insight review articles The diversity-stability debate Kevin Shear McCann 1205 Docteur Penfield Avenue,Department of Biology,McGill University,Montreal,Quebec,Canada H3A 1B1 There exists little doubt that the Earth's biodiversity is declining.The Nature Conservancy,for example,has documented that one-third of the plant and animal species in the United States are now at risk of extinction. The problem is a monumental one,and forces us to consider in depth how we expect ecosystems,which ultimately are our life-support systems,to respond to reductions in diversity.This issue-commonly referred to as the diversity-stability debate-is the subject of this review,which synthesizes historical ideas with recent advances.Both theory and empirical evidence agree that we should expect declines in diversity to accelerate the simplification of ecological communities. e now realize that the world's flora and complex communities,constructed from many predators fauna are disappearing at rates greater and parasites,prevented populations from undergoing than the mass extinction events whose explosive growth.Hisideas wereclosely akin to MacArthurs collapses punctuate the fossil record.It who reasoned that multiplicity in the number of prey and is also true that species invasions have predator species associated with a population freed that been elevated to unprecedented rates accompanying the population from dramatic changes in abundance when one increased globalization of our world5.These high rates of ofthe prey or predator species declined in density. extinction and invasion put ecosystems under enormous These early intuitive ideas were challenged by the work of stress,making it critical that we understand how the loss, Robert May in 1973.May turned to mathematics to rigor- or addition,of a species influences the stability and ously explore the diversity-stability relationship.By using function of the ecosystems we rely on.We are,in a very linear stability analysis on models constructed from a statis real sense,deconstructing the Earth under the implicit tical universe (that is,randomly constructed communities assumption that ecosystems have evolved the ability to with randomly assigned interaction strengths),May found withstand such assault without collapse. that diversity tends to destabilize community dynamics. Several advances in the diversity-stability debate form a Other ecologists,using similar approaches,found results conceptual thread that suggests that diversity can be expect- that were consistent with this hypothesis The results ed,on average,to give rise to ecosystem stability.This does were puzzling,as real ecosystems were undoubtedly not infer that diversity is the driver of this relationship. complex and diverse.The results also seemed to counter the Instead,diversity can be regarded as the passive recipient of ideas of Elton',Odum'and MacArthur.Yodzisheight- important ecological mechanisms that are inherent in ened this paradox by showing that models structured from ecosystems.One promising mechanism that has been pro- compiled food-web relationships,with plausible interac- posed recently is that weakly interacting species stabilize tion strengths,were generally more stable than randomly community dynamics by dampening strong,potentially constructed food webs.Although the early food-web data destabilizing consumer-resource interactions.Empirical that Yodzis structured his models around were incomplete, descriptions of the distribution of interaction strengths these data reflected real feeding relationships.Yodzis'result in real communities are consistent with this theory.If this indicated that interaction strength was probably crucial to is true then,all elsebeingequal,decreasing biodiversity will stability;but the exact reason for this remained elusive.If be accompanied by increases in average interaction diversity and stability were positively correlated,as early strengths within ecosystems,and a concomitant decrease in empirical evidence had indicated,then more had to be ecosystem stability. happening than simply increasing the number of species and the number of pathways.Something fundamental was Historical perspectives of the diversity-stability debate missing from the early arguments. The relationship between diversity and stability has fasci In the remainder of this paper I review recent investiga- nated ecologists.Before the 1970s,ecologists believed that tions of the diversity-stability debate.I first discuss a change more diverse communities enhanced ecosystem stability. in perspective that is beginning to allow us to unravel this A strong proponent of this view was Charles Elton,who long-standing problem and then review two different lines argued that"simple communities were more easily upset of investigation.One approach has searched for a general than that of richer ones;that is,more subject to destructive diversity-stability relationship,and a second,more mecha- oscillations in populations,and more vulnerable to nistic approach has sought a relationship between food-web invasions".In fact,both Odum'and Eltonarrived at similar structure and stability. conclusions based on the repeated observation that greatly simplified terrestrial communities are characterized by Changing perspectives more violent fluctuations in population density than Muchofecologicaltheoryisbased on theunderlyingassump- diverse terrestrial communities.For example,invasions tion of equilibrium population dynamics.Although this most frequently occur on cultivated land where human assumption is aesthetically pleasing,in that it suggests the influence had produced greatly simplified ecological balanceofnatureis infinitely precise,an alternativeand viable communities,and outbreaksofphytophagous insectsoccur ecological perspectiveexists.As real populations are variable, readily in boreal forests but are unheard of in diverse it is possible that the persistence of complex communities tropicalforests.Theseobservationsled Elton'to believe that depends to some degree on population fluxes (that is,the 228 2000 Macmillan Magazines Ltd NATURE|VOL 40511 MAY 2000 www.nature.com
insight review articles 228 NATURE | VOL 405 | 11 MAY 2000 | www.nature.com We now realize that the world’s flora and fauna are disappearing at rates greater than the mass extinction events whose collapses punctuate the fossil record1–3. It is also true that species invasions have been elevated to unprecedented rates accompanying the increased globalization of our world4,5. These high rates of extinction and invasion put ecosystems under enormous stress, making it critical that we understand how the loss, or addition, of a species influences the stability and function of the ecosystems we rely on. We are, in a very real sense, deconstructing the Earth under the implicit assumption that ecosystems have evolved the ability to withstand such assault without collapse. Several advances in the diversity–stability debate form a conceptual thread that suggests that diversity can be expected, on average, to give rise to ecosystem stability. This does not infer that diversity is the driver of this relationship. Instead, diversity can be regarded as the passive recipient of important ecological mechanisms that are inherent in ecosystems. One promising mechanism that has been proposed recently is that weakly interacting species stabilize community dynamics by dampening strong, potentially destabilizing consumer–resource interactions6 . Empirical descriptions of the distribution of interaction strengths in real communities are consistent with this theory. If this is true then, all else being equal, decreasing biodiversity will be accompanied by increases in average interaction strengths within ecosystems, and a concomitant decrease in ecosystem stability. Historical perspectives of the diversity–stability debate The relationship between diversity and stability has fascinated ecologists. Before the 1970s, ecologists believed that more diverse communities enhanced ecosystem stability7–9. A strong proponent of this view was Charles Elton8 , who argued that “simple communities were more easily upset than that of richer ones; that is, more subject to destructive oscillations in populations, and more vulnerable to invasions”. In fact, both Odum7 and Elton8 arrived at similar conclusions based on the repeated observation that greatly simplified terrestrial communities are characterized by more violent fluctuations in population density than diverse terrestrial communities. For example, invasions most frequently occur on cultivated land where human influence had produced greatly simplified ecological communities, and outbreaks of phytophagous insects occur readily in boreal forests but