For Keystones Identifying Keystone Species Is Difficukt--But Essential to Understanding Bow Loss of Species Will Affect Ecosystems Mary E

Home , Keystone species

Challenges in the Quest for Keystones Identifying keystone species is difficukt--but essential to understanding bow loss of species will affect ecosystems Mary E. Power, David Tilman, James A. Estes, Bruce A. Menge, William J. Bond, L. Scott Mills, Gretchen Daily, Juan Carlos Castilla, Jane Lubchenco, and Robert T. Paine

any ecologists believe that Defining keystones all species were not cre- A keystone species is M ated equal. For example, We offer a definition of keystone it is well known that the most abun- one whose effect is that has been expanded from the dant species play a major role in original usage of Paine (1969), in controlling the rates and directions large, and which keystone referred to a species of many community and ecosystem that preferentially consumed and held processes. These dominant species disproportionately in check another species that would are often crucial for the mainte- otherwise dominate the system. To large relative to better reflect its current use, we de- Mary E. Power is a professor in the fine a keystone species as one whose Department of Integrative Biology, Uni- its abundance impact on its community or ecosys- versity of California, Berkeley, CA tem is large, and disproportionately 94720. David Tilman is a professor in large relative to its abundance. the Department of Ecology, Evolution, nance of thejr communities, be- To develop a more operational and Behavior, University of Minnesota, cause they typically provide the definition for keystone species, one St. Paul, MN 55108. James A. Estes is a major energy flow and the three- must define the strength of the effect wildlife biologist in the National Bio- dimensional structure that supports of a species on a community or ecological Service, Institute of Marine Sci- system trait. This measure, which we ence, University of California, Santa and shelters other organisms call community importance (CI), is Cruz, CA 95064. Bruce A. Menge is a (Ashton 1992, Dayton 1985, Duran professor in the Department of Zool- and Castilla 1989, Gentry and the change in a community or eco- ogy, Oregon State University, Corvallis, Dodson 1987, Paine and Suchanek system trait per unit change in the OR 97331. William J. Bond is a profes- 1983, Strong 1977). abundance of the species. (Our ap- sor doctor in the Department of Botany, Many experiments, however, have proach is a generalization of the con- University of Cape Town, Rondebosch demonstrated that some less abun- cept of community importance in 7700 South Africa. L. Scott Mills is an dant species, often called keystone Mills et al. [1993].) In mathematical assistant professor in the Wildlife Biol- species, also have strong effects on terms, ogy Program, School of Forestry, Uni- versity of Montana, Missoula, MT 59812. communities and ecosystems (e.g., Gretchen Daily is Bing Interdisciplinary Paine 1969). Keystone species differ CI = [d(trait)/dp][l/(trait)] Research Scientist, Department of Bio- from dominant species in that their logical Science, Stanford University, effects are much larger than would where p is the proportional abun- Stanford, CA 94305. Juan Carlos Castilla be predicted from their abundance. dance (in most cases, proportional is a full professor and marine biology Ambiguity in the use of the term biomass relative to the total biomass head in Facultad de Ciencias Biologicas, keystone and the lack of an opera- of all other species in the commu- Pontificia Universidad Catolica de Chile, tional definition have led to criticism nity) of the species whose abundance Casilla 114-D, Santiago, Chile. Jane of its continued application in re- is modified. Trait refers to a quanti- Lubchenco is a distinguished professor search and policy contexts (Mills et tative trait of a community or eco- in the Department of Zoology, Oregon State University, Corvallis, OR 97331. al. 1993, Simberloff 1991). In this system. Potential community or eco- Robert T. Paine is a professor in the article we clarify the keystone con- system traits include productivity, Department of Zoology, NJ-15, Univer- cept, discuss its relevance to man- nutrient cycling, species richness, or sity of Washington, Seattle, WA 98195. agement processes, and suggest ad- the abundance of one or more func- 0 1996 American Institute of Biological ditional research that needs to be tional groups of species or of domi- Sciences. performed. nant species. Experiments that evalu-

September 1996 609 a proportional abundance of species i a 15: before it was deleted. If a species has an effect in direct proportion to its 10- abundance, CI, would be 1 (if, after the species deletion, the community 5- or ecosystem characteristic de- creased) or -1 (if the characteristic increased). If species i is a keystone, the absolute value of CI, is much cr 15- greater than 1. $b Although the frequency distributions of community importance val- 10- ues for species in natural communities are unknown, several shapes seem 5- plausible (Figure 1). In some communities, the distribution may be 0, II - 0 + close to normal, with its mean near Community importance zero (Figure la).Three experimental studies of interaction strength have Figure 1. Possible frequency distribu- found, however, that although the tions of community importance values majority of species in the guilds or for all species in a given community. assemblages studied had impacts Positive values occur when a community close to zero, a few species exerted characteristic decreases after a species is strong effects (Figure 2).Paine (1992) deleted; in the absence of a mutualist, for instance, the target dominant species measured the impacts of seven inver- would also decrease. Negative values tebrate grazers on a rocky intertidal occur when a community characteristic kelp sporeling assemblage; Fagan and increases after removal of a species, as Hurd (1994) studied impacts of a would be the case if the characteristic preying mantid on more than 12 were the abundance of another species orders or families of arthropod prey and the first species were a consumer of in an old field; and Rafaelli and Hall that species. Community importance (CI) (1992) studied impacts of predatory values may be normally distributed birds, fish, and invertebrates on ma- around zero (a), in which case most species would have immeasurably small rine invertebrates in mudflats and effects, and keystones would be rare. (b) mussel beds. Fagan and Hurd (1994) In some communities, the CI distribu- did not resolve prey to species, and tion may have several modes, with key- Rafaelli and Hall (1992) could ma- stone species falling into modes that are nipulate predators such as shorebirds sufficiently far from zero. only as groups of species. By con- trast, Paine’s (1992) was a pairwise study, but only because he measured ate the community importance of a the impacts of each consumer on a species by changing its abundance simplified reference state composed should proceed long enough for in- almost entirely of a single, competi- direct effects to become evident. The tively superior prey species. Never- full derivative is used here, rather theless, these studies show the feasi- than a partial derivative, because it bility of using experimental field includes all the direct and indirect approaches to estimate interaction effects of the species. strength. In practice, it is difficult to mea- How are interaction strengths and sure the effects of small changes in community importance values of species abundance. More commonly, species related? Paine’s interaction an attempt is made to study a spe- strength (Paine 1992)was computed cies’ impacts by removing it entirely. as [(t - t,) /to] (l/n),where t, is the If it can be removed, then abungance of the prey in the presence of the consumer, t, is prey abundance in the absence of the consumer, and n is the number of where tN is a quantitative measure of consumer individuals stocked in ex- the trait in the intact community or perimental arenas. This measure dif- ecosystem, t, is the trait when spe- fers from our index of community cies i has been deleted, and p, is the importance, [(t, - to) / tN] (l/pl),in

610 BioScience Vol. 46 No. 8 correlated, and if the distributions of cies are not always of high trophic interaction strengths documented in status. Third, keystone species can these studies prove widespread, com- exert effects, not only through the munity importance should also com- commonly known mechanism of con- monly be distributed as in Figure 1 b. sumption, but also through such Clearly, the variance, skew, and num- interactions and processes as compe- ber of modes of any such distribu- tition, mutualism, dispersal, polli- tions are of great ecological impor- nation, disease, and by modifying tance. The greater the variance and habitats and abiotic factors (as “key- skew, the more species have CI val- stone modifiers”; Bond 1993, Mills

