1. Some types of flexible foraging behaviours were incorporated into ecological thought in the 1960s, but the population dynamical consequences of such behaviours are still poorly understood.
2. Flexible foraging-related traits can be classified as shifts in general and specific foraging effort, and shifts in general and specific defense.
3. Many flexible foraging behaviours suggested by theory have received very little empirical attention, and empirical techniques used to compare the magnitudes of behavioural and non-behavioural responses to predation are likely to have overestimated the behavioural components.
4. Adaptively flexible foraging in theory causes significant changes in the forms of consumer functional responses and generates a variety of indirect interactions. These can alter fundamental ecological processes, such as co-existence of competitors, and top-down or bottom-up effects in food webs.
5. Many aspects of flexible foraging are still largely unknown, including the issues of how to represent the dynamics of such phenotypically plastic traits, how flexible traits in multiple species interact, what types of adaptive movements occur in metacommunities, and how adaptive behaviours influence evolutionary change.
6. Population dynamics in large food webs may be less dependent on behavioural flexibility than in small webs because species replacement may preempt some potential types of behavioural change within species.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Trophic relationships are an integral and essential part of every natural community and ecosystem. Naturalists have long been aware that foraging traits (behaviours as well as other forms of individual plasticity) are flexible, and may depend on the condition of the forager and the abundances of various foods, predators and competitors. Given this, it is surprising that considerations of adaptive foraging were almost entirely missing from ecological theory about predator–prey interactions and food webs for several decades after Lotka and Volterra independently laid the foundations for that theory. This lack of theory was reflected in a similar lack of quantitative empirical studies. In 1980, the only quantitative models in this subject area dealt with switching behaviour (Murdoch 1969) and diet choice (MacArthur & Pianka 1966; Schoener 1971; Charnov 1976) in consumer species. Work on diet choice was largely restricted to short-term behavioural studies that did not consider the consequences for population or community dynamics.
Flexible foraging is inextricably related to flexible defense, because many foraging decisions affect the vulnerability of the foragers to their own predators. Before 1980, the implications of adaptive defense had been almost totally ignored. Much of my own research over the past 30 years has been devoted to building a theoretical framework for understanding the population and community dynamical implications of adaptively flexible foraging and defense, and this is now an active and growing research topic in ecology (for reviews see Bolker et al. 2003 and Werner & Peacor 2003). Nevertheless, there are, at best, very few systems in which we have an empirically-supported model of community dynamics that includes quantitative descriptions of the main types of flexible foraging. In addition, empirical quantification of some of the foraging-related effects that have been studied may have provided a biased estimate of the importance or magnitude of the effects. The present article has several goals: (1) to advance a general framework for classifying the impacts of adaptive foraging and defense in food webs; (2) to review what is known about some of the most important consequences of these adaptations in systems with a small number of species; and (3) to suggest what areas of future theory and empirical work are most needed.
A classification of adaptively flexible foraging traits with a summary of current knowledge
The vast majority of adaptive feeding and defensive traits fall into the following four categories: (1) traits that determine the absolute foraging time or effort; (2) traits that determine the relative intakes of different foods; (3) traits that determine the total amount of predator exposure; (4) traits that determine the relative exposure to different types of predators. These four categories may be linked in different ways, but increasing the relative foraging or defense directed at one type of food or predator usually entails decreased intake of other foods or increased exposure to other predators, respectively. General intake is usually correlated with general risk. Thus, there are a limited number of trait categories and tradeoffs that constitute a large fraction of the flexible responses that might influence population dynamics. These few categories of traits result in a proliferation of indirect interactions in food webs, as the behaviour of one species will generally be a function of the behaviours as well as the densities of several other species (Abrams 1992a,b; Abrams, Matsuda, & Harada 1993). In addition, different functional forms of tradeoffs can in some cases reverse the direction of the response to a change in food or predator abundance (Abrams 1984). Finally, the level of ‘specificity’ of foraging and defense varies greatly between different traits and systems. The net result is that the range of indirect effects between species in the food web increases enormously. Because these indirect connections increase in number with greater numbers of species in a food web, it can become much more difficult to determine how a given environmental perturbation will alter the density of a particular species. Nevertheless, understanding the basic impacts of these categories in the context of systems with 3–5 species is a prerequisite to understanding their impacts in larger food webs.
In much of this article, I will refer to ‘behaviours’ rather than ‘traits’, but most of the phenomena discussed also apply to modalities of adaptive change other than behaviour. These include developmental plasticity, evolution, and species replacement within communities. I begin with a short review and summary of current knowledge about the four categories of flexible foraging traits proposed above.
