12083 770..784 Methodological Individualism in Ecology James Justus*y Methodological individualism has a long, successful, and controversial track record in the social sciences. Its record in ecology is much shorter but proving as successful and con- troversial with so-called individual-based models. Distinctions and debates about meth- odological individualism in social sciences clarify the commitments of this general, in- dividualistic approach to modeling ecological phenomena and show that there is a lot recommending it. In particular, a representational priority on individual organisms yields a cogent albeit deflationary account of ecological emergence and helps reveal how quite disparate models and theories in ecology might be unified. 1. Introduction. Cross-pollination between theorizing in biological and social sciences has a long and fruitful history, most of which explores con- nections between rational choice and evolutionary theory. The usual focus is how evolutionary concepts and principles illuminate models of the eco- nomic world and vice versa. Evolutionary economics exemplifies the for- mer; evolutionary game theory, the latter. Recent philosophical work targets the same confluences: evolutionary origins of the social contract (Skyrms 1996), risk aversion, and analogies between fitness and utility (Okasha 2011), among others. Evolution, however, is but part of biology; ecology is another. And an ecological perspective is as indispensable to evolu- tionary theorizing as evolution is to ecology. Fruitful cross-fertilizations of social-scientific and ecological theorizing should therefore be expected, but these surprisingly remain almost entirely unexplored. *To contact the author, please write to: Philosophy Department, Florida State University, Tallahassee, FL 32306; e-mail: jjustus@fsu.edu. yFor helpful feedback, thanks to Terrence Hill, Jay Odenbaugh, Joan Roughgarden, Carl Salk, Paul Teller, Michael Weisberg, Eric Winsberg, and audiences at “Values and Norms in Modeling” in Eindhoven (June 2012); the PSA meeting in San Diego (November 2012); joint Institut D’Histoire et de Philosophie des Sciences et des Techniques–Florida State University philosophy of biology workshop in Paris (July 2013); and the meeting of the International Society for the History, Philosophy, and Social Studies of Biology in Montpellier (July 2013). Philosophy of Science, 81 (December 2014) pp. 770–784. 0031-8248/2014/8105-0006$10.00 Copyright 2014 by the Philosophy of Science Association. All rights reserved. 770 Connections between methodological individualism (MI) and individual- based modeling are part of that untold story. In the social sciences MI consti- tutes a constraint on how social phenomena should be conceptualized and represented. It means different things to different thinkers, but privileging indi- viduals in explanations, representations, and theories of higher-level phenom- ena is a common denominator (Heath 2010). In ecology, MI finds clear ex- pression in a new modeling strategy: individual-based modeling (see Grimm and Railsback 2005). Just as expected utility calculations and actions of indi- vidual agents constitute the preferred level of analysis according to MI in social sciences, behaviors and physiologies of individual organisms function simi- larly for individual-based ecological models (IBMs). Describing and defend- ing this analogy is this paper’s first task. Evaluating how much light it sheds is the second. Debates about MI in the social sciences reveal important insights about how individual-based ecological modeling should be understood. For example, when properly construed, MI need not commit to population- or community-level properties ontologically reducing to individual-level proper- ties, or any metaphysical view about what marks the “real.” MI does require explanatory and representational priority on individuals, and the rationale un- derlying this requirement buttresses the role of IBMs in ecological theoriz- ing. Section 2 describes significant differences between classical and individual- based ecological modeling. Section 3 considers confusions about what MI in- volves and the particular way MI aligns with reductionism. Sections 4 and 5 clarify how a representational priority on individuals yields a cogent albeit deflationary account of ecological emergence and can underpin the unification of disparate models and theories in ecology. 2. Classical versus Individual-Based Ecological Modeling. Almost every introduction to ecological modeling begins with the canonical Lotka-Volterra one-predator, one-prey community model: dV dt ¼ ½prey growth rate� 2 ½capture rate of prey per predator�P; ð1aÞ dP dt ¼ ½predator births per capture�P 2 ½predator death rate�; ð1bÞ where Vand P designate prey and predator abundances, respectively. Besides the pedagogical virtues of starting simple before tackling more sophisticated representations, equation (1) exemplifies standard modeling strategy in ecol- ogy. Basic units are biological populations, and ecological processes such as predation and competition are typically represented with differential or dif- ference equations at the population level. When available, information about METHODOLOGICAL INDIVIDUALISM IN ECOLOGY 771 properties of individual organisms is statistically aggregated into population- level state variables representing birth, death, capture rates, and so on. More of- tenthannotitisunavailable,inwhichcasestatevariablessimplyabstractaway from individual-level details. The approach has several advantages. It situates ecological modeling in a mathematically precise, rigorous, and well-developed