Skip to content

Criteria of Individuation

Prime #
761
Origin domain
Philosophy
Subdomain
metaphysics and ontology → Philosophy
Also from
Computer Science & Software Engineering, Law & Governance, Biology
Aliases
Individuation Criteria, Principles of Individuation

Core Idea

Criteria of individuation are the rules a system fixes for what makes something one entity, and they are the shared structure beneath three questions that any system maintaining an inventory of countable things is forced to answer. First, composition: when does a collection of parts count as one whole rather than as many — what relation, at what threshold, confers oneness on a heap (the unity_test)? Second, co-reference: when do two presentations count as the same already-individuated entity rather than as two resembling ones — what operational criterion resolves any candidate pair as same-or-distinct (the identity_test)? Third, persistence-kind: among the categories a thing satisfies, which one fixes what it takes to remain the same thing over time — which category an instance cannot survive losing (the identity_providing_kind)? The prime names the umbrella: the principles a system uses to carve its domain into countable individuals, of which composition, co-reference, and persistence-kind are the three load-bearing sub-criteria.

The defining commitment — the one each sub-criterion inherits — is that there is no individuation-independent inventory of individuals waiting to be read off the world. The criteria partly constitute what is counted: a different unity test yields a different population of wholes, a different identity test yields a different count of distinct entities, a different identity-providing kind yields a different verdict on what survives a change. To individuate is to choose (or discover) the rules, and the choice fixes the answer to "how many things are there, which two are the same, and which one persisted?" The structural signature distinguishes individuation from mere description (which assigns properties to already-individuated things), from classification (which sorts already-individuated things into kinds), and from measurement (which summarizes over them): all three presuppose that the individuals already exist, while individuation is the prior act that produces the individuals the others operate on. This is the load-bearing line against the nearest non-child neighbor, essentialism. Essentialism asks what defining properties make a thing the kind of thing it is — a qualitative, modal question about natures, often discoverable a posteriori. Individuation asks the numerical questions — how many ones, which two are the same one, what one must remain to stay one — and answers them with operational tests (a unity threshold, a sameness criterion, a cease-to-exist test) that bare essentialism never supplies. An essence can be perfectly specified while the count is still open: knowing water is H₂O fixes no answer to "how many bodies of water are in this watershed," and knowing the essence of person does not by itself tell a registry which two records co-refer. Individuation is essentialism's missing arithmetic.

The single most consequential fact the prime names is the three-stage dependency: the sub-criteria run in a fixed order, and each presupposes the last. Composition must fix what counts as one whole before co-reference can ask whether two presentations are the same whole; co-reference must settle sameness before persistence can ask whether the same whole survived a change. Get the order wrong — resolve identity over collections whose unity boundaries differ across sources, or assess persistence over an entity whose unity was never fixed — and the question is ill-posed, not merely hard. What criteria_of_individuation provides as a prime is the recognition that "what is one thing here?" decomposes into exactly these three rules, that the rules are constitutive rather than descriptive, and that they compose into a pipeline — carve wholes, re-identify them, track their persistence — that any inventory-bearing system runs whether it names it or not.

How would you explain it like I'm…

One Thing Or Many?

Look at a pile of LEGO bricks. Is that ONE thing, or LOTS of things? And if you see your toy car today and tomorrow, how do you know it's the SAME car? Criteria of Individuation are the rules we use to decide what counts as one thing, when two things are really the same thing, and when something stays itself even after it changes.

What Makes One Thing

Whenever you count things, you secretly use rules. First, when do a bunch of parts make ONE whole instead of many parts, like when does a heap of sand become 'a sandcastle'? Second, when are two things you see actually the SAME one thing, not two look-alikes? Third, what can a thing change and still be itself, versus what change would make it stop being that thing? Criteria of Individuation are exactly these rules for carving the world into countable things. The surprising part: the rules you pick actually change the answer to 'how many are there?'