Relative foraging effort
Behaviours related to relative foraging effort were the first aspects of adaptive foraging to gain widespread recognition. Such behaviours include habitat choice, search image formation, and other aspects of foraging mode that increase encounter with one prey at the expense of encounters with others. Theory regarding such behaviours began in the 1960s, with Murdoch’s work on predator switching behaviour (1969), and MacArthur & Pianka’s (1966) seminal work on diet choice. There was an initial burst of interest in how switching altered the outcome of interactions between one predator and two or more prey species, which was mainly concerned with its effect on system stability and co-existence of competing prey species (Murdoch & Oaten 1975; Oaten & Murdoch 1975; Roughgarden & Feldman 1975). The impact of switching on apparent competition between prey was mentioned by Holt (1977), but was not given a detailed treatment until later (Holt & Kotler 1987; Abrams & Matsuda 1996, 2003, 2004; Abrams 1999; Křivan & Eisner 2006). Switching usually reduces, and may eliminate apparent competition. Adaptive diet choice theory generated a large subfield in behavioural ecology (reviewed in Stephens and Krebs 1986), but it failed to be incorporated into models of the dynamics of interacting species until much later (Gleeson & Wilson 1986; Fryxell & Lundberg 1994; Ma, Abrams, & Brassil 2003).
Switching and diet choice based on energy and handling time are not the only categories of change in relative foraging effort. A wide variety of costs and benefits may be associated with increased relative foraging, and it is likely that only a fraction of the possible combinations of costs and benefits have been modelled. One benefit structure that has been examined is adaptive choice when foods are nutritionally complementary or essential (sensu Leon & Tumpson 1975). This generally produces ‘anti-switching’ (i.e., increased effort directed towards rarer resources; Abrams 1987a). Such anti-switching was shown to be a potent driver of population cycles (Abrams & Shen 1989), and to lead to nonlinearity in competitive interactions between two consumers of two essential resources (Abrams 1987b). If resources are nutritionally complementary rather than essential, adaptive changes in foraging effort can change between switching and anti-switching depending on relative abundances, leading to complicated functional response shapes (Abrams 1987a). These possibilities remain unexplored in the empirical realm, in spite of a renewed interest in nutritional interactions under the rubric of stoichiometry (e.g. Sterner & Elser 2002). Toxins in foods and gut capacity constraints in foragers have been shown by theoretical work to have major impacts on foraging that cannot simply be described as switching or anti-switching (Abrams 1989, 1990); none of these appears to have received much empirical attention either (but see Beckerman 2005).
Surprisingly, general foraging effort was not formally explored until after a considerable literature on switching had developed. Early work includes Comins & Hassell (1979) and Abrams (1982). General foraging effort consists of responses that balance intakes of all foods with costs of feeding, including metabolic expenditures, intake of toxins, and risk of mortality due to predators and other factors. General effort may include decisions affecting search strategy, movement speed, and choice of foraging location. Balancing costs and benefits could produce acceleration or deceleration, changes in curvature, and even declines in the functional response to total prey (food) density (Abrams 1982, 1989). However, there has been little empirical attention to these possibilities or their population dynamical consequences. Subsequent articles have repeated the earlier view that type-3 functional responses only result from switching (e.g. van Baalen et al. 2001). Early work in this area (Abrams 1982, 1984) noted that, when foraging was costly, increased food abundance could in either increase or decrease foraging effort. Reviews of empirical studies in the 1990s came to different conclusions about the relative frequencies of these two responses, with Abrams (1991a) finding equal frequencies of both responses in a previous work, while Werner & Anholt (1993) arguing that reduced effort with increased food abundance was the only pattern observed. Decreased effort in response to greater abundance represents a potential mechanism for a type 2 functional response, but studies of functional responses have failed to distinguish this adaptive mechanism from the constraint of a fixed handling time.
In many cases, the cost of increased foraging is greater exposure to predators (Abrams 1984; Lima & Dill 1990). In this case, general foraging effort is linked to defense, which is discussed below.
As Bednikoff (2006, p. 306) noted, ‘Food is generally good for the forager, but not if the forager is dead.’ It is somewhat surprising that the connection between food intake and mortality risk only slowly gained attention after more than a decade of work on diet choice. Following a suggestion by Charnov, Orians, & Hyatt (1976), adaptive anti-predator behaviour was first incorporated into models of interacting species nearly 30 years ago (Comins & Hassell 1979; Abrams 1982, 1983, Abrams 1984; Sih 1984). Very similar models were presented and labelled ‘the ecology of fear’ by Brown, Laundre, & Gurung (1999). In most of this work it was assumed that there was a tradeoff, such that increasing defense entailed reduced food intake. Some of the fundamental ecological consequences of general risk-reward tradeoffs were laid out in some of the earliest work on the subject. One fundamental implication of adaptive defense by prey is to make the predator’s functional response depend on its own density (Abrams 1982, 1984, 1991a,b, 1995). As a result, predation rates should not be proportional to predator density (Abrams 1993), and the predator–prey system should often be more stable (Abrams 1984). General defense was also shown to be capable of modifying competitive interactions between predators and apparent competition between prey species that share predators (Charnov, Orians, & Hyatt 1976; Abrams 1984, 1987c). Predator-dependent functional responses have been observed (Skalski & Gilliam 2001), but apparently there has been no attempt to determine whether adaptive defense was the underlying cause.
