From the Neutral Theory to a Comprehensive and Multiscale Theory of Ecological Equivalence François Munoz, Philippe Huneman

From the neutral theory to a comprehensive and multiscale theory of ecological equivalence François Munoz, Philippe Huneman

To cite this version:

François Munoz, Philippe Huneman. From the neutral theory to a comprehensive and multiscale theory of ecological equivalence. Quarterly Review of Biology, University of Chicago Press, 2016, 91 (3), pp.321-342. 10.1086/688098. hal-01968460

HAL Id: hal-01968460 https://hal.archives-ouvertes.fr/hal-01968460 Submitted on 2 Jan 2019

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Volume 91, No. 3 THE QUARTERLY REVIEW OF BIOLOGY September 2016

FROM THE NEUTRAL THEORY TO A COMPREHENSIVE AND MULTISCALE THEORY OF ECOLOGICAL EQUIVALENCE

François Munoz Université de Montpellier, UMR-AMAP, TA A-51/PS2 34398 Montpellier cedex 5, France French Institute of Pondicherry Pondicherry 605001, India e-mail: [email protected]

Philippe Huneman Université Paris I Sorbonne, IHPST 75006 Paris, France e-mail: [email protected]

keywords neutral theory, community ecology, ecological equivalence, causal explanation, null hypothesis, philosophy of ecology

abstract The neutral theory of biodiversity assumes that coexisting organisms are equally able to survive, reproduce, and disperse (ecological equivalence), but predicts that stochastic ﬂuctuations of these abilities drive diversity dynamics. It predicts remarkably well many biodiversity patterns, although substantial evidence for the role of niche variation across organisms seems contradictory. Here, we discuss this apparent paradox by exploring the meaning and implications of ecological equivalence. We address the question whether neutral theory provides an explanation for biodiversity patterns and acknowledges causal processes. We underline that ecological equivalence, although central to neutral theory, can emerge at local and regional scales from niche-based processes through equalizing and stabilizing mechanisms. Such emerging equivalence corresponds to a weak conception of neutral theory, as opposed to the assumption of strict equivalence at the individual level in strong conception. We show that this duality is related to diverging views on hypothesis testing and modeling in ecology. In addition, the stochastic dynamics exposed in neutral theory are pervasive in ecological systems and, rather than a null hypothesis, ecological equivalence is best understood as a parsimonious baseline to address biodiversity dynamics at multiple scales.

321

Introduction on whether neutral theory provides an ex- INCE the publication of The Unified Neu- planation for observed species diversity pat- S tral Theory of Biodiversity and Biogeography terns, and on how to test its expectations. (Hubbell 2001), neutral theory has been Our primary objective here is to clarify the very influential but also very controversial explanatory nature of the theory. in ecology. Across the numerous models The basic idea of neutral theory is that that have been elaborated, neutral theory numerous and repeated “microscopic” indi- focuses on stochastic variations in individual vidual stochastic variations result, over space birth and death rates to predict long-term and time, in specific patterns of species rel- trends in the composition of communities. ative abundances. Central to the debate on Furthermore, the basic concepts of this the- neutral theory is the divergence between ory are closely related to the neutral theory “weak” and “strong” interpretations of the of molecular evolution, which has also been theory (Bell 2001). The “strong” interpreta- influential and controversial among evolution considers that neutral theory is refuted tionary biologists from the 1960s (Kimura if strict fitness equivalence is not met at the 1983). At the heart of both approaches is individual level. Conversely, in the “weak” the fundamental equivalence assumption interpretation, variation in individual fit- that biological variation among organisms ness can be counterbalanced and neutral does not reflect any difference in their abil- patterns of biodiversity can emerge without ity to survive, reproduce, and disperse (Bell strict individual equivalence. In this case, 2001). Species coexistence is then a dynam- the fact that an observed pattern does not ical equilibrium driven by the stochastic reflect the influence of biological differ- variations in speciation and extinction at ences does not mean that individual dynam- the regional scale, and in basic life, death, ics are actually neutral, but it indicates that and dispersal dynamics at the local scale the observed pattern does not convey a sig- (called “neutral dynamics”). The equiva- nature of nonneutral processes. In this re- lence assumption is traditionally opposed to gard, the strong interpretation emphasizes an exclusive explanatory role of niche vari- basic neutral dynamics at the individual ation and competitive exclusion in ecology level, while the weak interpretation empha- (Whitfield 2002), and to the correlated idea sizes the emergence of neutral patterns. of natural selection in evolution (Mikkelson Here, we analyze the mechanistic under- 2005). standing of neutral dynamics in the dual The neutral theory has shown a remark- interpretations of neutral theory, and show ably good heuristic value to predict pat- that the issue more broadly relates to what terns of species abundance distributions explanation, causality, and hypothesis test- (SAD) in a number of ecosystems (Hub- ing mean in community ecology. In this bell 1997, 2001; Bell 2001; Chave 2004). regard, even a decade and a half after Hub- Encouraged by this apparent robustness, bell’s book, some basic questions are still many studies have applied neutral models hard to answer: Does neutral theory pro- to analyze community dynamics in a variety vide an explanation for real community dy- of ecosystems (Volkov et al. 2003; Latimer namics? If so, is this explanation causal? Is et al. 2005; Chave et al. 2006). At the same the neutral theory a null hypothesis against time, variation in survival and reproductive which to analyze observed patterns? Are abilities is often observed in real communi- there instead signatures of both neutral and ties and apparently violates the equivalence nonneutral processes in the composition assumption (Purves and Turnbull 2010). of ecological communities? The nature of The fact that the theory works well in terms explanation is clearly a central issue, since of resulting species-abundance patterns de- answers to the above questions may rely spite the violation of its basic assumption is on what we mean by “signatures” (from a an apparent paradox (Gewin 2006). A de- pattern-oriented point of view) and “causal cade and a half after the work of Hubbell explanation” (from a process-oriented point (2001), debates and disagreements persist of view). A basic aim of this paper is then

This content downloaded from 131.156.224.067 on August 11, 2016 01:02:07 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). September 2016 Neutral theory and ecological equivalence 323 to delineate more clearly the nature of eco- the traditional dichotomy of neutral and logical equivalence, central to the neutral niche-based biodiversity dynamics should theory, so as to better assess the status of be abandoned, and be replaced by a more causation within the theory. We will show comprehensive theory of ecological equiva- that the weak and strong interpretations of lence. When analyzing biodiversity dynam- neutral theory imply different conceptions ics in terms of ecological equivalence, the of ecological equivalence, which entail a pro- focus is no longer on the immediate and found epistemic divide. We will emphasize local effects of biological differences, but that ecological equivalence can be found rather on the level and scale at which these at multiple scales under the influence of differences matter to explain biodiversity both neutral and nonneutral dynamics and, patterns. Figure 1 summarizes the overall therefore, that ecological equivalence can logic and organization of the paper. Table 1 be part of a comprehensive theory of biodi- provides a glossary of the basic concepts of versity dynamics. this manuscript. This paper will first address the nature of neutral dynamics, and whether the neutral theory can provide a mechanistic framework An Explanatory Theory of for biodiversity dynamics and species coex- Unspecified Individual Interactions istence. We will especially focus on the re- To understand the role of the neutral lationship between neutral processes and theory in ecology, one should first focus on neutral patterns. The second step will ques- the kind of explanation it provides for bio- tion the nature and status of ecological diversity. In neutral communities, species equivalence in terms of equalizing and sta- abundances change with time according to bilizing mechanisms, and how it should be stochastic birth, death, and dispersal events, used to investigate biodiversity dynamics plus speciation events in a large-scale biogeo- from local communities to a regional set of graphical metacommunity (Hubbell 2001). species. We will discuss the philosophical This framework is analogous to the neutral nature of the dualism in neutral and non- theory of population genetics, where species neutral views, and highlight that spatial and abundance is replaced by allele frequency temporal scales are critical aspects of the and speciation by mutation (Kimura 1983). link between emerging patterns of ecologi- In finite communities, stochastic fluctua- cal equivalence and possibly nonneutral un- tions of species abundances (so-called eco- derlying processes. We will then distinguish logical drift, analogous to random genetic several levels at which mechanisms can be drift in population genetics) lead to random appealed to when one analyzes biodiversity extinctions. Without any influx of migrants, patterns, and reconsider the notion of null a local community will undergo species loss hypothesis against which to test the role of until only one species survives (fixation). niche differences. Actually, a major gap exists Likewise, at the regional scale, the metacom- currently in ecology between hypothetico- munity will lose species until fixation, unless deductive approaches where neutral “ran- new species are created by speciation. Al- dom” models represent null hypotheses that though the stochastic variation in speciation, should be falsified in favor of alternative mod- migration, and local birth-death dynamics els representing the contribution of niche yields an unpredictable community compo- differences and integrative approaches that sition at any given time, statistical patterns recognize a genuine role for neutral dynam- such as species abundance distributions can ics and intend to characterize their contri be predicted at equilibrium depending on bution to community dynamics, possibly in the balance between speciation, migration, combination with other niche-based pro- and drift. In the model of Hubbell (2001) cesses. The fourth section will distinguish with point speciation in the metacommu- these perspectives and show that they per nity, the regional balance of speciation and tain to different conceptions of hypothesis drift leads to an equilibrium mean number testing. From this analysis, we will show that of species (Ewens 1972) and to a log-series