analytic framework that has proved fruitful in many sciences: dynamic systems theory. Theorems and derivational techniques within this framework constitute powerful tools for analyzing systems in other domains—classical and fluid mechanics, chem- ical kinetics, and many more—and are thought to do the same for ecology (see Justus 2008b). Second, abstracting from the individual level not only accom- modates the relative paucity of available ecological data on individual organ- isms and their interrelations; it better facilitates general theorizing about pop- ulation and community dynamics. Rather than become mired in sundry details about individuals, focusing on state variables seems to offer the best prospects for uncovering generalizations that hold across different configurations of in- dividuals into populations and communities. Without the abstraction, it seems unlikely that highly regarded fruits of ecological theorizing such as the com- petitive exclusion principle or island biogeography theory would have been found. Classical ecological models are typically highly idealized and unrealistic. For example, prey only die by predators consuming them and predators only increase by converting prey into births in equation (1). More sophisticated mod- els better avoid such conspicuously unrealistic assumptions, but ignoring in- dividuals means sacrificing maximal realism. This need not detract from the insights the models provide. Besides the tractability simplifying assumptions afford, they also help make representations of complex systems cognitively evaluable, especially in ecology (Odenbaugh 2005). Aggregating or abstract- ing from individual-level details also seems reasonable since organisms of a single species share most properties in virtue of that fact. It is increasingly be- ing appreciated, however, that classical models overlook two significant driv- ers of ecosystem dynamics: variability between organisms, even of the same age and species, and locality of interactions, both spatially and temporally. Classical models are blind to these features; IBMs are not. This significant advantage was recognized early by supporters of IBMs: “The assumption that all individuals are equally affected by increasing population density disagrees with empirical evidence . . . and leads theoretical population biology into a blind alley” (Łomnicki 1978, 473). Organisms rather than populations are the basic units of IBMs, and or- ganismal behavior, development, reproduction, and interactions are repre- sented explicitly at the level at which they occur. A set of variables represents each individual, and a set of equations represents relations between them and abiotic factors such as precipitation, temperature, and soil acidity. These vari- 772 JAMES JUSTUS ables and equations are intended to reflect the detailed physiologies and spa- tially specific interactions usually absent from classical models and, crucially, their variance across individuals. Biologically, organisms of a species are not created equal, and IBMs capture this important fact. And with spatial and tem- poral dynamics explicitly represented, the effects, for instance, of neighborhood relations and the locality of environmental gradients on overall ecosystem dy- namics—these individuals of those species are adjacent to this individual, that region is sunny and dry while this one is shaded and moist, and so on—can thereby be investigated. Admittedly, this attention to detail has a cost. IBMs contain hundreds, sometimes thousands, of variables and equations linking them and are therefore analytically intractable, only solvable by simulation. Despite this computational hurdle, some ecologists believe that integrating the individual-level information lost with state variables lends a crucial advan- tage to individual-based modeling. A recent comprehensive introduction to IBMs, for example, acclaims, “Individual-based models have demonstrated the potential significance of individual characteristics to population dynamics and ecosystem processes. Even in its infancy, individual-based modeling (IBM) has changed our understanding of ecological systems” (Grimm and Railsback 2005, xiii). Others are much less impressed. Roughgarden (2012) suggests, “IBMs as defined by Grimm and Railsback seem primarily applicable only to very large organisms such as vertebrates and trees, and even then might be worthwhile only for special applications where the individuals are each specifically identified, tagged and tracked.” To adjudicate this emerging controversy, a much clearer understanding is needed of the modeling strategy IBMs employ, its epistemic credentials, and its potential deficiencies. Fortunately, IBMs instantiate a general individu- alistic approach to representing the world labeled ‘methodological individ- ualism’ (MI) that has been applied extensively and fruitfully in other sci- ences, particularly social sciences, and over which much ink has been spilled. Extant debates about its legitimacy offer insights about how individual- based ecological modeling should be conceptualized. But as Kincaid (1997, 13) aptly observes, debates about MI have been “long on rhetoric and short on clarity.” The next section navigates through potential confusions to pin- point the defensible sense in which IBMs realize MI. 3. MethodologicalAtomism,Individualisms,andReductionism. MI is but one of numerous individualisms. Defensibly construed, it does not require ontological individualism, the idea that only individuals exist and aggre- gates, collections, and other wholes do not, or do not somehow exist “over and above” their individual