Perhaps the most important ecological consequence of adaptive adjustment of general defense is that it can produce a variety of indirect effects between separated trophic levels on a relatively short time-scale. This fact, first noted in Abrams (1984), is now usually discussed under the rubric of ‘trait mediated indirect effects/interactions’ (Abrams 1995; Peacor & Werner 1997). Such indirect effects seem to provide a good explanation for Schoener’s (1993) finding that there was not much difference between the time-course of direct effects and indirect effects following perturbations to real food webs. Recent reviews of experimental studies of responses to predation risk (Werner & Peacor 2003; Bolnick & Preisser 2005; Preisser, Bolnick, & Benard 2005) have all suggested that indirect effects of predators on their prey’s resources are often large in magnitude. There is much less evidence for the reverse effect of the abundance of the prey’s food on the fitness of the predator (reviewed by Werner & Peacor 2003; Bolnick & Preisser 2005), but this seems likely to be a reflection of the scarcity of experiments designed to look for such effects. The asymmetry in scientific attention between these downward and upward effects is surprising, because the often studied behavioural effect of predators on their prey’s resource is always expected to be positive. The upward effect of resources on predator may be either positive or negative (Abrams 1984, 1991a,b, 1995; Werner & Anholt 1993), and, as noted above, we still have insufficient data to say which of these directions in more likely. The ability to induce antipredator behaviour using predator cues has led to much greater empirical attention to the population dynamical consequences of behavioural responses to predators than to food.
There is little doubt that the many short-term experiments that have used cues to measure the population level consequences of defense have overestimated those effects (Abrams 2008a). This is in a large part because their time-course was too long to avoid the mixture of effects from behavioural change and density change, while also being too short for all of the density mediated effects to be fully manifested. Because behaviour changes faster than population densities, behavioural effects represent a much larger portion of the initial phase of the transient response of the system (Werner & Peacor 2003; Abrams 2008a,b). Even with this bias, the results of many experiments (Werner & Peacor 2003) make a strong case for the existence of sizable behavioural effects and for the need to quantify foraging/predation related behaviours as part of any predictive food web model.
Theoretical studies of defenses that decreased exposure to one predator at the expense of increased exposure to others began in the 1990s (Matsuda & Namba 1989, 1991; Matsuda, Hori, and Abrams 1994; Matsuda, Hori, & Abrams 1996). Specific defense was shown to often allow co-existence of competing predators, and to be capable of producing a mutualistic interaction between them (Matsuda, Abrams, & Hori 1993). These effects require that defense be effective against only a subset of the potential predators, and that defenses are mutually exclusive (or at least interfere with each other). There have been empirical examples of such defenses (e.g. Soluk & Collins 1988; Soluk 1993; Soluk & Richardson 1997). However, the specificity of defense and the tradeoff relationships between different types of defense remains unexplored in most species known to have some form of adaptive defense.
This rather cursory overview of current knowledge of the four basic types of flexible foraging and defense leads to a consideration of areas where more work is needed. It should be clear from the above that there is much that remains to be learned about the basic effects and their importance in different species, different habitat types, and different food web configurations. However, there are a number of research topics that extend our understanding in new directions, and that have only recently begun to be examined either theoretically or empirically. Some of these are summarized below.
Research topics for the future
Multiple interacting adaptively flexible traits
The examples mentioned above have mostly considered single adaptive traits that affect interspecific interactions. These can produce a trait mediated indirect effect between two species which both interact directly with the species whose trait changes. However, most food webs are much larger than those considered in the previous section, and it is very likely that adaptive changes in multiple species will propagate indirect effects rapidly, potentially to species that are topologically rather distant in the web. The possible food web effects introduced by flexible foraging of two or more species may be illustrated by considering Fig. 1, which shows a hypothetical food web consisting of five species distributed among three trophic levels. This is both far simpler than natural systems and significantly more complicated (i.e., more species and more links) than the models that have been used in the majority of theoretical analyses. One of the many questions that might be asked about such a web is what impact reduced harvesting of the top predator is likely to have on the absolute and relative abundances of its two prey on the middle level. Even without considering adaptive behaviour, this requires accurate estimates of the functional responses of all of the consumer species.