RoDortlonai biomass of species ues with unusually high absolute et al. 1993). values. Such species would be key- Both diversity and trophic-level Figure 3. Total (collective) impact of a stones, with a disproportionate considerations suggest that keystone species (absolute value of community effect on the composition and/or species are most likely to occur near impact x proportional abundance of a functioning of communities and eco- the top of the food chain. Top preda- species: lCI,l x p,)versus its proportional systems. Our intuition and limited tors typically have high per capita abundance, p,. Points representing a spe- experience suggest that only a small effects and low collective biomass, cies whose total impact is proportional proportion of the species in most relative to lower trophic levels. Nev- to its abundance would fall along the communities are likely to be key- ertheless, keystones may occur at diagonal line X = Y. Keystone species stones. other trophic levels. For example, have effects that exceed their propor- It is premature to prescribe nu- certain plant species may be key- tional abundances by some large factor. They also have total effects that exceed meric thresholds for applying the stone resources for pollinators or some threshold. Therefore, although a keystone designation, but with more dispersers if they flower or fruit in rhinovirus that made wildebeests sneeze data and development of the theory, times of scarcity (e.g., Didymo- (V,) might have a total effect that far one could choose quantitative crite- panax; Worthington 1982). Soil exceeded that expected from its low bio- ria. Two conditions should be ful- cyanobacteria and endolithic lichens mass, it would not be a keystone species filled. Keystone species would have may be keystone producers in the if the total effect fell below the thresh- absolute values of CI that were much Negev Desert. They fix nitrogen and old. On the other hand, a distemper greater than 1, and the absolute value support snails, whose grazing breaks virus (V,) that killed lions or wild dogs of the total (collective) effect of the down rock and creates soil (Shachak might have a collective effect of suffi- cient magnitude for keystone species species on its ecosystem ( I( t,- t,)/t,l ) and Steinberger 1980, Shachak et al. designation. Pisaster (P), sea otters (0), would also have to be great enough 1987). The community impacts of the predatory whelk Concholepas (C), to be detectable in typically noisy Negev cyanobacteria and lichens and freshwater bass (B) have large, and natural systems and to profoundly appear large relative to their small disproportionately large, impacts on influence the structure and dynamics biomass. their communities. Trees (T),giant kelp of these systems. Figure 3 depicts our Species whose primary impact on (K), prairie grass (G), and reef-building view of the relationship of keystone the community is not primarily corals (Cr),which dominate community and dominance status to the abun- trophic can also be keystones. Pos- biomass, would have total impacts that dance of species and their total ef- sible examples include keystone are large, but not disproportionate to fects on their communities or ecosys- modifiers (Mills et al. 1993), also their biomass. Positions of letters desig- nating keystone and dominant species tems. known as “ecosystem engineers” on this figure represent educated guesses. (Lawton and Jones 1995): beavers, Quantitative values that should be pre- Case studies which swamp forests and meadows scribed for thresholds of absolute total (Jenkins and Busher 1979, Naiman collective impact (vertical position re- Since the publication of Paine’s et al. 1986, Pollock et al. 1995); quired for keystone status) and factors (1966,1969)papers establishing the gophers and leaf cutter ants, whose by which keystone effects should exceed importance of top-down influences tunnels pipe water through hillslopes a species’ proportional abundance (dis- by starfish in rocky intertidal com- (Elmes 1991, Montgomery and tance above the line X = Y required for munities and the broader notion of Dietrich 1995); and badgers, whose keystone status) may vary with the community trait (e.g., species richness, bio- keystone species, there have been mounds maintain diversity in prairie mass of other species or guilds, primary many published examples in a broad floras (Platt 1975). Although such productivity, nutrient or soil retention, array of ecosystems, taxa, trophic species would not have been consid- albedo) under consideration. levels, and ecological processes ered keystones in Paine’s original (Table 1).These case studies, which formulation, they meet our criteria, have recently been reviewed by Bond because their impacts are obviously erence state” communities. Simpli- (1993), Mills et al. (1993), and important and are typically dispro- fication, however, increases the abil- Menge et al. (1994), make several portionate to their abundance. ity of ecologists to measure interaction important points. First, keystone spe- strength by reducing environmental cies, as we have defined them, have Approaches noise. been demonstrated or suggested to If interaction strength and com- occur in all of the world’s major Keystone species can be detected munity importance are positively ecosystems. Second, keystone spe- through a variety, or better, a combi-

September 1996 61 1 Table 1. Demonstrated or likely keystone species or guilds and their mechanisms of action. Ecosystem Citation(s) Keystone species or guild Target of direct effect Mechanism of effect Evidence

Marine Rocky intertidal Paine 1966, 1974 Pisaster ochraceus (predatory starfish) mussels consumption experimental, comparative Menge 1976 Nucella lapillus (predatory snail) mussels consumption experimental Hockey and Branch 1984 Haematopus spp. (black oystercatchers) limpets consumption comparativeexperimental Castilla and Duran 1985, Concholepas concholepas (predatory snail) mussels consumption Duran and Castilla 1989 Rocky subtidal Estes and Palmisano 1974 Enhydra lutris (sea otter) sea urchins consumption comparative Pelagic May et al. 1979 Balaenoptera spp. (baleen whales) krill consumption historical reconstruction Springer 1992 Theragra chalcogramma (walleye pollock) zooplankton, smaller fish consumption historical reconstruction Coral reef Hay 1984 herbivorous fish, sea urchins seaweeds consumption experimental, comparative Carpenter 1988, 1990 Diadema antillarum (herbivorous sea urchin) seaweeds consumption experimental, comparative Hughes et al. 1987 D. antillarum (herbivorous sea urchin) marine plants consumption experimental, comparative Birkeland and Lucas 1990 Acanthaster plunci (coral-eating starfish) corals consumption comparative Hixon and Brostoff Stegastes fasciolatus (territorial algivorous schooling parrotfish protection of seaweeds experimental 1996 damselfish) and surgeonfish within territories from heavy grazing Soft sediment Van Blaricom 1982 Urolophos halleri, Myliobatis californica amphipods consumption, disturbance experimental (carnivorous rays) Oliver and Slattery 1985 Eschrichtius robusta (gray whales) amphipod mats consumption, disturbance comparative Oliver et al. 1985 E. lutris (sea otters) bivalves consumption comparative Kvitek et al. 1992 E. lutris (sea otters) bivalves consumption, disturbance experimental, comparative Freshwater Lakes and ponds Brooks and Dodson 1965 Alosa pseudoharengus (planktivorous fish) zooplankton consumption comparative historical reconstruction Zaret and Paine 1973 Cichla ocellaris (piscivorous fish) prey fish consumption comparative Power and Gregoire 1978 harbor seals salmonid fishes consumption comparative Carpenter et al. 1985, Micropterus salmoides (piscivorous fish) planktivorous fish consumption experimental, Mittelbach et al. 1996 comparative Morin 1981,1983 Notophthalmus viridescens (salamander) anuran tadpoles consumption experimental

Rivers and streams Power et al. 1985 Micropterus salmoides and Micropterus algivorous minnows consumption experimental, punctatus (piscivorous bass) comparative Naiman et al. 1986 Castor canadensis (beaver) trees consumption, comparative habitat modification Cooper 1988 Oncorhynchus mykiss (predatory trout) benthic invertebrates, consumption experimental, anuran larvae Comparative Power 1990 0. mykiss, Hesperoleucas symmetricus invertebrates and fish fry consumption experimental (predatory steelhead, omnivorous minnow) Terrestrial GrassI ands Tansley and Adamson 1925 Oryctolagus cuniculus (rabbit) herbs and grasses consumption experimental, comparative Sinclair 1979 Rinderpest ungulate grazers disease epidemic comparative Huntly and Inouye 1988 Geomys bursarius (pocket gophers) underground plant tissue consumption, burrowing comparative, experimental Cantor and Whitham 1989 Thomomys bottae (pocket gopher) aspen roots consumption experimental, comparative Arctic marsh Kerbes et al. 1990 Chen caerulescens (lesser snow goose) grasses and sedges consumption experimental, comparative Woodlands Laws 1970 Lonodonta africana (elephants) trees consumption comparative Anoplolepis custodiens (seed-dispersing ant) seeds of proteaceous plants seed dispersal comparative Terborgh 1986 Ficus spp. (fig trees) vertebrates resource provision comparative Cox et al. 1991 Pteropus spp. (flying foxes) large-seeded fruits seed dispersal comparative McLaren and Peterson 1994 Canis lupus (wolves) moose consumption comparative, historical reconstruction Desert Shachak et al. 1987 Euchondrus (snails) lichens consumption, comparative, rock weathering, experimental soil formation Brown and Heske 1990 Dtpodymys spp. (kangaroo rats) seeds consumption experimental Tundra, taiga, Laine and Niemela 1980 predatory ants birch trees consumption comparative, or alpine experimental Bryant 1981 Lepus amerrcanus (snowshoe hares) trees consumption experimental, comparative Huntly 1987 Ochotona prrnceps (pika) subalpine vegetation consumption experimental nation of approaches. These ap- field experiments, and adaptive man- impacts. Each approach has distinct proaches include natural history ob- agement (sensu Walters 1986) that advantages and limitations. servation, historical reconstruction, extracts information from ecosys- The original work on the key- comparative studies, manipulative tem changes that follow large human stone concept was compelling in that

