other ecological changes (Bell et al. 1977).

Despite the obvious challenges posed by studying a highly plastic process such as growth, it is encouraging that several recent studies have been able to account for important abiotic and biotic environmental factors such as ambient temperature, eutrophication and conspecific biomass (e.g.Swain et al. 2007; Thomas & Eckmann 2007; Neuheimer & Taggart 2010). Even after adjusting for certain environmental factors, these studies suggest residual trends in size-at-age that are the result of selection pressures from fishing (Swain et al. 2007; Thomas & Eckmann 2007; Neuheimer & Taggart 2010).

Conclusions and Future Directions

A promising avenue for future investigations is to build on the approach taken by Swain et al. (2007, 2008) and study selection differentials. The response R of a trait to selection S (the difference in the trait in the parents of the next generation compared to the trait distribution in the same cohort at birth) is h2S, where h2 is the heritability that typically takes values around 0.2–0.3 for life-history traits, at least in gadoids and salmonids (Gjedrem 1983; Carlson & Seamons 2008). Heritabilities in this range imply that the selection differential S is three to five times stronger than the evolutionary response R (comparing one generation with the next), and S may therefore be easier to quantify, including the separation of effects of fishing from effects of the environment. Another advantage is that selection differentials can be quantified by cohort. Individuals belonging to the same cohort often experience similar environmental conditions, hence comparison within a cohort reduces the influence of environmental variation on phenotypically plastic traits.

However, there are two inherent problems when studying phenotypic selection differentials. First, one has to measure a given trait for fish of different ages. For size this may be accomplished through back-calculations of size-at-age from scales, otoliths or other structures (as in Swain et al. 2007). The second challenge is that the sampling method should ideally be unselective for all types of fish sampled. For gears with baits there is likely a correlation between acquisition and catchability. If growth rate decreases in the population, one may catch fewer and fewer fish but the samples (collected with baited gear) could be dominated by the fastest growing fish along the entire time-series, meaning that the change in growth could be underestimated or not detected at all. Most fishing gear, even the types used in research surveys, are selective on size, condition, satiation, activity or swimming speed, and arriving at the correct conclusions may turn out to be sensitive to any deviations from random sampling.

Another promising method is the process-based bioenergetics method that fits process parameters to individual growth trajectories (Mollet et al. 2010). The data requirements are similar to that of Swain et al. (2007), and it will be interesting to see how the method performs when used to study temporal change and when applied to other species.

In summary, studying fishing-induced evolution of growth rate requires careful consideration of the many processes involved in resource acquisition and allocation. As with all aggregated phenomena, a deeper understanding requires that the intertwined sub-processes are teased apart and studied in more detail. Accounting for all the components will, at least in field studies, remain impossible. However, this difficulty should not dissuade us from trying to identify the main components. One important point from our review is that one should distinguish between size, which describes a state, and growth, which is a multitude of processes leading to a given size. This includes acknowledging and correcting for plasticity and environmental effects and simultaneously accounting for changes in other important life-history traits, particularly maturation schedule and reproductive investment. For example, developing methods to apply reaction norms to traits other than maturation and thereby incorporating effects of environmental variables on growth will likely help detect evolutionary changes. By involving other approaches we can hope to expand the range of suitable data also for species and stocks where environmental data are otherwise unavailable, and broaden the scope of studies of fishing-induced evolution of growth. Examples of these other approaches include stable isotope analyses from scales and otoliths or oceanography modelling for hind-casting environmental conditions and environmental monitoring of potential prey or predator species. Because growth integrates many dimensions of environmental influence with physiological function, it will be exciting and necessary to see how the phenotypic view of life history changes compares with genetic data when those become available.

Many of the mechanisms that can lead to increased growth rate due to fishing-induced evolution entail a riskier life and could increase natural mortality rates (e.g.Jørgensen & Fiksen 2010). Growth evolution will affect both species ecology and fisheries economics but the devil is in the detail: the consequences will depend on which traits are adapting and how. While increased individual growth is sometimes seen as beneficial for the productivity of a fish stock, the positive effects could be counteracted by a larger loss through predation and other sources of natural mortality (Swain 2011). And while slow growth means smaller fish, these adaptations are likely to help individuals survive to reproduce and populations persist in an environment dominated by fishing. Earlier maturation may also have a positive influence on stock reproduction, although this does not necessarily propagate to an increase in harvestable biomass or catches (Enberg et al. 2009). Whether productivity and viability will increase or decrease will thus depend on the evolution of multiple life-history traits, and no thorough analysis exists yet. Many challenges of demonstrating and interpreting growth evolution and its consequences lie ahead, and an increased understanding requires that multiple biological disciplines together colour the rich picture of intertwined mechanisms and the complex layers of confounding factors.


We thank J. Hard for help with the salmonid literature and Ray Hilborn and anonymous referees for constructive comments on the manuscript. Funding was provided by the Bergen Research Foundation, the Research Council of Norway, the EU Marie Curie Research Training Network FishACE, EU FP7 Integrated Project MEECE, and the Academy of Finland.