Figure 1. Conceptual Diagram of the Present Paper, with Emphasis on the Central Ecological Equivalence Assumption The left part shows the basic motivations of the neutral theory, which translates in the middle part into a framework of ecological equivalence and biodiversity dynamics at local and regional scales. The right part introduces the main epistemic aspects that are discussed in the paper. species abundance distribution. In addition, dance distributions, it may still be mislead- the relative abundances in a local commu- ing to refer to them as neutral processes, in nity depend on the balance of local drift and the sense that a process is classically defined immigration from this metacommunity. Al- as a determinate (i.e., specific) cause of some ternative models have considered different outcomes, which is often characterized in speciation and migration processes, and thus terms of the typicality of some interactions predicted varying biodiversity patterns (e.g., (Ellis 1999; Dowe 2009). Chave and Leigh 2002; Rosindell et al. 2010). Given that a central motivation of neutral Apart from these variations, any neutral models has been to suggest new explanations model assumes that there is no influence of of biodiversity patterns, let us now consider biological differences between species on in- the explanatory difference between compet- dividual dynamics. ing neutral and niche theories. Within the Conversely, a niche-based theory of spe- philosophical literature on explanation, ex- cies coexistence (in short, “niche theory”) plaining can be seen as either providing uni- claims that species can coexist or not de- fying schemes (Friedman 1974; Kitcher 1976) pending on their niche properties and on or characterizing causal relations (Salmon specific ecological interactions such as com- 1984). These views are not necessarily exclu- petition or mutualism. Therefore, whereas sive (e.g., Strevens 2004 for a conciliation), niche theory is a theory of specified inter- but in ecology a primary objective is to find actions and determinate processes, neutral “mechanistic explanations” to observed pat- theory is a theory of unspecified interactions terns, which is a causalist conception of ex- and stochastic dynamics. In the context of planation (McGill and Nekola 2010). This population genetics, natural selection is anal‑ conception is indeed central to niche-based ogous to niche-based processes as a theory coexistence theory, where mechanisms of of specified interactions between individu- resource use and interactions at the individ- als having distinct genotypes. Even though ual level, depending on niche differences, neutral dynamics yield specific species abun- determine whether species can coexist or

Table 1 Glossary of key concepts Term Deﬁnition Neutral biodiversity pattern A biodiversity pattern classically summarizes variation in species number and relative abundances, within and between communities. A neutral biodiversity pattern represents an equilibrium state resulting from neutral dynamics. Neutral dynamics A combination of immigration, speciation, and ecological drift driving community and metacommunity composition over time, based on an assumption of ecological equivalence. Neutral model Mechanistic model of individual life, death, reproduction, and dispersal events, under the assumption of per capita ﬁtness equivalence. These basic stochastic events collectively yield neutral dynamics.

Strict ecological equivalence All of the individuals have identical prospects of living, dying, reproducing, and dispersing, which implies that they have equal fitnesses. They can display distinct ecological properties, but equalizing mechanisms then yield fitness equality (Figure 2a). Emerging ecological All of the populations of coexisting taxa show positive growth rates from low density, equivalence meaning that there is no better competitor. Strict equivalence implies emerging equivalence, but not the reverse. Nonequal individual fitness can be counterbalanced by stabilizing mechanisms yielding emerging equivalence (Figures 2b and 2c). not (MacArthur and Levins 1967). The Conversely, neutral dynamics are not word “cause” here purports to a view of a causal in the same sense that niche-based “difference maker” (Lewis 1973; Woodward processes are. Even though the relative spe- 2003; Hall 2004; Menzies 2004; Waters cies abundance distributions can be deter- 2007), namely, a cause is a difference (e.g., mined by a given set of neutral parameters, changing the value of an input variable) that such as the fundamental biodiversity num- makes a difference (e.g., changing the value ber of Hubbell (2001), the species composi- of an output variable). In the case of niche- tion is variable and changes from a replicate based species coexistence, the difference in neutral community to another. If species A ecological characteristics between two spe- realizes the peak of the abundance distri- cies makes a difference in resulting species bution, by definition no biological prop- relative abundances. A species is expected erty of A is a reason for its dominance in a to become more abundant than other spe- neutral community; which means that, if we cies in a local community if it displays some design a replicate community and consider attributes conferring better abilities for sur‑ the resulting biodiversity pattern under the vival and reproduction in the local envi assumption of ecological equivalence, we ronment (Shipley et al. 2006). The niche will get a similar relative abundance distri- theory then provides an explanation based bution without A necessarily being a domi- on determinate causal processes at the indi- nant species. The explanatory target of the vidual level (e.g., competition, predation, or neutral theory concerns the relative species mutualism). In population genetics, to be abundance and resulting diversity patterns selected likewise means to be there because at local and regional scales. It explains of a difference in individual fitness due to these patterns in the sense that it provides some genotypic attributes (Sober 1993; a way to generate them while minimizing Abrams 2007; Brandon and Ramsey 2007; the appeal to the influence of numerous Dietrich and Millstein 2008). Therefore, in biological properties. niche theory, the difference—between spe- As a consequence, neutral and niche cies, alleles, or genotypes—makes a differ- theories cannot be considered two rival hy- ence upon the identity of species, alleles, or potheses that would stand on the same level genotypes present at equilibrium, so that because their explanatory targets are par- it provides causal relations in the sense of tially different. A major reason for the suc- difference-making. cess of neutral theory in community ecology

This content downloaded from 131.156.224.067 on August 11, 2016 01:02:07 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). 326 THE QUARTERLY REVIEW OF BIOLOGY Volume 91 is that it predicts realistic patterns of biodi- esting biology from happening or complex versity, expressed via distributions of species ecological interactions from taking place abundances that were debated in macro- among individuals” (Hubbell 2001:6). The ecology for a long time, such as the log- fuzzy limits clearly reflect the fact that the series distribution (Fisher et al. 1943; Preston assumption represents some effective eco- 1948; Bell 2001). This quest for general logical equivalence. In the following, we laws has long been centered on species di- will address how ecological equivalence so versity patterns (essentially SAD), which do broadly defined can be reached in differ- not integrate the nature of species biologi- ent ways, thus entailing different concep- cal differences. A basic requisite of neutral tions of neutrality. Quite independently of theory is that these patterns should be in- the neutralist perspective, Chesson (2000) dependent of species properties. However, proposed a general conceptual framework if a niche model predicts a species diversity of diversity maintenance in ecological com- pattern, it may also explain why some spe- munities, linking individual fitness and cies are more represented than others based niche properties to community-level pat- on some biological properties (Grime 1998; terns of coexistence. Here we will use this Shipley et al. 2006). Therefore, other pat- framework to explore and discuss the mean- terns of diversity based on species properties ing of ecological equivalence. It is based should help detect the alternative effect of on the long-term per capita population niche-based processes. Trait-based perspec- growth rate of a putative invader i, de- − tives to community ecology (McGill et al. noted as r i, which initially occurs at low 2006b) and ecophylogenetics (Mouquet et al. density in a local community. Invasibility 2012) have conveyed novel tests of niche- here represents the ability of a species to based processes by analyzing the diversity of increase from low abundance in the pres- species traits instead of relying solely on spe- ence of the resident species of the com- cies taxonomy and abundances. Although munity. If there is resource limitation in the neutral theory does not say where a a system with a resident s and an invader specific species stands in the abundance dis- i, the per capita population growth rate of tribution, this question could be answered μ μ by niche theory, since niche properties then i is −r = b i − s , where µ represents the i i b b explain the success of a specific species in a ( i s ) specific environment (Chase 2005). We will mean per capita growth rates of species i and keep this clarification in mind while turning s in the absence of resource limitation, and to a second step for a specification of what b is the rate at which these per capita growth the ecological equivalence assumption can rates decrease as resources decrease (Ches- integrate. son and Huntly 1997; Chesson 2000). k. = µ./b. (. = i or s) is then a measure of the relative average fitness of species at population The Origin(s) of Ecological level, such that a species with the largest k Equivalence will be the winning competitor. This basic The neutral theory explains and predicts model cannot lead to stable coexistence, as − biodiversity patterns at community and/or r i will be positive for one species only. regional scales, based on the assumption of Conversely, in a context of niche parti- ecological equivalence, also termed neutral- tioning between coexisting species, the per ity assumption. Basic per capita ecological capita population growth rate of i can be − ρ equivalence is defined in a very broad sense written as r i = bi(ki − ks) + bi(1 − )ks, where by Hubbell: “I use neutral to describe the ρ is a parameter of niche overlap of the two assumption of per capita ecological equiva- species, that is, the proportion of resources lence of all individuals of all species in a tro- they both use. With ρ < 1 (niches do not phically defined community. This is a very completely overlap), the growth rates can unrestrictive and permissive definition of be positive for both species i and s, and neutrality because it does not preclude inter- competitive exclusion by a top competitor

