The end of the line: An ameliorative approach to the concept of pollution

Introduction

Pollution is a central concept in environmental science, law and public discourse. And yet, when we look closer, the concept remains vague. What exactly do we mean when we call something pollution? The Oxford Handbook of Environmental Ethics, defines it: ‘In the simplest sense, pollution consists of energy or substances that are released into the environment and that cause harm to humans or other living organisms’ (Elliott, 2017). Particulate matter is typically considered pollution, and so is an oil spill, but is overtourism a form of pollution, too? Surprisingly, little philosophical literature exists analysing the concept of pollution and how it relates to the phenomena it seeks to describe.

Concepts, as we understand them, are mental and linguistic representations of phenomena. Concepts carry normative weight and can be more or less adequate in capturing certain phenomena or cases as pollution. They also play an epistemic role. Besides being merely descriptive, they define how a phenomenon should be studied. In our article, we describe how pollution has been primarily characterised in terms of thresholds and studied accordingly. We believe that this is inadequate, both in capturing phenomena that can be considered pollution and in guiding scientific and policymaking efforts. We therefore argue for the conceptual engineering of pollution to better reflect phenomena and guide science and policymaking.

Our revision might appear more metaphysical than functional, in that we assume there is a range of underlying phenomena that the concept of pollution should track more adequately. We do not deny this. However, we hold that metaphysical and functional considerations are interlinked rather than opposed. Which phenomena count as pollution is partly constituted by how we conceptualise it, and better concepts guide better practice. Hence, ours is also an ameliorative project. Our aim is not to discover a hidden essence of pollution, but to revise the concept so that it more adequately reflects the full range of phenomena that warrant concern and is better able to guide science and policy towards just and adequate responses.

Fundamentally, pollution is scientifically rooted in the concept of assimilative capacity, which captures the idea that a specific environment can handle a certain amount of pollutants before harm occurs. Assimilative capacity became institutionalised when tolerable amounts of pollutants were integrated into environmental laws (Liboiron, 2016; Holling, 1973). This threshold-based framework has been systematically analysed in both the philosophy of science and environmental studies: Feinberg (1984) explicitly names ‘the threshold of harm’ as a philosophical account of pollution. Cranor (1993) provides a philosophical account of how threshold-based evidentiary standards are embedded in environmental law. Liboiron (2016) reconstructs the framework's three core premises from its regulatory history. Thompson (2024) provides a parallel account from food ethics, noting that the use of agricultural chemicals differs from industrial pollution in a morally significant way. Pesticides are deliberately applied, not by-products, which means the benefit-risk calculation is built into the regulatory framework from the outset, and thresholds encode prior decisions about acceptable risk rather than neutral scientific discoveries. Institutionalisation is precisely what makes threshold thinking a theory of pollution rather than merely a regulatory heuristic. Our use of ‘threshold theory’ draws on this body of work. The theory defines what pollution ‘is’ (harm above threshold), not only how it is measured. It rests on three interrelated premises that shape scientific and legal understandings of environmental harm (Liboiron, 2016). First, it posits that a certain level of pollutants is acceptable, based on an ecosystem's assimilative capacity. Second, it defines pollution as the moment when harm occurs, not just when harmful substances are present. Third, this harm must be observable, measurable and traceable through scientific methods. This privileges harms that are punctual, immediate and scalar, leaving out harms that are slow, relational or emergent.