Rules For Counting Ones

Criteria of Individuation are the rules a system uses to carve its domain into separate countable individuals, and they cover three linked questions. Composition: when do parts add up to one whole rather than many? Co-reference: when are two appearances the same already-counted entity rather than two similar ones? Persistence-kind: which category fixes what a thing must keep to remain the same thing over time? The key commitment is that there's no individual-count just sitting in the world waiting to be read off; the rules you choose partly constitute what gets counted, so a different rule yields a different population. This is not the same as essentialism, which asks what defining properties make something the kind of thing it is. Individuation asks the numerical questions and answers them with operational tests, the arithmetic that essentialism leaves out.

 

Criteria of individuation are the principles a system fixes for what makes something one entity — the shared structure beneath three questions any inventory-bearing system must answer. Composition asks when a collection of parts counts as one whole rather than many (the unity test). Co-reference asks when two presentations are the same already-individuated entity rather than two resembling ones (the identity test). Persistence-kind asks which of the categories a thing satisfies fixes what it takes to remain the same thing over time (the identity-providing kind). The defining commitment is that there is no individuation-independent inventory of individuals waiting to be read off the world: the criteria partly constitute what is counted, since a different unity test yields a different population of wholes, a different identity test a different count, a different persistence-kind a different verdict on what survives. This separates individuation from description (which assigns properties to already-individuated things), classification (which sorts them), and measurement (which summarizes them) — all three presuppose the individuals exist. It also separates individuation from essentialism, its nearest neighbor: essentialism asks what defining properties make a thing its kind (a qualitative, modal question), while individuation asks the numerical questions and answers them with operational tests. Crucially the three sub-criteria run in a fixed dependency order — carve wholes, then re-identify them, then track persistence — and getting the order wrong makes the question ill-posed rather than merely hard.

Structural Signature

the domain of parts and presentations to be carved — the composition criterion (one-whole-from-parts) — the co-reference criterion (same-entity-across-presentations) — the persistence-kind criterion (what-fixes-identity-over-time) — the constitutive invariant (the criteria partly determine the inventory) — the ordered-dependency invariant (compose, then re-identify, then persist)

Criteria of individuation are present when each of the following holds:

  • A domain to be carved (the raw material). A field of parts, presentations, states, or stages that does not arrive pre-sorted into countable individuals — cells and tissues, records and observations, subsidiaries and branches, stages of a life, tokens in memory — over which an inventory of individuals must be imposed.
  • A composition criterion (the unity rule). A rule fixing when parts cohere into one whole rather than many: the unifying relation (causal coupling, a boundary, common governance, a shared scope) and the threshold on it that confers oneness. This is the unity_test sub-criterion.
  • A co-reference criterion (the identity rule). A rule fixing when two presentations refer to the same already-individuated entity: the kind at issue, the operational sameness test, and its equivalence regime. This is the identity_test sub-criterion.
  • A persistence-kind criterion (the sortal rule). A rule fixing which category an instance belongs to for as long as it exists and so supplies its persistence criteria — separated from the roles and phases it merely occupies for a while by the cease-to-exist test. This is the identity_providing_kind sub-criterion.
  • The constitutive invariant (the load-bearing fact). The criteria do not read off a pre-existing inventory; they partly make it. Different criteria yield different counts, different sameness verdicts, and different persistence judgments over the identical raw material, so disputes about "how many" or "the same" resolve into disputes about which criteria are in force.
  • The ordered-dependency invariant (the pipeline). The sub-criteria are order-sensitive: composition fixes the wholes, co-reference re-identifies those wholes, persistence tracks them through change — and running them out of order (identity over unsettled unity, persistence over unsettled identity) makes the question ill-posed.

The components compose into a single object — the rule-set by which a system turns an uncarved domain into an inventory of countable, re-identifiable, persisting individuals — and it is the constitutive-plus-ordered pairing that generates everything downstream: that the inventory is an artifact of the criteria, and that the three sub-criteria must be resolved in sequence or not at all.