This content downloaded from 131.156.224.067 on August 11, 2016 01:02:07 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). September 2016 Neutral theory and ecological equivalence 327 can be avoided. The model can be gener- ance, the right term of Equation (1) is equal alized to multispecies assemblage, such as to 0 (no stabilization), and the left term must be kept close to 0 to allow long-term − b (1 − ρ)D species coexistence. The model then pro- −r = b (k − k ) + i , (1) i i i n − 1 vides an instance of neutral dynamics. As Ad- ler et al. noted, “[n]eutral models are simply − where n is the number of species, k is the the special case where species have equiva- average fitness of resident species, and D lent fitness and there are no stabilizing, is a positive constant. Coexistence will be niche-based processes” (Adler et al. 2007: possible for a set of species i when their −r 96). The theory is thus robust to trait varia- i values are positive, and long-term coexis- tion among species regarding some aspects tence will occur if the −r values are close to of their biology, as far as they result in simi- i 0. This is a form of ecological equivalence lar levels of fitness (Figure 2a). expressed at the level of population dynamics, which differs from Hubbell’s (2001) primary assumption of fitness equivalence stabilizing mechanisms of individuals. This population-level equiv- Conversely, several studies showed that alence can be achieved either by reducing patterns of community diversity quite simi- the magnitude of per capita fitness differ- lar to those predicted by the neutral theory ences between species in the left term of can be found even when fitness is uneven Equation (1) (equalization; Figure 2a) or across species (Chave et al. 2002; Allouche if processes overcome these differences in and Kadmon 2009; Doncaster 2009). In the second stabilizing term of Equation (1) these cases, the fitness difference in the left (stabilization; Figures 2b and 2c; Adler et al. term of Equation (1) is no longer equal to 2007). We will consider this formalism to 0 and the model does not conform to the explore the limits of the basic fitness equiv- equivalence assumption as defined by Hub- alence assumption in the neutral theory bell (2001). However, the stabilizing term (extent of equalizing mechanisms), and of Equation (1) can still be large enough the robustness of its predictions when fit- to overcome the effect of fitness differ- ness equivalence is violated (extent of sta- ences. Specifically, the stabilizing term can bilizing mechanisms). be increased by niche differences between species, which allows species coexistence in communities at equilibrium (MacArthur equalizing mechanisms and Levins 1967). In particular, niche dif- Ecological equivalence has been cap- ferences between species make competi- tured by the concept of fitness invariance in tion stronger within species than between the framework of Hubbell: “By fitness invari- species, so that the growth rate ri can de- ance I mean that there are different trade-off crease when population density of species combinations of life-history traits that con- i increases (negative density dependence). fer equivalent per capita relative fitnesses Negative density dependence can, therefore, on the species exhibiting them” (Hubbell prevent competitive exclusion by maintain 2001:322). Even when ecological variation ing the coexisting species at low density exists among individuals and species, the (Levine and HilleRisLambers 2009). An- assumption holds whenever the lifetime other factor contributing to the stabilizing survival and fecundity probabilities are in- term is differences in colonization abilities variant in mathematical expectation among (Mouquet and Loreau 2003; Figure 2c). individuals. Averaging out these differences In this case, a less competitive species can amounts to equalizing fitness. Therefore, survive in a community because more com- equivalence can stem from aggregated in- petitive species display lower colonizing trinsically nonneutral differences between ability, so that they cannot establish every- organisms (Hubbell 2006; Doncaster 2009). where and occupy all of the communities Under Hubbell’s definition of fitness invari- at the same time (Hubbell 1979; Tilman

Figure 2. Patterns and Spatial Scales of Ecological Equivalence The ellipses represent local communities including a set of individuals (circles). The motif and the shading of the circles represent fitness variations between individuals. The three figures exemplify situations where ecological equivalence is expected at local and/or regional scales. The curves represent some putative species abundance distributions at local and regional scales, in which species abundances depend on their niche properties (dashed line) or not (solid line). The figures then basically stress when and how species abundances are not influenced by niche differences at local and regional scales. In the strong interpretation of neutral theory (a), effective fitness equivalence of individuals within communities leads to niche-independent dynamics. Variation in ecological characteristics among species is possible, insofar as they level out and yield

1994). Coexistence of many species is thus pretation, while it is a consequence of neu- possible within communities despite a com tral and possibly nonneutral dynamics in petitive hierarchy (Mouquet and Loreau the weak interpretation of the theory. Here 2003). Here stabilization occurs in the stabilization can counterbalance the in- overall set of communities connected by fluence of fitness differences and generate migration (the metacommunity), and can patterns of relative abundances close to that maintain species equivalence in the local of neutral models (Chave et al. 2002; Al- community despite the asymmetry in spe- louche and Kadmon 2009; Doncaster 2009). cies fitness. In this case, the pattern is considered to be neutral in the weak interpretation, but the dynamics are nonneutral in the sense of the emergence of ecological equivalence strong interpretation. It entails that a basic Therefore, a per capita ecological equiv- neutral model of speciation, migration, and alence can be reached via two ways in drift does not necessarily represent actual Chesson’s framework. Doncaster (2009) in- dynamics in the weak interpretation. troduced the concept of realized fitness to A central role in the weak interpretation stress that “species must achieve ecological is granted to dispersal limitation, which can equivalence at their coexistence equilib- act as both an equalizing and a stabilizing rium, which is defined by equal realised fit- factor (Holyoak and Loreau 2006:1373). It ness for all” (Doncaster 2009:2). Therefore, is equalizing because dispersal limitation various mechanisms can make the global limits the number of competitors coexist- state of a community close to ecological ing at a given time and hence the extent of equivalence through the stabilization and effective fitness differences (Hurtt and Pa- equalization terms of Equation (1). The cala 1995). It is stabilizing because locally role of equalization and stabilization is cen- dispersed propagules tend to generate spa tral to the distinction between the “weak” tial clustering of populations, which in- and “strong” interpretations of neutral the- creases intraspecific competition compared ory. The strong interpretation is a mecha- to interspecific competition. In addition, nistic view of the neutral theory, where the dispersal couples local and regional species fitness equivalence must actually be met at dynamics and requires moving from the the individual level, with or without the in- individual-level conception of equivalence fluence of equalizing mechanisms. In this to a larger-scale emerging conception, as case, equivalence is explanatory of the pro- understood in the weak interpretation. cesses that yield the biodiversity patterns. Acknowledging the pervasive effect of dis Conversely, “the weak version recognizes persal limitation in ecological communities that the . . . [neutral theory] is capable is, therefore, a key to understanding the rel of generating patterns that resemble those evance of ecological equivalence when ex- arising from survey data, without acknowl- plaining patterns of biodiversity from local edging that it correctly identifies the under- to regional scale. lying mechanism responsible for generating In this context we can specify the reason these patterns” (Bell 2001:2418, emphasis why the term process does not mean the same added). Therefore, ecological equivalence thing in neutral and niche-based models. On entails neutral dynamics in the strong inter- one hand, niche-based models address the

equivalent fitness. In a broader perspective of emerging ecological equivalence, as understood in the weak interpretation of neutral theory, there is no effect of biological differences on species relative abundances thanks to stabilizing mechanisms. In the case of local stabilizing mechanisms (b), fitness differences are counterbalanced by mechanisms such as negative frequency dependence within communities, which prevent the most competitive species to dominate. In the case of regional stabilizing mechanisms (c), fitness differences within local communities are counterbalanced by the dynamics between communities as, e.g., when there is a competition-colonization tradeoff. Stabilization allows local and/or regional ecological equivalence despite actual fitness differences between individuals (variation of motif and shading of individuals).

This content downloaded from 131.156.224.067 on August 11, 2016 01:02:07 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). 330 THE QUARTERLY REVIEW OF BIOLOGY Volume 91 determinate effect of elementary ecological What is the Scale of processes based on specific, individual-level Ecological Equivalence? biological differences. These differences con- In order to clarify at which spatial and stitute the basic ecological nonequivalence. temporal scale ecological equivalence can On the other hand, neutral theory represents occur, one needs to characterize the role stochastic biodiversity dynamics based on of species ecological attributes for their some species ecological equivalence, but it persistence at nested spatial scales. A tra- does not preclude a role of niche-based dif- ditional top-down scheme of niche-based ferences in yielding this ecological equiva- processes represents successive ecological lence. Ecological equivalence can itself be filters determining how individuals from the result of different processes—here, based a regional source may establish and sur- on equalizing and stabilizing mechanisms. vive in a local community (Lortie et al. Therefore, when one actually wants to talk of 2004; Cornwell and Ackerly 2009; Figure 3 “neutral processes,” there is an equivocation: left). However, such a top-down approach “the neutral processes” either mean a set of to niche-based processes is challenged by equalizing and stabilizing mechanisms yield- the idea of ecological equivalence, which ing ecological equivalence, or the dynamics can emerge from stabilizing and equaliz- occurring under the condition of ecological ing processes operating at a finer spatial equivalence. With respect to this equivoca- scale (Figure 3 right). In addition, Ches- tion, Doncaster wrote that “neutral patterns son (2000) highlighted that the balance of need not imply neutral processes” (Don- equalizing and stabilizing mechanisms can caster 2009:2). change according to the spatial and tem- Thus we have used the Chesson’s frame- poral scale. Therefore, Holyoak and Lo- work to address the nature of ecological reau stressed that ecological equivalence equivalence from the individual to popula- “occurs because species that have spatial or tion level. Because individual niche differ- temporal niche partitioning become equiv- ences can play a role in stabilization, the alent in their competitive abilities at some way ecological equivalence is reached in this spatial or temporal scales” (Holyoak and framework does not necessarily conform to Loreau 2006:1370). Several hypotheses ex- the patterns predicted by neutral models ist regarding the scales at which equalizing considering strict equivalence at the indi- and stabilizing mechanisms can yield eco- vidual level. But whatever the mathemati- logical equivalence: cal nature of the pattern is, any ecological Local-scale equalization (Figure 2a). Hub- equivalence emerging from equilization and bell (2001) and Leibold and McPeek (2006) stabilization prevents a determinate influ proposed that local guilds of functionally ence of niche differences on the species equivalent species are included in a matrix relative abundances. In this regard, it ba- of heterogeneous habitats inhabited by other sically emphasizes the role of unspecified guilds (see also Leibold 2008). In this case, interactions and stochastic dynamics, which niche variation can exist between individuals conforms to the nature of neutral theory of distinct communities, while the individ- (see the section, An Explanatory Theory of uals within local communities are ecologi- Unspecified Individual Interactions). The cally equivalent (Walker 2007). The ancient focus on emerging ecological equivalence concept of ecological guild and the mod- thus allows extending the scope of original ern notion of a functional group represent neutral models while retaining their epi such hierarchy of ecologically similar organ- stemic specificity. Given that the processes isms embedded in dissimilar groups (Wilson considered at the source of ecological equiv- 1999). A point of reference for this theory is alence may not occur at the same scale, this the guilds of canopy tree species in wet ev- leaves open the issue of the scale at which ergreen tropical forests: many species, often ecological equivalence can occur, a question more than 100 species in a 1 ha patch of that we now address. forest, appear to be functionally equivalent