are unheard of in diverse tropical forests. These observations led Elton8 to believe that complex communities, constructed from many predators and parasites, prevented populations from undergoing explosive growth. His ideas were closely akin to MacArthur9 , who reasoned that multiplicity in the number of prey and predator species associated with a population freed that population from dramatic changes in abundance when one of the prey or predator species declined in density. These early intuitive ideas were challenged by the work of Robert May10 in 1973. May turned to mathematics to rigorously explore the diversity–stability relationship. By using linear stability analysis on models constructed from a statistical universe (that is, randomly constructed communities with randomly assigned interaction strengths), May found that diversity tends to destabilize community dynamics. Other ecologists, using similar approaches, found results that were consistent with this hypothesis11,12. The results were puzzling, as real ecosystems were undoubtedly complex and diverse. The results also seemed to counter the ideas of Elton7 , Odum8 and MacArthur9 . Yodzis13 heightened this paradox by showing that models structured from compiled food-web relationships, with plausible interaction strengths, were generally more stable than randomly constructed food webs. Although the early food-web data that Yodzis structured his models around were incomplete, these data reflected real feeding relationships. Yodzis’ result indicated that interaction strength was probably crucial to stability; but the exact reason for this remained elusive. If diversity and stability were positively correlated, as early empirical evidence had indicated, then more had to be happening than simply increasing the number of species and the number of pathways. Something fundamental was missing from the early arguments. In the remainder of this paper I review recent investigations of the diversity–stability debate. I first discuss a change in perspective that is beginning to allow us to unravel this long-standing problem and then review two different lines of investigation. One approach has searched for a general diversity–stability relationship, and a second, more mechanistic approach has sought a relationship between food-web structure and stability. Changing perspectives Much of ecological theory is based on the underlying assumption of equilibrium population dynamics. Although this assumption is aesthetically pleasing, in that it suggests the balance of nature is infinitely precise, an alternative and viable ecological perspective exists. As real populations are variable, it is possible that the persistence of complex communities depends to some degree on population fluxes (that is, the The diversity–stability debate Kevin Shear McCann 1205 Docteur Penfield Avenue, Department of Biology, McGill University, Montreal, Quebec, Canada H3A 1B1 There exists little doubt that the Earth’s biodiversity is declining. The Nature Conservancy, for example, has documented that one-third of the plant and animal species in the United States are now at risk of extinction. The problem is a monumental one, and forces us to consider in depth how we expect ecosystems, which ultimately are our life-support systems, to respond to reductions in diversity. This issue — commonly referred to as the diversity–stability debate — is the subject of this review, which synthesizes historical ideas with recent advances. Both theory and empirical evidence agree that we should expect declines in diversity to accelerate the simplification of ecological communities. © 2000 Macmillan Magazines Ltd
insight review articles Figure 1 The Ecotron experiment creates model multitrophic community assemblages containing plants,herbivores,parasitoids and decomposers in 16 different chambers. The Ecotron is an ambitious attempt to bridge the scale between field communities and laboratory experiments.(Photographs show the inside of an Ecotron chamber and a technical service corridor between two banks of chambers;courtesy of the Centre for Population Biology,Imperial College at Silwood Park.) fairly regular waxingand waning ofa population's density).Such back- densities move further away from extremely low or high densities. ground population variability,whether driven by biotic or abiotic This is a broad definition,including equilibrium and non- processes,can provide species with the opportunity to respond differ- equilibrium dynamics as well as subsuming the definition of entially to their environment.In turn,these differential species permanences(a population is considered permanent if the lower responses weaken the destructive potential ofcompetitive exclusion. limit to its density is greater than zero).Because the definition ofgen- Because such variability can significantly change our understand- eral stability implies decreased variability (owing to greater limits on ing of ecological interactions,ecologists have begun to relax density),it is closely related to field measurements of stability,which equilibrium-based measures of stability.A recent theoretical tend to rely on variability in population or community densities as a analysis7 has shown that population fluctuations,driven by compe- measure of stability.One can also extend equilibrium resilience to a tition,can actually promote the persistence of large numbers of less biologically restrictive form by defining resilience as the return competing phytoplankton communities on a minimal number of time after a perturbation to an equilibrium or a non-equilibrium limiting resources(but greater than two resources).Coexistence was attractor (Table 1).In a nonlinear system there is no reason to believe found to rely on the fluctuation in population densities,while that an equilibrium that attracts weakly in a local setting (near the community-level densities(the summation of the competing plank- equilibrium)also attracts weakly far away from the equilibrium, ton densities)varied little.We will see that a similar relationship wherethe issueofaspecies'permanence isresolved .For theremain- appears in diversity-stability experiments.Here,too,the evidence der of the paper,unless stated otherwise,the definitions of general points to variable population densities that sum to produce a stability and variability will be used to consider empirical and relatively constant biomass at the community level. theoretical results on the relationship between diversity and stability under a common framework. Definitions of stability Definitions ofstability in ecology can be classified generally into two The search for a general diversity-stability relationship categories(Table 1)-stability definitions that are based on a sys- In 1982,David Tilman began a long-term study to delineate experi- tem's dynamic stability,and stability definitions that are based on a mentally the relationship between diversity and stability in plant system's ability to defy change(resilience and resistance in Table 1). communities.The undertaking involved four grassland fields at Despite the breadth of definitions,ecological theory has tended Cedar Creek Natural History Area,Minnesota,divided into over 200 traditionally to rely on the assumption that a system is stable if,and experimental plots,and gathered information on species richness, only if,it is governed by stable equilibrium dynamics(that is,equilib- community biomass and population biomass through time.The rium stability and equilibrium resilience).As discussed in the results of this and other extensive studies converge on the finding that previous section,these are strong assumptions with no a priori diversity within an ecosystem tends to be correlated positively with justification.In fact,the variable nature of population dynamics plant community stability (that is,decreased coefficient ofvariability foundbothin fieldandin laboratoryexperimentshasledexperimen- in community biomass)20-23.At the same time,diversity seems to talists to use measures of variability as indices of a system's stability. show littleinfluenceon populationvariability.Thebasicarguments This discontinuity between stability experiments and equilibrium- for a positive relationship between diversity and stability for primary based theory has made it difficult to unite theory and experiment in producers at the community level have been classified into two,not the diversity-stability debate. mutually exclusive,hypotheses called the averaging effect2and the More general definitions of stability exist.In Table 1,general negative covariance effect2s(see Table 2 for the underlying logic stability is defined such that stability increases as population behind these ideas).In essence,these hypotheses argue that diversity NATURE|VOL 40511 MAY 2000www.nature.com 2000 Macmillan Magazines Ltd 229