constituents. Quine’s forceful arguments not- withstanding, methodological commitments within scientific modeling, or epistemic inquiry generally, do not necessitate a particular thread in the tangle of metaphysical issues concerning ontology (see Azzouni 2004). METHODOLOGICAL INDIVIDUALISM IN ECOLOGY 773 Nor is MI the same as methodological atomism, the idea that wholes are only explainable by properties of individuals, not their interrelations. This view—sometimes attributed to Hobbes’s account of how the civil state emerges from “solitary,” presocial creatures in the state of nature (Heath 2010)—would disallow even binary, pairwise interactions invoked in mi- croeconomics to explain macroeconomic patterns (Hoover 2010). Interac- tions between individuals and their environments that drive ecosystem dy- namics are similarly at the core of ecological IBMs. This inclusivity prompts a clarification and raises important issues. I will first give the clarification. Within social sciences, MI prioritizes individual properties and actions, interactions they catalyze, and ultimately the under- lying intentional states motivating both. These supply the base from which all social phenomena can be accounted for according to MI. Methodological atomism is more restrictive: the origin and content of these intentions must be individualistically pure. They might arise from individual-level selective history, developmental processes, percolating quantum indeterminacy, or some- thing entirely sui generis, but broader sociological, political, cultural, or other nonindividualistic factors are prohibited. For example, aspirations of increas- ing the general welfare of others or fear of societal decay, both of which ref- erence extra-individual social factors, cannot be responsible for individuals’ intentions and subsequent actions for methodological atomists. If they were, higher-level factors would contaminate the unadulterated lower-level indi- vidualistic basis that supposedly accounts for social phenomena. Similarly, interactions cannot be invoked to account for social phenomena, unless per- haps they follow trivially from the properties of individuals atomistically con- ceived. Invoking associative or neighborhood effects caused (even partially) by broader social dynamics is therefore disallowed. Despite assertions from some of its critics,1 plausible versions of MI are unencumbered by this utterly implausible view of human psychology. Be- sides representing individuals explicitly, MI endeavors to represent them re- alistically. Social, political, and cultural factors obviously influence individ- uals’ beliefs, intentions, and actions, and MI can correctly accommodate that fact.2 The lesson is the same in individual-based ecological modeling. Prop- erties organisms possess reflect selection histories that depend on particular 1. Sen (2009, 244), for example, endorses the characterization of MI as “all social phe- nomena must be accounted for in terms of what individuals think choose and do” and then says, “There have certainly been schools of thought based on individual thought, choice and action, detached from the society in which they exist” (emphasis added). Only the former defensibly characterizes MI, and it differs from and does not entail the latter. 2. Professed advocates of MI too frequently ignore these influences by uncritically en- dorsing unrealistic accounts of human motivation and rationality. I follow Heath (2010) in sharply distinguishing MI from such views. 774 JAMES JUSTUS past environments and fitness landscapes, both of which population dynam- ics influence. Physiologically onerous male traits of many species—the pea- cock, Irish elk, marvelous spatuletail—display the selective power of such dynamics. And apart from evolutionary considerations, ecological dynamics exhibit similar dependencies. Organisms’ properties, physiologically and be- haviorally, can differ dramatically based on status and pedigree in popula- tions, particularly as cognitive sophistication and social complexity increase. In a congress of orangutans, for example, alpha and beta males of similar age and heredity look and act very differently. Beyond the origin of individual properties and intentional content, re- alistically representing even binary interactions between individuals also re- quires rejecting the stringency of methodological atomism because these interactions are often shaped by extra-individualistic factors. For example, broad structural features of populations can partially determine which con- specifics organisms associate with and how, something that would not fol- low from their dubiously conjectured atomistic properties. But this prompts a difficult question: If representing binary interactions with these influences is permissible for MI, are higher-order n-ary relations as well? Realism seems to again dictate yes, at least for some n > 2. In humans and simians, for exam- ple, instances of complex, higher-order relationships abound. That less cogni- tively endowed organisms sometimes exhibit similar complexity is less appre- ciated. Fire ants (Solenopsis invicta) in North America typically outcompete native ant species to local extinction. But if even low densities of parasitoid flies from their native range are present, Solenopsis consumes less, grows to smaller size, and competes less effectively against native ants (Mehdiabadi and Gilbert 2002). Accurately representing these interactions seems to re- quire ternary relationships, and it is an open empirical question whether and how often higher-order relations are necessary to represent systems across different scientific