Figure 3. Multiscale Biodiversity Dynamics and Hypothesis Testing Contrasting representation and hypothesis testing of biodiversity dynamics in a top-down framework of niche-based ecological filters from the regional scale (left, adapted from Lortie et al. 2004), and in a bottom-up framework from individual dynamics to emerging patterns of ecological equivalence (right). In both cases, the varying ecological properties of local individuals (different symbols in the community ellipse) entail niche differences. On the left, the diversity of the properties observed in the community is the result of filtering processes depending on niche differences and selection of successful individuals from upper-level pools. On the right, tradeoff in niche dimensions can result in fitness equalization, or niche differences can overcome the effect of competition to yield emergent neutral patterns of community structure (local ecological equivalence). Larger scale mechanisms of stabilization (e.g., source-sink metacommunity dynamics) can further generate neutral patterns of biodiversity at the regional scale, even if niche-based processes predominate in local communities (see Figure 2). These contrasting frameworks involve different approaches for hypothesis testing. In the top-down framework, the uppermost-level pool is given and random communities are generated by assuming the absence of filters (pattern-generating null models). In the bottom-up framework, mechanistic models of niche-based and neutral dynamics are compared to their neutral counterparts without niche difference (nullification), and their predictions are tested at local and regional scales. and display a neutral pattern of relative homogenous environmental context where species abundances at very local scale. The coexisting species display similar prospects spatial extent of habitats occupied by guilds of living, reproducing, and dispersing, while varies according to the nature and function environmental variation selects different of the organisms (e.g., coral reef communi- compositions in separate communities. Dis- ties, mycorrhizal networks in soil). In this re- persal limitation contributes to local equal- gard, ecological equivalence is restricted to a ization by limiting the probability that many

This content downloaded from 131.156.224.067 on August 11, 2016 01:02:07 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). 332 THE QUARTERLY REVIEW OF BIOLOGY Volume 91 competitors are found in the community at 2014). A critical issue here is to charac- the same time (Hurtt and Pacala 1995). In terize ecological objects at a level (e.g., by an evolutionary perspective, the interplay of delineating guilds) and a scale (local or neutral and niche-based dynamics can yield regional) that comply or not with an the emergence of distinct guilds along envi- equivalence assumption. In addition, con- ronmental gradients (Scheffer and van Nes sidering the way ecological equivalence 2006; Vergnon et al. 2009). and niche-based dynamics are intertwined Local-scale stabilization (Figure 2b). Local across scales uncovers a tension between top- stabilizing mechanisms can prevent compet- down (from regional to local) and bottom- itive dominance and maintain species-rich up (from local to regional) approaches to local communities. This is the core idea of the biodiversity dynamics (Figure 3). To this negative density dependence model, which extent, the question of how to handle eco- predicts that the competitive advantage of logical equivalence connects to Ricklefs’ a species decreases as its density increases (2008) challenge about community ecology. because it is penalized by stronger intraspe- He argued that working at the community cific competition. The underlying mecha level is not appropriate because it does not nism may be related to limited resources take into account the influence of regional (Levine and HilleRisLambers 2009), includ- dynamics. The way niche-based and neutral ing limited pollinator availability (Chesson dynamics drive regional biodiversity in turn and Warner 1981; Gigord et al. 2001) or to influence the dynamics of local communi- species-specific parasites or predators (Louda ties receiving immigrants (Mouquet and et al. 1990; Hatcher et al. 2006; Johnson et al. Loreau 2003). In our case, we highlight the 2012). As mentioned above, dispersal limi potential for confusion in any discussion of tation also contributes to negative density the status of the neutral theory that implic- dependence and local stabilization because itly focuses on the local scale. clusters of individuals of the same species Furthermore, the question of scale con- are formed close to the parents and increase cerns not only community assembly in intraspecific competition (Holyoak and Lo- geographic space, but also evolution in mul- reau 2006:1373). tidimensional niche space. Scheffer and van Large-scale stabilization (Figure 2c). Much Nes (2006) predicted that guilds of equiv- research in ecology still emphasizes that alent species can evolutionarily emerge in variation in local performance relates to distinct regions of niche space, as a result of niche differences between coexisting spe- both niche-based and neutral processes. As cies (Grime 1998; Shipley et al. 2006). In a consequence, “[w]hen considering pairs situations where such nonneutral processes of competitors, or species-poor assemblages, dominate locally, niche variation across spe- competitive divergence is expected, but cies can still average out across communities when considering entire, species-rich com‑ dispatched in heterogeneous habitats, lead- munities, convergence among subsets of the ing to a larger-scale pattern of ecological community can generate sets of nearly com‑ equivalence (Lavin et al. 2004; Pueyo 2006; petitively equivalent species” (Holt 2006:532). Muneepeerakul et al. 2008). Dispersal lim- Up to this point we have analyzed differ- itation acts as a mechanism of large-scale ences proper to the explanatory structure stabilization by avoiding the predominance of neutral theory in ecology: first, the ex- of better competitors or better colonizers planatory nature and the predictions of across communities (Mouquet and Loreau neutral models versus niche models and, 2003). This form of stabilization can create second, the processes yielding ecological neutral patterns of biodiversity at the re- equivalence either locally or regionally (Fig- gional scale, but not locally. ures 2 and 3). The role of equalizing and Therefore, addressing niche-based and stabilizing mechanisms is central here, and neutral dynamics is fundamentally contin- models of ecological equivalence can ex- gent upon the spatial and temporal scales tend beyond neutral models based on indi- of interest (Chase 2014; Garzon-Lopez et al. vidual fitness equivalence (see the section,

Emergence of Ecological Equivalence). In models perform “randomization of ecolog- what follows, we discuss the status of neutral ical data or random sampling from a known theory regarding its role as a possible null or imagined distribution. The null model hypothesis or a parsimonious baseline for is designed with respect to some ecological community and metacommunity ecology or evolutionary process of interest. Certain in such a broad perspective of ecological elements of the data are held constant, and equivalence. others are allowed varying stochastically to create new assemblage patterns. The randomization is designed to produce a pat- Is the Neutral Theory a Null or tern that would be expected in the absence a Parsimonious Hypothesis? of a particular ecological mechanism” (Go- The success of the neutral theory since telli and Graves 1996:3–4). Emphasis is put Hubbell (2001) has primarily relied on its here on designing a randomization pro- heuristic ability to account for some pat- cedure that generates a variety of virtual terns of diversity, even though these pat- communities complying with a null hypoth- terns had been interpreted as niche-driven esis, under which some processes are not for decades. Based on the fit of species abun- involved. The alternative hypothesis hence dance distributions, some authors com states that these neglected processes are ac- pared the predictive ability of neutral versus tually involved, and the null hypothesis is niche-based models, yielding conflicting rejected when the observed patterns signifi- results and much debate (Chave et al. 2002; cantly deviate from the randomized commu- McGill 2003). A critical point is whether nities, provided that the empirical data are such comparison allows concluding in favor relevant enough to allow discernment. This of one or the other theory. Therefore, much approach is closely related to the top-down subsequent research turned to define neu- conception of ecological filters (Figure 3 tral models as a basis for hypothesis testing left), where the constraints represent a re- (Bell 2001; Gotelli and McGill 2006). Now gional background context from which we that we have circumscribed the explana- assemble null communities by using a ran- tory specificity of neutral theory and of the domization scheme. There has been much central equivalence assumption in a scale- debate on how to design appropriate null dependent framework of equalizing and sta‑ models (Gotelli and Entsminger 2001, 2003; bilizing mechanisms, we aim to clarify the Gotelli and Ulrich 2012; Ulrich and Gotelli status of the neutral theory for hypothesis 2013). In order to detect the effect of niche- testing after one decade of conflicting views. based processes in the alternative hypothe- In this regard, we will investigate the mean- sis, classical randomization schemes shuffle ing of the neutralist claim that neutral mod- species biological attributes or community els explain biodiversity patterns (Bell 2001; members under the null hypothesis that the Hubbell 2001), against the idea that neu- distribution and abundances of species is tral models should rather be used as null independent of local environmental condi- hypotheses for understanding these patterns tions and species attributes. A major issue is (Holyoak et al. 2006; McGill et al. 2006a). the nature of the processes represented by such a randomization scheme: Do randomization procedures comply with a specific conflicting conceptions of model of community dynamics? And, if not, the null hypothesis should community dynamics be simulated Beyond the context of neutral theory, directly? two diverging conceptions exist in ecology Nullification of model parameters. In the sec- on what a null hypothesis is: ond sense, a “null hypothesis” represents Generation of null patterns. Gotelli and the influence of a set of processes but Graves (1996) promoted an instrumental- nullifies their parameters. To understand ist treatment of the null hypothesis based this, think of a process governed by several on pattern-generating null models. These parameters, such that the null instance of

