Homeostatic property cluster theory

Homeostatic property cluster theory (HPC theory) is an account of natural kinds in the philosophy of science, introduced by Richard Boyd as a non-essentialist alternative to views that define kinds by necessary and sufficient conditions. On this account, many natural kinds are characterised by clusters of properties whose co-occurrence is maintained by causal mechanisms, even though no single property need be shared by every member and the boundaries of the kind may be vague.

Boyd introduced the account in a 1988 essay on moral realism, where the illustrative cases were moral goodness and health, and afterwards developed it chiefly with biological species in mind. It became the most widely discussed alternative to essentialism in debates over natural kinds and was adopted across the philosophy of biology, psychiatry, and the social sciences.

The account has been criticised for the difficulty of individuating the causal mechanisms it requires, for failing to specify what fixes membership in a kind, and for matching poorly the classifications that scientists actually use. P. D. Magnus has argued that being a natural kind is neither necessary nor sufficient for being an HPC, and that the account is better understood as describing the ontological basis of some natural kinds rather than defining natural kinds as such. Matthew Slater has proposed a refinement, the stable property cluster (SPC) account, which replaces the mechanism requirement with a more general notion of stability.

Background

Traditional essentialism about natural kinds holds that each kind is defined by a real essence statable in terms of necessary and sufficient conditions for membership. This approach faces difficulties when applied to biological species. Species members vary considerably in their properties; no single genetic, morphological, or behavioural trait is shared by all and only the members of a given species, and species boundaries are often not sharp. Boyd argued that these features are not defects of biological taxonomy but reflect the genuine structure of the natural world, and that an account of natural kinds was needed that could accommodate them.

Development

Boyd first stated the account in "How to Be a Moral Realist" (1988), where it appears in the course of an argument for moral realism rather than as a thesis about biology. The section setting it out gives eleven numbered conditions on a "homeostatic property cluster", illustrates them with the terms healthy and is healthier than, and then proposes that moral goodness is such a cluster, unified by psychological and social mechanisms – a position Boyd called "homeostatic consequentialism". Biological species enter as the paradigm case in the same discussion.

Boyd restated and extended the account in a series of later papers, with the fullest treatment for biology in his 1999 chapter "Homeostasis, Species, and Higher Taxa". He returned to the position in 2010, arguing that if higher taxa are to be treated as HPC kinds then the traditional definition of monophyly must be revised, since the emergence of a taxon-defining property cluster does not always correspond to a unique speciation event. A late statement, responding to critics, appeared in 2019.

Magnus observes that the account has no locus classicus: it was given piecemeal across papers spanning twenty-five years, in most of which Boyd invokes HPCs in order to establish something else as the paper's central topic. One consequence, Magnus argues, is that the account as popularly received differs from Boyd's own.

Theory

Core formulation

As Magnus summarises it, "an HPC is a structured repetition of properties in the world, maintained by an underlying causal process". Boyd's formulation has several components:

1. There is a family of properties that are contingently clustered in nature, in the sense that they co-occur in an important number of cases.
2. This co-occurrence is "at least typically, the result of what may be metaphorically (sometimes literally) described as a sort of homeostasis". Either the presence of some properties tends to favour the presence of others, or there are underlying mechanisms or processes that maintain the cluster, or both.
3. The homeostatic clustering of the properties is causally important: significant effects are produced by the joint occurrence of many of the properties together with the underlying mechanisms.
4. The kind term has no analytic definition. Rather, the property cluster together with the mechanisms that maintain it provide a natural definition. Which properties and mechanisms belong in the definition is an a posteriori theoretical question, not an a priori conceptual one.

Imperfect homeostasis is normal: some members of a kind may display some but not all of the properties in the cluster, and some but not all of the relevant mechanisms may be operating. Cases of extensional indeterminacy – borderline cases where no rational consideration dictates whether something falls under the kind – are an expected and irremovable feature, not a sign that the classification is defective.

The accommodation thesis

Central to Boyd's account is what he calls the accommodation thesis: that the naturalness of natural kinds consists in the contribution that reference to them makes to successful induction and explanation. Kinds are natural insofar as they allow scientists to "accommodate" their classificatory practices to the causal structure of the world. In Boyd's formulation, "the naturalness of natural kinds consists in their aptness for induction and explanation". The properties in an HPC cluster are thus required to be projectable: knowing that something belongs to the kind licenses inferences about properties not yet observed.

Magnus notes that this makes the accommodation thesis an answer to what he calls the taxonomy question – what distinguishes natural kinds from arbitrary categories – while HPCs themselves answer the ontology question of what holds particular kinds together.

Vagueness and historicity

Two consequences of the HPC account distinguish it from essentialist views of natural kinds.

First, HPC kinds are inherently vague. Because no single property is individually necessary for kind membership, the boundary of the kind cannot be drawn sharply. Boyd argued that no refinement of usage that replaces a vague kind term with a more precise one can preserve the naturalness of the kind: any such refinement would either treat as important distinctions that are irrelevant to causal explanation or ignore similarities that are important to it.