domains. This presents a potential impediment: although it also seems entirely arbitrary to posit that an n exists above which MI is abandoned, as n increases, recognizing n-ary relations surely departs from an individual-level focus. Whether this recognition would necessitate rejecting MI depends on the ultimate origin of those relations (see below). Such permissiveness raises another concern: it ostensibly precludes re- ducing higher-level phenomena. ‘Methodological individualism’ and ‘reduc- tionism’ are frequently considered synonymous labels (Kincaid 1997). At the very least, reductionism is taken to require MI. But if MI condones appro- priating extra-individualistic information when modeling systems, this seems to preclude rather than facilitate reduction. An unnecessarily strong conception of MI underlies these concerns. Un- like methodological atomism, the epistemic priority MI places on individuals does not necessitate that particular representations of systems be individual- METHODOLOGICAL INDIVIDUALISM IN ECOLOGY 775 istically pure, free from any influence of extra-individualistic factors. Such a stringent criterion is tantamount to stipulating that mathematical proofs con- tain logical derivations for every proposition they utilize, however founda- tional or well established. For proofs in all but a small subset of issues in the most foundational fields, the goal is not to establish theorems from the axi- omatic ground up. Rather, specific proofs establish new theorems conditional on others, some of which may be controversial or even actively doubted. IBMs function similarly for reductionist programs. They establish how properties and interactions between individuals, which in turn may depend on input from extra-individualisticsources,canaccountforhigher-levelphenomena.Whether the phenomena can be fully accounted for ultimately depends on whether this input can itself be accounted for individualistically. In cases with such in- put, IBMs therefore facilitate a kind of conditional reduction: conditional on the relevant extra-individualistic features being reducible, IBMs show how re- ductionism’s front can advance. A genuine exception to MI and reductionism only obtains if they cannot. Whether these features are ultimately reducible is an exceedingly difficult context-dependent and in part empirical question that the current state of sci- entificknowledgeoftenonlygivesdimglimpsesof.Occasionally,itisrelatively clear. An institution’s culture may shape the beliefs, intentions, and actions of agents within it and influence how and with whom they interact. But tracing an institution’s origin back to a charter among individual founders and the sub- sequent history of individual decisions responsible for its culture presumably eliminates this higher-level influence. There is likely a reminder, of course, in the legal and governmental framework without which charters could not ex- ist, but no insurmountable obstacle appears to preclude iteratively adopting precisely the same approach, at least in principle. Practical priorities may make such monumental undertakings an unwise allocation of limited resources, but pragmatic considerations do not challenge the plausibility of these social en- tities and phenomena ultimately being accounted for individualistically. In- dividualistic accounts of how monetary institutions emerge from bartering interactions as much more efficient exchange mechanisms are a widely her- alded MI success (see Nozick 1977). Social phenomena that are not purposive artifacts of human planning, such as widely shared fairness and equality norms, pose a more formidable challenge. They likely predated legal and political institutions and thereby elude the MI strategy just described. They do not elude, however, a wide range of “bottom-up” modeling techniques based on replicator dynamics.3 Replicator dynamics constitute a simple model of evolution in which popu- lations change through differential reproduction, or differential imitation for 3. See Skyrms (1996) for an accessible introduction. 776 JAMES JUSTUS cultural evolution. The approach does not necessitate that individuals be rep- resented explicitly, but, intriguingly, when individual-level dynamics are in- cluded, modeling results often become more plausible. Alexander and Skyrms (1999), for example, analyze potential evolutionary origins of the widespread sense of distributive justice with two-player divide-the-dollar bargaining games.Eachplayerformulatesabottom-linedemandabouthowadollarshould be distributed. If demands jointly exceed $1, no one gets anything; otherwise, each gets what was demanded. Game-theoretic experiments overwhelmingly result in fair, 50-50 splits, which classical rational choice theory cannot ex- plain. Standard replicator dynamics, which assume that individuals randomly pair to bargain from a very large population, offers a better explanation. Fair division goes to fixation more often than not, and the farther other divisions are from fair, the lower their probability. But by far the best results emerge when locality is imposed on interactions. When individuals bargain only with their eight-member Moore neighborhood rather than randomly, fair division almost always goes to fixation, which closely fits empirical facts, and it fixes much more rapidly. This provides a striking glimpse of how social and moral norms might emerge from individual-level dynamics. Much more remains to be explored with these techniques, but they have already uncovered tantalizing suggestions about how lower-level dynamics, even among rudimentary crea- tures, might generate primitive social norms (Skyrms 1996), representational information (Skyrms 2010), category formation (Skryms 2010), and others. The next section explains how emergence occurs similarly in ecology. Of course, endorsing MI does not require committing to this modeling strat- egy’s universal applicability or success. But its successes to date demonstrate that the dependency of some individual-level properties and interactions on higher-level factors poses interesting questions for future research, rather than revealing insurmountable obstacles to reduction. A final clarification: MI’s ambition is often overstated. MI maintains that nonindividualistic explanations are often inferior, incomplete, and therefore too often inadequate, but they are not thereby specious or necessarily useless. Regarding explanation, for instance, MI need not commit to the misguided view that wholes are only explainable by properties of individuals and their interactions with each other and the environment. This is only attractive on an implausible view that there is a uniquely correct explanation, or kind of expla- nation, for a given explanandum. The same holds for model representation. Explanations and models serve a variety of ends, and some well-established ends—enhancingunderstanding,empiricaladequacy,discoveringunderlying mechanisms, or expanding a theory’s scope—are served in a variety of ways. “Science has room for both lumpers and splitters,” as Sober (1999, 551) deftly put it. Having emerged from a conceptual thicket of potential confusion with a clearer understanding of MI, its plausibility and merits in ecology can be as- METHODOLOGICAL INDIVIDUALISM IN ECOLOGY 777 sessed. Section 4 proceeds from the most plausible and least controversial points of support to more contentious issues of emergence. 4. Supervenience, Compositionality, and Emergence in Ecology Supervenience.—Although methodological doctrines do not entail meta- physical ones, the latter can bear on the former. If one domain fails to super- vene on another, for example, the latter presumably cannot be an adequate methodologically individualistic basis for the former. Since one domain’s facts would not determine those of another, some facts of the nonsuper- vening domain would be unascertainable via the other. MI has been criti- cized in the social sciences on this ground (Epstein 2012), but no alleged obstacle there threatens the claim that population and community facts su- pervene on properties of individual organisms, their interrelations, and re- lations with the environment. Compositionality.—Undergirding supervenience in ecology is another un- controversial claim: individual organisms constitute biological populations and communities.4 This compositionality removes a gap frequently thought to pose insuperable obstacles to reduction in other contexts: the inability to bridge fundamentally different domains. As lodged against reductive aspirations, such gaps have been maintained for connections between mind and body, psychol- ogy and neurobiology, first- and third-person perspectives, classical and mo- lecular genetics, and others. And even when these domains are taken to be systematically related, the nature of the relationship is usually conceptualized as instantiation, realization, or simple correlation, rather than mereologically. Ecology’s purview fortunately does not span fundamentally different domains, so it happily avoids an intractable “hard problem.”5 Emergence.—Supervenience and compositionality are uncontroversial in ecology. Emergence is not. Though far from consensus, some population-, 4. This uncontroversial claim is largely unrelated to contentious issues about how pop- ulation concepts should be characterized (see Millstein 2009). The latter primarily con- cern what kind and how strong relations must be between conspecifics (causally, genea- logically, or spatiotemporally) for a defensible population concept to apply, rather than whether populations are composed of conspecifics. Indeed, all definitions of population concepts in that debate affirm compositionality. Note that an analogous compositionality claim is also plausible for ecosystems: they are composed of individual organisms and their abiotic environments. 5. One might think that this is too quick. Perhaps compositionality can fail when domains are not considered fundamentally different. Kincaid (2004, 302) suggests that some social entities, such as ‘society’, might contain nonindividuals, such as ‘capital’. The institutions of ‘money’ and ‘property’ might also be candidates. Irrespective of whether these should be considered proper societal parts or simply by-products of interactions between parts, biological populations and communities seem to contain no counterparts that challenge compositionality. 778 JAMES JUSTUS community-, and ecosystem-level properties are considered emergent and thus impervious to reductive analysis (see Drake et al. 2007 for a sampling). A survey is impossible here, but one prominent, well-studied candidate for ecological emergence that has garnered philosophical attention illustrates underappreciated merits of MI outside the social sciences and reveals particu- larly clearly how IBMs can contribute to reductive (and deflationary) accounts of putatively “emergent” properties: community stability. The concept has a long history, tracing back at least to Aristotle’s ideas about a balance of nature, if not earlier (Egerton 1973). It was only formu- lated in precise terms, however, in the mid-twentieth century with seminal work by Robert MacArthur and Robert May, among others (see Justus 2008a, secs. 1–4). Their characterizations differed, and