This content downloaded from 131.156.224.067 on August 11, 2016 01:02:07 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). 334 THE QUARTERLY REVIEW OF BIOLOGY Volume 91 the process is obtained when all the param- the section, Emergence of Ecological Equiv- eters are set to 0. Whether or not setting pa- alence). In any case, emerging ecological rameters to 0 in the model influences the equivalence constitutes a null hypothesis ability to predict patterns similar to the ob- of the pattern-generation kind, resulting served one allows falsifying the null hypoth- from both neutral and nonneutral dynam- esis, based on a statistical criterion of model ics through equalization and stabilization. comparison (Burnham and Anderson 2002). In this conception, the role of the neutral Contrarily to the pattern-generation con- theory “is then restricted to providing the ap- ception, the nullification conception of null propriate null hypothesis when evaluating pat- hypothesis represents explicit community terns of abundance and diversity. Even this dynamics. relatively modest role, however, involves To what extent can the neutral theory revising the comparative approach to ecol- be employed as a null hypothesis (Holyoak ogy” (Bell 2001:2418, emphasis added). et al. 2006; McGill et al. 2006a)? On one Holt concurred: “In its ‘weak’ form, neutral hand, the equivalence assumption can be theory at the very least provides the appro- considered the nullification kind, such as priate null model for evaluating patterns in the relative fitness parameters wi of all spe- comparative data sets” (Holt 2006:531, em- cies are set to 0. If you consider Equation phasis added). Since the weak interpreta- (1), the null hypothesis would be then tion resorts to neutral pattern generation, given by nullification of fitness differences it implies that the null hypothesis here is only (equalization), which means that the not defined by parameter nullification. terms from the stabilizing function are 0: Even so, the weak interpretation still does it is the interpretation given by Adler et al. not comply with a randomization way of when they see neutrality as a “special case generating null patterns. Patterns result- where . . . there are no stabilizing, niche- ing from dispersal and local birth-death based processes” (Adler et al. 2007:96). dynamics are indeed sensibly different Thus, the nullification approach resorts to from that of shuffled compositions, as Bell the strong interpretation of the neutral the- underlined: “[s]tatistical null hypotheses ory. Under this interpretation, we see how based on randomization are not appropri- the neutral model can be a null hypothe- ate for evaluating ecological patterns that sis: instantiating the random processes that stem from species distributions, because take place when parameters ruling the in- local dispersal readily gives rise to spatial fluence of any biological difference are set patterns” (Bell 2001:2418). Therefore, sim- to 0. In this regard, the main motivation of ulation of emerging ecological equivalence “nearly neutral models” (Ohta 1992; Zhou should be preferred to randomization of and Zhang 2008; Noble et al. 2011) is to community composition. A difficulty of such provide an alternative model of fitness dif- a pattern-generating model is that it needs ferences such that individual equivalence to estimate the parameters of the dynamics is obtained by simply constraining the pa- from the data (Gotelli and McGill 2006). rameters. In addition, because the null hy- In the perspective of using neutral the- pothesis explicitly considers no influence ory as a null hypothesis, we have related of biological differences, any ecological emerging ecological equivalence, and thus equivalence emerging due to stabilizing the weak interpretation of the theory, to mechanisms in fact pertains to the alterna- the first kind of null hypothesis, i.e., pat- tive hypothesis. tern generating, and the strong interpreta- On the other hand, we have underlined tion of neutral theory, which assumes strict that the weak interpretation of neutrality is equivalence at individual level to the sec- based on emerging neutral patterns, what- ond kind of null hypothesis, i.e., parameter ever the fine-scale dynamics are. We have nullification. In both cases, a critical issue integrated in this perspective patterns of is whether a model of neutral dynamics pro ecological equivalence emerging at the pop vides a well-defined null hypothesis. Spe ulation level in Chesson’s framework (see cifically, we need to clarify what role we

This content downloaded from 131.156.224.067 on August 11, 2016 01:02:07 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). September 2016 Neutral theory and ecological equivalence 335 would expect for niche differences in hierarchy of alternative hypotheses is needed null and alternative hypotheses. Several to address the extent of the influence of sta- authors have claimed that the hypothetico- bilization and niche differences at multiple deductive approach should begin with the scales in space and time. The role of stabiliz- examination of some maximally parsimo- ing processes must then be identified in the nious model of community dynamics by alternative hypotheses as, for instance, we eliminating nonexplanatory components expect that density-dependent effects would related to an additional influence of niche generate nonrandom spatial distributions differences. Holyoak and Loreau empha- of conspecific and heterospecific individu- sized that we have the choice “of whether als (Comita et al. 2010). In the context of to appeal to parsimony and select a [neu- the weak interpretation, complying with a tral community model] . . . or to accept a pattern of neutral dynamics at a given scale more complex niche model” (Holyoak and entails that niche differences do not con- Loreau 2006:1372). But, as noted above, tribute to explain it: it is a contraposition of ecological equivalence can emerge even the hypothesis that niche differences should when niche-based processes influence indi- influence biodiversity patterns, but not a vidual dynamics because of equalizing and validation of the hypothesis that everything stabilizing mechanisms (Figure 2). In this is neutral. Therefore, the two approaches regard, emerging ecological equivalence is are not incompatible, but involve a differ- not a well-specified null hypothesis when ent specification of the null hypothesis. As the alternative hypothesis should identify a we will see below, the way the hypotheses role for specified niche dimensions. are specified determines the conclusion to In addition, we have underlined that the reach from their acceptation or rejection. predominance of neutral and niche-based processes is basically scale-dependent (see the section, What is the Scale of Ecological falsification versus parsimony Equivalence?). It is then crucial to keep in We have seen above that neutral theory mind that null hypotheses must be defined can be used as a null reference, but also in a certain context and at a certain level. that the nature of the null and alternative Hence, added to the alternatives of pattern hypotheses is scale-dependent and differs generation and nullification in hypoth- between the weak and stong interpreta- esis testing, we have to consider the level tions. Yet, apart from these distinctions, a and scale at which neutrality is envisaged. more general methodological issue raised Conflicting takes on the neutral theory by neutral models in ecology is whether we then reflect conflicting views on whether should concentrate on rejecting a null neu- a “strong” neutral model would be an tral hypothesis (falsification), or we could explanation for patterns extended from accept it if the predicted patterns are con- individual-level ecological equivalence or sistent with observed ones (parsimony): a “weak” model would provide an emer- Falsification. In the hypothetico-deductive gent null reference against which to de- approach, the fact that the null hypothesis is tect a large-scale imprint of niche-based not falsified does not mean that it is true. As processes (Araújo and Rozenfeld 2014). In Rosindell et al. stated, “[i]t does not follow this regard, aknowledging emerging eco- that species-specific qualities are absent in logical equivalence is required to address the real world; instead, it suggests that their the scaling of niche-based processes from effects do not penetrate the foggy lens of the the scale of individuals bearing distinct bio- summary statistics being studied. This might logical attributes to larger-scale patterns of explain how a neutral model can effectively biodiversity. fit data from a non-neutral world” (Rosin- The focus thus shifts from testing the dell et al. 2011:342). Biodiversity patterns neutral theory against a single niche-based may not be informative enough to allow alternative to testing ecological equivalence discriminating the contributions of neutral across scales. In the strong interpretation, a and nonneutral dynamics, as they could in-