fairly regular waxing and waning of a population’s density). Such background population variability, whether driven by biotic or abiotic processes, can provide species with the opportunity to respond differentially to their environment. In turn, these differential species responses weaken the destructive potential of competitive exclusion. Because such variability can significantly change our understanding of ecological interactions6,14–19, ecologists have begun to relax equilibrium-based measures of stability. A recent theoretical analysis17 has shown that population fluctuations, driven by competition, can actually promote the persistence of large numbers of competing phytoplankton communities on a minimal number of limiting resources (but greater than two resources). Coexistence was found to rely on the fluctuation in population densities, while community-level densities (the summation of the competing plankton densities) varied little. We will see that a similar relationship appears in diversity–stability experiments. Here, too, the evidence points to variable population densities that sum to produce a relatively constant biomass at the community level. Definitions of stability Definitions of stability in ecology can be classified generally into two categories (Table 1) — stability definitions that are based on a system’s dynamic stability, and stability definitions that are based on a system’s ability to defy change (resilience and resistance in Table 1). Despite the breadth of definitions, ecological theory has tended traditionally to rely on the assumption that a system is stable if, and only if, it is governed by stable equilibrium dynamics (that is, equilibrium stability and equilibrium resilience). As discussed in the previous section, these are strong assumptions with no a priori justification. In fact, the variable nature of population dynamics found both in field and in laboratory experiments has led experimentalists to use measures of variability as indices of a system’s stability. This discontinuity between stability experiments and equilibriumbased theory has made it difficult to unite theory and experiment in the diversity–stability debate. More general definitions of stability exist. In Table 1, general stability is defined such that stability increases as population densities move further away from extremely low or high densities. This is a broad definition, including equilibrium and nonequilibrium dynamics as well as subsuming the definition of permanence18 (a population is considered permanent if the lower limit to its density is greater than zero). Because the definition of general stability implies decreased variability (owing to greater limits on density), it is closely related to field measurements of stability, which tend to rely on variability in population or community densities as a measure of stability. One can also extend equilibrium resilience to a less biologically restrictive form by defining resilience as the return time after a perturbation to an equilibrium or a non-equilibrium attractor (Table 1). In a nonlinear system there is no reason to believe that an equilibrium that attracts weakly in a local setting (near the equilibrium) also attracts weakly far away from the equilibrium, where the issue of a species’ permanence is resolved18. For the remainder of the paper, unless stated otherwise, the definitions of general stability and variability will be used to consider empirical and theoretical results on the relationship between diversity and stability under a common framework. The search for a general diversity–stability relationship In 1982, David Tilman began a long-term study to delineate experimentally the relationship between diversity and stability in plant communities. The undertaking involved four grassland fields at Cedar Creek Natural History Area, Minnesota, divided into over 200 experimental plots, and gathered information on species richness, community biomass and population biomass through time. The results of this and other extensive studies converge on the finding that diversity within an ecosystem tends to be correlated positively with plant community stability (that is, decreased coefficient of variability in community biomass)20–23. At the same time, diversity seems to show little influence on population variability22. The basic arguments for a positive relationship between diversity and stability for primary producers at the community level have been classified into two, not mutually exclusive, hypotheses called the averaging effect24 and the negative covariance effect25 (see Table 2 for the underlying logic behind these ideas). In essence, these hypotheses argue that diversity insight review articles NATURE | VOL 405 | 11 MAY 2000 | www.nature.com 229 Figure 1 The Ecotron experiment creates model multitrophic community assemblages containing plants, herbivores, parasitoids and decomposers in 16 different chambers. The Ecotron is an ambitious attempt to bridge the scale between field communities and laboratory experiments. (Photographs show the inside of an Ecotron chamber and a technical service corridor between two banks of chambers; courtesy of the Centre for Population Biology, Imperial College at Silwood Park.) © 2000 Macmillan Magazines Ltd
insight review articles (species richness)increases stability at the community level because Table 1 Definitions of stability diverse plant communities respond differentially to variable back- Term Definition ground processes.The differential responses of populations sum, Definitions of dynamic stability through time,to give stable community dynamics. rete measure that considers a system stable if it returns to Ifdiversity and stabilityare positivelycorrelated,then both theaver- its equilbrium after a small perturbation away from the aging and negative covariance effect predict that population variance equlibrium.A stable system.therefore.has no variability in the absence of perturbations. has to scale as a function ofmean population densities in a precise way General stability A measure which assumes that stability increases as the lower (see Table 2).Tilmanhas used these predictions to show that his field Eimit of population densiy moves further away from zero.Under experiments are consistent with the interpretation that increasing non-equilibrium dynamics.such limits to population dynamics diversity increases community stability.Although this is a clever nerally imply a decrease in population variance (see variability definition below). combination of theory and experiment,it cannot be used to infer that The variance in population densities over time,usually measured diversity is responsible directly for stability Asa counter example, as the coefficient in variation.Common in experimental tests of no correlation was found between diversity and stability at the cross- stability. ecosystem scale2.Other experiments have found that the positive Definitions of resilience and resistance stability diversity-stability correlation is not a pure species effect (that is,a Equtbrium resisence A measure of stablty that assumes system stability increases as time required to return to equilibrium decreases after a diversityeffect),and have indicated that ecosystem function and stabil- perturbation.A rapid response means that a system recois ammcrtmapeeiocbt rapidly back to its equilbrium state. General resilience A measure of stablity that assumes system stability increases as In a similar manner,plant community stability and productivity return time to the equilibrium/non-equilibnum solution decreases after a perturbation A raoid response means that a svstem in European grasslands were shown to be tightly coupled to the ecoils rapidly back to its/non-equilibrium state functional diversity of mutualistic arbuscular mycorrhizal fungi A measure of the degree to which a variable changes after a (AMF).In this system,large fluctuations in plant biomass were perturbation.Frequently used as a discrete measure that associated with low diversities of AMF,whereas more constant assesses a community's abiity to resist invasion (that is,if an invader fails,the community resists invasion). biomass and greater productivity accompanied high AMF diversities.This study highlights that higher-level interactions, which are inherent in food webs (for example,microbial interac- communities within a controlled setting).The main advantage of tions,herbivory and predation),are of great importance in microcosms is that the experiments can easily be manipulated and understanding the relationship between the diversity and