What It Is Not

  • Not the unity_test alone. The unity test is one of the three sub-criteria — the composition rule, fixing one-whole-from-parts. Criteria of individuation is the umbrella that also includes co-reference and persistence-kind. Naming only unity answers "are these parts one thing?" but leaves "is this the same thing as that?" and "what does this thing have to keep being to exist?" untouched.
  • Not the identity_test alone. The identity test is the co-reference sub-criterion, resolving same-or-distinct over already-individuated entities. It presupposes that composition has fixed what counts as one entity, and it says nothing about which category supplies persistence. It is one leg of the tripod, not the whole.
  • Not the identity_providing_kind alone. The identity-providing kind is the persistence sub-criterion, fixing which category an instance cannot survive losing. It presupposes that the entity has been individuated and re-identified; it is the third leg, governing survival-through-change, not composition or co-reference.
  • Not classification. Classification sorts already-individuated entities into kinds (what type is this?). Individuation is the prior act that produces the entities classification then sorts — you must have carved a countable individual before you can ask what kind it is. Classification presupposes individuation; it cannot supply it.
  • Not essentialism. Essentialism names necessary defining properties broadly. Individuation isolates the specifically numerical questions — oneness, sameness, persistence — and supplies operational tests (unity threshold, sameness criterion, cease-to-exist test) that bare essentialism lacks. Not every essential property is individuating, and individuation includes composition and co-reference questions essentialism does not address.
  • Not ontology as a whole. An ontology is a full specification of categories and relations for a domain. Criteria of individuation is one load-bearing distinction within an ontology — the rules by which it counts individuals — not the whole map. The ontology is the territory's specification; individuation is its counting discipline.
  • Common misclassification. Treating an inventory of individuals as a fact read off the world rather than as an artifact of the criteria that produced it — and so disputing "how many things are there" as if it had a criterion-independent answer. Catch it by asking which composition, co-reference, and persistence rules are in force: where two parties' counts diverge, they almost always hold different individuation criteria over the same domain, and the dispute dissolves once the criteria are named.

Broad Use

Criteria of individuation, read as the rule-set that carves a domain into countable individuals, recur wherever a system maintains an inventory of things it can count, re-identify, and track. In ontology engineering the three sub-criteria are made first-class simultaneously: OntoClean tags classes with whether they carry a unity criterion (integral whole, plural whole, or none), an identity criterion (what fixes sameness), and a rigidity / identity-providing meta-property (what an instance cannot survive losing) — and the methodology's central result is that hierarchies smuggle in incompatible individuation assumptions that cause inferential errors when unmasked. In biology, defining a countable organism for a colonial species, a holobiont, or a modular plant requires all three at once: a unity criterion (metabolic, reproductive, or genetic integration) to fix what is one organism, an identity criterion to re-identify it across observations, and a persistence kind to say what it must remain through metamorphosis or cell turnover — and the same domain yields different counts under different criteria (an aspen clone is one genetic individual and thousands of trunks; a Portuguese man-o'-war is one siphonophore and many zooids). In fundamental physics the question reaches its sharpest form: identical particles are indistinguishable in principle, so the classical identity criterion ("the same one if it has the same trajectory") fails — quantum statistics (Bose–Einstein, Fermi–Dirac) presuppose that permuting two electrons yields no new state, which is to deny them a haecceity, a this-ness over and above their shared properties; the inventory of "how many distinct particles" is fixed by the symmetrization rule, not read off labels the particles do not carry. In law, legal and corporate personhood is an individuation regime: a unity rule (which subsidiaries consolidate into one entity), an identity rule (which records refer to the same party for liability), and a persistence kind (the corporation survives changes of officers, name, and address while a partnership may dissolve on a partner's exit). In database and information systems, the three criteria are engineered explicitly: a primary key is a stipulated identity criterion (two rows are the same entity iff they share a key), entity resolution and master-data management reconstruct co-reference where no shared key exists, transaction and document boundaries fix unity, and the chosen primary kind fixes persistence — together determining how many distinct customers, patients, or documents the system believes exist. In accounting, what counts as one transaction is a unity rule (a single sale, or a sale plus its shipping and tax as one composite entry?), reconciliation is an identity rule (which bank line and which ledger line are the same event), and the reporting entity is a persistence kind — and audit failures are individuation failures: one economic event split into two records (double-count) or two events merged into one. In software and type systems, object identity (reference, structural, value, content-hash equality), aggregate boundaries, and class-for-life versus temporary state all instantiate the three rules, and the classic bugs are individuation errors — a "set" that double-counts (broken co-reference), a role modeled as a type (broken persistence kind). In logic and set theory the criteria become axioms: Leibniz's identity of indiscernibles is an identity criterion (no two distinct things share all properties), the axiom of extensionality is a unity-and-identity rule for sets (a set just is its members; two sets with the same members are one set), and a urelement-versus-set distinction is a persistence-kind choice. In historiography, periodization fixes the unity of an event (battle, campaign, war), source criticism fixes whether two chronicles name the same event, and the genre of the entity fixes what counts as its continuation. Across all of these the recurring fact is identical: a domain that does not arrive pre-individuated, three rules that carve it into countable individuals, and an inventory that is the joint product of those rules rather than a reading of the world.