612 BioScience Vol. 46 No. 8 it employed two strong approaches- Much of the work on putative levels of organization, and diverse the experimental and the compara- keystone species has, partly by ne- taxonomic groups. Ideally, experi- tive method (Paine 1966). Experi- cessity, been descriptive in charac- mental demonstrations of keystone mental removal of a species is the ter. The importance of natural his- effects would come from manipula- most convincing way of determining tory observations and intuition in tions of single species; in practice, interaction strength, but it has logis- identifying keystone species cannot these manipulations can be hard to tic limitations. An exhaustive ex- be overstated. Inferences based solely achieve. For instance, exclosures may perimental analysis based on com- on descriptions can, however, be mis- exclude more than one member of a 9 munity manipulation would require leading. For example, a predator- guild or trophic level. If exclusion C n!/r!(n - r)! treatments in a prey interaction may appear unim- produces a dramatic change, it will community containing n different portant if the prey is rare in the not be obvious whether the unmani- * species (Wootton 1994), a prohibi- predator’s diet. This rarity could pulated condition is maintained by a tive number in most cases. This prob- arise, however, if the prey is so vul- single keystone species, or by a group lem can be partly overcome by com- nerable to the predator that it has of species with similar effects. In bining experiments with modeling already been depleted by the time the some cases, relative impacts of single approaches such as path analysis system is first observed. Such prey species are unknown, yet groups of (Schemske and Horvitz 1988, Sokal may rebound dramatically when species are known to have impacts and Rohlf 1981, Wootton 1994). predators are removed (Estes 1995, that are disproportionately large rela- Path analysis, a sequence of multiple Huffaker and Kennett 1959, Paine tive to their collective biomass regressions and correlations struc- 1966, Power 1990). Understanding (Brown and Heske 1990, Power tured by an a priori hypothesis and management of potential key- 1990). Some combination of spe- (Wootton 1994), holds great prom- stone species has also often followed cies-by-species manipulations and ise because it requires manipulating a descriptive, narrative approach, natural history detective work (e.g., only one or two strongly interacting based on a series of sequentially for- documenting the dietary preferences, species and then monitoring the re- mulated and revised hunches about feeding rates, and performances over sponses of a potentially large num- how the world works locally. This is various environmental conditions of ber of other community members. a promising approach if combined possible keystone consumers) is ne- An added benefit is that it quantifies with experimental (adaptive) man- cessary to distinguish keystone spe- both direct and indirect interaction agement (Walters 1986). cies effects from strong collective strengths. A second logistic limita- An increasing number of large- impacts of guilds or trophic levels tion of manipulative experiments is scale “natural experiments” are oc- (“diffuse predation” in Menge et al. that they are typically more restricted curring through massive human 1994). Although the clearest appli- in scope than are observational stud- habitat alteration and associated bio- cation of the keystone concept is to ies. It is usually not apparent how far diversity loss. Where such impacts single species, detection of what pro- results obtained from isolated field are unavoided or have already oc- visionally may be called “keystone experiments can be generalized to curred, ecologists should capitalize guilds” is often a useful step, both other spatial, temporal, or biotic on them to assess the influence and for advancing scientific understand- contexts. Finally, social, ethical, and prevalence of apparent keystone spe- ing and for management. technical factors may limit the ex- cies (e.g., Sparks et al. 1990, Another challenge is to determine tent to which some species and com- Terborgh 1986).For example, much the time required to assess the im- munities of interest can be manipu- of what we know about keystone pacts of changes in species’ abun- lated. species has come from studying the dances. The effects of a particular Comparative studies (of habitats results of overhunting or overfishing species perturbation may require a in which densities of species of inter- of sea otters (e.g., Estes et al. 1978), long period of time to manifest them- est vary) overcome many of the limi- of baleen whales (May et al. 1979), selves. The best known and most tations of the experimental approach, and of walleye pollock (Springer compelling examples of keystone but they inherently involve a loss of 1992). Unfortunately, poor knowl- species come from manipulative ex- rigor, given that many factors (in edge of the structure and dynamics periments (Carpenter et al. 1985, addition to the one of interest) may of natural ecosystems before mas- Paine 1974, Power et al. 1985) or differ among disparate sites and that sive human impacts often limits our from spatial or temporal contrasts of larger study sites are increasingly ability to understand changes. This habitats in which the purported key- difficult to replicate (Carpenter situation has been aggravated by the stone species were present or absent 1989). However, a combination of tragic loss of knowledge of indig- (Estes and Palmisano 1974, Owen- comparative and experimental ap- enous peoples of their own natural Smith 1988). Indirect effects in proaches can be powerful (e.g., ecosystems as they are displaced by aquatic communities often manifest Menge et al. 1994, Paine 1966),with large-scale development schemes. themselves more rapidly than in ter- the comparative observations sug- restrial systems (Estes 1995), due in gesting both the hypotheses to be Challenges some cases to the more rapid turn- tested experimentally, and, subse- over times of aquatic autotrophs. In quently, the factors that may deter- Identifying keystone species is some, perhaps most, terrestrial sys- mine the generality of experimental fraught with difficulty. It requires tems, total responses to keystones results. bridging temporal and spatial scales, may require more time than is avail-