Second, HPC kinds are historical. The property cluster that defines a kind is not individuated extensionally (by its current membership) but like a historical entity: certain changes over time in the property cluster or in the underlying mechanisms preserve the identity of the kind, while others do not. This means that the properties determining membership in a kind can shift over time while the kind retains its identity, a feature Boyd took to be essential for handling biological species, which evolve.

Boyd noted that the HPC account is self-applying: "The kind natural kind is itself a natural kind in the theory of our inferential practice."

Applications

Biology

Biological species were Boyd's paradigm case for HPC kinds. Members of a species typically share a cluster of morphological, behavioural, physiological, and genetic properties, but none of these properties is individually necessary or sufficient for membership. The co-occurrence of these properties is maintained by homeostatic mechanisms including gene flow within populations, shared developmental pathways, and common ecological constraints. Boyd argued that species are "paradigmatic natural kinds, their historicality and lack of sharp boundaries notwithstanding".

The species case also illustrates the accommodation thesis. Biologists have proposed competing species concepts – the biological species concept based on interbreeding, phylogenetic species concepts based on shared ancestry, and ecological species concepts based on adaptive zone – each serving different explanatory projects. Boyd argued that this plurality is expected on the HPC account: different ways of demarcating species correspond to different accommodation demands and can be equally legitimate, since each tracks causally relevant structure for its own purposes.

The extension of the account to biological kinds was developed further by Robert Wilson, Matthew Barker and Ingo Brigandt, who argued that the view remained underdeveloped and was often misrepresented by its critics.

Psychiatry and medicine

The most developed application outside biology has been to psychiatry. Kenneth Kendler, Peter Zachar and Carl Craver argued in 2011 that psychiatric disorders are best understood as mechanistic property cluster (MPC) kinds, a variant of Boyd's account. Rejecting essentialist, socially constructed, and merely "practical" models of psychiatric classification, they held that MPC kinds are "defined not in terms of essences but in terms of complex, mutually reinforcing networks of causal mechanisms" operating at multiple levels, and that, like species, disorders form populations with central paradigmatic and more marginal members. A 2026 review in JAMA Psychiatry revisited the proposal as a response to the long-standing problems of heterogeneous diagnostic categories, high comorbidity, and limited interrater reliability in successive editions of the DSM.

The account has also been applied to disease more generally. Chloé de Canson has argued that because disease pathophysiologies are mechanisms, the property clusters that individual diseases form are homeostatic in Boyd's sense, and that the standard objections to HPC theory should not deter its application to disease.

Other proposed applications

Beyond biology, the HPC account has been discussed in relation to kinds in psychology and chemistry. In the philosophy of language, J. T. M. Miller argued in 2021 that individual words (considered as types rather than tokens) are HPC kinds, drawing an explicit analogy with biological species. The question of whether and how the account extends beyond biology has been a recurring theme in the literature.

Reception

Ian Hacking, responding to Boyd in 1991, called the theory "the best recent contribution to the doctrine of natural kinds", and argued that it gave a natural explanation of older ideas in the tradition of Mill and Russell. In the same discussion Hacking objected that HPC kinds should not be equated with Wittgensteinian family resemblances, which were coined in deliberate opposition to terms naming biological species, and resisted extending the account to human kinds.

Magnus reports that by 2010 Richard Samuels and Michael Ferreira could write that philosophers of science had "reached a consensus – or as close to consensus as philosophers ever get – according to which natural kinds are Homeostatic Property Clusters". Ereshefsky and Reydon likewise describe HPC theory as "becoming the received view of natural kinds in the philosophy of science", listing Paul Griffiths, Robert Wilson, Ingo Brigandt and Ron Mallon among philosophers who have adopted it, and noting that it has also been taken up within biology itself.

Criticisms and refinements

Scope of the account

Magnus (2014) argued that the HPC account is often misconstrued as a definition of natural kinds, when it is better understood as describing the ontological basis of some of them. He distinguished two questions: the taxonomy question (what distinguishes a natural kind from an arbitrary category?) and the ontology question (what features of the world hold a particular kind together?). The HPC account, Magnus argued, answers only the ontology question, and only for a subset of natural kinds.

The argument turns on cases that pull apart in each direction. Electrons are natural kinds but not HPCs: their properties are not maintained by homeostatic mechanisms but hold as a matter of nomic necessity. Conversely, some trivial HPCs – clusters of properties held together by causes – do not constitute natural kinds in any scientifically useful sense. Magnus concluded that Boyd's own formulation already accommodates this point, since Boyd described HPCs as "a class of natural kinds" and "an important family of natural kinds" rather than equating the two.

The mechanism problem

Several critics have raised concerns about the role of causal mechanisms in the HPC account. Craver (2009) argued that individuating mechanisms is itself partly conventional: "human perspectives and conventions enter into judgments about how mechanisms should be typed and individuated". If the mechanisms that sustain a property cluster cannot be objectively delineated, then the kinds they define may depend partly on human convention, undermining the realist credentials of the account. Craver noted that "one can be led to lump or split the same putative kind in different ways depending on which mechanism one consults".

