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:
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. 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. 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. 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.