Assuming that reduced harvesting increases predator density (which may not be true; Abrams 2002, 2009; Abrams & Matsuda 2005), one must then determine whether the resulting greater predator population has a larger negative impact than the positive indirect impact resulting from decreased competition from a less abundant competitor. The addition of adaptive behaviours adds several important processes that may affect this conflict between direct and indirect influences. Each of the arrows describing flows of energy from one species to another is likely to be affected by every one of the five populations.
Here is one of many potential pathways of effect that might arise from flexible foraging, assuming that it is safer for prey species 1 to forage for resource 1 than for resource 2. The initially larger numbers of predators can cause prey (consumer) 2 to be less active and also to consume relatively more of the safer resource 1. As a result, individuals of resource 1 hide more. This causes prey 1 to increase its general foraging activity. The resulting increased exposure of prey 1 causes the predator to increase its relative foraging on prey 1, with a concomitant decrease in foraging on prey 2. This may be more than sufficient to decrease the net predation pressure on prey 2, in spite of the initial increase in predator density. What this account makes clear is that accurate prediction of changes in density is likely to require a reasonably good quantitative understanding of how the foraging and defensive behaviours of each species changes in response to the densities and behaviours of other species. This is most likely to emerge from targeted behavioural studies of various subsets of the five species involved. The considerations discussed for this hypothetical web are likely to be important for predicting the consequences of human impacts on natural webs; some of these are discussed in a recent review (Heithaus et al. 2008) of the consequences of reducing top predator populations in marine webs.
Some of the theory for systems having interacting adaptive foraging/defense in two species was developed in the early 1990s. Abrams (1992b) showed that foraging-risk tradeoffs in the two middle species of a four-species food chain could result in positive effects of increased density of at least one of the predators on their prey’s per capita growth rate, and negative effects of increased prey density on predator per capita growth rate. In addition, the density of the bottom species affected the per capita growth rate of the top species on a behavioural time-scale. In fact, all four species’ densities appeared in each species’ per capita growth rate function. Abrams & Matsuda (1993) investigated a model in which a switching predator consumed two prey, each of which adjusted its defense adaptively. The result was often negative switching, rather than the positive switching expected in the absence of flexible prey defense. Yamauchi & Yamamura (2005) have studied the dynamics of this system when all traits change via evolution. There are no doubt many other, as yet undiscovered, effects that come about from the interaction of flexible foraging in groups of three or more interacting species. The effects predicted by Abrams (1992b) or Abrams & Matsuda (1993) do not appear to have been examined in any field or laboratory system.
Adaptive foraging and defense based upon movements in metacommunities
Both the discussion of Fig. 1 and most previous theory has implicitly assumed that all populations occur in a single well-mixed system. The nature of the tradeoffs involved in both diet choice and adaptive risk-taking typically differ when those tradeoffs are based on the choice of location in a spatially heterogeneous area. Most theory about adaptive behaviour in heterogeneous habitats has been based on metapopulation models (Levins 1969). Because the initial theory for single-species metapopulations used presence/absence rather than population size to characterize a particular patch, it was fundamentally inconsistent with the framework developed (and required) to understand the details of food web interactions.
More recent work has often adopted a description that includes population size and has included several species; models of interacting species using this framework are now discussed under the rubric of metacommunities (see Holyoak, Leibold, & Holt 2005). Two of the major determinants of dynamics in metacommunities are the rate of movement between patches, and the patch choice of those individuals that do move. Food, predator, and conspecific densities within a patch are likely to be major determinants of patch choice. Although adaptive movement has rarely been incorporated into metacommunity models, the studies that have been done show that it can greatly alter population dynamics and species composition (Schwinning & Rosenzweig 1990; Heithaus 2001; Abrams 2000, 2007; Abrams, Cressman, & Křivan 2007; Amarasekare 2007). Patch choice can be considered a trait that often involves tradeoffs that are linked to foraging, but it has specific features that differ from behaviours that affect foraging within homogeneous systems. For example, Abrams (1999) and Abrams & Matsuda (2004) discuss some of the similarities and differences between switching when the foods in question are located in the same or different patches.