that pluralism continues to- day. But among the diversity of notions, one definition initially developed by David Tilman (1999) to assess the stability of grassland communities is widely entrenched, particularly among field ecologists. His concept, labeled ‘tem- poral stability’, focuses on how population biomasses in communities vary over time. In particular, temporal stability decreases as biomass variability increases, the motivating idea being that if two series of abundances are plot- ted across time, the stabler one exhibits less fluctuation. Measuring temporal stability therefore requires empirical measurement of biomasses over some period. More precisely, let Bi be a random variable designating the biomass of species i in an n-species community C assayed during some time period, and let Bi̅ designate the expected value of Bi for that period. Temporal stabil- ity St of C is then StðCÞ ¼ ½mean species biomass�ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ½biomass variance� þ ½biomass covariance� p ¼ ∑ni¼1B̄iffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ∑ni¼1VarðBiÞ þ ∑ i ¼ 1; j ¼ 1 i ≠ j n CovðBi; BjÞ r ; ð2Þ where Var and Cov designate variance and covariance, respectively. With this equation, St precisely characterizes ecological stability in terms of bio- mass constancy of the community’s species. Precision is too often lacking when putatively emergent phenomena are discussed, so St affords a rare op- portunity. With this account of stability in mind, Maclaurin and Sterelny (2008, sec. 6.4) argue that it constitutes an emergent property, immune to the reduc- tive craft of MI. Their contrast is labeled an “individualistic” view of com- munities in which abundances and distributions of organisms are largely con- METHODOLOGICAL INDIVIDUALISM IN ECOLOGY 779 trolled by abiotic factors, not interorganismal interactions. As embodied by seminal figures such as Henry Gleason and Robert Whittaker, this view re- flects a long-standing theoretical orientation in ecology according to which ostensibly community-level properties simply derive from patterns of abiotic factors impacting different species; lower-level interactions are causally neg- ligible or fail to produce anything novel at the community level. The marked difference between this austere conception of individualism and MI should be clear. While MI countenances interactions between individuals in accounts of higher-level phenomena, the conception Maclaurin and Sterelny describe corresponds most closely with methodological atomism (see sec. 3). Such a constraint would seem to preclude canonical cases of past reduction—ther- modynamics to statistical mechanics, Newtonian mechanics to relativity the- ory, and so on—reductions that depend indispensably on lower-level entities interacting in specific ways. Championing ecological emergence based on the implausibility of ecological atomism would therefore be a victory too easily achieved. Marginal controversy exists, but as Maclaurin and Sterelny them- selves document, few contemporary ecologists doubt that biological interac- tions such as competition, predation, obligate mutualism, and others shape the composition and structure of communities.6 Nor is any conceptual difficulty posed for MI by the substantive propo- sitions Maclaurin and Sterlney (2008) associate with emergence, that com- munities are “real” or that there are causally efficacious properties usually described at the community level.7 These are open questions that depend on empirical data, and both contentions are consistent with MI. Tilman’s own extensive grassland studies have attempted to demonstrate, for instance, that community stability is responsible for and/or produced by diversity, another community-level property (Tilman, Reich, and Knops 2006). And Sterelny (2001) himself convincingly argues for the reality of communities (at least in the past) based on extensive analysis of paleoecological data. These are reasonable contentions, but they do not establish anything meriting the label ‘emergence’, which Maclaurin and Sterelny characterize as the idea that com- munity properties “are not just an extrapolation of” (2008, 113) or are “not simple reflections of ” (120) the properties of individuals. Emergence so con- ceived, they claim, challenges the reductive view that “all important system- 6. Note that even “individualistic” communities in Maclaurin and Sterelny’s sense can be highly temporally stable. Without further restrictions, presumably on interspecific co- variances, nothing about St prohibits or renders it unlikely that biomass means are high (numerator) and variances and covariances are low (denominator) in communities gov- erned primarily by abiotic factors of the kind described. Variable St may therefore poorly indicate the higher-level causal integration they take emergence to require. 7. “Communities are real, causally important ecological systems if they have emergent or ensemble properties” (Maclaurin and Sterelny 2008, 113). 