But this scalar model fails to capture the normative and relational character of the phenomenon of pollution. Indeed, as we will make clear, pollution is always for someone or something, embedded in specific cultural, sociological, and ecological contexts. Threshold models do implicitly assume a target, typically the ecosystem or human health in aggregate, but this target remains abstract. However, to say that pollution is always ‘for someone’ is to insist that the relevant standpoint be made explicit: whose health is at risk, whose environment is affected, and whose interests are deemed acceptable to sacrifice. The threshold model's implicit target obscures rather than illuminates these questions. Moreover, saying that a river is polluted is to condemn a situation as harmful, something that ought to be remedied and prevented. Indeed, pollution has never been understood purely in scientific terms. From Latin polluere, pollution initially meant desecration or defilement. Until the Industrial Revolution, the concept was often understood in a moral rather than a material sense. Anthropologist Mary Douglas framed cultural constructions of defilement as a symbolic phenomenon in which the idea of dirt has less to do with hygiene or objective contamination than with the maintenance of a particular social order. Her ‘matter out of place’ (Douglas, 2015: 44) reveals pollution as not an inherent property, but as a symbolic transgression of social boundaries. For instance, a shoe is clean on a foot, but when left on a table, it is dirty. It becomes pollution because it has crossed the line between proper and improper behaviour, stressing a normative component (Douglas, 2015: 49–50). Douglas thus challenges a purely scientific, material understanding of pollution by highlighting its role as a cultural construct that societies use to define what is impure, dangerous or impermissible. We should stress that, in Purity and Danger, Douglas was primarily theorising about ritual defilement and cultural taboo, and clearly distinguished them from environmental pollution in the scientific sense. In her essay ‘Environment at Risk’, she addresses the intersection of this symbolic framework with environmental hazards (Douglas, 1999; Thompson, 2024).

We argue that the scalar conception of pollution is conceptually inadequate and risks obscuring its complexity. We contend that pollution is best conceived as a thick, multi-dimensional concept whose evaluative and descriptive aspects cannot be separated, necessitating ameliorative conceptual engineering. Our contribution is fourfold. First, we demonstrate that threshold models face three primary issues: vagueness, discreteness and monotonicity. Second, we argue that exclusive reliance on thresholds fosters pernicious reification and ontological flattening. Third, we develop a multi-dimensional model of pollution that encompasses scalar, symbolic, political, ecological and corporeal aspects. Fourth, we apply this model to concrete cases to demonstrate its value in guiding both research and policy. Our critique targets the conceptualisation of pollution embedded in threshold models and not merely their (regulatory) operationalisation. Moreover, we believe that enriching the concept of pollution with relational, cultural and political dimensions provides new tools for environmental science and more equitable policy directions.

Throughout this article, we treat pollution and harm as linked but not identical. The concept of pollution designates a specific kind of undesirable environmental state, one that is harmful, or credibly threatens to harm organisms, communities or ecosystems, and that has different dimensions. We use ‘harm through pollution’ where we wish to foreground the causal dimension of this relationship, to distinguish the state of pollution from its effects. This distinction matters because it is collapsed in threshold theory.

Three problems of threshold theory

According to Liboiron (2021), threshold theory is based on three premises. First, it posits that a certain level of pollutants is acceptable given an ecosystem's assimilative capacity. Second, it defines pollution as the moment when harm occurs, not just when harmful substances are present. Third, this harm must be observable, measurable and traceable through scientific methods. The first problem is that pollution is a vague concept. Indeed, this first premise posits that pollutants are tolerable up to a certain threshold, marking the moment a phenomenon becomes pollution. This resembles the Sorites paradox, which arises from vague predicates like ‘heap’ or ‘tall’ (Williamson, 1996: 4). Pollution is a vague term, with clear non-cases (a pristine river is not polluted), and clear cases (a river saturated with arsenic and mercury is polluted). However, no accumulation in itself magically yields pollution. However, at some point, a river does indeed become polluted.