One important aspect of movement in metacommunities and other spatially heterogeneous systems that differentiates it from phenotypic tradeoffs is the role of attraction or avoidance of conspecifics and of other species that are neither food nor predator. The predominance of the type-2 functional responses in empirical studies (Jeschke, Kopp, & Tollrian 2004) argues that conspecific attraction (or attraction to other prey species) should be a potent mechanism for reducing risk. It has also been shown to be a potent mechanism for producing temporal and spatial instability in the food web (Schwinning & Rosenzweig 1990; Abrams 2007). Conspecific attraction may also at least temporarily increase foraging intake in some situations with patchy resources (Stamps 2001). If conspecifics are abundant, predators are likely to be satiated, and pose less of a risk. However, over time, being in a high density area reduces food intake and is likely to attract more predators, requiring movement by the prey. Different assumptions about whether the predator exhibits saturating functional responses have led to very different conclusions about the impact of patch choice in predator prey systems (cf. Křivan 1997 with Abrams 2007). Conspecific avoidance may also occur. Such behaviours in prey species are likely to be adaptive when predators are quickly attracted to high density areas and have functional responses that do not display pronounced saturation. The effects of such prey over-dispersion on predator prey interactions has yet to attract much scientific attention.
The dynamics of behavioural change
The underlying message derived from previous research (Abrams 1999, 2003; Ma, Abrams, & Brassil 2003) is that different, plausible assumptions about the time-course of behavioural change can yield very different population dynamics. This section briefly considers different ways in which some of the behaviours described above have been incorporated into models, and stresses how little is known about this topic. However, this discussion is not aimed exclusively at modelers. The fact that very different models of the same behaviour can both be termed ‘plausible’ is a reflection of the relative lack of empirical investigation of anything more than the most basic aspects of the dynamics of the behaviours. The most common approach in theoretical work thus far assumes instantaneous quasi-optimization of fitness. The optimization is based on a tradeoff that depends on the population densities (and perhaps behaviours) of other species. If the behavioural trait is a quantitative one (e.g. amount of time spent foraging, which can be represented by a continuous variable), the model involves determining the fitness maximizing value of the trait, and substituting it into the model. If the trait is qualitative (e.g. occupy patch 1 or patch 2; ignore or include a particular food) then the optimum is often an ‘all or none’ rule; occupy patch 1 if it is even slightly more rewarding than patch 2. Some authors have incorporated this type of step function into dynamic models (e.g. Křivan 1997, 2007), but such non-differentiable functions usually complicate the analysis of the model and are biologically implausible. As a result, it has been more common to represent such qualitative choices using an S-shaped approximation to the step function (e.g. Fryxell & Lundberg 1994, 1998; van Baalen et al. 2001). However, both of these approaches are based on an instantaneous response to changes in the densities of all organisms that influence fitness. The S-shaped function implies that some individuals make mistakes in their assessment of current conditions. It also assumes that equal proportions make mistakes when conditions are getting better and when conditions are getting worse, and that performance does not improve over time if conditions remain constant.
The above approaches fail to incorporate two important features of adaptively flexible behaviours. The first is that all behavioural changes involve some time-lags. It usually takes time to detect a change in conditions, and it may take time to learn to effectively employ or learn a new feeding behaviour, or to become familiar with a newly occupied patch. Individuals that react quickly are likely to make more mistakes, and pay the cost of a greater frequency of changes. A second feature ignored by ‘instantaneous fitness maximization’ models is that the magnitude of the fitness gain from altered behaviour should influence both the rate of the behavioural change, and the ultimate proportion of individuals that make the correct decision. Both of these features are aspects of behavioural dynamics, which are ignored when optimization approaches are used. The most important problem here is that the dynamics of behavioural change may prevent individuals of all species from achieving a fitness maximizing state.
Some of the implications of behavioural dynamics in foraging contexts have been investigated for the case of switching and diet-choice behaviours (Abrams 1999; Abrams & Matsuda 2003, 2004; Ma, Abrams, & Brassil 2003). For both behaviours, the differences from the standard optimization approach have the potential to greatly alter system dynamics, generally producing more complex cycles or chaos, and often expanding the range of parameters characterized by instability. On the other hand, lags can actually reduce the amplitude of cycles (e.g. increasing the minimum densities) in systems that would cycle in the absence of lags.
It is fair to say that relatively little is known about the consequences of different types of behavioural dynamics, even in theory. The work reviewed in the previous paragraph has left out population structure (and hence life history), and has assumed no costs associated with the process of changing behaviour. The rate of changing behaviour was assumed to be an increasing function of how rapidly instantaneous fitness changed with a unit change in the behaviour (see e.g. Abrams & Matsuda 2004). The adequacy of this representation has not been examined empirically. Just adding a cost to the process of behavioural change makes it likely that multiple behavioural types will be maintained in a population (Abrams, unpub.) Building better models requires more empirical work on the dynamics of change in foraging and defensive behaviours.