780 JAMES JUSTUS level behavior can be explained, and can only be explained, by explaining the behavior of the parts” (2008, 120). Before turning to whether temporal stability is emergent in the intended sense, one might think that once causally efficacious properties are described at a higher level, emergence has been established. This criterion is much too weak. Car crashes offer a straightforward example. In analyzing a crash, a car’s total mass is (correctly) represented as causally efficacious. It ac- counts, for instance, for how far a car penetrated a storefront, propelled a pedestrian, and so on. But the total mass is simply the sum of, and therefore simply reflects, the masses of car components. Causal efficacy at a higher level therefore provides a poor guide to emergence. And note that beyond such straightforward cases, the same conclusion holds for more complex higher-level properties. The car’s propulsive causal capability, for example, reflects its parts and their interactions, and explanations of that capability would thereby derive from behaviors of its parts. What is missing from Maclaurin and Sterelny’s analysis is an account of how temporal stability is any different. If emergence amounts to anything, it precludes reductive analyses of higher-level properties in terms of lower-level properties. The calculation is obviously more complicated than summing car part masses, but in an explicit formula equation (2) indicates exactly how a community-level property reflects population-level properties, namely, bio- mass means, variances, and covariances. Once population-level details are set, not only is this community-level property fixed, but its precise magnitude can be calculated, and the contributions different population biomasses make to it can be inspected. Population biomasses, in turn, exhibit a much simpler, aggregative relationship with the biomasses of individual organisms com- posing them.8 It therefore seems that temporal stability is not emergent. It can be reductively accounted for by the properties and behavior of a com- munity’s parts. 5. Reductive Unification in Ecology. In trade-offs between breadth and depth, MI is usually faulted for aligning too rigidly with depth. But scientific history shows that this charge is frequently unfair. Finding underlying con- nections or commonality is often the means to greater generality; reduction is sometimes the conduit of unification. There are several unificatory targets in ecology. Perhaps the most prized is the elusive integration of classical population-community ecology (see sec. 2) and ecosystem ecology. Biological entities are the primary focus of the for- 8. In fact, at least two types of IBMs, so-called fixed-radius neighborhood and zone of influence models, have been developed to study how the distribution of sizes of indi- vidual organisms, their density in areas, and physiological differences might influence populations’ overall biomass (see Grimm and Railsback 2005, sec. 6.7). METHODOLOGICAL INDIVIDUALISM IN ECOLOGY 781 mer, and its models address issues concerning species composition, intra- specific dynamics, and interspecific interactions between coexisting species. The abiotic environment has little, and rarely an explicit, role in these models. In contrast, energy and material flows through biotic interactions and abiotic mechanisms are the principal focus of ecosystem ecology. For instance, trac- ing cycles of nitrogen as it is processed by organisms and flows through phys- ical channels is one of many such targets of analysis. Given their contrasting perspectives, models developed in each subfield differ markedly. The causal relations represented are quite dissimilar, and even the variables considered are usually completely disparate. The pros- pects of synthesizing such divergent modeling approaches therefore seem dim, and, unsurprisingly, relatively few ecologists have attempted. Loreau’s (2010) recent and cogent attempt, however, not only is one of the most prom- ising but also showcases MI’s salience for ecological modeling. Loreau tries to integrate several models in population-community and ecosystem ecology, which involves making connections between in- tratrophic-level species diversity and ecosystem functioning, interactions in food webs and ecosystem functioning, indirect mutualism and nutrient cy- cling, population stability and ecosystem stability, and many others. Survey- ing such extensive and mathematically sophisticated results is impossible here, but fortunately there is a manageable core. The edifice’s foundation is a theoretical insight about how population-community and ecosystem per- spectives can be bridged that Loreau then applies, elaborates, and scales in different contexts to affect broader integration. And, crucially, the connec- tion concerns the mass and energy budgets of individual organisms. Before giving the relevant equation, Loreau explains why such a bridge should ex- ist: Ecosystem ecology and eco-physiology share the concepts of mass and energy budgets as tools for understanding the acquisition, allocation, and disposal of materials and energy in the metabolism and life cycle of both organisms and ecosystems. On the other hand, growth and reproduction are the two processes at the individual level that are responsible for pop- ulation growth, and these processes place high demands on energy and ma- terials in the metabolism of individual organisms. Thus, the unification of population and ecosystem approaches should be rooted in the ecophysiol- ogy of organisms, in particular, in the constraints that govern the acquisi- tion,allocation, and disposal of materials and energy. (2010, 9; emphasis added) Loreau believes that the following equation captures the insight: P ¼ εðC − mBÞ > 0; ð3Þ 782 JAMES JUSTUS where P designates total organismal production (tissue growth and repro- duction), C designates energy consumption, B designates organismal biomass, μ designates mass-specific basal metabolic rate, and ε designates production efficiency beyond basal metabolism, that is, the efficiency with which con- sumed energy is converted into tissue growth or reproductive output beyond what survival requires. If P > 0, organisms consume enough energy to grow and/or reproduce. When aggregated