Thresholds attempt to resolve this vagueness by stipulating a limit using concepts such as assimilative capacity. However, there is no sharp cut-off to be found: the indeterminacy lies in the concept itself. At the same time, in practice, regulators and policymakers face the distinct challenge of drawing operational boundaries for laws and policies. Environmental laws, therefore, apply the term pollution flexibly, depending on the context (Nagle, 2009), creating a practical indeterminacy that compounds conceptual vagueness. Because the concept has no sharp cut-off, regulators have to make decisions where to ‘draw the line’, thereby illustrating Heather Douglas’ point that technical decisions and criteria inevitably embed values (Douglas, 2004). When considering vagueness, thresholds appear not as neutral discoveries of nature but as decisions that carry normative weight. This is precisely what makes vagueness a problem for threshold theory. The model presents itself as scientifically neutral and objective, yet drawing the line is always a normative decision. The appearance of neutrality conceals the value-laden character of what are, in fact, choices about acceptable risk. Take, for example, plastics. Unlike other pollutants, plastics break down into micro-plastics, which persist indefinitely and circulate through ecosystems and the bodies of organisms (Kaandorp et al., 2020). Here, the concept of assimilative capacity becomes untenable because self-purification is not possible, and plastics continue to accumulate. Plastics, like other pollutants such as radiation or carcinogens, have been argued to admit no ‘safe dose’ (Liboiron, 2021: 61), and the effects on ecological and human health are unclear (Koelmans et al., 2017). One might object that threshold theory can accommodate substances like plastics simply by setting the threshold at zero. While we take this objection seriously, we argue that it undermines the explanatory power of thresholds. Indeed, if the threshold is set at zero for any substance with no safe dose, the scalar logic and the idea of assimilative capacity are abandoned entirely. The defining commitments of threshold theory, that there is a tolerable level, that harm begins at a specific point, and that quantities below that point are presumed harmless, are then untenable. Hence, it is a departure from rather than a refinement of the idea of thresholds. For these reasons, many have already called for treating plastics as inherently hazardous rather than threshold-based pollutants (Rochman et al., 2013a).

This demonstrates that a concept of pollution, resting on assimilation and modelled as a scalar property, is strained. Without a ‘safe dose’, the very idea of (acceptable) pollution starts losing its meaning. The fact that thresholds cannot capture all cases strengthens Douglas' definition of a moral transgression rather than a quantitative measure. Plastics both materially violate established orders and permeate bodies and environments at the cellular level. They also constitute such orders by providing containment and stability, and preventing pollution, for instance, in food storage. The very qualities that make plastics socially and materially useful also enable the forms of harm that undermine the underpinnings of threshold theory.

Secondly, pollution also faces a problem of discreteness. The second premise posits that pollution is the moment when harm occurs, not merely the presence of harmful substances. Feinberg (1984) analyzes pollution as a phenomenon defined by crossing a harm threshold. At the regulatory level, the same structure applies: standard toxicological practice defines a threshold as the exposure level above which harmful effects are observed and below which they are not. For instance, the US Clean Water Act embeds this idea in law by defining toxic pollutants strictly in terms of the observable harms, such as disease or physiological malfunctions, they cause under exposure (Bureau of Ocean Energy Management, 1972).

The discreteness premise operates at three levels simultaneously. It is a philosophical claim about the nature of pollution-related harm, as a toxicological standard, and as a legal definition. As long as this presence is below the threshold level, it is not considered pollution because it is presumed harmless. Here, one could object that this discreteness characterises regulatory practices rather than the underlying concept of pollution. However, we contend that threshold models do more than stipulate regulatory thresholds as a pragmatic measure: they conceptualise pollution as a discrete event (the crossing of a line). As Liboiron (2016) remarks, threshold frameworks define pollution as the moment when harm occurs, not the ongoing presence of harmful substances. This claim goes beyond how pollution should be regulated. It is a claim about what pollution ‘is’, namely a phenomenon causing discrete and punctual harm, arising at specific points rather than gradually. It is embedded in risk assessment practices that formalise an acceptable exposure to pollutants, with the decisive marker being when harm has occurred. However, plastics act as endocrine disruptors that interfere with the hormonal and endocrine systems by mimicking hormones that bind to receptors. These disruptors do not act exactly as trespassers, unlike other pollutants, but rather as endogenous messengers gone awry.