stability of replicated.Nonetheless,the issue of how scale influences outcome whole ecological communities.The complexity of whole ecological looms over microcosm experiments-can we extrapolate results to communities-the basis from which Odum,Elton and MacArthur the whole ecosystem?Ambitious experimental set ups such as that formed their diversity-stability hypotheses-cannot manifest itself currently underway in the Ecotron(Fig.1)are attempting to bridge in experiments that focus on single trophic levels the gap between the complexity of real field communities and the Field tests at the scale of the food web are few in number.But one simplicity oflaboratory or greenhouse experiments. thorough examination tested seven different stability-diversity The evidence that has emerged from microcosm experiments, criteria in the grazing ecosystem of the Serengeti under naturally regardless ofscale and system type(that is,terrestrial oraquatic),has variable conditions (that is,strong seasonal changes).Of these seven tended to agree that diversity is positively related to ecosystem stabili- stability measures,five were positively related to diversity whereas ty.In addition,and consistent with field experiments on plant two were unrelated to diversity.The study found that greater diversity communities,experiments using aquatic microcosms have shown reduced the relative magnitudes of fluctuations in productivity that population-level variation is relatively uninfluenced by diversity, induced by seasonal change.Although a relationship between whereas community-level variance tends to decrease with increased stability and diversity exists within the Serengeti,the evidence again diversity35.Two ideas have been advanced in explanation of these points to the importance of functional species in understanding this findings.One explanation is that increasing diversity increases the relationship.For example,the grazing-tolerant plant species have a odds that at least some species will respond differentially to variable disproportionately large role in the Serengeti community dynamics conditionsand perturbations.Thesecondisthat greater diversity by preventing herbivores from dramatically reducing plant biomass. increases the odds that an ecosystem has functional redundancy by The paucity of field tests at the scale of the food web reflects the containing species that are capable of functionally replacing impor- fact that such experiments require an enormous undertaking.As an tant species Taken together,these two notions have been called alternative,ecologists have approached this problem by investigating the insurance hypothesis(Table 2).This idea has been extended to how diversity influences stability and function within a multitrophic suggest that the greater the variance in species'responses contained setting in controlled microcosm experiments (often referred to as in a community then thelower the species richness required to insure bottle experiments as they are attempts to create realistic ecological the ecosystem.As with the averagingand negative covariance effect, Table 2 General diversity-stability theory Theory Underlying logic Averaging effect Assume covariances between species are zero and variance (s)in abundance of individual species/in a plant community is equal to cmf,where c and z are constants and m,is the mean density of species i.Given that all k species of a community are equal in abundance and sum to m (that is.m=m/k), then the coefficient of variation (CV)of community abundance can be determined as: CV=100s/m=100e/ For the casez>1,increasingk(species number)decreases the varlation in biomass for the plant community. Negative-covariance effect If covariances between species (say.species a and b)are negative (that is.cov(a.b)<0).then the variance in the abundance of two species S=s2+s2+2c0va,b) will be less then the sum of the individual variances (that is,s+s),and so will decrease overall biomass variance in the plant community. Insurance effec An ecosystem's ability to buter perturbations.loss in species and species invasions is dependent on the redundancy of the species having important as on the abilty of the species in the community to respond differentially to perturbations.n extended to suggest that the greater the variance of species'responses in a community then the lower the species richness requred to buffer an ecosystem. 230 2000 Macmillan Magazines Ltd NATURE VOL 40511 MAY 2000 www.nature.com
(species richness) increases stability at the community level because diverse plant communities respond differentially to variable background processes. The differential responses of populations sum, through time, to give stable community dynamics. If diversity and stability are positively correlated, then both the averaging and negative covariance effect predict that population variance has to scale as a function of mean population densities in a precise way (see Table 2). Tilman20 has used these predictions to show that his field experiments are consistent with the interpretation that increasing diversity increases community stability. Although this is a clever combination of theory and experiment, it cannot be used to infer that diversity is responsible directly for stability26,27. As a counter example, no correlation was found between diversity and stability at the crossecosystem scale26. Other experiments have found that the positive diversity–stability correlation is not a pure species effect (that is, a diversity effect), and have indicated that ecosystem function and stability are more directly related to functional diversity (for example, graminoids or grasses, nitrogen-fixing legumes and other herbs)27–30. In a similar manner, plant community stability and productivity in European grasslands were shown to be tightly coupled to the functional diversity of mutualistic arbuscular mycorrhizal fungi (AMF)31. In this system, large fluctuations in plant biomass were associated with low diversities of AMF, whereas more constant biomass and greater productivity accompanied high AMF diversities. This study highlights that higher-level interactions, which are inherent in food webs (for example, microbial interactions, herbivory and predation), are of great importance in understanding the relationship between the diversity and stability of whole ecological communities. The complexity of whole ecological communities — the basis from which Odum, Elton and MacArthur formed their diversity–stability hypotheses — cannot manifest itself in experiments that focus on single trophic levels. Field tests at the scale of the food web are few in number. But one thorough examination32 tested seven different stability–diversity criteria in the grazing ecosystem of the Serengeti under naturally variable conditions (that is, strong seasonal changes). Of these seven stability measures, five were positively related to diversity whereas two were unrelated to diversity. The study found that greater diversity reduced the relative magnitudes of fluctuations in productivity induced by seasonal change. Although a relationship between stability and diversity exists within the Serengeti, the evidence again points to the importance of functional species in understanding this relationship. For example, the grazing-tolerant plant species have a disproportionately large role in the Serengeti community dynamics by preventing herbivores from dramatically reducing plant biomass. The paucity of field tests at the scale of the food web reflects the fact that such experiments require an enormous undertaking. As an alternative, ecologists have approached this problem by investigating how diversity influences stability and function within a multitrophic setting in controlled microcosm experiments (often referred to as bottle experiments as they are attempts to create realistic ecological communities within a controlled setting). The main advantage of microcosms is that the experiments can easily be manipulated and replicated33. Nonetheless, the issue of how scale influences outcome looms over microcosm experiments — can we extrapolate results to the whole ecosystem? Ambitious experimental set ups such as that currently underway in the Ecotron (Fig. 1) are attempting to bridge the gap between the complexity of real field communities and the simplicity of laboratory or greenhouse experiments. The evidence that has emerged from microcosm experiments, regardless of scale and system type (that is, terrestrial or aquatic), has tended to agree that diversity is positively related to ecosystem stability34–39. In addition, and consistent with field experiments on plant communities, experiments using aquatic microcosms have shown that population-level variation