This content downloaded from 131.156.224.067 on August 11, 2016 01:02:07 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). 336 THE QUARTERLY REVIEW OF BIOLOGY Volume 91 differently be generated by a niche-based ity locally and regionally despite differences or by a neutral model (Chave et al. 2002; in competitive and dispersal between spe- Purves and Pacala 2005; Chisholm and Pa- cies. Indeed, no organism can disperse in- cala 2010). Then using a neutral model as stantaneously and homogeneously across a a null hypothesis implies that only rejection whole region. Holt (2006:531) thus under- of the null hypothesis makes sense, that is, lined that dispersal limitation is pervasive we know in this case for sure that something in communities and may account for most else than this neutral model should be pro- neutral dynamics and patterns found in posed for explaining the pattern of interest species-rich ecosystems. Stochastic fluctua (McGill et al. 2006a). tions of birth and death rates are also ubiq‑ Parsimony. On the other hand, a central uitous and influence species population dy- concern in biology is to favor parsimonious namics because of finite-size effects (Lande explanations (Occam’s razor; Sober 1981). et al. 2003). Therefore, a reasonable stand- In this regard, if two models can explain a point is that dispersal and population sizes given pattern, one should keep the most of any species are axiomatically limited, so parsimonious one. A great deal of research that neutral dynamics should be acknowl- in ecology is devoted to investigating alter- edged everywhere, but still with a varying rel- native explanatory models according to a ative importance compared to niche-based balance between their goodness of fit and processes (Gravel et al. 2006; Leibold and their complexity (Burnham and Anderson McPeek 2006; Munoz et al. 2014). 2002), given that too complex models may Acknowledging the pervasiveness of neu pick out too much noise in the data. Then, if tral dynamics ultimately leads to an inte- both a niche-based and a neutral model can grative perspective, which forces one to explain an observed pattern equally well, overcome dualities in the interpretation the neutral model would be the best choice of the neutral model as a null hypothesis, as it provides a more parsimonious explana- and in the acceptation or refutation of the tion (Holyoak and Loreau 2006). In the tra- theory. The recognition that stochastic pro- dition of macroecology and biogeography, cesses of birth, death, and immigration are the neutral theory thus provides the most ubiquitous and can predominate in some simple explanation of large-scale diversity contexts has led some authors to plead for patterns (Bell 2001). This parsimony crite- a more comprehensive approach, merging rion has led some researchers to conclude into one general model the effects of both that neutral theory predicts robust emerg- niche-based and neutral processes (Gravel ing patterns and is a relevant basis to assess et al. 2006; Holt 2006; Holyoak and Loreau the pervasive influence of dispersal lim- 2006; Adler et al. 2007; Vellend 2010). The itation at large spatial scales (Leigh 2007). neutral theory then becomes a component If the neutral theory is a parsimonious hy- of a more general theory. In the logic of pothesis of biodiversity dynamics, it is not to model selection mentioned above, one can be falsified, on the contrary, it is a baseline then conclude that niche-based processes from which to start. contribute to some observed pattern if they improve the goodness-of-fit with a limited increase in complexity (Burnham and from parsimony to integration: Anderson 2002). Such an approach aims beyond the niche-neutral divide at disentangling the signatures of neutral A key argument for using neutral theory and niche-based processes from their com- as a parsimonious baseline is the general- bined effect (e.g., Adler et al. 2007; Don- ity of the basic mechanisms. We have men- caster 2009). tioned that dispersal limitation is both a key In this perspective, the neutral theory is no equalizing factor determining the equilib- more a null hypothesis but a proper parsimo- rium composition of neutral communities nious hypothesis for explaining biodiversity. in the strong sense, and a stabilizing factor It holds that the mechanisms responsible allowing biodiversity to stay close to neutral- for neutral biodiversity patterns are indeed

This content downloaded from 131.156.224.067 on August 11, 2016 01:02:07 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). September 2016 Neutral theory and ecological equivalence 337 of the sort Hubbell described (i.e., ecolog- search is and what are the perspectives ical drift, etc.), and a neutral model is ac- opened by this recent paradigm. cordingly a reference parsimonious model among a set of more complex models further incorporating the effect of niche differences. situation of the neutral Along this line, neutral ecology according to theory in levins’ triangle Hubbell (2005) did what neutral molecular Wennekes et al. (2012) claimed that the evolution did regarding evolutionary theory perspectives appropriate to niche and neu- (Veuille 2000): it shifted the burden of proof tral theories are different, in the sense Levins to the extent that the patterns against which (1966) famously distinguished between pos- one will check whether selection is acting sible model-building strategies based on the (e.g., when doing a McDonald-Kreitmann fact that a strategy cannot simultaneously test on genomic sequences) are no longer fulfill generality, precision, and realism as the expected patterns under selection, but distinct epistemic values. Neutral theory the patterns expected under the hypothesis would aim for generality and niche theory of neutral dynamics alone. Therefore, the for realism, which means that they provide roles of neutral and of niche-based theories complementary rather than conflicting per- are actually not symmetrical, as neutral the- spectives. In this viewpoint it is difficult to ory provides a baseline, parsimonious model make sense of the claim that neutrality could against which to test more complex theories. be a null hypothesis—something Wennekes Differences from the neutral predictions are et al. (2012) acknowledged, calling it a “base- then not a cause for rejecting a null hypoth- line” model rather than a null model. Hence esis, but rather a call for complexifying the they see the neutral theory as an explanation model by adding some parameters initially that philosophically can be seen in an instru- taken as null. mentalist perspective. In the view defended here, it is the very interpretation of ecological equivalence, as Discussion well as the perspective that the research We have identified several fundamental strategy adopts on it, which determines the epistemic divides that explain conflicting status of neutral theory. Whereas Wennekes takes on the role and use of neutral theory et al. (2012) focused on a global opposition in ecology. The strong and weak interpre- between realism and generality, we have in- tations of ecological equivalence differ in sisted on the level- and scale-dependence of the way basic mechanisms or emergent pat- neutral and niche-based explanations. Un- terns are considered, respectively. These derstanding the stabilizing mechanisms that interpretations relate to distinct concep- yield ecological equivalence in a community tions of ecological equivalence as defining pertains to realism, whereas taking ecologi- a null hypothesis, either as a nullification cal equivalence as a macroscopic emergent or a pattern-generating hypothesis, respec- property may illuminate the reasons for very tively. Furthermore, apart from the weak- general patterns in nature. These are two strong distinction, considering the neutral possible positions in Levins’ scheme, but ac- theory as a null reference appeals to the cording to the fine-grained distinctions we question of whether the aim is to falsify a have drawn between the strong and weak neutral model or rather to accept it as a interpretations, the variety of epistemic dis- parsimonious explanation when data match tinctions between neutral and niche theories the predictions. Beyond such a dichotomy, may increase. Finally, we have highlighted we have stressed that the neutral theory that neutrality can either emerge regionally is, by essence, a theory of pervasive and from stabilizing mechanisms in the meta- ubiquitous stochastic dynamics related to community, or locally in the community limited dispersal and population size. From from equalizing and stabilizing mechanisms this analysis of the neutral theory, we con- (Figure 2): hence in order to compare niche sider now what its place in ecological re- and neutral theories, the degree of generality

This content downloaded from 131.156.224.067 on August 11, 2016 01:02:07 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). 338 THE QUARTERLY REVIEW OF BIOLOGY Volume 91 of models seems to be less epistemically rele- and dispersion leading to larger-scale stabi- vant than the status of scales. lization, as in the model by Mouquet and Loreau (2003). In this regard, depending on the role of the stabilizing mechanisms, ecological equivalence across scales a neutral model based on strict individual To wrap up the view of the epistemological ecological equivalence may not provide ap- status of neutral theory and ecological equiv- propriate expectations of the species rela- alence proposed throughout the paper, let tive abundances. In contrast, the concept us recall our previous claims. First, there is a of emerging ecological equivalence goes distinction between ecological equivalence beyond a particular neutral model, and con- as a cause in the strong interpretation of cerns the fact that biological attributes do neutral theory, and as an emerging pattern not make a difference in the relative abun- in the weak interpretation (see the section, dances at local and/or regional scales. When The Origin(s) of Ecological Equivalence). authors say that there might be several ways Second, neutral and niche-based processes toward neutral patterns, that in other words are entangled over a hierarchy of spatial one can have neutral patterns with nonneu- and temporal scales, and yield emerging tral processes (e.g., Doncaster 2009:3), they equivalence at multiple scales (see the sec- talk about this second version of a neutral tion, What is the Scale of Ecological Equiv- model, i.e., more generally, they address the alence?). Third, neutral theory is a theory status of neutrality as an attractor for many of ubiquitous stochastic dynamics of birth, processes taking place at several scales in death, dispersal limitation, and regional space and time. We must move then from speciation. This implies that neutral theory a neutral theory of stochastic individual dy- considers basic components of actual biodi- namics to a broader theory of ecological versity dynamics and, as such, it is a compo- equivalence across scales. nent of a more integrative theory (see the As such, a core aspect of a theory of eco- section, Is the Neutral Theory a Null or Par- logical equivalence is that biological differ- simonious Hypothesis?). From these claims, ences between species are not explanatory a major challenge for modern ecology is to at all: hence we do not have here a causal go beyond the niche and neutral divide, and explanation in the sense of difference to consider the nature and consequences of making, as emphasized previously (in the ecological equivalence. We should also go section, An Explanatory Theory of Un- beyond the divide of the weak and strong in- specified Individual Interactions), and no terpretations, which should be bypassed by genuine causal process. Then in a perspec- a focus upon integrating ecological equiva- tive of emerging ecological equivalence, lence in parsimonious models. differences between species do not make The significance of stabilizing mecha- a difference, but for a reason other than nisms leading to emerging ecological equiv- strict fitness equivalence: here, the differ- alence depends on the way the influence of ences play a role in stabilizing mechanisms individual biological attributes on one hand, so that, at an emergent level, they no longer and of stochastic dynamics on the other make a difference. This aspect may be frus- hand, propagates across scales to shape mac- trating for ecologists who long aimed at roscopic patterns. As both an equalizing and finding causal pathways of niche-based dy- stabilizing mechanism, dispersal limitation namics propagating over scales. However, is a central aspect of this multiscale perspec- addressing ecological equivalence is crucial tive on ecological equivalence. Because most to understand whether and how biological organisms are dispersal-limited, emerging differences matter to explain emergent equivalence is expected to be pervasive in patterns, because these emergent patterns ecological systems, not only when fitness in turn influence the availability of immi- equivalence occurs in local communities, grants in regional pools of species (Mou- but also via a combination of competition quet and Loreau 2003; Lessard et al. 2012).