Plastics and endocrine disruptors produce diffuse, accumulative harms rather than punctual ones (Rochman et al., 2013b). Micro- and nano-plastics accumulate throughout the food chain in environments where their polymers and additives can bind to organic material or contaminants, forming an eco-corona. This thin molecular coating surrounds them when they enter an environmental medium, like seawater or soil, and affects how they interact with organisms (Galloway et al., 2017). Moreover, the toxic effects of endocrine disruptors unfold across lifespans and generations, rather than as punctual events (Langston, 2010). Harm through pollution is thus not a singular, isolated event. It is systemic, relational and cumulative. It is better conceived as a process rather than an event, in line with process ontology, which holds that the right way to understand the living world is as dynamic processes rather than discrete entities (Nicholson and Dupré, 2018: 11). The direct implications are twofold: at the level of the concept and at the level of regulatory practices, threshold theories mischaracterise harm, which is not (only) punctual, and hence, undermine the definition of pollution as the moment when harm occurs and is measured. More importantly, threshold models may make specific harms invisible to policy and regulation, which rely on event-based thresholds. Elliott (2011) documents this structural invisibility in his analysis of endocrine disruption. The framework's inability to register diffuse, cumulative harm is not a failure of measurement technology but a consequence of the conceptual commitments built into threshold theory itself.

Thirdly, threshold theories face a problem of monotonicity. The third premise, that harm must be observable, measurable, and traceable, does not, by itself, entail monotonicity. Rather, monotonicity is an additional assumption: a monotonic relationship between dose and harm is built into threshold frameworks as a background condition. A relation is monotonic if it consistently goes in one direction: as x increases, y never decreases. In contrast, non-monotonicity means that ‘the slope of the dose-response curve changes sign from positive to negative or vice versa at some point along the range of doses examined’ (Vandenberg et al., 2012). Threshold models, however, assume that as exposure to pollutants increases, harm also increases. Monotonicity reduces complex and relational harms into a single scalar variable, enabling thresholds that are simple to define, communicate and enforce. This reductive clarity, rather than superior accuracy, makes monotonic relations appear objective in regulatory practice. Two transitions require some further explanation. First, the move from regulatory practice to the concept of pollution. Because threshold theory has been institutionalised in environmental law and scientific guidelines (Liboiron, 2016), the model's assumptions have become embedded in the concept itself. As such, it shapes not only how pollution is regulated but also how it is understood and investigated. Second, the move from conceptual commitments to ontological claims. When a concept defines pollution as a scalar, dose-dependent property, it carries implicit ontological commitments about what pollution is, namely, a quantity rather than a relation. Monotonicity also makes pollution ontologically scalar. Harm becomes a matter of degree on a single continuum. However, pollution cases challenge this assumption. Some monomer additives can have a significant impact at minute doses, while having little to no effect at high doses (Vandenberg et al., 2012). In principle, such patterns could be modelled by fitting multiple S-curves at different orders of magnitude. In practice, regulatory science rarely incorporates these complexities, instead privileging simplified threshold models. This gap between empirical variability and regulatory practice raises questions about the adequacy of thresholds as representations of harm. A large body of research has shown, for instance, that bisphenol-A (BPA) can exert biological effects at very low doses (vom Saal et al., 2007), while higher doses may produce weaker effects due to receptor down-regulation and desensitisation, consistent with its hormone-like action (Liboiron, 2021: 92–93). Endocrine disruptors with non-monotonic dose–response curves suggest that harm cannot be attributed to a single level of exposure, as effects can appear at infinitesimal doses, disappear at intermediate levels, and reappear at higher levels. Furthermore, dose intervals would not resolve the deeper challenge posed by endocrine disruptors: their effects are context-dependent, vary across endpoints, and are difficult to capture within standardised ranges. An objection is that more sophisticated models, such as multi-curve dose–responses, could capture these cases. However, these models still presuppose that harms remain scalar and commensurable. The point we make is that non-monotonicity exposes not only the limits of current models but also those of scalar frameworks: harms are not always additive or linear, but rather relational and processual.