across individuals, this determines whether populations grow or shrink. Equation (3) therefore links factors typically in ecosystem ecology’s purview (available energy in organisms’ environments and organ- ismal physiology) with individual growth and, scaling up, population growth. Through this link, familiar models of population ecology, such as the logistic equation and equation (1) from above, can incorporate principles of and mod- els developed within ecosystem ecology. The two domains of models being reconciled both largely ignore individ- ual variation. Loreau follows suit to affect their integration. He recognizes, however, that the mathematical “convenience” (2010, 13) of this unrealistic ab- straction constitutes a nontrivial limitation, one ecologists working with IBMs are unwilling to accept. In fact, equation (3) and related equations are common- place within IBMs, and spatially explicit representations of individual-level var- iability organismal production often yield significantly different results than when all individuals are idealized as equal (see Grimm and Railsback 2005, sec. 7). Loreau does not pursue individual-based modeling, but the crucial in- sight undergirding his expansive integrative project nevertheless stems from the representational import of the individual level. 6. Conclusion. As a relatively new modeling strategy in ecology, individual- based modeling does not benefit from the mathematically sophisticated tech- niques developed to analyze classical ecological models over several decades. Instead, IBMs opt for enhanced specificity and the greater realism it furnishes by making individual organisms the explanatory and representational priority. In so doing, individual-based modeling draws from a general, individualistic approach to understanding the world with a long tradition in the social sci- ences known as ‘methodological individualism’. This paper has argued that it is as fruitful in ecology as it is there. REFERENCES Alexander, J., and B. Skyrms. 1999. “Bargaining with Neighbors: Is Justice Contagious?” Journal of Philosophy 96:588–98. Azzouni, J. 2004. Deflating Existential Consequence. New York: Oxford University Press. Drake, J., M. Fuller, C. Zimmerman, and J. Gamarra. 2007. “Emergence in Ecological Systems.” In From Energetics to Ecosystems: The Dynamics and Structure of Ecological Systems, ed. N. Rooney, K. McCann, and D. Noakes, 157–84. New York: Springer. METHODOLOGICAL INDIVIDUALISM IN ECOLOGY 783 Egerton, F. 1973. “Changing Concepts of the Balance of Nature.” Quarterly Review of Biology 48: 322–50. Epstein, B. 2012. “Agent-Based Modeling and the Fallacies of Individualism.” In Models, Simu- lations, and Representations, ed. P. Humphreys and C. Imbert, 115–44. New York: Routledge. Grimm, V., and S. Railsback. 2005. Individual-Based Modeling and Ecology. Princeton, NJ: Princeton University Press. Heath, J. 2010. “Methodological Individualism.” In Stanford Encyclopedia of Philosophy, ed. Edward N. Zalta. Stanford, CA: Stanford University. http://plato.stanford.edu/archives/spr2009 /experiment/. Hoover, K. 2010. “Idealizing Reduction: The Microfoundations of Macroeconomics.” Erkenntnis 73:329–47. Justus, J. 2008a. “Complexity, Diversity, Stability.” In A Companion to the Philosophy of Biology, ed. S. Sarkar and A. Plutynski, 321–50. Malden, MA: Blackwell. ———. 2008b. “Ecological and Lyapunov Stability.” Philosophy of Science 75:421–36. Kincaid, H. 1997. Individualism and the Unity of Science. New York: Rowman & Littlefield. ———. 2004. “Methodological Individualism and Economics.” In The Elgar Companion to Economics and Philosophy, ed. J. B. Davis, A. Marciano, and J. Runde, 299–314. Chelten- ham: Elgar. Łomnicki, A. 1978. “Individual Differences between Animals and the Natural Regulation of Their Numbers.” Journal of Animal Ecology 47:461–75. Loreau, M. 2010. From Populations to Ecosystems: Theoretical Foundations for a New Ecological Synthesis. Princeton, NJ: Princeton University Press. Maclaurin, J., and K. Sterelny. 2008. What Is Biodiversity? Chicago: University of Chicago Press. Mehdiabadi, N., and L. Gilbert. 2002. “Colony-Level Impacts of Parasatoid Flies on Fire Ants.” Proceedings of the Royal Society of London B 269:1695–99. Millstein, R. 2009. “Populations as Individuals.” Biological Theory 4:267–73. Nozick, R. 1977. “An Austrian Methodology.” Synthese 36:353–92. Odenbaugh, J. 2005. “Idealized, Inaccurate but Successful: A Pragmatic Approach to Evaluating Models in Theoretical Ecology.” Biology and Philosophy 20:231–55. Okasha, S. 2011. “Optimal Choice in the Face of Risk: Decision Theory Meets Evolution.” Phi- losophy of Science 78:83–104. Roughgarden, J. 2012. “Individual Based Models in Ecology: An Evaluation; or, How Not to Ruin a Good Thing.” Unpublished manuscript, PhilSci Archive. http://philsci-archive.pitt.edu/9434/1 /RoughgardenPSA2012IBMLecture.pdf. Sen, A. 2009. The Idea of Justice. Cambridge, MA: Harvard University Press. Skyrms, B. 1996. Evolution of the Social Contract. Cambridge: Cambridge University Press. ———. 2010. Signals: Evolution, Learning, and Information. New York: Oxford University Press. Sober, E. 1999. “The Multiple Realizability Argument against Reductionism.” Philosophy of Science 66:542–64. Sterelny, K. 2001. “The Reality of Ecological Assemblages: A Palaeo-ecological Puzzle.” Biology and Philosophy 16:437–61. Tilman, D. 1999. “The Ecological Consequences of Biodiversity: A Search for General Principles.” Ecology 80:1455–74. Tilman, D., P. Reich, and J. Knops. 2006. “Biodiversity and Ecosystem Stability in a Decade-Long Grassland Experiment.” Nature 441:629–32. 784 JAMES JUSTUS