Macroecological pluralism

Abstract

Theory-based and model-based reasoning shape macroecology: one derives models from general principles, the other builds models around data. We argue these are not rival traditions but interacting positions along a continuum of formalization. This pluralism is macroecology’s strength: by allowing theories and models to exchange roles as sources of constraint, synthesis and innovation, the field can confront the complexity of natural systems and generate explanations that are cumulative and predictive.

Introduction

“The ideal end of science is not unity but absolute plethora” ¹

Macroecology was introduced as a powerful methodological approach for investigating ecological patterns, offering a broad and integrative perspective on natural systems. The term was first introduced in 1989² to define an empirical and operational research agenda focused on “big questions” in ecology and evolution. At its core, the field seeks to identify the processes that link ecological phenomena across various spatial and temporal scales, from the behaviors of individual organisms and local populations to large-scale patterns of dispersal, speciation, and extinction. This approach renewed interest in broad-scale empirical analyses that had been prominent in the 1950s and 1960s across ecology, comparative physiology, evolutionary biology, and early biogeography^3,4, but that had later declined with the rise of experimental and increasingly localized ecological studies⁵.

By the late 1990s^6–8, macroecology was increasingly viewed as a non-experimental, statistical approach. It sought to connect species’ population dynamics and local interactions, typically studied at small scales, with speciation, extinction and range dynamics that unfold across broader spatial and temporal scales. This shift reflected a growing recognition that large-scale ecological patterns cannot be understood as straightforward extrapolations of processes observed at local scales^6–9. When systems are aggregated across many populations, environments, and species, the relationships observed locally may change in form, and new regularities can emerge that are not apparent at smaller scales⁹. The challenge, therefore, lies not only in the complexity of ecological components but in the difficulty of translating their behavior across scales. Macroecology, therefore, adopted a distinct non-reductionist perspective, treating large-scale patterns not as straightforward aggregations of fine-scale processes but as outcomes of dynamics operating at higher levels of organization that generate broad patterns of biodiversity^7,9. This perspective naturally encouraged conceptual integration with macroevolution, whose focus on speciation and extinction processes operates at comparable scales^8,10,11.

Together, this emphasis on scale integration and cross-disciplinary synthesis positions macroecology not only as a methodological approach, but as a scientific perspective^12,13. Following perspectivist accounts of science¹², perspectives are structured ways of representing and investigating the world, defined by particular questions, scales of analyses, and criteria for what counts as explanation. From this standpoint, macroecology adopts a broad view¹⁴. It seeks general patterns across large taxonomic, temporal and spatial scales⁹, deliberately setting aside fine-scale detail in order to reveal large-scale regularities. In doing so, it shapes what counts as a relevant unit of analyses, what data are assembled, and what forms of explanation are pursued. It cuts across traditional disciplinary boundaries, blurring lines between fields that were once treated as distinct^13,15. Yet within this shared perspective, there has been no single view about how explanation should be structured. Researchers have differed in the degree to which theoretical principles should precede model construction, and how strongly model development should be constrained by explicit theoretical commitments^14,16. These differences have given rise to distinct, though related, research traditions¹⁷ within the macroecological perspective.

One view regards the primary goal of macroecology as the identification of fundamental principles or general laws governing the structure and function of complex ecological systems¹⁴. In this tradition, large-scale patterns are explained by deriving models from overarching theoretical commitments and testing their predictions against empirical data. A contrasting view emphasizes the construction and comparison of models grounded in empirical patterns to uncover the mechanisms underlying broad-scale ecological and evolutionary dynamics^15,18. Here, explanation develops through iterative model constructions and evaluation, with broader generalizations emerging as models are confronted with empirical evidence.

These orientations differ in how scientific reasoning is organized and are often described as top-down and bottom-up modes of inquiry¹⁹. In top-down approaches, reasoning proceeds from general principles to specific models and empirical tests. In bottom-up approaches, reasoning begins with model construction and pattern analysis, from which more general insights are progressively articulated. Both rely on theoretical assumptions, but they differ in how explicitly and formally those assumptions constrain model development^{14,16,20–22}. We refer to these traditions as theory-based and model-based macroecology (Fig. 1a).

Fig. 1. The macroecological perspective and its research traditions.

a the macroecological perspective is represented metaphorically as a magnifying glass, emphasizing its role as a structured way of viewing ecological systems at broad spatial, temporal, and taxonomic scales. Within this shared perspective, two research traditions differ in how theory constrains modelling. Theory-based macroecology is characterized by strong. top-down constraints, first-principle commitments, and models that operationalize and evaluate theory. Model-based macroecology emphasizes phenomenon-driven model construction, in which theoretical commitments emerge though modelling and remain open to revision as models are compared with empirical systems. b Formalization as a continuum in macroecological inquiry. The gradient arrow represents the degree of explicit theoretical formalization, ranging from theory-based to model-based orientations. Despite differences in how theoretical commitments are structured, both traditions operate within a shared macroecological reasoning space, encompassing theoretical principles, representational models, models of data, empirical comparison, generalization, and prediction.