In Is a Little Pollution Good for You? Elliott (2011) analyses the example of hormesis. Hormetic dose–response curves represent the phenomenon where low doses of a substance are beneficial and high doses harmful. He gives the paradigmatic case of Phosphon, a growth retardant and inhibitor. Although Phosphon inhibits growth at high doses, low concentrations of the chemical stimulate growth through a compensatory biological response. Hormesis presents a paradox in the context of pollution: if a substance is beneficial at low doses, can we really deem the mere presence of that substance a form of pollution? If we answer positively, we must accept that low doses, which are beneficial, count as pollution and abandon the idea that pollution is inherently harmful. If we answer negatively, we still define pollution partly by dose, but the threshold model's framework no longer holds: we can no longer simply say that above threshold X, a substance is a pollutant. We would instead have to say that below X the substance is beneficial, and above Y it is harmful. This is a fundamentally different conceptual structure that surrenders the monotonic assumption on which threshold approaches depend. Moreover, as Elliott argues, these questions cannot be resolved scientifically: determining whether a low-dose beneficial exposure should be permitted or regulated as pollution requires normative judgements about acceptable risk and the distribution of bodily effects across populations. This confirms our central claim that the evaluative and descriptive dimensions of pollution are inseparable even at the level of dose–response modelling.

Non-monotonicity reveals harm isn’t captured by universal dose-based laws like Paracelsus' maxim. The ‘one pollutant, one threshold’ model collapses when harm is context-dependent rather than dose-dependent. Plastics exemplify this multi-dimensionality. Their harms span scales (macro/micro), media (land, water, air), and mechanisms (ingestion, toxicity, endocrine disruption). Yet, threshold theory persists because it serves regulatory purposes, despite being conceptually inadequate. By ‘conceptual inadequacy’ we mean that the threshold framework, taken as an account of what pollution is, fails to accommodate the full range of phenomena we recognise as pollution, mischaracterises the temporal and relational nature of pollution-related harm, and generates systematic blind spots in both science and policy (Boudia and Jas, 2014). At the same time, threshold theory is a deeply entrenched model that resists revision even in the face of counterevidence, and regulations still adhere to it (Demeneix et al., 2020). For instance, a guideline on the environmental risk assessment of medicinal products for human use refers to toxicity data in these terms: ‘The lowest no observed adverse effect level (NOAEL) from the available repeated-dose toxicity studies (oral administration and chronic treatment are preferred, the use of other routes of administration should be justified) should be recalculated to a chronic no observed adverse effect concentration (NOAEC)’ (European Medicines Agency, 2006).

This exclusion is not only an epistemic but also an ethical limitation, since diffuse and cumulative harms from pollution are rendered invisible to science and policy. It also conflates two conceptualizations of such harm as additive and commensurable, while pollution-related harm is also relational and discontinuous. If we persist in treating pollution as if threshold theory could account for it, we reduce it to what is most convenient for regulation, while excluding the harms that pose the greatest challenges to human and ecological health. Moreover, by treating pollution as a single, commensurable quantity, threshold theory flattens diverse phenomena into a uniform category. We will link this to the notion of ontological flattening.

Pernicious reification and ontological flattening

Mary Douglas observed, ‘there is no such thing as absolute dirt: it exists in the eye of the beholder’ (Douglas, 2015: 2). Douglas develops a complex theory of why certain cultural practices are elevated to the status of taboo, irrespective of any apparent harm or risk associated with their violation. Her primary concerns are ritual defilement and the maintenance of social order, and she was, at the time, careful to distinguish this from environmental pollution in the scientific sense. In ‘Environments at Risk’ (Douglas, 1999), which more directly bridges these domains, she argues that science has come to serve a culturally defined role once played by religion, and that what counts as an environmental risk reflects symbolic and institutional assumptions, not only empirical facts. As she puts it: ‘Just as transgression became the salient point of reference for blaming for disaster in the Bible, and sin in the history of Christianity, in our secular, scientific world, risk has become the convenient, conspicuous blame term that all parties connive to promote’ (Douglas, 1999: 227–228). The parallel is instructive for our argument about reification. Just as ‘sin’ once appeared as a self-evident property of actions rather than a culturally constructed verdict, the scalar threshold presents itself as a neutral discovery about nature rather than a normative judgement about acceptable harm. Reification in pollution science is not, therefore, a contingent failure of rigour. It follows a deeper pattern by which cultures stabilise moral accountability through categorical claims that appear to have objective, natural authority. The threshold does not merely measure pollution, but it performs the same work that concepts of transgression and taboo have always performed. It draws a line between the permitted and the condemned, while concealing the fact that drawing it is a social act.