Even when instantaneous quasi-optimization is a good approach to modeling behaviour, the nature of departures from the optimal strategy can have surprisingly large impacts on dynamics. Abrams & Matsuda (2003) adopted a model in which behavioural switching was instantaneous with errors, but the accuracy of the ultimate choice between prey depended on the difference between their densities rather than the ratio (which is the usual assumption). This meant that choice of the better option was very inaccurate when both population sizes were low, but the ratio of more to less abundant was large. Reduced switching at low densities was found to greatly alter the dynamics of systems with unstable dynamics and to alter the indirect effects between prey in those systems. The dynamics of the two prey, which would be synchronized in models with switching based on optimization, become desynchronized. This in turn increased the magnitude of apparent competition between prey, and the likelihood of exclusion of the slower growing prey. Similarly, relatively small changes in the dynamics of prey preference were shown to make large changes in population dynamics (cf. Abrams (1999) and Abrams & Matsuda (2004)).
If there are transitions between a fixed number of discrete strategies, the change in their relative prevalence in the population may often be modelled by the same methods used to describe patch selection in metapopulation models (see below). Abrams (2000) showed that different rates of between-patch movement in a system with temporal variation in patch quality can make a huge difference for the net intake of the consumer. Paradoxically, intake often decreases with faster movement due to increased overexploitation of patches. A variety of different models for the dynamics of patch choice have been proposed (e.g. Bernstein, Auger, & Poggiale 1999; Alonzo 2002; Armsworth & Roughgarden 2005). Abrams (2007), Abrams, Cressman, & Křivan (2007) review previous approaches, and show that the details make a large difference for a model in which two competitors move adaptively based on local fitness in a two patch environment. Amarasekare (2007) has shown that the type of movement rule is an important determinant of the outcome of interactions in a three-species, three-patch system with intraguild predation, and with movement based on population densities rather than fitness.
The community-level implications of flexible foraging and defense
The work reviewed above shows that the structure of any community model that incorporates flexible foraging and defense will differ considerably from one that does not. Given the variety of questions about community structure and responses to perturbations that could arise, knowing the appropriate structure of a model is considerably more important than any particular prediction about the impact of behaviour on a particular community phenomenon. However, some impacts on larger-scale ecological processes are frequently of interest, and some of these have been considered by researchers working on flexible foraging. Here, I will only discuss the effects of adaptively flexible foraging on top-down and bottom-up effects and on food web stability.
Top-down and bottom-up effects refer to responses of the abundance of higher or lower trophic levels to perturbations on the highest or lowest trophic level. Traditionally, mortality applied to a particular trophic level is expected to decrease the abundance of that level, to decrease the abundances of all higher levels, increase the abundance of lower levels that are an odd number of levels removed from the focal level, and decrease the abundance of lower levels that are an even number of levels removed. All of these expectations are frequently violated in food chain models having adaptive adjustment of food intake and predation risk (Abrams 1992b,c, 1995, 2004, 2005; Abrams & Vos 2003). If the consumers have linear functional responses, it is possible to deduce conditions required for different directions of response to each mortality rate in a three-species chain, given knowledge of the curvatures of the trade-off functions. For example, if species decrease foraging in response to increase food abundance, and if that decreases their exposure to predators, the bottom-up effect of increase food growth on the predator population will often be negative, rather than the positive effect that is expected in models without a foraging-defense tradeoff. See Abrams & Vos (2003) or Abrams (2005) for a more complete review of these effects. It is likely that some of the counter-intuitive responses to perturbations have frequently been noted in predator addition or removal experiments (reviewed in Sih et al. 1985) are due to flexible foraging. The impacts of adaptive behaviour by the middle-level species in a three-species food chain are very similar to the impacts of shifts in species composition on the middle level of a multi-species food web (Leibold et al. 1997).
Stability is one of the major aspects of food-web dynamics. Once again, flexible behaviours have been shown to affect stability and to produce what are often regarded as counter-intuitive responses. Most often adaptive behaviour is proclaimed to be a stabilizing process (e.g. Ives & Dobson 1987; Uchida, Drossel, & Brose 2007), and this is expected from the predator-dependence of the functional responses of predators eating adaptively defending prey. It is also expected in those cases where foraging effort increases with food density (Abrams 1982). However, adaptively flexible foraging certainly has the potential to be destabilizing as well. The possibility of type-2 responses (which are destabilizing) arising from adaptive foraging was noted above (see Abrams 1982, 1984, 1995), as was the possibility of cycles or complex dynamics due to anti-switching (Abrams & Shen 1989). Lags in switching responses may also increase instability (Abrams 1999; Abrams & Matsuda 2004). Abrams & Matsuda (1997) predicted that adaptively flexible and costly prey defense towards a predator having a type-2 response often results in population cycles in systems that would not cycle (or would cycle with a much shorter period) in the absence of flexible defense. Yoshida et al. (2003) later demonstrated that such cycles could occur as the result of rapid evolution in the prey (alga) in a laboratory predator-prey (rotifer-alga) system, where the better defended clones had slower resource uptake. Thus, it is unlikely that adaptive foraging provides a general explanation for community stability.