is relatively uninfluenced by diversity, whereas community-level variance tends to decrease with increased diversity35. Two ideas have been advanced in explanation of these findings. One explanation is that increasing diversity increases the odds that at least some species will respond differentially to variable conditions and perturbations37–39. The second is that greater diversity increases the odds that an ecosystem has functional redundancy by containing species that are capable of functionally replacing important species37–39. Taken together, these two notions have been called the insurance hypothesis (Table 2). This idea has been extended to suggest that the greater the variance in species’ responses contained in a community then the lower the species richness required to insure the ecosystem40. As with the averaging and negative covariance effect, insight review articles 230 NATURE | VOL 405 | 11 MAY 2000 | www.nature.com Table 1 Definitions of stability Term Definition Definitions of dynamic stability Equilibrium stability A discrete measure that considers a system stable if it returns to its equilibrium after a small perturbation away from the equilibrium. A stable system, therefore, has no variability in the absence of perturbations. General stability A measure which assumes that stability increases as the lower limit of population density moves further away from zero. Under non-equilibrium dynamics, such limits to population dynamics generally imply a decrease in population variance (see variability definition below). Variability The variance in population densities over time, usually measured as the coefficient in variation. Common in experimental tests of stability. Definitions of resilience and resistance stability Equilibrium resilience A measure of stability that assumes system stability increases as time required to return to equilibrium decreases after a perturbation. A rapid response means that a system recoils rapidly back to its equilibrium state. General resilience A measure of stability that assumes system stability increases as return time to the equilibrium/non-equilibrium solution decreases after a perturbation. A rapid response means that a system recoils rapidly back to its equilibrium/non-equilibrium state. Resistance A measure of the degree to which a variable changes after a perturbation. Frequently used as a discrete measure that assesses a community’s ability to resist invasion (that is, if an invader fails, the community resists invasion). Table 2 General diversity–stability theory Theory Underlying logic Averaging effect24,25 Assume covariances between species are zero and variance (si 2 ) in abundance of individual species i in a plant community is equal to cmi z ,where c and z are constants and mi is the mean density of species i. Given that all k species of a community are equal in abundance and sum to m (that is, mi= m/k), then the coefficient of variation (CV) of community abundance can be determined as: CV = 100s/m = 100(c/k) 1/z For the case z > 1, increasing k (species number) decreases the variation in biomass for the plant community. Negative-covariance effect25 If covariances between species (say, species a and b) are negative (that is, cov(a,b) < 0), then the variance in the abundance of two species s2 (a+b) = sa 2 + sb 2 + 2cov(a,b) will be less then the sum of the individual variances (that is, sa 2 + sb 2 ), and so will decrease overall biomass variance in the plant community. Insurance effect37–40 An ecosystem’s ability to buffer perturbations, loss in species and species invasions is dependent on the redundancy of the species having important stabilizing roles, as well as on the ability of the species in the community to respond differentially to perturbations. Increasing diversity increases the odds that such species exist in an ecosystem. This idea has been extended40 to suggest that the greater the variance of species’ responses in a community then the lower the species richness required to buffer an ecosystem. © 2000 Macmillan Magazines Ltd
insight review articles alization of the ecosystem as a linear food chain with the view that food webs are highly interconnected assemblages' 2-45 characterized by recurrent food-web structures (for example,omnivory and apparent competition).Because combinations of competition and predation can represent these common food-web structures,the use of simple food-web modules has been advocated"to explore the repercussions ofthese ubiquitous species interactions. Several model investigations have grown out of this approach to show that natural food-web structures can,indeed,enhance ecosystem stability7.These food-web models are extensions of a bioenergetic consumer-resource modelthat constrains parameters to empirically determined relationships ofbody size.The approach is R2-low akin to the dynamic modelling of a population's energy budgets through time,and has the important consequence of placing food-web models within a biological universe with reasonable con- straints operating on energy flow between any consumer-resource interaction-a feature that is considerably different from a statistical universe.The result is that increasing diversity can increase food-web R2-high stability under one condition:the distribution of consumer-resource interaction strengths must be skewed towards R1 weak interaction strengths.I will refer to this as the weak-interaction effect (Table 3),and to connect this to general diversity-stability Figure 2 Consumer-resource interactions.a,A simple food-web diagram depicting a theory I briefly discuss the stabilizing mechanisms behind this effect. strong consumer-resource interaction (C-R.)coupled to a weak consumer-resource Two general stabilizing mechanisms underlie the weak-interac- interaction (C-R).b,Consumption rates by consumer Cof R,for two different densities tion effect.First,the weak-interaction effect generates negative of R.Because the resources negatively covary,the actual consumption response is covariances and promotes community-level stability.Second,these qualitatively similar to a combination of these two curves(dashed circles and line).In negative covariances ensure that the weak interactors dampen the the presence of R resource R,is less influenced by consumption when at low destabilizing potential of strong consumer-resource interactions. densities. These mechanisms can be best understood with a simple example. Figure 2a depicts a simple food-web interaction in which a strong consumer-resource interaction (C-R)is coupled to a weak which are intimately related,the insurance hypothesis does not infer consumer-resource interaction (C-R,).Being a weakly interacting that diversity actively promotesstability. species,Risan inferior competitor whoseability to persist ismediat- In summary,the results indicate that within an ecosystem,diver- ed by the top predator.This food-web relationship ensures that the sity tends to be correlated positively with ecosystem stability.This resources negatively covary.For example,R,is released from compet- correlation does not necessarily extend to population-level stability. itive limitation to flourish whenever R is suppressed by high Much work is still required to determine the driver of the positive densities of consumer C.This occurs because R,is weakly coupled to diversity-stability relationship;however,it seems that community- Cand so is not strongly influenced by high densities ofC.In this man- level stability is dependent on the differential response of species or ner,the weak interaction drives the differential responses ofspecies. functional groups to variable conditions,as well as the functional We can use the knowledge of this negative covariance to deter- redundancy of species that have important stabilizing roles.I now mine qualitatively the consumption rate of C on its preferred turn to a separate approach that has not focused on diversity,but has resource,R.Figure 2b depicts C's consumption rate on R under two concentrated instead on understanding the implications ofcommon different densities of R,assuming an optimally foraging,type II food-web structures on stability. multispecies functional response (for full details,see refs 49,51). High densities of R,reduce the overall consumption rates on R. Food-web structure and stability Because the resources negatively covary,then for low densities of R In an important theoretical contribution,Chesson and Huntley" we expect C's consumption rates to fall on the R,-high curve in Fig.2 showed that diversity cannot be maintained by variation alone.Rather, (the lower dashed circle).Similarly,for high densities ofR we expect maintenance of diversity requires the two following components:the