With this idea in mind, we can understand pollution as a thick concept (Williams, 1985). Like ‘cruel’ or ‘brave’, thick concepts demonstrate how evaluation is inseparable from description, unlike thin concepts like ‘good’ or ‘bad’, which only express evaluation. Pollution indicates the physical state of an environment being contaminated while simultaneously condemning this state as harmful. Saying that a ‘river is polluted but it is not bad’ would strike most people as incoherent, because the evaluative component is integral to the concept.

Although threshold approaches capture and define pollution in metrics and variables, seemingly objectively, they are not innocent. They also carry normative judgements. Regulatory thresholds such as mercury concentrations in fish embody past decisions about harm. Over time, these numbers become reified as neutral cut-offs, obscuring the judgements that initially led to their creation. For these reasons, an overreliance on thresholds is problematic: the risk is that pollution, a complex and multi-dimensional concept, is reduced to a narrow technical measure.

Engineering pollution as a relational dimensional concept

Building on everything we have established so far, we now propose a revised concept of pollution with five dimensions. Before doing so, we wish to be explicit about our approach to conceptual engineering and about the requirements we apply to evaluate competing conceptions of pollution. Our approach combines metaphysical and functionalist commitments. On the metaphysical side, following Haslanger (2013), we assume that there is a range of phenomena in the world, including endocrine disruption, PFAS accumulation and oil spills, that warrant being grouped under the concept of pollution, and that a better concept will track these phenomena more adequately. On the functionalist side, following Thomasson (2020), we hold that the revised concept should also be more useful. It should guide scientific inquiry, coordinate policy and support environmental justice more effectively than the threshold model. We believe these commitments go together. A concept that better tracks the phenomenon's relational, multi-dimensional character will also be more practically adequate. We believe an adequate conception of pollution should at least have the following characteristics. First, it should distinguish pollution from mere toxicity or harm. Second, it should preserve the idea of pollution as a thick concept. Third, it should accommodate both paradigm and borderline cases while avoiding trivial cases, such as water. Fourth, it should guide transdisciplinary inquiry and policy in ways that are sensitive to context and structural injustice. We showed in the first through third sections that the threshold model fails on all four counts. We will show in the fourth through fifth sections that our five-dimensional framework meets them.

We approach it by recognising that these five dimensions encompass the minimal set of relations without which the concept risks being flattened. They are distilled from legal and scientific frameworks (scalar thresholds), anthropological accounts (symbolic transgressions), environmental justice and discard studies (political issues), ecology (ecosystem and systemic disruption), and feminist/post-humanist theory (corporeality). The framework is deliberately non-stipulative: it is grounded in past and current uses but organised in a way that resists reduction to a single axis. Also, this framework is flexible: in some instances of pollution, certain dimensions will be more salient than others.

Hence, we do not consider these five dimensions to be necessary and sufficient in the usual sense. No single dimension is by itself sufficient to deem something to be pollution. For instance, scalar harm alone does not distinguish pollution from toxicity, and symbolic transgression alone does not distinguish pollution from taboo violation. These five dimensions are better understood as a set of lenses for diagnosing pollution: the more dimensions are present and instantiated, the more appropriately something can fall under the concept of pollution. As such, borderline cases call for careful assessment and contextual judgement rather than decisions made on metric measurements alone.

The multi-dimensionality allows us to capture a range of phenomena that demand urgent attention under the heading of pollution. It is, we argue, useful in both science and policy. It also enables us to distinguish between ‘proper’ uses of environmental pollution that warrant scientific investigation and action and improper ones. Our concept has ontological advantages, as it reflects more closely which phenomena are considered pollution in an environmental context. It has epistemic advantages, as it points towards ways of studying pollution that extend beyond thresholds to include contexts and relationships. Additionally, it has normative advantages, as it can guide policy not only in preventing direct harms but also in addressing structural injustices.