September 1996 623 able for scientific observation. The characterize keystone consumers in role under one context but a full impact of top predator removal other systems. nonkeystone role in other contexts, from tropical forest ecosystems takes Traits that predicted keystone spe- even though the specific locations decades to centuries to become ap- cies were not, however, clearly evi- were sometimes only tens of meters parent and considerably longer to dent in a survey of well-studied ma- apart (Menge et al. 1994). Table 2 ripple through different elements of rine and freshwater keystone species summarizes various types of evidence the community (Dirzo and Miranda (Menge et al. 1994). Preferentialpre- documenting other context depen- 1990, Terborgh 1986). Brown and dation on dominant species appeared dencies that potentially affect the Heske’s (1990) demonstration of the both in systems that had or lacked keystone status of species or the im- critical role of heteromyid rodents as keystone predation (Menge et al. pacts of guilds that may include key- desert granivores (Table 2)took more 1994). Eleven other possible traits of stone species. than ten years of experimental main- predators, prey, or habitats were We know little about the causal tenance before the strong effects ap- surveyed, but none consistently dis- factors that underlie the variation in peared. Funding for the research had tinguished systems with keystone impacts of particular species in dif- ended, and the now well-known find- interactions. In general, we are pes- ferent settings. Figure 4 illustrates ings were largely a result of the au- simistic about developing what Steve possible context dependencies for thors’ interest and persistence.’ Carpenter has called “A Field Guide annual plants, sea urchins, and fresh- Calculating the interaction strength to the Strong interact or^,"^ based on water fish, whose respective impacts and related community importance of species traits alone. Field guides typi- and status as keystone or dominant a particular species to evaluate its po- cally have range maps. For strong species change with time since dis- tential keystone status requires link- interactors, we would need range turbance (Figure 4a), deletion of ing the action of individuals through maps of the variation in their im- predators (Figure 4b), or ecosystem their populations to community- and pacts, not only across geographic productivity (Figure 4c). For ex- ecosystem-level effects. It also demands space but also across gradients of ample, riverine fishes play keystone monitoring responses of potentially disturbance, productivity, physical roles as top predators in food chains diverse groups, a task that challenges factors, and abundances of other that control algal biomass in rivers, the breadth of most scientists’ taxo- species. In short, we need to under- but only following scouring winter nomic experience. These challenges stand better how context affects spe- floods (Figure 5; Table 2). More are among the greatest in ecology to- cies interaction strength if we are to quantitative field studies and devel- day; but overcoming them does not predict the roles particular species opment of theory are needed before guarantee a general result, because the may play in a particular context. we can understand, let alone predict, impact of a particular species is poten- how species interaction strengths will tially context dependent. Context dependency change in various contexts or across ranges of conditions. Nevertheless, Identifying keystone species a An increasing body of evidence sug- testable hypotheses can be formu- gests that keystone species are con- lated from trends that may occur priori by their traits text dependent. That is, keystone with diversity, trophic position, and Given the difficulties of identifying species are not necessarily dominant time for which species have been keystone species and the short time controlling agents in all parts of their associated. remaining if we are to apply this range or at all times, but instead play knowledge to their conservation, it keystone roles only under certain Diversity. In Paine’s original (1969) would be useful if such species could conditions. Along the Oregon coast, demonstration, keystone species af- be identified a priori (i.e., before the original keystone species Pisaster fected community diversity. The con- experimental removal or extinction). ochraceus occupies an unambiguous verse may sometimes be true-com- Are there traits that make species keystone role on wave-exposed rocky munity diversity may affect keystone likely to play a keystone role? Paine’s headlands (Menge et al. 1994), the status. The more species that are Pisaster is a keystone predator be- “context” in which Paine (1966, trophically similar to a species in the cause it preferentially consumes and 1974) originally demonstrated the food web (or functionally similar to suppresses mussels, which in the ab- keystone concept. In more wave-shel- a species in the interaction web; sence of this starfish can be domi- tered habitats, however, the impact Menge and Sutherland 1987), the nant space holders. Estes’ otters are of Pisaster predation was weak or greater the chance that deleting that active and mobile, and they feed vo- nonexistent (Menge et al. 1994). In species would cause compensatory raciously on sea urchins, potentially sheltered areas, prey input rates were increases in species functionally simi- destructivegrazers (Esteset al. 1978). low, and at one site, periodic, unpre- lar to it (Frost et al. 1995). This argu- These traits-high consumption rates dictable sand burial, not starfish pre- ment suggests that loss of species di- relative to prey production and dif- dation, was the overwhelming force versity may thrust more of the ferential impacts on potential domi- eliminating mussels from the lower remaining species into keystone roles nant species-would seem likely to shore. Thus, in a rocky intertidal (Chapin et al. 1995, Lawton and Brown habitat, Pisaster occupied a keystone 1993, Tilman and Downing 1994). ‘J. H. Brown, 1994, personal communica- Support for this hypothesis is un- tion. University of New Mexico, Albuquer- 2S. Carpenter, 1994, personal communica- folding on South Pacific islands, que, NM. tion. University of Wisconsin, Madison, WI. where archaeological excavations

614 Bioscience Vol. 46 No. 8 Table 2. Context dependency in keystone effects, with demonstrated or suspected causal factors.

Habitat (citation) Species (type of organism) Context dependency Factor underlying context dependence

~~ Marine

New England rocky Nucella lapillus (carnivorous gastropod) keystone in low turbulence areas, not wave forces intertidal (Menge 1976) in high turbulence areas

New England rocky Littorina littorea (herbivorous gastropod) keystone on permanently submerged change in competitive ability of intertidal (Lubchenco 1978) substrata, but not on periodically algal food submerged substrata

New Jersey soft bottom Callinectes sapidus (carnivorous crab) keystone in low turbulence areas, wave forces ! (Peterson 1979) not in high turbulence areas

Chilean rocky intertidal Concholepas concholepas keystone in high turbulence areas, wave forces, vulnerability of prey (Castilla 1981) (carnivorous gastropod) not in low turbulence areas or where sea squirt prey dominate

Oregon rocky intertidal Pisaster ochracelrs (carnivorous starfish) keystone on wave exposed headlands, prey mortality from sand burial (Menge et al. 1994) nonkeystone in wave-sheltered areas

California kelp beds Strongylocentrotus franciscanus keystone in areas with little drift kelp, degree of herbivory on locally (Harrold and Reed 1985) (herbivorous sea urchin) nonkeystone in areas with much growing food, of “donor control” drift kelp

California salt marsh Cuscuta salina (parasitic plant) keystone effect strongest where variation in host traits and (Penningsand Callaway in press) salicornia host most dominant availabilities

Freshwater

Wisconsin lake Stizostedion vitreum, keystones when phosphorus inputs are phytoplankton productivity and (Carpenter 1992, Micropterus salmoides, low to moderate, not when phosphorus species composition Lathrop and Carpenter 1992)* Esox lucius (piscivorous fish) inputs are high

California rivers (Power 1995) Onchorynchus mykiss, Hesperoleucas keystones following scouring winter overwinter mortality of predator- symmetricus (invertebrate-eating fish) floods, not during drought years resistant primary consumers

California rivers (Power 1992) 0. mykiss, H. symmetricus keystones over boulder-bedrock habitat structure (invertebrate-eatingfish) substrates, not over gravel

Southeastern US ponds Notophthalmus viridescens keystone in presence of Siren, not in prey density (Fauth and Resetarits 1991) (carnivorous salamander) absence of this predator

Swedish lakes (Diehl 1992) Perca fluviatilis (carnivorous fish) keystone in absence of macrophytes, habitat structure weaker effects with macrophytes

Terrestrial

South African shrublands Anoplolepis custodiens keystone in sclerophyll shrublands, not presence of alternative seed (Bond 1984) (seed-dispersing ant) in other shrublands, grasslands, dispersers and savannah

African savannah Loxodonta afrrcana (elephants) keystone in fire-disturbed or sparse prey (tree) size (Dublin 1990) woodlands, nonkeystone in dense, undisturbed woodlands

South Pacific Islands Pteropus spp. (large frugivorous bats) keystone on islands where large presence of alternative seed (Cox et al. 1991, frugivorous birds have been exter- dispersers Rainey et al. 1995) minated, probably not where they remain

West African villages Lassa virus (agent of lethal human potential keystone where humans density and habitat use (Garrett 1994) hemorrhagic fevers) contact the African brown rat of animal reservoir of virus (Mastomysnatalensis)

Islands, Gulf of California Metepiera anzonrca, spiders can suppress herbivores unless weather, which determines avail- (Polis et al. in press) Argrope argentata (spiders) parasitized by pornpilid wasps ability of wasp’s adult food

Southwestern US meadows Thomomys bottae gophers suppress aspen invasion of physical refuge for prey from (Cantor and Whitham 1989) (root-eating pocket gophers) meadows except on rocky outcrops burrowing herbivore

(Brown and Heske 1990) Dipodomys spp. (seed-eating kangaroo rats may prevent transition rainfall mediated rates of plant kangaroo rats) from shrubland to grassland only near recruitment, growth, survival, biogeographic zone of transition between and outcomes of competition two vegetation types