consumption rates to be on the R,-low curve in Fig.2(the upper existence offluxor variability in ecosystems;and populations capable of dashed circle).Piecing these functions together we see that the asyn- differentially exploiting this flux or variability.Regardless of the source chrony in resource densities drives a sigmoid-shaped response that is of the variability(for example,whether spatially or temporally generat- qualitatively similar to what ecologists refer to as a type III functional ed),their results indicate that coexistence requires that populations response.This has the non-equilibrium effect of releasing the prey must be released,either directly or indirectly,from the limiting (R )from strong consumptive pressures when it is at low densities, influences of species interactions such as predation and competition. and thereby the weak interaction dampens the oscillatory potential Species interactions,therefore,must be important in maintaining and of the strong C-R interaction.Consistent with the above discussion promoting persistence in diverse communities in spite of,and perhaps is the fact that investigators have found that donor control(in which a because of,the variability that underlies ecosystems.Several more consumer responds numerically to a resource but has no influence on specific models can be included under this general framework,and all resource dynamics)also promotes community stability54.Donor reveal that flux interacting with specific biotic,nonlinear responses can control can generate differential responses of species by allowing promote persistence.Wenowturn toaset offood-webmodels that species using these resources to disconnect themselves from fluxes have shown how persistent,complex ecosystems can be an outcome of that are inherent to the community. this combination offlux and density-dependent food-web interactions I have discussed the weak-interaction effect within the context of (that is,competitiveandpredatory influencesthat varywith density). relatively simple food-web modules.Does the effect operate for real, complex communities with enormous numbers of direct and The weak-interaction effect indirect interactions?It is still too early to tell,but Kokkoris et al. Over the past decade,ecologists have begun to replace the conceptu- followed the distribution of interaction strengths as competitive NATURE|VOL 40511 MAY 2000www.nature.com 2000 Macmillan Magazines Ltd 231
which are intimately related, the insurance hypothesis does not infer that diversity actively promotes stability. In summary, the results indicate that within an ecosystem, diversity tends to be correlated positively with ecosystem stability. This correlation does not necessarily extend to population-level stability. Much work is still required to determine the driver of the positive diversity–stability relationship; however, it seems that communitylevel stability is dependent on the differential response of species or functional groups to variable conditions, as well as the functional redundancy of species that have important stabilizing roles. I now turn to a separate approach that has not focused on diversity, but has concentrated instead on understanding the implications of common food-web structures on stability. Food-web structure and stability In an important theoretical contribution, Chesson and Huntley41 showed that diversity cannot be maintained by variation alone. Rather, maintenance of diversity requires the two following components: the existence of flux or variability in ecosystems; and populations capable of differentially exploiting this flux or variability. Regardless of the source of the variability (for example, whether spatially or temporally generated), their results indicate that coexistence requires that populations must be released, either directly or indirectly, from the limiting influences of species interactions such as predation and competition. Species interactions, therefore, must be important in maintaining and promoting persistence in diverse communities in spite of, and perhaps because of, the variability that underlies ecosystems. Several more specific models can be included under this general framework, and all reveal that flux interacting with specific biotic, nonlinear responses can promote persistence6,14,17. We now turn to a set of food-web models that have shown how persistent, complex ecosystems can be an outcome of this combination of flux and density-dependent food-web interactions (that is, competitive and predatory influences that vary with density). The weak-interaction effect Over the past decade, ecologists have begun to replace the conceptualization of the ecosystem as a linear food chain with the view that food webs are highly interconnected assemblages42–45 characterized by recurrent food-web structures (for example, omnivory and apparent competition). Because combinations of competition and predation can represent these common food-web structures, the use of simple food-web modules has been advocated46 to explore the repercussions of these ubiquitous species interactions. Several model investigations have grown out of this approach to show that natural food-web structures can, indeed, enhance ecosystem stability6,47–49. These food-web models are extensions of a bioenergetic consumer–resource model50 that constrains parameters to empirically determined relationships of body size. The approach is akin to the dynamic modelling of a population’s energy budgets through time, and has the important consequence of placing food-web models within a biological universe with reasonable constraints operating on energy flow between any consumer–resource interaction — a feature that is considerably different from a statistical universe. The result is that increasing diversity can increase food-web stability under one condition: the distribution of consumer–resource interaction strengths must be skewed towards weak interaction strengths. I will refer to this as the weak-interaction effect (Table 3), and to connect this to general diversity–stability theory I briefly discuss the stabilizing mechanisms behind this effect. Two general stabilizing mechanisms underlie the weak-interaction effect. First, the weak-interaction effect generates negative covariances and promotes community-level stability. Second, these negative covariances ensure that the weak interactors dampen the destabilizing potential of strong consumer–resource interactions. These mechanisms can be best understood with a simple example. Figure 2a depicts a simple food-web interaction in which a strong consumer–resource interaction (C–R1) is coupled to a weak consumer–resource interaction (C–R2). Being a weakly interacting species, R2 is an inferior competitor whose ability to persist is mediated by the top predator. This food-web relationship ensures that the resources negatively covary. For example, R2 is released from competitive limitation to flourish whenever R1 is suppressed by high densities of consumer C. This occurs because R2 is weakly coupled to C and so is not strongly influenced by high densities of C. In this manner, the weak interaction drives the differential responses of species. We can use the knowledge of this negative covariance to determine qualitatively the consumption rate of C on its preferred resource, R1. Figure 2b depicts C’s consumption rate on R1 under two different densities of R2, assuming an optimally foraging, type II multispecies functional response (for full details, see refs 49, 51). High densities of R2 reduce the overall consumption rates on R1. Because the resources negatively covary, then for low densities of R1 we expect C’s consumption rates to fall on the R2-high curve in Fig. 2 (the lower dashed circle). Similarly, for high densities of R1we expect consumption rates to be on the R2-low curve in Fig. 2 (the upper dashed circle). Piecing these functions together we see that the asynchrony in resource densities drives a sigmoid-shaped response that is qualitatively similar to what ecologists refer to as a type III functional response. This has the non-equilibrium effect of releasing the prey (R1) from strong consumptive pressures when it is at low densities, and thereby the weak interaction dampens the oscillatory potential of the strong C–R1 interaction. Consistent with the above discussion is the fact that investigators have found that donor control (in which a consumer responds numerically to a resource but has no influence on resource dynamics) also promotes community stability45,48. Donor control can generate differential responses of species by allowing species using these resources to disconnect themselves from fluxes that are inherent to the community. I have discussed the weak-interaction effect within the context of relatively simple food-web modules. Does the effect operate for real, complex communities with enormous numbers of direct and indirect interactions? It is still too early to tell, but Kokkoris et al.52 followed the distribution of interaction strengths as competitive insight review articles NATURE | VOL 405 | 11 MAY 2000 | www.nature.com 231 a b C R1 R1 R2 Consumption rate of C R2-low R2-high Figure 2 Consumer–resource interactions. a, A simple food-web diagram depicting a strong consumer–resource interaction (C–R1) coupled to a weak consumer–resource interaction (C–R2). b, Consumption rates by consumer C of R1 for two different densities of R2. Because the resources negatively covary, the actual consumption response is qualitatively similar to a combination of these two curves (dashed circles and line). In the presence of R2, resource R1 is less influenced by consumption when at low densities. © 2000 Macmillan Magazines Ltd