In such an integrative perspective, neu- which gave rise to a more explanatory, pow- tral theory does not need to be proved or erful, and encompassing “nearly neutral” disproved. It is a parsimonious baseline theory with relaxed assumptions on fitness model from which to build any refined equality (Ohta 1992). The controversies and model of biological interactions in order to developments that we have reviewed in this understand the nature and extent of equiv- paper seem to attest that such a move has alence and nonequivalence in ecological also occurred in community ecology. systems. Third, the necessity of an integrative and multiscale framework of ecological equivalence exposes the fact that niche-based pro- Conclusions cesses cannot provide an exclusive causal First, characterizing the neutral theory as explanation of biodiversity dynamics in space a null hypothesis (or not) seems too broad and time. Community ecology has under- a characterization to capture what is epis- gone a profound paradigm shift with neu- temologically at stake when using neutral tral theory, not only by providing a robust models. Mechanisms of species coexistence theory based on ecological equivalence, but are multiple and nested in spatial scales and also by connecting local ecological dynam- in niche dimensions, each of them defined ics to regional biogeographical and evolu- by some salient processes, and these scales tionary dynamics. Beyond the neutral-niche are at the same time decoupled and inter- divide, a crucial role of a theory of ecologi- acting (Figures 2 and 3). This was captured cal equivalence will be to solve the tension through the idea that the contribution of between the bottom-up and top-down per- stabilizing and equalizing mechanisms to spectives on this relationship—as they were neutral patterns of biodiversity depends on described in the section, What is the Scale spatial scale (see the section, What is the of Ecological Equivalence? and illustrated in Scale of Ecological Equivalence?). Here Figure 3—and to show how emergent pat- “neutrality” includes many more models terns of biodiversity can in turn influence than the neutral theory sensu Hubbell large-scale biogeographical and evolution- (2001). In this sense, ecological equivalence ary dynamics. This would make ecological is not a null hypothesis on the absence of equivalence the cornerstone of a compre- niche-based processes, since it may explic- hensive theory of the emergence and reg- itly refer to an entangled set of generating ulation of biodiversity dynamics at multiple processes, including niche-based ones. scales. In this regard, it will no longer be a Second, therefore, the neutral theory phenomenological theory of emergent pat- encompasses, as explanatory, the processes terns, but a theory of the explanatory signifi- that are involved in establishing a resulting cance of ecological equivalence. or emerging ecological equivalence (see Finally, future research should help un- the section, An Explanatory Theory of Un- derstand and predict the robustness and specified Individual Interactions). To this resilience of biodiversity dynamics in the extent it is no longer the case that biologi- face of ongoing environmental changes. cal differences (between species) make no In this regard, the urgent need for predic- difference, amounting to a noncausal ex- tive ecology requires taking into account planation—but these differences result in the mechanisms coupling local and larger- a pattern of no difference that can itself be scale biodiversity dynamics (Mouquet et al. explanatory of biodiversity. Therefore, we 2015). By considering the influence of have to acknowledge that neutrality is not so equalizing and stabilizing mechanisms in neutral since it encompasses more than in- space and time, multiscale modeling of dividual ecological equivalence. This exten- neutral and niche-based dynamics should sion and, at the same time, the weakening of help forecast the rate and extent of biodi- the original neutral theory parallels the fate versity changes and the possible cascades of of the neutral theory in molecular evolution, environmental alterations.

acknowledgments Cyrille Violle, the editors, and the two reviewers for We warmly thank Pierre Couteron, Michael Dietrich, their insightful comments and suggestions. This work Sébastien Dutreuil, John Huss, Michael Strevens, is part of the project Explabio ANR-13-BSH3-0007.

REFERENCES

Abrams M. 2007. How do natural selection and ran- Academy of Sciences of the United States of America dom drift interact? Philosophy of Science 74:666–679. 107:15821–15825. Adler P. B., HilleRisLambers J., Levine J. M. 2007. A Comita L. S., Muller-Landau H. C., Aguilar S., Hub- niche for neutrality. Ecology Letters 10:95–104. bell S. P. 2010. Asymmetric density dependence Allouche O., Kadmon R. 2009. A general framework shapes species abundances in a tropical tree com- for neutral models of community dynamics. Ecol- munity. Science 329:330–332. ogy Letters 12:1287–1297. Cornwell W. K., Ackerly D. D. 2009. Community as- Araújo M. B., Rozenfeld A. 2014. The geographic scal- sembly and shifts in plant trait distributions across ing of biotic interactions. Ecography 37:406–415. an environmental gradient in coastal California. Bell G. 2001. Neutral macroecology. Science 293:2413– Ecological Monographs 79:109–126. 2418. Dietrich M. R., Millstein R. L. 2008. The role of causal Brandon R. N., Ramsey G. 2007. What’s wrong with processes in the neutral and nearly neutral theo- the emergentist statistical interpretation of nat- ries. Philosophy of Science 75:548–559. ural selection and random drift? Pages 66–84 in Doncaster C. P. 2009. Ecological equivalence: a realis- The Cambridge Companion to the Philosophy of Biol- tic assumption for niche theory as a testable alter- ogy, edited by D. L. Hull and M. Ruse. Cambridge native to neutral theory. PLOS ONE 4:e7460. (United Kingdom): Cambridge University Press. Dowe P. 2009. Causal process theories. Pages 213–233 Burnham K. P., Anderson D. R. 2002. Model Selection and in The Oxford Handbook of Causation, edited by H. Multimodel Inference: A Practical Information-Theoretic Beebee, C. Hitchcock, and P. Menzies. Oxford Approach. Second Edition. New York: Springer. (United Kingdom): Oxford University Press. Chase J. M. 2005. Towards a really unified theory for Ellis B. 1999. Causal powers and laws of nature. Pages metacommunities. Functional Ecology 19:182–186. 19–34 in Causation and Laws of Nature, edited by Chase J. M. 2014. Spatial scale resolves the niche ver- H. Sankey. Dordrecht (The Netherlands): Kluwer sus neutral theory debate. Journal of Vegetation Sci- Academic Publishers. ence 25:319–322. Ewens W. J. 1972. The sampling theory of selec- Chave J. 2004. Neutral theory and community ecol- tively neutral alleles. Theoretical Population Biology ogy. Ecology Letters 7:241–253. 3:87–112. Chave J., Leigh E. G. Jr. 2002. A spatially explicit neu- Fisher R. A., Corbet A. S., Williams C. B. 1943. The tral model of β-diversity in tropical forests. Theoret- relation between the number of species and the ical Population Biology 62:153–168. number of individuals in a random sample from Chave J., Muller-Landau H. C., Levin S. A. 2002. Com- an animal population. Journal of Animal Ecology paring classical community models: theoretical 12:42–58. consequences for patterns of diversity. American Friedman M. 1974. Explanation and scientific under- Naturalist 159:1–23. standing. Journal of Philosophy 71:5–19. Chave J., Alonso D., Etienne R. S. 2006. Comparing Garzon-Lopez C. X., Jansen P. A., Bohlman S. A., Or- models of species abundance. Nature 441:E1. donez A., Olff H. 2014. Effects of sampling scale Chesson P. 2000. Mechanisms of maintenance of spe- on patterns of habitat association in tropical trees. cies diversity. Annual Review of Ecology and Systemat- Journal of Vegetation Science 25:349–362. ics 31:343–366. Gewin V. 2006. Beyond neutrality—ecology finds its Chesson P., Huntly N. 1997. The roles of harsh and niche. PLOS Biology 4:e278. fluctuating conditions in the dynamics of ecologi- Gigord L. D. B., Macnair M. R., Smithson A. 2001. cal communities. American Naturalist 150:519–553. Negative frequency-dependent selection main- Chesson P. L., Warner R. R. 1981. Environmental vari- tains a dramatic flower color polymorphism in the ability promotes coexistence in lottery competitive rewardless orchid Dactylorhiza sambucina (L.) Soò. systems. American Naturalist 117:923–943. Proceedings of the National Academy of Sciences of the Chisholm R. A., Pacala S. W. 2010. Niche and neutral United States of America 98:6253–6255. models predict asymptotically equivalent species Gotelli N. J., Entsminger G. L. 2001. Swap and fill abundance distributions in high-diversity eco- algorithms in null model analysis: rethinking the logical communities. Proceedings of the National knight’s tour. Oecologia 129:281–291.