*S. R. Carpenter, 1992, personal communication. Unlversity of Wisconsin, Madison, WI. ' have documented that extinction of Figure 4. Three scenarios that suggest a frugivorous and nectar-feeding birds that interaction strengths, and hence followed human settlement (Stead- keystone or dominant status, can change man 1995). On Mangaia, Cook Is- for a given species under different cir- cumstances. (a)Successional changes in lands, for example, all of the avian the dominance and total impact of an- frugivores and nectar-feeders have nual herbs following fire in South Afri- been extirpated (Steadman and Kirch can savannah: Immediately after fire, 1990), leaving one species of flying annuals sprout and make up most of the fox (Pteropus' toganus) as the last plant biomass (AEarly).Over time, woody volant vertebrate pollinator and seed shrubs and tree seedlings reinvade and b / disperser capable of carrying large- make up increasing proportions of the seeded fruits (Rainey et al. 1995). total community biomass. The annuals On Guam, where flying foxes have at this stage (A,,,,,,) strongly determine also been nearly eliminated by hu- sites at which the later successional plants man hunting, sampling efforts de- can colonize, both positively (if annuals provide safe sites, such as more favor- tected not a single vertebrate-dis- able microclimates for survival of seed- persed seed, whereas comparable lings) and negatively, if annuals com- efforts on Samoa, where bats are still pete with woody seedlings. During the abundant, revealed much more ver- third stage, annuals disappear (biomass tebrate seed dispersal (Pierson et al. becomes undetectable; ALate).(b) Changes in press). Because seed disperser (and in the status of consumers with addition pollinator) guilds on isolated tropi- or deletion of predators that can control cal islands are depauperate to begin them: Sea urchins, when suppressed by with, and further impoverished by otters (UOtter)have low abundance, and low per capita effects if, for example, /.fFMesoFEU human impacts, species in these their kelp food has begun to escape in guilds may play crucial keystone roles size (enlarge in girth so urchins cannot Proportional biomass of species in maintaining plant diversity (Cox easily sever their fronds). If otters are et al. 1991, Elmqvist et al. 1992). deleted, urchins may increase in num- A counter-argument, however, can bers and collective biomass to the point at which they can denude kelp forests and be made that as human impacts de- maintain so-called urchin barrens (UK+). Whether their impacts are disproportion- grade ecological communities, key- ate to their biomass will depend on their per capita feeding rates and the extent to stones and the ecological organiza- which they graze drift versus attached kelp. Grazing drift kelp (UDrLft)greatly reduces tion they maintained may both be urchins' impacts on local communities even where they are abundant, because local lost, leaving dysfunctional remnant permanent habitat structure is not altered. (c) Possible impacts of visually feeding assemblages of those species that predatory fish along a productivity gradient. In oligotrophic communities, predatory fish may effectively suppress prey, with cascading impacts on other species or happen to be able to survive in the trophic levels (FOllgo).In mesotrophic communities, fish may be less effective per highly altered environments. When capita in controlling prey, if prey have faster population growth or more refuges, species loss following human impacts due to proliferating aquatic vegetation (FMeso).In eutrophic communities, fish may is not incremental but massive, rem- be even less effective as predators if algal blooms reduce visibility, and fish biomass nant species seem less likely to take may be reduced by periodic kills due to oxygen depletion by vegetation (FEU). on keystone roles, because the community architecture once maintained by species interactions has also col- terns may vary, perhaps systemati- have come together recently. Is there lapsed (Paine 1995). cally, among ecosystem types. For a relation between the length of spe- example, inverted trophic pyramids cies associations and the strength of Trophic position. Above, we have (of biomass) are common in aquatic their interactions? For example, are contrasted species that exert strong systems. Does this imply that aquatic recently added species more likely to effects by virtue of their large bio- plant species, or the herbivores that play keystone roles in communities mass (dominant species) with key- consume them, may more commonly and ecosystems than those with long stone species, whose strong effects play keystone roles than their coun- histories of association? There is emanate from their per capita (orper terparts in terrestrial systems? some evidence, both paleontological biomass) impact. The distribution of and contemporary, for new species dominant species versus keystone Time. Species living together in na- being strong interactors. The paleo- species may vary across trophic lev- ture may have radically different his- ecological literature provides evi- els. Energy flow considerations dat- tories of association with one an- dence that biotic interchanges often ing back to Lindeman (1942) suggest other. At one extreme, an interaction are followed by abnormally high rates that basal species, which have more may be old and include enough of of extinction in the recipient biotas biomass, might have lower per bio- the interacting species' evolutionary (Vermeij 1991, Webb 1985, but see mass effects, whereas keystone spe- histories to encompass the develop- Lindberg 1991). Many well-docu- cies may be more prevalent at higher ment of those traits that determine mented contemporary examples trophic levels. This hypothesis awaits the nature of their interactions. At come from the literature on inva- tests in real ecosystems, where pat- the other extreme are species that sions by alien species (e.g., Bailey

616 Bioscience Vol. 46 No. 8 strategy that takes into account potential surprises from small inter- ventions or changes. In particular, the keystone concept shows how:

the loss of some species of b low abundance may have sur- prisingly dramatic effects; the preservation of a species of concern may depend on the distribution and abundance of other species with which the target has no recognized interaction; and, conversely, the loss of a species, such as a top carnivore, may reverberate to affect members of seemingly disparate guilds, such as plants or decomposers.

These insights from the keystone Figure 5. Interactions among (a) juvenile steelhead (Onchorhynchus mykiss), (b) concept suggest three key points re- chironomid larvae (Pseudochironomus richardsonii), and (c) algae (Cfudophora, lated to policy and management. Nostoc, and Epithemia) can be strong, but vary between drought and flood years. First, land managers should care- fully consider the consequences of 1993, D’Antonio and Vitousek 1992, mentation has provided additional the loss of species for which no obvi- Kitchell and Crowder 1986, McDonald guidelines on the minimal areas and ous role in the ecosystem has been et al. 1990, Vitousek 1990, Zaret spatial network of preserves needed discovered. The keystone species con- and Paine 1973), which can have to minimize extinction risk for par- cept indicates the need for a design dramatic and widespread conse- ticular species (e.g., Soul6 et al. 1992). with a wide margin of safety for quences for communities. These too have been translated into managed lands to guard against the general sets of rules or procedures loss of those organisms with dispro- Application to preserve that are widely used by practitioners portionately high community impor- selection and management in biodiversity assessments (e.g., tance values. Second, introduced Mace and Lande 1991). alien species may, like keystone spe- The keystone concept has great rel- Far less attention has been paid to cies, have potential strong effects evance for identifying the most suit- the assessment of the critical ecologi- disproportionate to their biomass. able areas for biodiversity preserves. cal processes that maintain whole com- More commonly, however, intro- To date, areas to preserve have been munities or ecosystems (but see Leigh duced aliens may become dominant selected by comparing the species et al. 1993).Some, althoughnot all, of species in new habitats that lack the present in alternative areas and these processes are driven by keystone parasites, pathogens, predators, or choosing those areas that contain or critical species. Approaches to iden- competitors that controlled invaders the most diversity or the most irre- tifying critical ecological processes, in their native ecosystems. These in- placeable species (Pressey et al. 1993). species that may drive them, and their vaders, at the onset of invasion, can Systematists have developed meth- mode of action are, today, the most be relatively cryptic, and managers ods for including taxonomic unique- glaring omissions in the conservation can play keystone roles themselves ness when setting priorities for spe- biologist’s toolbox for selection and by eradicating such invaders before cies conservation (Vane-Wright et design of preserves. Methods are they become well established. al. 1991).Although these criteria are needed for detecting species likely to Finally, we note the lack of a well- valid, they are static. Natural com- be strong interactors, including poten- developed protocol for identifying munities are not museum collections. tial keystone species, in rapid bio- potential keystone species. We urge The diversity or particular species diversity surveys. that when a potential keystone role that conservation managers seek to If we can identify keystone species is suggested for a given species, ef- preserve may be lost if the dynamic in various ecosystems, it will be use- fort be directed toward obtaining real fragility of communities and ecosys- ful to set aside critical areas and to evidence for this hypothesis. The field tems contained within preserves is manage them so as to maintain these is littered with far too many untested not taken into account in managing keystones, instead of solely focusing anecdotal “keystone species.” them. Dynamics have been incorpo- on endangered local species or geo- rated into management models, but graphical hot spots of biodiversity. Future directions only at the level of single-species If local keystones cannot be identi- populations. Research on the popu- fied, the keystone concept points to The keystone concept has been in- lation consequences of habitat frag- the need for a cautious management voked for almost 30 years by ecolo-