Cappelen (2018) stresses that conceptual engineering may face an implementation challenge: How can revised concepts be adopted in practice and not remain theoretical speculations? Similarly, a common issue in conceptual engineering is the ‘topic continuity problem’, which suggests that there would be no such thing as conceptual revision. When we re-engineer a concept, we merely shift attention away to another concept and replace it (Kocurek, 2022). By situating our five-dimensional model within both scientific practice and public discourse, we aim to mitigate these objections: it revises the concept without severing it from existing usage. One might ask why not eliminate the concept and replace it with a cluster of more precise concepts or abandon it altogether? We retain the term pollution because it is deeply entrenched in scientific, legal and everyday discourse and, as a thick concept, eliminating it would mean losing its distinctive normative force.

What our five dimensions have in common is that they are all related to the idea of pollution as a relational concept. Each dimension brings into focus different relations between substances and bodies, communities and environments, symbolic orders and material processes, and thus resists ontological flattening. We assume that there is a range of phenomena in the world that our concept should adequately track. The five dimensions are designed to capture the different relational structures that characterise this range. This relational aspect has epistemological implications: no single discipline or method suffices to grasp pollution or pollutants in their complexity. Relationality also makes the ethical stakes explicit: to name something pollution is to mark it as undesirable for someone or something, and these standpoints must be acknowledged.

Pollution ‘re-engineered’: Advantages and examples

We have presented a five-dimensional relational framework for pollution that can guide both the study of pollution and the measures to be taken, improving on scalar interpretations in several ways. First, because our approach requires multiple overlapping dimensions rather than relying solely on dose–response relationships, substances do not ‘suddenly’ become pollution when they cross a certain threshold. Second, it addresses the fact that threshold approaches often conceal forms of harm that do not fit neatly into dose–response curves, such as the non-monotonic effects of endocrine disruptors, the cumulative impacts of micro-plastics, and the intergenerational epigenetic effects of toxic exposures. Third, our approach does justice to pollution as a thick concept, incorporating political and social dimensions, and revealing how environmental harms are always embedded in relations of power and structural injustice. Our multi-dimensional concept also suggests that pollution should be studied in a transdisciplinary manner, integrating insights from toxicology, ecology, anthropology, political science, philosophy and other relevant sciences.

An objection could be that we do not adequately address reification. Indeed, in the third section, we have argued that pollution, as a threshold, becomes reified, and that this is a problem, as thresholds are merely one aspect of the phenomenon. Admittedly, pollution in five dimensions is also an abstraction, and any abstraction risks reification if it is mistaken for the thing itself. We believe, however, with Whitehead, that abstraction is unavoidable and necessary in science (Whitehead, 2010). Rather than a metaphysical claim about reality, the five-dimensional framework is an epistemic tool. Its value lies in enabling more adequate reasoning and action across diverse contexts. It is also a plea to use abstractions that are as thick and multi-dimensional as possible. We hope to have provided a map that represents the territory as well as possible. We will now demonstrate how a concept of pollution incorporating the five dimensions can be applied to demarcate and study specific phenomena.

In our framework, per- and polyfluoroalkyl substances (PFAS) are a paradigmatic case of pollution, exemplifying why threshold approaches are insufficient. We can refer back to the possible objection that a threshold set near zero could accommodate such a case. Once again, that would be abandoning the core commitment to assimilative capacity in threshold theory, and it cannot account for the symbolic, political, ecological and corporeal dimensions of PFAS harm that our framework captures.

First, with respect to the scalar harm dimension, PFAS exhibit health effects at parts-per-trillion concentrations and persist indefinitely in bodies and environments, transforming supposedly ‘safe’ environments into chronically polluted ones (Buck et al., 2011; Grandjean and Clapp, 2015). PFAS also violate cultural boundaries (symbolic dimension) by infiltrating spaces of purity, such as breast milk, arctic ice and newborn blood. As forever chemicals, they transgress natural cycles of decay and renewal. PFAS is directly linked to patterns of environmental injustice, with military bases and industrial facilities disproportionately affecting Indigenous communities. PFAS become embedded in ecological processes, generating ‘novel ecosystems’ in the form of PFAS-adapted microbial communities that actively participate in new environmental relationships (Huang and Jaffé, 2019). PFAS demonstrate trans-corporeality by incorporating into biological processes, binding to serum proteins, and concentrating in the liver and kidneys. Transgenerational transfer through pregnancy and breastfeeding reveals species-specific and gendered vulnerability patterns. Our dimensional analysis establishes PFAS as a form of pollution, rather than merely a toxic substance, and points towards the transdisciplinary and environmental justice it demands.