insight review articles model communities were assembled.They found that as the Table 3 Food-web structure and stability theory assembly process progressed,larger permanent communities (that Theor Underlying logic is,with a lower limit above zero)attained lower mean interaction Weak interactions serve to limit energy flow in a strengths.They also found that communities with lower mean inter- ion and. action strength were more resistant to invasion.These results are destablizes the dynamics of food webs.In addition,the encouraging and indicate that the weak-interaction effect might weak interactions serve to generate negative covariances scale to the whole ecosystem.If the weak-interaction mechanism is between resources that enable a stabilizir ng effect at operating in real communities then the distributions of interaction the population and community level.The negative covarances ensure that consumers have weak strengths willbe skewed towards weak interactions in order that a few consumptive influences on a resource when the potentially excitable consumer-resource interactions are muted.I resource is at low densities.See text and Fia 2 tor urther clarification. now turn to empirical investigations of food-web structure and stability,first reviewing experiments concerned with the distribution Berlow7 suggested an additional influence of weak of interaction strengths in natural communities before examining interactions.Weak interactions in intertidal communities seem to be extremely variable in strength,and as a experiments that have investigated directly the influence offood-web structure on stability. structure.This community variablity in space can provide a canvas for species to respond differentially. and so may further promote the maintenance of Interaction strength and species invasions Although quantitative field estimates of interaction strength are still in the process of development,work by a few ecologists has enabled a the relative proportion of nonomnivorous damselbugs versus preliminary glimpse into the nature of the distributions of interac- omnivorous wolf spiders in arthropod assemblages of the Mount tion strength within real food webs3-57.The early data indicate Saint Helenblowdown zone.The results showed that increasing the unequivocally that distributions of interaction strength are strongly degree ofcommunity omnivory(that is,increasing the proportion of skewed towards weak interactions3-57.Nonetheless,these experi- wolfspiders)decreased variation in the population responses after an ments also highlight that the removal or addition of a single key aphicide application. species can have pronounced impacts on the dynamics and In an earlier investigation,de Ruiter et al.s investigated model persistence of the species in the enclosure or exclosure.For example, communities constructed from empirical estimates based on some experimental removal of the predatory starfish,Pisaster ochraceus, well-studied food webs from native and agricultural soils.Their resulted in greatly simplified lower-intertidal communities because results are consistent with the experiments on interaction strength the mussel,Mytilus californianus,competitively dominates all other and the weak-interaction effect discussed above,as most interactions sessile benthic organisms when freed from predation3. had only a negligible impact on community dynamics.Although Arecentexperimenthas confirmed the abundance ofweakinter- their results indicate that energetics are important in constraining actions in ecosystems,but showed that weak average interaction interaction strength,they found no positive correlation between strength in a rocky intertidal community tends to be correlated with feeding rates and community impact.This,too,is consistent with the high variability in interaction strength.In this study,variation in the weak-interaction effect.Similarly,experiments on both terrestrial magnitude of the weak interactions seemed to excite spatial variation and aquatic microcosms have tended to find that increasingthe num- in community structure.The variation in interaction strength may be ber of prey items enhances stability although one microcosm important in generating landscape-scale variation that promotes the experimentfound that theaddition ofan alternate prey destabilized maintenance ofdiversity,an area that demands further investigation. community dynamics.This last case can be reconciled with other It is important to know if these phenomena can be extended experiments as the alternate prey introduced was efficient fare for the beyond the scale of the enclosure/exclosure experiment.Species predator.In essence,the alternative prey energetically fuelled the invasions may be seen as the uncontrolled version ofspecies addition predator,and so the experiment may be viewed as evidence that a experiments.Similar to theexperiments described above,the current strong consumer-resource interaction is potentially destabilizing. evidence indicates that,although most species invasions have a weak impact on ecosystems,the occasional invasive species alters an Conclusion ecosystem profoundly.For example,a recent study used stable Taken together,recent advances indicate that diversity can beexpected, isotopes to document energy flow through food webs.Lakes that on average,to give rise to ecosystem stability.The evidence also were uninvaded by bass were compared with lakes that had just been indicates that diversity is not the driver of this relationship;rather, invaded,and the recently invaded lakes showed marked differences in ecosystem stability depends on the ability for communities to contain energy flow patterns(implying aseverely altered food-web structure) species,or functionalgroups,that are capable ofdifferential response. as well as rapid declines in forage fish diversity.These results All of these views are consistent with the ideas put forth by the influen- indicate that the addition of a single species can precipitate a form of tial figures of Odum',Elton",MacArthur and May.May's result ecosystem collapse that sends a wave of extinction through the reflectsthe fact that random distributions(that is,anull universe where ecosystem.Another noteworthy case concerns the introduction of dynamics are influenced by diversity alone),on average,do not create the large predatory fish,the Nile perch (Lates niloticus),in Lake the necessary tension between community members that forces differ- Victoriain the 1950s.Theaddition ofthe Nileperch was followedbya ential response and community stability.In a randomly constructed sequence of amazing ecological and genetic changes that culminated community,for example,strong interactions are not necessarily in a cascade of cichlid extinctions Overall,however,the invasion coupled to weak interactions that mute their destabilizing potential.In literature is harmonious with enclosure/exclosure experiments fact,onecanexpect that random communities will generally not create most invasions have a weak impact with infrequent occurrences ofan such couplings,and so tend to produce diverse communities with invasive species capable of precipitating monumental changes to an complex,wildly oscillatory dynamics.Furthermore,Odum,Elton and ecosystem. MacArthur recognized that real food webs contain a complex array of energetic pathways that can act as buffers against dramatic population Food-web structure and stability experiments explosions.Specifically,MacArthur's hypothesis 一that greater Some direct experimental tests of stability and food-web structure connectance drives community and ecosystem stability seems a exist.In a clever experimental manipulation,Fagantested commu- strong possibility provided most pathways are constructed from weak nity response to a perturbation (aphicide application)as a function interactionsthat mute the potentially destabilizing rolesofafewstrong of the degree of omnivory.Fagan accomplished this by controlling consumer-resource interactions. 232 ☆©2000 Macmillan Magazines Ltd NATURE VOL 40511 MAY 2000 www.nature.com