Theory and model traditions

In many philosophical accounts of science, theories are conceived as hierarchically structured frameworks that organize and explain observable phenomena within a defined domain²³. In this view, a theory functions as a hierarchical system linking broad principles to more specific models²⁴. General principles are progressively specified into more concrete theoretical assumptions, which in turn underpin individual models that can be evaluated against empirical data. This top-down organization, often associated with the semantic view of theories²⁵, characterizes what is commonly described as theory-based science^23,26,27.

A notable example of this top-down, theory-based approach in macroecology is the Metabolic Theory of Ecology (MTE)^28,29. MTE seeks to explain a wide range of ecological patterns by deriving mechanistic models from general physical and biological principles, particularly the scaling of metabolic rate with body mass and temperature³⁰. Its central premise is that metabolic processes, constrained by organismal size and biochemical kinetics, govern rates of growth, reproduction, and energy flux, thereby shaping ecological patterns across scales.

MTE has played an important role in generating a family of models that account for diverse ecological phenomena, including biomass production, ontogenetic growth, population growth rates and densities, interspecific interactions, species diversity, and a range of ecosystem processes³⁰. One illustrative example is a model derived from first principles to explain latitudinal diversity gradients^31,32. This model predicts a linear relationship between the logarithm of species richness and the inverse of environmental temperature, with an expected slope between −0.6 and −0.7, determined by metabolic activation energy. Because this quantitative prediction follows from first principles, empirical estimates of the richness-temperature slope can be used to test whether the predicted value emerges across datasets³¹.

As illustrated by the MTE example, theory-based inquiry proceeds from general principles to models whose predictions are tested against observations. While this top-down structure is often taken to support rigor and explanatory breadth, it sits within a long-standing philosophical dispute about whether scientific understanding primarily arises from unifying theories and general laws or from the construction and use of models tailored to particular phenomena^33–35. In the latter “model-first” tradition, theories may function less as axiomatic systems and more as toolboxes or organizing commitments, with models mediating between theoretical ideas and empirical data in ways that are not reducible to straightforward derivation from a unified set of general principles³³.

In macroecology, this model-based orientation becomes visible in approaches that treat models, rather than overarching theories, as primary vehicles of inquiry^34,36,37. Here, explanations begin with the construction of models designed to represent empirical systems, often drawing selectively on multiple theoretical ideas without committing to a single unifying framework. These models take diverse forms. Some are correlative, implemented under different inferential frameworks, while others are mechanistic, specifying processes explicitly through equations or simulations that may be deterministic or stochastic. Models are compared, refined, and extended across datasets, and theoretical commitments emerge gradually through this process rather than preceding it. In this more pragmatic view²⁵, models serve as explicit representations of ecological systems³⁶ and are developed to engage directly with empirical data, functioning as tools to explore and organize large-scale patterns.

More broadly, models serve different roles within scientific reasoning³⁸. Some are primarily interpretative³⁶, clarifying concepts, examining theoretical assumptions, or exploring how particular mechanisms might generate observed patterns without aiming to reproduce empirical systems in detail. Others are more directly representative³⁶, constructed to approximate empirical structure, often through correlative relationships, to enable comparison with data and support prediction. These roles do not form discrete categories but instead define a continuum that reflects how closely a model is tied to empirical evidence and how explicitly its theoretical commitments are articulated.

Mechanistic simulation models in macroecology^39,40 illustrate this continuum clearly. Such models can function as interpretative tools when used to explore how hypothesized processes or constraints generate large-scale patterns, and as representative models parametrized and constrained by empirical data for direct comparison with observations^41–43. By spanning this range, mechanistic models demonstrate how a single modelling approach can shift between interpretative and representative roles within the same scientific perspective.

Several influential developments in ecology align with model-based mode of inquiry^44–46 in which model construction does not proceed from a unified axiomatic theory³⁶. Such approaches have been criticized by those advocating greater formalization and structural coherence in ecological theory^16,23,24,47. While increased formalization is often taken to secure rigor and cumulative advance, it is difficult to demonstrate that formalization alone has driven ecological progress. Much of the field’s empirical growth has proceeded through iterative model-based research rather than through tightly unified theoretical systems^44–46. These patterns resonate with a long-standing philosophical dispute about how theory and models should be related in scientific inquiry, and about whether unified, axiomatic frameworks are necessary as either a prerequisite or an endpoint of explanatory progress^{34,36,38,48–50}.