Co-existence of competing species is another community-level property that is affected by adaptive behaviours. Early work focused on the role of a switching consumer in facilitating co-existence of its competing resources (Roughgarden & Feldman 1975). To the extent that adaptive foraging produces divergence in the diets of competing consumers (Abrams 1986), it is likely to promote co-existence. Switching by generalist consumers can also allow co-existence with competing specialist consumers in varying environments (Wilson & Yoshimura 1994; Egas, Dieckmann, & Sabelis 2004; Abrams 2006a). Adaptively flexible predator-specific defense by prey has been shown not only to promote predator co-existence, but to also result in mutualism between predators (Matsuda, Abrams, & Hori 1993). The topic of co-existence provides a good example of the fact that different mechanisms of adaptation can produce similar results. Vandermeer (1980) showed that changes in the species composition of a group of competing prey species could produce mutualism between predators, and De Roos et al. (2008) showed that adaptive shifts in the size structure of a single prey in response to two different size-specific predators could produce the same result. The mechanism for mutualism in all three cases is the same; one forager species produces adaptive shifts in prey defense that make the prey more vulnerable to the other forager.
The effects on community level properties reviewed above will only be ecologically significant if they are significant in magnitude and duration. Luttbeg, Rowe, & Mangel (2003) suggested that large behavioural effects on performance at one life stage could be reversed later in the life history. If this was the case generally, there might be less need to incorporate adaptively flexible foraging and defense into ecological thought. While such reversals are no doubt present in some systems, there is as yet no reason to believe that they happen more frequently for ecological effects driven by flexible foraging than for effects driven by inflexible mass-action interactions. As yet, we have no reason to believe that the magnitude of behaviourally mediated indirect effects or the behavioural changes brought about in direct interactions should be small relative to the comparable interactions in the absence of the behaviour. Work on laboratory systems that exhibit induced defenses have demonstrated large trait-mediated effects on trophic cascades in systems that persist for many generations of the species involved (e.g. van der Staap et al. 2007).
Evolutionary Implications of flexible foraging
Evolution of traits that are related to foraging should be intimately connected to foraging behaviour. The interplay of flexible foraging behaviour and the evolution of other foraging traits has received relatively little attention, but the few studies that have been done suggest that flexible foraging can significantly alter evolutionary outcomes (Matsuda & Abrams 1994; Abrams 2003, 2006b). Behaviours that either generate or suppress population cycles are particularly likely to alter the course of evolution of a trait affecting foraging traits (Abrams 1997). This is a simple result of the fact that variation in a nonlinear function about some equilibrium typically changes the mean value (e.g. Ruel & Ayres 1999).
Habitat selection behaviours determine the type of environment to which a mobile organism is exposed. Clearly, if an animal always chooses to occupy habitat 1, traits that determine the animal’s relative ability to exploit different types of resources occurring in habitats 1 and 2, will be fixed for those that maximize ability to exploit the resource in habitat 1. This suggests that habitat selection should increase the range of circumstances under which natural selection on relative resource exploitation traits is disruptive rather than stabilizing, and this is in fact the case (Abrams 2006b). This conclusion also applies to switching behaviour between prey; predators are expected to lose morphological adaptations for exploiting one prey if they consistently choose others. In a variable environment with two foods, the ability to switch can lead to evolution of as many as four different co-existing types from a single generalist ancestor (Abrams 2006b). There is suggestive evidence that switching may have promoted diversification in a guild of Mediterranean scrubland birds that switch between insect and fruit resources (Carnicer, Abrams, & Jordano 2008). A wide variety of evolutionary scenarios involving small numbers of interacting and phenotypically flexible species remain to be explored.
Why flexible foraging may be less important in determining community responses in larger food webs
As has been noted above, different mechanisms of adaptive change that involve the same tradeoff can usually be modelled in a similar manner and produce similar indirect effects (Abrams, Matsuda, & Harada 1993; Abrams & Vos 2003). In the context of larger food webs, this means that shifts in the relative abundances of two or more species on a given trophic level whose trophic characteristics fit a tradeoff relationship will usually have effects on trophic level dynamics that are similar to those that would be produced by adaptive behaviour if only one species were present. At the same time, shifts in species composition are likely to pre-empt or even reverse the direction of changes in the behaviours of within those species.