Noise pollution presents a borderline case that allows us to address directly whether our framework is restricted to substance-based pollution. Unlike PFAS, noise involves no dispersal of a physical substance through an environment. One might therefore wonder whether it belongs to the same category at all, or whether calling it pollution is merely a metaphorical extension of a concept. With respect to the scalar dimension, noise pollution has well-documented, quantifiable effects. Exposure to sound above 85 decibels causes incremental hearing damage, and chronic exposure to lower-level urban noise is associated with cardiovascular disease, sleep disruption and stress-related illness (Babisch, 2008; WHO Regional Office for Europe, 2011). Regulatory bodies set noise thresholds for residential areas, workplaces and areas near airports. Corporeally, noise does not merely affect hearing: it is incorporated into the body's stress–response systems. Chronic noise exposure elevates cortisol levels, disrupts circadian rhythms and affects fetal development (Selander et al., 2009, 2019). These are examples of trans-corporeality, as the boundary between the sonic environment and the body appears as permeable, and exposure is experienced differently depending on age, health status and prior exposure history.

The ecological dimension is also present. Anthropogenic noise disrupts animal communication: it masks birdsongs, disrupts cetacean navigation and communication, and interferes with mate selection and territorial signalling (Slabbekoorn and Peet, 2003; Weilgart, 2007). The political dimension is present too, as airports, motorways and industrial facilities are disproportionately located near low-income and minority communities, and standards such as ‘acceptable noise levels’ embed economic interests that systematically expose communities with little political power (Hajat et al., 2015). Noise also carries symbolic weight. Unwanted sound is widely understood as an intrusion and a disrespect for shared space. At the same time, what counts as ‘noise’ (as opposed to sound or music) is context-dependent and culturally variable in ways that material pollutants are not. A muezzin's call, a neighbours’ celebration, or industrial machinery may or may not constitute symbolic transgression depending on the cultural and relational context. This context-dependence does not disqualify noise as pollution but illustrates why the symbolic dimension must always be assessed alongside the others.

Considered across all five dimensions, noise pollution qualifies as genuine pollution under our framework. It is not merely a metaphorical extension of the concept. Its scalar, corporeal, ecological and political dimensions are present, and its symbolic dimension is present but context-sensitive. This case shows that what matters is not the medium of transmission but the relational structure of the harm. It matters whether something intrudes into bodies, disrupts ecological processes, transgresses symbolic orders and imposes asymmetric burdens. The water case discussed in the second section confirms this: lethal doses of water entail scalar harm without symbolic transgression, ecological disruption or political injustice, and fall outside the concept of pollution, demonstrating that the framework can distinguish pollution from mere toxicity.

Conclusion

We have engineered the concept of pollution as referring to a relational phenomenon instantiated when (i) scalar harms arise or credibly threaten to arise, together with (ii) symbolic boundary transgressions, (iii) ecological process disruptions, (iv) politically asymmetric impositions of risk/benefit, and (v) corporeal incorporations or invasions across multi-species bodies.

Our conceptual engineering is both ameliorative and practical. We have revised the concept, not to discard it altogether, but to make explicit the relational dimensions that ordinary, scientific and regulatory uses of pollution have implicitly presupposed and even reified. First, the dimensions can be used as a tool to delineate ‘proper’ pollution, which should be studied qua pollution. Second, it can guide transdisciplinary scientific research into pollutants. It helps intra-project communication between different disciplines. By adopting a dimensional approach to the concept of pollution, each discipline within a project can have its own nuances, and specific interpretations of pollution need not overlap completely. Moreover, the concept suggests something about how pollution should be studied. Projects investigating pollutants should not be confined to risk and threshold analyses to guide policymaking. The dimensional nature of pollution implies that pollutants should be examined from multiple perspectives, and that sociological, cultural and political aspects are not merely nice-to-have afterthoughts but essential to most scientific projects.