The application of the MTE to latitudinal diversity gradients, discussed above as an example of theory-based macroecology, illustrates this philosophical tension in practice. The first-principle model predicting a slope between -0.6 and -0.7 relating richness to inverse temperature³¹ shows limited agreement with empirical richness patterns across taxa and spatial scales^51,52 (Fig. 2). Within a strictly theory-based logic, such systematic discrepancies would not merely challenge a specific empirical prediction but would call into question the theoretical principles from which that prediction was derived. Because the same metabolic premises are invoked to explain a wide range of ecological patterns²⁸, persistent failure in one domain would, in principle, require reassessment of the broader theoretical structure itself.

Fig. 2. Comparison of Metabolic Theory Predictions with Empirical Richness Patterns.

This figure illustrates the variation in Ordinary Least Squares (OLS) slopes for different taxonomic groups across various geographical regions, as tested by Hawkins et al.⁵¹. The shaded region between −0.7 and −0.6 represents the range predicted by the Metabolic Theory of Ecology, within which the slopes are expected to align with theoretical models. However, empirical data rarely fit within this predicted range, highlighting discrepancies between metabolic theory predictions and observed large-scale richness patterns in real-world data.

In complex ecological systems, however, such discrepancies rarely lead to direct theory revision. Instead, responses typically occur at the level of model development. Researchers modify structure, incorporate additional processes or variables, narrow the conditions under which predictions are expected to hold, or reformulate the model’s scope. This recurring pattern highlights the difficulty of sustaining a purely top-down mode of explanation in systems characterized by multicausality and historical contingency.

In studies of latitudinal diversity gradients, for example, ecosystem productivity has frequently been found to account for variation in species richness more effectively than temperature alone across taxa and spatial scales⁵³. Models that incorporate productivity, rather than relying solely on temperature as emphasized by the MTE, tend to provide a closer empirical fit. Whether such model extensions constitute a comprehensive explanation remains debated. Although progress has been made in clarifying mechanistic links between productivity and diversity^54,55, the historical and contingent nature of large-scale patterns makes direct tests of fully articulated theories difficult. Under these conditions, inquiry often proceeds through exploration of alternative variables, refinement of model structure, and simulation of interacting processes to assess how large-scale patterns may arise. This recurrent pattern helps explain the frequent emergence of model-based strategies in macroecology. Yet this emphasis on model development is sometimes interpreted as a retreat from theory or as an invitation to unrestricted statistical association.

Such interpretations overlook the fact that even when models are not derived from a unified axiomatic framework, their construction reflects prior assumptions about causality, scale, and which processes are considered influential. Theoretical commitments may be less hierarchically organized⁵⁶, but they remain embedded in decisions about variables, mechanisms, and constraints⁵⁷. What sometimes appears as arbitrariness often reflects limited explicitness in how these assumptions are articulated rather than their absence⁵⁸.

At the same time, genuinely unconstrained model fitting rarely yields durable conceptual advances. Claims that rest solely on statistical fit, without coherent justification of model structure, are unlikely to withstand replication, comparison across systems, or continued empirical scrutiny. In macroecology, as in other sciences, cumulative evaluation, including replication and critical peer assessment, tends to filter out weakly grounded associations. The concern, therefore, is not that flexibility inevitably produces unreliable results, but that the assumptions guiding model construction must be clearly articulated and critically evaluated.

These concerns will become more visible with the rise of machine learning in macroecology⁵⁹. Highly flexible algorithms can achieve strong predictive performance while offering limited insight into the mechanisms generating large-scale ecological patterns⁶⁰. This contrast between predictive success and explanatory transparency sharpens an issue already present in model-based inquiry, namely, when accurate predictions are enough and when explicit theoretical articulations are required⁶¹. Machine learning thus represents a revealing case at the model-mediated end of the continuum, highlighting the need to distinguish predictive power from explanatory understanding.

More fundamentally, however, the central epistemic challenge in macroecology does not stem primarily from excessive model flexibility, but from underdetermination: the condition in which different theories and models remain compatible with the same empirical evidence⁶². Across the continuum of formalization (Fig. 1b), both unified theoretical frameworks and competing models can generate similar large-scale predictions, leaving multiple explanations compatible with the same empirical patterns. In systems characterized by historical contingency and multicausal processes, decisive tests are rare. Under such conditions, explanatory strength rests on consilience, in Whewell’s sense^63,64, as independent lines of evidence converge on a shared theoretical or model-based explanation. Across the continuum of formalization (Fig. 1b), it is this convergence that lends robustness to macroecological accounts.

Reasoning in macroecology

Macroecological inquiry operates within a shared process of scientific reasoning. This process reflects a nuanced interplay between theory, model, and data, rather than a strict hierarchy among them. Drawing on ideas from scientific perspectivism¹², such a framework provides a coherent way to integrate both theoretical and empirical elements across the macroecological perspective (Fig. 1b).