model communities were assembled. They found that as the assembly process progressed, larger permanent communities (that is, with a lower limit above zero) attained lower mean interaction strengths. They also found that communities with lower mean interaction strength were more resistant to invasion. These results are encouraging and indicate that the weak-interaction effect might scale to the whole ecosystem. If the weak-interaction mechanism is operating in real communities then the distributions of interaction strengths will be skewed towards weak interactions in order that a few potentially excitable consumer–resource interactions are muted. I now turn to empirical investigations of food-web structure and stability, first reviewing experiments concerned with the distribution of interaction strengths in natural communities before examining experiments that have investigated directly the influence of food-web structure on stability. Interaction strength and species invasions Although quantitative field estimates of interaction strength are still in the process of development, work by a few ecologists has enabled a preliminary glimpse into the nature of the distributions of interaction strength within real food webs53–57. The early data indicate unequivocally that distributions of interaction strength are strongly skewed towards weak interactions53–57. Nonetheless, these experiments also highlight that the removal or addition of a single key species can have pronounced impacts on the dynamics and persistence of the species in the enclosure or exclosure. For example, experimental removal of the predatory starfish, Pisaster ochraceus, resulted in greatly simplified lower-intertidal communities because the mussel, Mytilus californianus, competitively dominates all other sessile benthic organisms when freed from predation58. A recent experiment59has confirmed the abundance of weak interactions in ecosystems, but showed that weak average interaction strength in a rocky intertidal community tends to be correlated with high variability in interaction strength. In this study, variation in the magnitude of the weak interactions seemed to excite spatial variation in community structure. The variation in interaction strength may be important in generating landscape-scale variation that promotes the maintenance of diversity, an area that demands further investigation. It is important to know if these phenomena can be extended beyond the scale of the enclosure/exclosure experiment. Species invasions may be seen as the uncontrolled version of species addition experiments. Similar to the experiments described above, the current evidence indicates that, although most species invasions have a weak impact on ecosystems60, the occasional invasive species alters an ecosystem profoundly61–63. For example, a recent study61 used stable isotopes to document energy flow through food webs. Lakes that were uninvaded by bass were compared with lakes that had just been invaded, and the recently invaded lakes showed marked differences in energy flow patterns (implying a severely altered food-web structure) as well as rapid declines in forage fish diversity61. These results indicate that the addition of a single species can precipitate a form of ecosystem collapse that sends a wave of extinction through the ecosystem. Another noteworthy case concerns the introduction of the large predatory fish, the Nile perch (Lates niloticus), in Lake Victoria in the 1950s. The addition of the Nile perch was followed by a sequence of amazing ecological and genetic changes that culminated in a cascade of cichlid extinctions63. Overall, however, the invasion literature is harmonious with enclosure/exclosure experiments60 — most invasions have a weak impact with infrequent occurrences of an invasive species capable of precipitating monumental changes to an ecosystem. Food-web structure and stability experiments Some direct experimental tests of stability and food-web structure exist. In a clever experimental manipulation, Fagan64 tested community response to a perturbation (aphicide application) as a function of the degree of omnivory. Fagan accomplished this by controlling the relative proportion of nonomnivorous damselbugs versus omnivorous wolf spiders in arthropod assemblages of the Mount Saint Helen ‘blowdown zone’. The results showed that increasing the degree of community omnivory (that is, increasing the proportion of wolf spiders) decreased variation in the population responses after an aphicide application. In an earlier investigation, de Ruiter et al.65 investigated model communities constructed from empirical estimates based on some well-studied food webs from native and agricultural soils. Their results are consistent with the experiments on interaction strength and the weak-interaction effect discussed above, as most interactions had only a negligible impact on community dynamics. Although their results indicate that energetics are important in constraining interaction strength, they found no positive correlation between feeding rates and community impact. This, too, is consistent with the weak-interaction effect. Similarly, experiments on both terrestrial and aquatic microcosms have tended to find that increasing the number of prey items enhances stability66–68, although one microcosm experiment69 found that the addition of an alternate prey destabilized community dynamics. This last case can be reconciled with other experiments as the alternate prey introduced was efficient fare for the predator. In essence, the alternative prey energetically fuelled the predator, and so the experiment may be viewed as evidence that a strong consumer–resource interaction is potentially destabilizing. Conclusion Taken together, recent advances indicate that diversity can be expected, on average, to give rise to ecosystem stability. The evidence also indicates that diversity is not the driver of this relationship; rather, ecosystem stability depends on the ability for communities to contain species, or functional groups, that are capable of differential response . All of these views are consistent with the ideas put forth by the influential figures of Odum7 , Elton8 , MacArthur9 and May10. May’s result reflects the fact that random distributions (that is, a null universe where dynamics are influenced by diversity alone), on average, do not create the necessary tension between community members that forces differential response and community stability. In a randomly constructed community, for example, strong interactions are not necessarily coupled to weak interactions that mute their destabilizing potential. In fact, one can expect that random communities will generally not create such couplings, and so tend to produce diverse communities with complex, wildly oscillatory dynamics. Furthermore, Odum, Elton and MacArthur recognized that real food webs contain a complex array of energetic pathways that can act as buffers against dramatic population explosions. Specifically, MacArthur’s hypothesis — that greater connectance drives community and ecosystem stability — seems a strong possibility provided most pathways are constructed from weak interactions that mute the potentially destabilizing roles of a few strong consumer–resource interactions. insight review articles 232 NATURE | VOL 405 | 11 MAY 2000 | www.nature.com Table 3 Food-web structure and stability theory Theory Underlying logic Weak-interaction effect6,46–48 Weak interactions serve to limit energy flow in a potentially strong consumer–resource interaction and, therefore, to inhibit runaway consumption that destabilizes the dynamics of food webs. In addition, the weak interactions serve to generate negative covariances between resources that enable a stabilizing effect at the population and community level. The negative covariances ensure that consumers have weak consumptive influences on a resource when the resource is at low densities. See text and Fig. 2 for further clarification. Berlow57 suggested an additional influence of weak interactions. Weak interactions in intertidal communities seem to be extremely variable in strength, and as a result may drive spatial variability in community structure. This community variability in space can provide a canvas for species to respond differentially, and so may further promote the maintenance of diversity. © 2000 Macmillan Magazines Ltd