Clarity

Naming criteria of individuation separates two questions that practitioners across fields routinely fuse: how many things are there, and which are the same and persisting? — which feels like a question about the world — and which composition, co-reference, and persistence rules are we applying? — which is a question about the system's own conventions. The clarifying force of the prime is to convert "there are N organisms / customers / events here" into the testable claim "under these unity, identity, and persistence criteria there are N; under others there would be a different count," relocating an apparently factual dispute to the criteria that produced the count. From inside a system the inventory simply is the set of things, and the constitutive role of the criteria is invisible; the prime makes it explicit and forces the criteria to be stated rather than assumed.

The prime also clarifies a recurring class of confusions by exposing the order the sub-criteria must run in. Practitioners conflate the three constantly — they try to re-identify entities whose unity is unsettled ("is this the same company?" before fixing which subsidiaries are one company), or assess persistence over entities whose identity was never resolved, or treat a classification disagreement as an identity disagreement. Naming the umbrella makes the pipeline legible: composition first (carve the wholes), then co-reference (re-identify them), then persistence (track them through change). It also separates individuation cleanly from its three look-alikes — description, classification, and measurement — each of which presupposes that the individuals already exist. The clarifying move throughout is to show that "what is one thing here?" is not one question but three ordered, constitutive rules, and that most durable disputes about counting, sameness, and survival are disputes about which rules are in force, not about the world the rules are applied to.

Manages Complexity

Criteria of individuation compress an unbounded, case-by-case adjudication of "what is one thing?" into a maintained rule-set with three fixed parts. The complexity reduction is large because the prime replaces per-case argument with a reusable discipline: rather than relitigating, for every cluster of parts, whether it is one whole; for every pair of presentations, whether they co-refer; and for every change, whether the entity survived — the system fixes a unity rule, an identity rule, and a persistence kind once, and applies them many times. The decision moves from the case level to the rule level, and disagreements are escalated only when a rule itself is challenged, which converts an open-ended ontological negotiation into a bounded one over three named criteria.

The compression is also what licenses whole-level operations to run consistently. Once the three criteria are fixed, counting, persisting, merging, holding accountable, and destroying can all range over the same individuals, with parts and presentations inherited, rather than being redefined ad hoc on loose collections — and the deepest complexity gain is self-consistency: a single individuation regime guarantees that the system does not silently disagree with itself about how many things there are, which two are the same, and which persisted, because counting, re-identification, and persistence all draw on one rule-set. The prime further sorts the cost structure of getting the criteria wrong, and it does so per sub-criterion: a too-permissive unity rule pads the inventory with spurious wholes while a too-restrictive one fragments genuine ones; a too-loose identity rule false-merges distinct entities while a too-strict one false-splits one entity across phantoms; a mis-chosen persistence kind makes an entity vanish when a mere role ends. Because the prime names the three rules as discrete, tunable objects with directional error costs, an open-ended "why does this inventory keep contradicting itself?" reduces to three operating-point decisions, each with a known asymmetry.