At the core of this reasoning process, inquiry may begin either with general theoretical principles or directly with models. When theory plays a leading role, emphasis is placed in broad principles and conditions rather than on highly formalized structures²⁶. These principles guide the construction of models that translate abstract ideas into specific, testable formulations²³. Conversely, inquiry may also start from models themselves, which can motivate or constrain subsequent theoretical refinement^33,34. In either case, what varies is not the presence of theory, but the degree to which theoretical commitments are explicitly formulated and hierarchically organized, shaping how models are constructed, interpreted, and related to the natural world¹².

Model evaluation proceeds through hypothesis testing, understood here not as the verification or falsification of theories in isolation, but as an assessment of how well models align with empirical patterns. These empirical patterns may be derived from large-scale observations, comparative analyses, or controlled experiments conducted either in the field^65,66 or in silico. In theory-based approaches, experiments are often designed to test predictions derived from general principles, serving primarily an evaluative role. In model-based approach, experiments may instead help identify mechanisms, estimate parameters, or constrain model structure. In macroecology, hypotheses are better understood as model-derived expectations that organize empirical testing^33,46, rather than as propositions to be conclusively verified or refuted under a strictly positivist conception of science²⁶.

A range of inferential strategies may be used in this process, including deductive derivation of model predictions and inductive inference from data, frequently formalized through statistical or probabilistic methods. Importantly, how evidence is interpreted in model-data comparisons depends not only on statistical or methodological choices, but also on broader philosophical commitments, such as views on realism versus instrumentalism⁶¹ or on the role of underdetermination in inferring unobservable processes from empirical data⁶².

In macroecology, empirical data themselves require substantial abstraction and transformation. Rather than working with direct observations, researchers operate with models of data⁶⁷, such as species classifications, inferred geographic ranges, or aggregated measures of diversity. These constructions embed decisions about classification, spatial scale, and aggregation, meaning that comparisons between models and data involve matching one representation to another. Hypothesis testing, therefore, involves matching models with models of data, both shaped by prior assumptions that may be more or less explicitly articulated³³. The final stage of this process involves generalization and prediction, through which the scope, robustness, and transferability of models are evaluated across different ecological systems.

In this broader sense, theory should not be understood solely as a unified axiomatic framework positioned above models^26,34. Rather, it encompasses the organized structure of assumptions, representations, and inferential commitments that guide this entire process²³. What varies across macroecological research is the degree to which these commitments are made explicit and hierarchically formalized (Fig. 1b). Recognizing this broader conception of theory clarifies why pluralism in macroecology is not a weakness but a reflection of how explanation proceeds in complex systems.

A pluralistic scientific perspective

The difficulty of developing fully unifying theories in ecology is often interpreted as evidence that the field does not conform to a traditional image of science centered on comprehensive, axiomatized frameworks. Yet macroecology has progressed precisely by resisting a rigid hierarchy between theory, models and data. Its development reflects movements along a continuum in the degree to which theoretical commitments are formalized and hierarchically structured. At times, theory-based reasoning has provided broad constraints and unifying ambitions. At others, model-based strategies have generated insights by integrating patterns across scales, taxa, and datasets without requiring a fully articulated overarching structure. As Ian Hacking provocatively suggests, “the ideal end of science is not unity but an absolute plethora”¹. Pluralism, in this sense, is not simply the coexistence of two research traditions. It is recognition that explanatory strength in macroecology emerges from the dynamic interplay between degrees of formalization. The field need not aspire uniformly to axiomatic completeness, nor should it be ashamed of model-based approaches in pursuit of a more formal ideal. Rather, progress often arises when model-based investigations clarify mechanisms, reveal constraints, or expose regularities that later inform more general theoretical commitments.

From this perspective, debates that set formalism against pragmatism misrepresent the structure of macroecological inquiry. Theory-based and model-based reasoning are not competing endpoints but interacting positions along a gradient of formalization. Their relative influence shifts with the questions asked, the scale examined, and the evidence available. Macroecology’s strength lies not in resolving this pluralism into a single dominant mode, but in embracing it as a productive feature of scientific practice. By allowing theories and models to exchange roles as sources of constraint, synthesis, and innovation, the field is better equipped to confront the complexity of natural systems and to generate explanations that are both cumulative and predictive.

Author contributions

M.T.P.C. and J.A.D.F. conceived and designed the study and wrote the manuscript.

Data availability

No datasets were generated or analysed during the current study.

Competing interests

Author J.A.D.F. is an Associate Editor of npj-Biodiversity. J.A.D.F. was not involved in the journal’s review of, or decisions related to, this manuscript.