Gotelli N. J., Entsminger G. L. 2003. Swap algorithms Lande R., Engen S., Sæther B.-E. 2003. Stochastic Popu- in null model analysis. Ecology 84:532–535. lation Dynamics in Ecology and Conservation. Oxford Gotelli N. J., Graves G. R. 1996. Null Models in Ecol- (United Kingdom): Oxford University Press. ogy. Washington (DC): Smithsonian Institution Latimer A. M., Silander J. A. Jr., Cowling R. M. 2005. Press. Neutral ecological theory reveals isolation and Gotelli N. J., McGill B. J. 2006. Null versus neutral mod- rapid speciation in a biodiversity hot spot. Science els: what’s the difference? Ecography 29:793–800. 309:1722–1725. Gotelli N. J., Ulrich W. 2012. Statistical challenges in Lavin M., Schrire B. P., Lewis G., Pennington R. T., null model analysis. Oikos 121:171–180. Delgado-Salinas A., Thulin M., Hughes C. E., Beyra Gravel D., Canham C. D., Beaudet M., Messier C. Matos A., Wojciechowski M. F. 2004. Metacom- 2006. Reconciling niche and neutrality: the con- munity process rather than continental tectonic tinuum hypothesis. Ecology Letters 9:399–409. history better explains geographically structured Grime J. P. 1998. Benefits of plant diversity to ecosys- phylogenies in legumes. Philosophical Transactions of tems: immediate, filter and founder effects. Jour- the Royal Society B: Biological Sciences 359:1509–1522. nal of Ecology 86:902–910. Leibold M. A. 2008. Return of the niche. Nature 454: Hall N. 2004. Two concepts of causation. Pages 225– 39–40. 276 in Causation and Counterfactuals, edited by Leibold M. A., McPeek M. A. 2006. Coexistence of the J. Collins, N. Hall, and L. A. Paul. Cambridge niche and neutral perspectives in community ecol- (Massachusetts): MIT Press. ogy. Ecology 87:1399–1410. Hatcher M. J., Dick J. T. A., Dunn A. M. 2006. How Leigh E. G. Jr. 2007. Neutral theory: a historical per- parasites affect interactions between competitors spective. Journal of Evolutionary Biology 20:2075– and predators. Ecology Letters 9:1253–1271. 2091. Holt R. D. 2006. Emergent neutrality. Trends in Ecology Lessard J.-P., Belmaker J., Myers J. A., Chase J. M., Rah- and Evolution 21:531–533. bek C. 2012. Inferring local ecological processes Holyoak M., Loreau M. 2006. Reconciling empirical amid species pool influences. Trends in Ecology and ecology with neutral community models. Ecology Evolution 27:600–607. 87:1370–1377. Levine J. M., HilleRisLambers J. 2009. The impor- Holyoak M., Loreau M., Strong D. 2006. Neutral com- tance of niches for the maintenance of species di- munity ecology. Ecology 87:1368–1369. versity. Nature 461:254–257. Hubbell S. P. 1979. Tree dispersion, abundance, and Levins R. 1966. The strategy of model building in pop- diversity in a tropical dry forest. Science 203:1299– ulation biology. American Scientist 54:421–431. 1309. Lewis D. 1973. Causation. Journal of Philosophy 70: Hubbell S. P. 1997. A unified theory of biogeography 556–567. and relative species abundance and its application Lortie C. J., Brooker R. W., Choler P., Kikvidze Z., to tropical rain forests and coral reefs. Coral Reefs Michalet R., Pugnaire F. I., Callaway R. M. 2004. 16:S9–S21. Rethinking plant community theory. Oikos 107: Hubbell S. P. 2001. The Unified Neutral Theory of Bio- 433–438. diversity and Biogeography. Princeton (New Jersey): Louda S. M., Keeler K. H., Holt R. D. 1990. Herbivore Princeton University Press. influences on plant performance and competitive Hubbell S. P. 2005. Neutral theory in community ecol- interactions. Pages 413–444 in Perspectives on Plant ogy and the hypothesis of functional equivalence. Competition, edited by J. B. Grace and D. Tilman. Functional Ecology 19:166–172. San Diego (California): Academic Press. Hubbell S. P. 2006. Neutral theory and the evolution MacArthur R., Levins R. 1967. The limiting similarity, of ecological equivalence. Ecology 87:1387–1398. convergence, and divergence of coexisting spe- Hurtt G. C., Pacala S. W. 1995. The consequences of cies. American Naturalist 101:377–385. recruitment limitation: reconciling chance, his- McGill B. 2003. Strong and weak tests of macroecolog- tory and competitive differences between plants. ical theory. Oikos 102:679–685. Journal of Theoretical Biology 176:1–12. McGill B. J., Nekola J. C. 2010. Mechanisms in mac- Johnson D. J., Beaulieu W. T., Bever J. D., Clay K. 2012. roecology: AWOL or purloined letter? Towards a Conspecific negative density dependence and for- pragmatic view of mechanism. Oikos 119:591–603. est diversity. Science 336:904–907. McGill B. J., Maurer B. A., Weiser M. D. 2006a. Em- Kimura M. 1983. The Neutral Theory of Molecular Evo- pirical evaluation of neutral theory. Ecology 87: lution. Cambridge (United Kingdom): Cambridge 1411–1423. University Press. McGill B. J., Enquist B. J., Weiher E., Westoby M. 2006b. Kitcher P. 1976. Explanation, conjunction, and unifi- Rebuilding community ecology from functional cation. Journal of Philosophy 73:207–212. traits. Trends in Ecology and Evolution 21:178–185.

Menzies P. 2004. Causal models, token causation, and Scheffer M., van Nes E. H. 2006. Self-organized sim- processes. Philosophy of Science 71:820–832. ilarity, the evolutionary emergence of groups of Mikkelson G. M. 2005. Niche-based vs. neutral models similar species. Proceedings of the National Academy of ecological communities. Biology and Philosophy of Sciences of the United States of America 103:6230– 20:557–566. 6235. Mouquet N., Loreau M. 2003. Community patterns in Shipley B., Vile D., Garnier É. 2006. From plant traits source-sink metacommunities. American Naturalist to plant communities: a statistical mechanistic ap- 162:544–557. proach to biodiversity. Science 314:812–814. Mouquet N., Devictor V., Meynard C. N., et al. 2012. Sober E. 1981. The principle of parsimony. British Jour- Ecophylogenetics: advances and perspectives. Bio- nal for the Philosophy of Science 32:145–156. logical Reviews 87:769–785. Sober E. 1993. The Nature of Selection: Evolutionary The- Mouquet N., Lagadeuc Y., Devictor V., et al. 2015. ory in Philosophical Focus. Cambridge (Massachu- Predictive ecology in a changing world. Journal of setts): MIT Press. Applied Ecology 52:1293–1310. Strevens M. 2004. The causal and unification ap- Muneepeerakul R., Bertuzzo E., Lynch H. J., Fagan proaches to explanation unified—causally. Noûs 38: W. F., Rinaldo A., Rodriguez-Iturbe I. 2008. Neu- 154–176. tral metacommunity models predict fish diversity Tilman D. 1994. Competition and biodiversity in spa- patterns in Mississippi-Missouri basin. Nature 453: tially structured habitats. Ecology 75:2–16. 220–222. Ulrich W., Gotelli N. J. 2013. Pattern detection in null Munoz F., Ramesh B. R., Couteron P. 2014. How do model analysis. Oikos 122:2–18. habitat filtering and niche conservatism affect Vellend M. 2010. Conceptual synthesis in community community composition at different taxonomic ecology. Quarterly Review of Biology 85:183–206. resolutions? Ecology 95:2179–2191. Vergnon R., Dulvy N. K., Freckleton R. P. 2009. Niches Noble A. E., Temme N. M., Fagan W. F., Keitt T. H. versus neutrality: uncovering the drivers of diver- 2011. A sampling theory for asymmetric communi- sity in a species-rich community. Ecology Letters ties. Journal of Theoretical Biology 273:1–14. 12:1079–1090. Ohta T. 1992. The nearly neutral theory of molecular Veuille M. 2000. Genetics and the evolutionary pro- evolution. Annual Review of Ecology and Systematics cess. Comptes Rendus de l’Académie des Sciences—Series 23:263–286. III—Sciences de la Vie 323:1155–1165. Preston F. W. 1948. The commonness, and rarity, of Volkov I., Banavar J. R., Hubbell S. P., Maritan A. 2003. species. Ecology 29:254–283. Neutral theory and relative species abundance in Pueyo S. 2006. Diversity: between neutrality and struc- ecology. Nature 424:1035–1037. ture. Oikos 112:392–405. Walker S. C. 2007. When and why do non-neutral Purves D. W., Pacala S. W. 2005. Ecological drift in metacommunities appear neutral? Theoretical Pop- niche-structured communities: neutral pattern does ulation Biology 71:318–331. not imply neutral process. Pages 107–140 in Biotic Waters C. K. 2007. Causes that make a difference. Jour- Interactions in the Tropics: Their Role in the Mainte- nal of Philosophy 104:551–579. nance of Species Diversity, edited by D. F. R. P. Burs- Wennekes P. L., Rosindell J., Etienne R. S. 2012. The lem, M. A. Pinard, and S. E. Hartley. Cambridge neutral-niche debate: a philosophical perspective. (United Kingdom): Cambridge University Press. Acta Biotheoretica 60:257–271. Purves D. W., Turnbull L. A. 2010. Different but equal: Whitfield J. 2002. Neutrality versus the niche. Nature the implausible assumption at the heart of neutral 417:480–481. theory. Journal of Animal Ecology 79:1215–1225. Wilson J. B. 1999. Guilds, functional types and ecolog- Ricklefs R. E. 2008. Disintegration of the ecological ical groups. Oikos 86:507–522. community. American Naturalist 172:741–750. Woodward J. 2003. Experimentation, causal infer- Rosindell J., Cornell S. J., Hubbell S. P., Etienne ence, and instrumental realism. Pages 87–118 in R. S. 2010. Protracted speciation revitalizes the neu The Philosophy of Scientific Experimentation, edited by tral theory of biodiversity. Ecology Letters 13:716– H. Radder. Pittsburgh (Pennsylvania): University 727. of Pittsburgh Press. Rosindell J., Hubbell S. P., Etienne R. S. 2011. The Zhou S.-R., Zhang D.-Y. 2008. A nearly neutral model unified neutral theory of biodiversity and bioge- of biodiversity. Ecology 89:248–258. ography at age ten. Trends in Ecology and Evolution 26:340–348. Salmon W. C. 1984. Scientific Explanation and the Causal Structure of the World. Princeton (New Jersey): Prince Associate Editor: Mark Dale ton University Press. Handling Editor: Daniel E. Dykhuizen

This content downloaded from 131.156.224.067 on August 11, 2016 01:02:07 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).