Yukinori Onishi and Davide Serpico (2022) examined two attempts to reformulate the theory so as to avoid this difficulty: one that drops any reference to the causes of the clustering, and one that replaces homeostatic mechanisms with causal networks represented by causal graphs. They argued that the first is too thin to account for some inductive inference in science, while the second invites the same interest-relativity as the original, and concluded that human interest is likely to be an ineliminable part of any satisfactory account of natural kindness. Emma Tobin has argued that mechanisms face an overlapping problem parallel to the one long discussed for natural kinds themselves.

Fixing kind membership

Thomas Reydon (2009) argued that HPC theory cannot say what makes an entity a member of a kind in the first place. Because the theory identifies a kind by the properties that are found to cluster in its members, it "can only account for kinds the extensions of which have already been fixed independently by other means" – in the case of species, by an organism's location on a branch of the tree of life. A theory of natural kinds should also supply criteria by which kind membership can be determined, and on that count, Reydon held, the theory fails. His proposed repair adds to the definition of a kind the factor that makes its members explanatorily interesting, such as the capacity to perform a particular causal role, which allows functional kinds used in mechanistic explanation to count as kinds.

Mismatch with scientific classifications

Marc Ereshefsky and Reydon (2015) argued that HPC theory "neglects many kinds highlighted by scientific classifications while at the same time endorsing kinds rejected by science". They identified three types of scientific kind that fall outside its scope: non-causal kinds, such as the microbial species of the phylo-phenetic species concept, whose defining parameters include no causal mechanisms; functional kinds; and heterostatic kinds, whose members exhibit persistent differences rather than shared similarities.

The heterostatic case bears directly on Boyd's paradigm. Stable polymorphism – sexual dimorphism, insect metamorphosis, the life stages of a slime mould – occurs in virtually every biological taxon, yet HPC theory attends only to the similarities among members and the mechanisms producing them. Ereshefsky and Reydon argued that treating such variation as "conditionally specified dispositional properties", as Boyd and Wilson and colleagues had suggested, does not adequately capture it. In place of HPC theory they proposed an account built on classificatory programs, judged by criteria including internal coherence between a program's sorting principles and its motivating principles.

Stable property cluster account

Slater (2015) proposed the stable property cluster (SPC) account as a refinement of HPC theory. In deference to its Boydian origins, the account keeps the property cluster but, "[r]ather than emphasizing homeostasis or causal mechanism, the SPC account emphasizes the stability of a property cluster over the various ways stability may be maintained". Slater described the change as relinquishing causal language in favour of the idiom of stability, homeostatic causal mechanisms being only one way to achieve the stability that a cluster of properties may have.

He gave several reasons for the shift: that an account of kinds is better off without controversial commitments about the nature of causal mechanism; that homeostatic mechanisms are neither sufficient nor necessary for accommodating scientific practice; and that Boyd's notion of homeostasis is left vaguely defined, sometimes used literally and sometimes as a metaphor. The necessity argument turns on cases such as fundamental particles, which possess their properties as a matter of nomic necessity rather than through any sustaining mechanism, and on kinds above the species level: if the whole genus Drosophila is a natural kind, no propensity to interbreed holds its property cluster together, since its sibling species are reproductively isolated. Slater also suggested that biological taxa may exhibit a form of metastability, comparable to the slow spontaneous transition of diamond to graphite, such that stability on short timescales need not indicate homeostatic mechanisms at work. On the resulting picture, essentialist kinds are a limiting case within a single general account of natural kinds rather than a categorically different type of kind.

Eduardo Martinez (2017) defended the mechanism requirement against this refinement, arguing from two test cases in academic medicine that grounds are genuinely explanatory of scientific epistemic practices and that the SPC account should not do without them.

Causal-network accounts

Muhammad Ali Khalidi has proposed a related alternative on which natural kinds are nodes in causal networks. Taking the projectibility of kind terms as a starting point, he argues that kinds are not mere concatenations of properties but ordered hierarchies, whose instances stand to one another as causes and effects in recurrent causal processes, with clusters of core causal properties giving rise to clusters of derivative ones.

See also

• Natural kind
• Essentialism
• Philosophy of biology
• Species problem
• Family resemblance
• New riddle of induction

Further reading

• Boyd, Richard (1999). "Kinds, Complexity and Multiple Realization". Philosophical Studies. 95 (1–2): 67–98. doi:10.1023/a:1004511407133.
• Kornblith, Hilary (1993). Inductive Inference and Its Natural Ground: An Essay in Naturalistic Epistemology. Cambridge, MA: MIT Press. doi:10.7551/mitpress/3730.001.0001. ISBN 978-0-262-27746-4.
• Millikan, Ruth Garrett (1999). "Historical Kinds and the "Special Sciences"". Philosophical Studies. 95 (1–2): 45–65. doi:10.1023/A:1004532016219.
• Hacking, Ian (1991). "A Tradition of Natural Kinds". Philosophical Studies. 61 (1–2): 109–126. doi:10.1007/BF00385836.