September 1996 617 gists to interpret and publicize their macroscale community traits are important questions that remain findings in a variety of ecosystems. likely to be causally linked, and un- about keystones more quickly than Nevertheless, ecologists still lack the raveling their connections is another would exclusive focus on the traits empirical basis needed to detect, in- important avenue of research needed of keystone species alone. terpret, and predict general patterns for understanding and preserving Among the important unanswered in the occurrence of keystone species natural ecosystems.) Although we questions are the following: How or to apply the concept for manage- tend to favor per biomass measures, are interaction strengths distributed ment. We do not yet have quantita- the investigator needs to decide among species in various communi- tive data with which to position spe- whether per capita measures are more ties? Are keystone species common? cies on Figures 3 and 4. Few if any useful or more feasible for specific Are communities structured by key- studies indicate how community systems and questions. The spatial stone species common? Are keystone importance is distributed among spe- and particularly the temporal scope species more prevalent in some types cies in communities. Generally, com- of monitoring following the manipu- of communities or ecosystems than munity importance should be corre- lation will strongly affect the esti- in others (aquatic versus terrestrial, lated with field measurements of mate of a species’ community impor- ancient versus recently assembled, interaction strengths, as measured in tance. Therefore, the researcher must diverse versus depauperate)? Are the studies of keystone consumers by use his or her best judgment about taxonomically unique species more Paine (1992), Fagan and Hurd the temporal and spatial scales over or less likely to be keystone species (1994),and Rafaelli and Hall (1992; which most of the important feed- than species with close contempo- Figure 2). Quantitative data on backs occur. The investigator must rary relatives? Is our present focus nontrophic keystone species are even also decide whether impacts or re- on keystone consumers at high more scant than on keystone con- sponses of species, rather than of trophic levels warranted, or are we sumers. Mutualists, such as pollina- groups of species, must be isolated. overlooking species at lower trophic tors or seed dispersers, are most likely Detailed resolution of pairwise in- levels that play other cryptic but to have keystone effects if they inter- teractions on a species-by-species critical keystone roles? We hope that act with many species that depend basis is important for detailed un- natural ecosystems remain intact long on the services provided (Gilbert derstanding of the mechanisms of enough for such questions to be ad- 1980) or if they strongly affect the community interactions, but it can- dressed, but this outcome depends criti- performance of a species that is quan- not be a first priority when assessing cally on accelerating the feedback be- titatively or qualitatively important potential keystones under most tween science and management. in a system. Studies of both kinds of biodiversity triage scenarios. keystone mutualists exist, but none Methods for rapid assessment of Acknowledgments has gone far beyond the anecdotal in potential keystones would be an ex- documenting community- or ecosys- tremely useful addition to the con- Our thanks to Hal Mooney for orga- tem-level impacts (Bond 1993,1994). servation biologist’s toolbox. As the nizing the 1995 Keystone workshop In general, species with nontrophic database on demonstrated keystone in Hilo, Hawaii; to Adrian Sun for effects may be most important if species grows, it should be mined for preparing the figures; to Thomas they affect the performance and patterns that may forecast likely key- Elmqvist, Lauri Oksanen, Sergio population dynamics of species that stones by their attributes, or con- Navarrete, Karina Nielsen, Bill are potential dominants, as demon- texts in which species with certain Rainey, Cathy Pfister, Lennart strated for keystone consumers. Test- attributes are likely to play strong Persson, Gary Polis, Don Strong, ing this hypothesis deserves more roles. Both community importance James Thompson, and the Berkeley attention. and interaction strength (and there- Ecolunchgroup for constructive com- In the effort to refocus the term fore the status of species as keystone ments and/or references; and to the keystone for ecological research, and or dominant species) are context de- United Nations Environmental to make it more useful for policy pendent, simply because perfor- Programme Global Biodiversity As- makers concerned with preserving mances of organisms change with sessment Initiative and the National biodiversity, we confront a tradeoff variation in their environments. AI- Science Foundation (DEB-9319924 between flexibility and rigor. The though this ecological truism ensures [Ecology]) for support. community importance index offered that results from specific field stud- in this article is quantified in an ies will be difficult to generalize, References cited objective and generally repeatable examination of the nature of context Ashton PS. 1992. Species richness in plant com- fashion, within the constraints im- dependencies may lead to more fun- munities. Pages 3-22 in Fiedler PL, Jain KS, posed by noisy natural ecosystems. damental generalizations. Keystone eds. Conservation biology. London (UK): Yet it also can be tailored to a con- status depends not only on the prop- Chapman and Hall. Bailey EP. 1993. Introduction of foxes to Aleu- siderable degree by the investigator. erties of that species with a dispro- tian Islands: history, effects on avifauna, He or she chooses, based on the portionate influence, but also on the and eradication. US Fish and Wildlife Ser- natural history of the system and the species with which it interacts and vice Resource Publication 193: 1-53. purpose of the study, which commu- the physical arena containing and BirkelandC, Lucas JS. 1990. Acanthasterplanci: nity trait (e.g., species richness, al- constraining these interactions. Ap- major management problem of coral reefs. Ann Arbor (MI): CRC Press. bedo, arthropod biomass, nutrient preciating this aspect of the keystone Bond WJ. 1993. Keystone species. Pages 237- retention) is to be monitored. (These concept will lead to answers to the 253 in Schulze ED, Mooney HA, eds. Eco-