A useful system for illustrating the relationships between adaptive behaviour and species replacement is the case of the ‘diamond’ food web in which two ‘consumer’ species share a common predator and a common resource. The two consumers must have a tradeoff in their abilities to exploit the resource and to avoid the predator if they are to co-exist in this situation. This means that perturbations to either the top or bottom species change the relative abundances of the two mid-level species, and may cause one to exclude the other. In general, harvesting the predator will decrease the more predator-resistant consumer and increase the better resource exploiter. The resulting changes in the mean exploitation rate and mean vulnerability to predators is the same as would occur in a 3 species food chain in which the middle species had a linear tradeoff that was identical to the line connecting the strategies of the two consumer species in the diamond web (Abrams & Matsuda 1997; Abrams & Vos 2003). Similarly, an increased carrying capacity of the resource causes an increase in the more predator resistant species in the diamond web with two consumer species, and increases the level of predator resistance in the adaptive middle species of the three-species chain. Allowing each of the two mid-level species in the diamond web to have an adaptive tradeoff between foraging and defense makes little difference to these predictions. The change in densities of two co-existing species on the middle trophic level in response to a top-down or bottom-up perturbation implies that both predator and resource densities at equilibrium remain the same, provided both species are only limited by predation and food and the functional and numerical responses are linear (Leibold et al. 1997). As a result, there is no change in the equilibrium values of the foraging/defense trait in either species.
Adaptive foraging/defense on one trophic level (whether via behaviour or species replacement) generally alters the responses of other trophic levels to environmental perturbations.
Consider the impact of increased mortality on the predator in either the diamond web or a three-species chain with an adaptive tradeoff between foraging and predation risk. Both the diamond web and the adaptive three-species chain can exhibit a ‘hydra effect’, meaning that a greater per capita harvest rate applied to a species increases its equilibrium or average density (Abrams & Matsuda 2005). Both of these differ from a system in which there is a single inflexible species on the middle level. Fig. 2 illustrates the difference between the predator mortality vs. predator abundance relationships between systems with adaptive and inflexible mid-level species. Panel a shows the relationship for an adaptive middle trophic level, while panel b shows the relationship when there is a single inflexible prey species whose foraging and predator vulnerability traits are the average of the values of the two prey in panel a. Clearly, trait flexibility on the prey trophic level, whether behavioural or due to species replacement, has a major effect on this relationship. Leibold et al.’s (1997) review of empirical studies of top-down and bottom-up effects in aquatic systems suggests that species replacement usually results in a larger magnitude of adaptive response in the trophic level than does phenotypic flexibility within a species. This is expected, because several species are likely to encompass a wider range of phenotypes than are behavioural variants within a species.
In spite of the above discussion, flexible foraging is known to have important consequences in some models of large food-webs, and too little work has been done to make sweeping generalizations. Switching based on relative prey abundance is the behaviour most often been incorporated into multi-species models. McCann et al. (McCann, Hastings, & Huxel 1998; Rooney et al. 2006) have incorporated switching behaviour into a number of analyses, but have not used these to compare dynamics with different types of switching. The section on specific foraging above has reviewed some of the studies that have found impacts of switching on both food web structure and stability. Flexible foraging may also explain some other non-dynamic aspects of large food-webs, such as the relationships of feeding links to relative body size (Beckerman, Petchey, & Warren 2006; Petchey et al. 2008).
There is little doubt that adaptively flexible foraging and defence behaviours have the potential to alter the population, community, and evolutionary dynamics of the species that comprise food webs. One could argue that this makes foods webs hopelessly complex. It clearly adds to the list of processes that an ecologist must consider in predicting the population/community consequences of environmental perturbations. However, the fact that four types of adaptive traits comprise the vast majority of those that seem likely to influence food web structure implies that the task of understanding the possible impacts is not one of unmanageable proportions. I have suggested five specific topics that are currently under-studied and that could contribute greatly to our understanding of the nature of all four of these categories of adaptive change. The similarities in the impacts of species replacement and within-species adaptive behaviour suggest that behavioural and community ecology can both benefit and learn from each other. This similarity also means that many of the responses of abundance of trophic levels in large food webs may be predicted from the analyses of three- and four-species models that include adaptive behaviour. Predicting the future abundances of individual species will often involve a higher level of uncertainty than one would like. However, the level of uncertainty can only be decreased by including processes, such as adaptive foraging, that are almost always present in natural systems, but frequently ignored in models. Concentrating exclusively on problems that do not involve predicting the future populations of particular species is simply not an option if we are to have effective conservation biology and resource management.
While the complexity of the scientific task that lies ahead is uncertain, understanding adaptively flexible behaviours is certain to be essential for understanding the responses of species and ecosystems to the current rapid rate of environmental change.
I thank the Natural Sciences and Engineering Research Council of Canada for financial support. I thank Andrew Beckerman and Owen Petchey for organizing the E.S.A. Symposium that led to this article, and for their encouragement during the process of revising this paper.