618 BioScience Vol. 46 No. 8 system function and biodiversity. Berlin habitatcomplexity. Ecology 73: 1646-1661. Journal of Experimental Marine Biology (Germany): Springer-Verlag. Dublin HT, Sinclair ARE, McGlade J. 1990. and Ecology 113: 39-59. . 1994. Do mutualisms matter? Assess- Elephants and fire as causes of multiple Huntly NJ. 1987. Effects of refugingconsumers ing the impact of pollinator/disperser dis- stable states in the Serengeti-Mara wood- on pikas (Ochotonaprinceps) on subalpine ruption on plant extinctions. Philosophical lands. Journal of Animal Ecology 59: 1147- meadow vegetation. Ecology 68: 274-293. Transactions of the Royal Society of Lon- 1164. Huntb N, Inouye RS. 1988. Pocket gophers in don B Biological Sciences 344: 83-90. Duran LR, Castilla JC. 1989. Variation and ecosystems: patterns and mechanisms. Brooks JL, Dodson SI. 1965. Predation, body persistence of the middle rocky intertidal Bioscience 38: 786-792. size and composition of plankton. Science community of central Chile, with and with- Jenkins SH, Busher PE. 1979. Castor canadensis. 150: 28-35. out human harvesting. Marine Biology 103: Mammalian Species 120: 1-8. Brown JH, Heske EJ. 1990. Control of a desert- 555-562. Kerbes RH, Kotanen PM, Jefferies RL. 1990. grassland transition by a keystone rodent Elmes GW. 1991. Ant colonies and environ- Destruction of wetland habitats by lesser guild. Science 250: 1705-1707. mental disturbance. Symposia of the Zoo- snow geese: a keystone species on the west Bryant JP. 1981. Phytochemical deterrence of logical Society of London 63: 15-32. coast of Hudson Bay. Journal of Applied snowshare hare browsing by adventitious Elmqvist T, Cox PA, Rainey WE, Pierson ED. Ecology 27: 242-258. shoots of four Alaskan trees. Science 250: 1992. Restricted pollination on oceanic is- Kitchell JF, Crowder LB. 1986. Predator-prey 1705-1 707. lands: pollination of Ceiba pentundra by interactions in Lake Michigan: model pre- Cantor LF, Whitham TG. 1989. Importance of flying foxes in Samoa. Biotropica 24: 15- dictions and recent dynamics. Environmen- belowground herbivory:pocket gophers may 23. tal Biology of Fishes 16: 205-21 1. limit aspen to rock outcrop refugia. Ecology Estes JA. 1995. Top-level carnivores and eco- Kvitek RG, Oliver JS, DeGange AR, Anderson 70: 962-970. system effects: questions and approaches. BS. 1992. Changes in Alaskan soft-bottom Carpenter RC. 1988. Mass mortality of a Car- Pages 151-158 in Jones CG, Lawton JH, prey communities along a gradient in sea ibbean sea urchin: immediate effects on com- eds. Linking species and ecosystems. New otter predation. Ecology 73: 413-428. munity metabolism and other herbivores. York: Chapman and Hall. Laine KJ, Niemela P. 1980. The influence of Proceedings of the National Academy of Estes JA, Palmisano JF. 1974. Sea otters: their ants on the survival of mountain birches Sciences of the United States of America 85: role in structuring nearshore communities. during an Oporinia autumnata (Lepi- 511-514. Science 185: 1058-1060. doptera, Geometridae) outbreak. Oecologia . 1990. Mass mortality of Dzadema Estes JA, Smith NS, Palmisano JF. 1978. Sea 47: 39-42. antillarum. I. Long-term effects on sea ur- otter predation and community organiza- Lathrop RC, Carpenter SR. 1992. Zooplank- chin population-dynamics and coral reef tion in the western Aleutian Islands, Alaska. ton and their relationship to phytoplank- algal communities. Marine Biology 104: Ecology 59: 822-833. ton. Pages 127-150 in Kitchell JF, ed. Food 67-77. Fagan WF, Hurd LE. 1994. Hatchdensityvaria- web management. New York: Springer- Carpenter SR. 1989. Replication and treatment tion of a generalist arthropod predator: Verlag. strength in whole-lake experiments. Ecol- population consequences and community Laws RM. 1970. Elephants as agents of habitat ogy 70: 453-463. impact. Ecology 75: 2022-2032. and landscape change. Oikos 21: 1-15. . 1992. Destabilization of planktonic Fauth JE, Resetarits WJ Jr. 1991. Interactions Lawton JH, Brown VK. 1993. Redundancy in ecosystems and blooms of blue-green algae. between the salamander Siren intermedia ecosystems. Pages 255-270 in Schulze ED, Pages 461-481 in Kitchell JF, ed. Food web and the keystone predator Notophthalmus Mooney HA, eds. Biodiveristy and ecosys- management. New York: Springer-Verlag. viridescens. Ecology 72: 827-838. tem function. Berlin (Germany): Springer- Carpenter SR, Kitchell JF, Hodgson JR. 1985. Frost TM, Carpenter SR, Ives AR, Kratz TK. Verlag. Cascading trophic interactions and lake pro- 1995. Species compensation and comple- Lawton JH, Jones CG. 1995. Linking species ductivity. Bioscience 35: 634-639. mentarity in ecosystem function. Pages 224- and ecosystems: organisms as ecosystem Castilla JC. 1981. Perspectivas de investigacion 239 in Jones CG, Lawton JH, eds. Linking engineers. Pages 141-150 in Jones CG, en estructura y din mica de comunidades species and ecosystems. New York: Lawton JH, eds. Linking species and ecosys- intermareales rocosas de Chile Central. 11. Chapman and Hall. tems. New York: Chapman and Hall. Depredadores de alto nivel trafico. Medio Garrett L. 1994. The coming plague. New Leigh EG Jr., Wright SJ, Herre EH, Putz FE. Ambiente 5: 190-215. York: Farrar, Straus and Giroux. 1993. The decline of tree diversity on newly Castilla JC, DuranLR. 1985. Human exclusion Gentry AH, Dodson CH. 1987. Diversity and isolated tropical islands: a test of a null from the rocky intertidal zone of central biogeography of neotropical vascular epi- hypothesis and some implications. Evolu- Chile: the effects on Concholepas con- phytes. Annals of the Missouri Botanical tionary Ecology 7: 76-102. cholepas (Gastropoda). Oikos 45: 391-399. Garden 74: 205-233. Lindberg DR. 1991. Marine biotic interchange Chapin FS 111, Lubchenco J, Reynolds H. 1995. Gilbert LE. 1980. Food web organization and between the northern and southern hemi- Biodiversity effects on patterns and pro- conservation of neotropical diversity. Pages spheres. Paleobiology 17: 308-324. cesses of communities and ecosystems. In 11-34 in Soul6 ME, Wilcox BA, eds. Con- Lubchenco J. 1978. Plant species diversity in a Mooney HA, Lubchenco J, Dirzo R, Sala servation biology. Sunderland (MA): marine intertidal community: importance OE, eds. Global biodiversity assessment, Sinauer. of herbivore food preference and algal com- United Nations Environmental Programme. Harrold C, Reed DC. 1985. Food availability, petitive abilities. American Naturalist 112: Cooper SD. 1988. The responses of aquatic sea urchin grazing, and kelp forest commu- 23-39. insects and tadpoles to trout. Internationale nity structure. Ecology 66: 1160-1 169. Mace GM, Lande R. 1991. Assessing extinc- Vereinigung fur theoretische und ange- Hay ME. 1984. Patterns of fish and urchin tion threats: toward a reevaluation of IUCN wandte Limnologie, Verhandlungen 23: grazing on Caribbean coral reefs: are previ- threatened species categories. Conservation 1698-1 703. ous results typical? Ecology 65: 446-454. Biology 5: 148-157. Cox PA, Elmqvist T, Pierson ED, Rainey WE. Hixon MA, Brostoff WN. 1996. Succession May RM, Beddington JR, Clark CW, Holt SJ, 1991 Flying foxes as strong interactors in and herbivory: effects of differential fish Laws RM. 1979. Management of multi- South Pacific island ecosystems: a conserva- grazing on Hawaiian coral-reef algae. Eco- species fisheries. Science 205: 267-277. tion hypothesis. Conservation Biology 5: logical Monographs. 66: 67-90. McDonald ME, Crowder LB, Brandt SB. 1990. 448-454 Hockey PAR, Branch GM. 1984. Oystercatch- Changes in Mysis and Pontoporeia popula- D’Antonio CM,Vitousek PM. 1992. Biological ers and limpets: impacts and implications. tions in southeastern Lake Michigan: a re- invasions by exotic grasses, the grass-fire Ardea 72: 199-200. sponse to shifts in the fish community. Lim- cycle, and global change. Annual Review of Huffaker CB, Kennett CE. 1959. Ten-year study nology and Oceanography 35: 220-227. Ecology and Systematics 23: 63-87. of vegetational changes associated with bio- McLaren BE, Peterson RO. 1994. Wolves, Dayton PK 1985 The structure and regulation logical control of Klamath weed. Journal of moose, and tree rings on Isle Royale. Science of some South American kelp communities. Range Management 12: 69-82. 266: 1555-1558. Ecological Monographs 55: 447-468. Hughes TP, Reed DC, Boyle ML. 1987. Her- Menge BA. 1976. Organization of the New Diehl S 1992 Fish predation and benthic com- bivory on coral reefs: community structure England rocky intertidal community: role munity structure: the role of omnivory and following mass mortality of sea urchins. of predation, competition and environmen-

September 1996 619 tal heterogeneity. Ecological Monographs Pierson ED, Elmqvist T, Rainey WE, Cox PA. ecological risks of biotechnology. Stoneham 46: 355-393. In press. Effects of tropical cyclonic storms (MA): Butterworth-Heinemann. Menge BA, Sutherland JP. 1987. Community on flying fox populations on the South Pa- Soul6 ME, Alberts AC, Bolger DT. 1992. The regulation, variation in disturbance, com- cific islands of Samoa. Conservation Biol- effects of habitat fragmentation on chapar- petition, and predation in relation to envi- ogy. ralplants andvertebrates. Oikos 63: 39-47. ronmental stress and recruitment. Ameri- Platt WJ. 1975. The colonization and forma- Sparks RE, Bayley PB, Kohler SL, Osborne LL. can Naturalist 130: 730-757. tion of equilibrium plant species associa- 1990. Disturbance and recovery of large Menge BA, Berlow EL, Blanchette CA, tions on badger disturbances in a tall-grass floodplain rivers. Environmental Manage- Navarrete SA, Yamada SB. 1994. The key- prairie. Ecological Monographs 45: 285- ment 14: 699-709. stone species concept: variation in interac- 305. Springer AM. 1992. A review: walleye pollock tion strength in a rocky intertidal habitat. Polis GA, Hurd SH, Jackson T. In press. The in the North Pacific-how much difference Ecological Monographs 64: 249-287. role of natural history in the design and do they really make? Fisheries Oceanogra- Mills LS, Soul6 ME, Doak DF. 1993. The key- evaluation of experiments on land commu- phy 1: 80-96. stone-species concept in ecology and con- nities. In Resetarits WJ Jr., Bernard0 J, eds. Steadman DW. 1995. Prehistoric extinctions of servation. Bioscience 43: 219-224. Experimental approaches in ecology and Pacific Island birds: biodiversity meets Mittelbach GG, Turner AM, Hall DJ, Tettig JE, evolution. Oxford (UK): Oxford University zooarchaeology. Science 267: 1123-1 131. Osenberg CW. 1996. Perturbation and re- Press. Steadman DW, Kirch PV. 1990. Prehistoric silience in an aquatic community: a long- Pollock MM, Naiman RJ, Erickson HE, extinction of birds on Mangaia, Cook Is- term study of the extinction and reintroduc- Johnston CA, Pastor j,Pinay G. 1995. Bea- lands, Polynesia. Proceedings of the Na- tion of a top predator. Ecology 76: ver as engineers: influences on biotic and tional Academy of Sciences of the United 2347-2360. abiotic characteristics of drainage basins. States of America 87: 9605-9609. Montgomery DR, Dietrich WE. 1995. Hydrolgic Pages 117-126 in Jones CG, Lawton JH, Strong DR. 1977. Epiphyte loads, tree falls, and processes in a low-gradient source area. eds. Linking species and ecosystems. New perennial forest disruption: a mechanism Water Resources Research 31: 1-10. York: Chapman and Hall. for maintaining higher tree species richness Morin PJ. 1981. Predatory salamanders reverse Power G, Gregoire J. 1978. Predation by fresh- in the tropics without animals. Biotropica 4: the outcome of competition among three water seals on the fish community of Lower 215-21 8. species of anuran tadpoles. Science 212: Seal Lake, Quebec. Journal of the Fisheries Tansley AG, Adamson RS. 1925. Studies of the 1284-1286. Research Board of Canada 35: 844-850. vegetation of the English chalk. 111. The . 1983. Predation, competition, and the Power ME. 1992. Habitat heterogeneity and chalk grasslands of the Hampshire-Sussex composition of larval anuran guilds. Eco- the functional significance of fish in river border. Journal of Ecology 13: 177-223. logical Monographs 53: 119-138. food webs. Ecology 73: 1675-1688. Terborgh J. 1986. Keystone plant resources in Naiman RJ, Melillo JM, Hobbie JE. 1986. . 1990. Effects of fish in river food webs. the tropical forest. Pages 330-344 in Soul6 Alteration of North American streams by Science 250: 811-814. M, ed. Conservation biology: the science of beaver (Castor canadensis). Ecology 67: . 1995. Floods, food chains, and ecosys- scarcity and diversity. Sunderland (MA): 1254-1 289. tem processes in rivers. Pages 52-60 in Jones Sinauer. Oliver JS, Slattery PN. 1985. Destruction and CG, Lawton JH, eds. Linking species and Tilman D, Downing JA. 1994. Biodiversity and opportunity on the sea floor: effects of gray ecosystems. New York: Chapman and Hall. stability in grasslands. Nature 367: 363- whale feeding. Ecology 66: 1965-1975. Power ME, Matthews WJ, Stewart AJ. 1985. 365. Oliver JS, Kvitek RG, Slattery PN. 1985. Wal- Grazing minnows, piscivorous bass and Van Blaricom GR. 1982. Experimental analysis rus disturbance: scavenging habits and stream algae: dynamics of a strong interac- of structural regulation in a marine sand recolonization of the Bering Sea benthos. tion. Ecology 66: 1448-1456. community exposed to oceanic swell. Eco- Journal of Experimental Marine Biology Power ME, Sun A, Parker G, Dietrich WE, logical Monographs 52: 283-305. and Ecology 91: 233-246. Wootton JT. 1995. Hydraulic food chain Vane-Wright RI, Humphries CJ, Williams PH. Owen-Smith RN. 1988.Megaherbivores. Cam- models: an approach to the study of food 1991. What to protect? Systematics and the bridge (UK): Cambridge University Press. web dynamics in large rivers. BioScience 45: agony of choice. Biological Conservation Paine RT. 1966. Food web complexity and 159-167. 55: 235-254. species diversity. American Naturalist 100: Pressey RL, Humphries CJ, Margules CR, Vane- Vermeij, GJ. 1991. Anatomy of aninvasion: the 65-75. Wright RI, Williams PH. 1993. Beyond op- trans-Arctic interchange. Paleobiology 17. . 1969. A note on trophic complexity portunism: key principles for systematic re- Vitousek PM. 1990. Biological invasions and and community stability. American Natu- serve selection. Trends in Ecology and ecosystem processes: Towards an integra- ralist 103: 91-93. Evolution 8: 124-128. tion of population biology and ecosystem . 1974. Intertidal community structure: Raffaelli D, Hall SJ. 1992. Compartments and studies. Oikos 57: 7-13. experimental studies on the relationship predation in an estuarine food web. Journal Walters C. 1986. Adaptive management of between a dominant competitor and its prin- of Animal Ecology 61: 551-560. renewable resources. New York: MacMillan. cipal predator. Oecologia 15: 93-120. Rainey WE, Pierson ED, Elmqvist T, Cox PA. Webb SD. 1985. Late Cenozoic mammal dis- . 1992. Food-web analysis through field 1995. The role of flying foxes (Pteropodidae) persals between the Americas. Pages 49-85 measurement of per capita interaction in oceanic island ecosystems of the Pacific. in Stehli FG, Webb SD, eds. The great Ameri- strength. Nature 355: 73-75. Pages 47-62 in Racey PA, Swift SM, eds. can biotic interchange. New York: Plenum Paine RT, Suchanek TH. 1983. Convergence of Ecology, evolution and behaviour of bats. Press. ecological processes between independently Symposium of the Zoological Society of Wootton JT. 1994. Predicting direct and indi- evolved competitive dominants: a tunicate- London nr 67. Oxford (UK): Clarendon rect effects: an integrated approach using musselcomparison. Evolution37: 821-831. Press. experiments and path analysis. Ecology 75: Pennings SC, Callaway RM. In press. Impact of Shachak M, Steinberger Y. 1980. An algae- 151-165. a parasitic plant on the structure and dy- desert snail food chain: energy flow and soil Worthington A. 1982. Population sizes and breed- namics of salt marsh vegetation. Ecology. turnover. Oecologia 146: 402-41 1. ing rhythms of two species of manakins in Peterson CH. 1979. Predation, competitive ex- Shachak M, Jones CG, Granot Y. 1987. Her- relation to food supply. Pages 213-226 in clusion, and diversity in the soft-sediment bivory in rocks and the weathering of a Leigh EG Jr., Rand AS, Windsor DM, eds. benthic communities of estuaries and la- desert. Science 236: 1098-1099. The ecology of a tropical forest. Washington goons. Pages 233-264 in Livingston RJ, ed. Simberloff D. 1991. Keystone species and com- (DC):Smithsonian Institution Press. Ecological processes in coastal and marine munity effects of biological introductions. Zaret TM, Paine RT. 1973. Species introduction systems. New York: Plenum. Pages 1-19 in Ginzberg LR, ed. Assessing in a tropical lake. Science 182: 449-455.

BioScience is the monthly magazine for biologists in all fields. It includes articles on research, policy, computers, and education; news features on developments in biology; book reviews; meetings calendar. Published by the American Institute of Biological Sciences, 1444 Eye St. NW, Suite 200, Washington, DC 20005; 202/628-1500. 1996 membership dues, including BioScience subscription: Individual $60.00/year; Student $35.00/year. 1996 Institutional sub- 620 scription rates: $165.00/year (domestic), $193.00/year (foreign). Bioscience Vol. 46 No. 8