Abstract Reasoning

The criteria-of-individuation pattern licenses several substrate-independent moves. Decompose "what is one thing?" into three ordered rules: faced with any inventory question, the reasoner separates the composition question (one-whole-from-parts), the co-reference question (same-across-presentations), and the persistence question (what-fixes-identity-over-time), and resolves them in that order — because each presupposes the last, and conflating them is the commonest source of ill-posed individuation disputes. Treat the inventory as criterion-relative: any count of individuals is partly constituted by the rules that produced it, so the reasoner reads "there are N of them" as "there are N under these criteria" and, when two counts diverge, looks first for a criteria mismatch rather than a factual disagreement. Locate a dispute on the right leg: when parties disagree about sameness, the reasoner checks whether they actually disagree about unity (what counts as one entity) or about the kind (which category is at issue) before checking the identity criterion itself, because a unity or kind mismatch masquerades as an identity-test disagreement and cannot be fixed by tuning the identity test. Price each criterion's error asymmetry: rather than seeking a single "correct" individuation, the reasoner sets each sub-criterion's operating point by weighing its directional costs — spurious-versus-fragmented wholes, false-merge-versus-false-split, role-mistaken-for-kind — because the right criteria are dictated by cost asymmetry, not by a criterion-independent fact. And run the cease-to-exist and out-of-band checks: to fix the persistence kind, ask whether the instance survives losing a category (if so it is a role, not the identity-providing kind); to validate co-reference, measure sameness against a source independent of the criterion being tested — portable diagnostics that travel unchanged from a database to a taxonomy to a legal register.

Knowledge Transfer

Because criteria of individuation are the bare rule-set for carving a domain into countable individuals, a discipline built around them in one field transfers to any other by re-identifying the domain, the unity rule, the identity rule, and the persistence kind — and the prime's reach is the reach of that one tripod across substrates. The OntoClean discipline of stating all three criteria as first-class meta-properties transfers from ontology engineering to clinical microbiome research (treating host-plus-microbiota as the individuated unit of analysis, which is a joint choice of unity and persistence kind) and to corporate accounting (consolidation as a unity rule, the reporting entity as a persistence kind, party-matching as an identity rule), because in each the move is identical: name the three rules and check they are mutually compatible. The entity-resolution discipline from databases — fix the kind, fix the sameness test, fix the equivalence regime, then audit false-merge and false-split rates — transfers verbatim to epidemiology (building longitudinal cohorts from heterogeneous records), to historiography (deciding when two chronicles name the same battle), to taxonomy (molecular tests for "same individual"), and to accounting reconciliation (matching a bank line to a ledger line), because all face the same co-reference problem over a domain whose unity must already have been fixed. The sortal discipline of the cease-to-exist test — separate the identity-providing kind from the roles and phases an instance merely occupies — transfers from metaphysics to software domain modeling (make the underlying kind the type and demote Customer/Employee to roles, so deactivating an account does not annihilate the person) and to law (the natural or corporate person persists across every role), because the failure mode (role mistaken for kind, entity vanishes when the role ends) is identical across substrates. The transfer even reaches the formal sciences, where the criteria appear not as engineering heuristics but as axioms: a set theorist's axiom of extensionality and a database designer's primary key are the same move — stipulate the sameness criterion that makes two presentations one entity — and the identity of indiscernibles a logician debates is the limiting case the physicist confronts empirically when two electrons share every property yet the formalism must still say how many there are. In every transfer the practitioner runs the same three-stage diagnosis — carve the wholes by a unity rule, re-identify them by an identity rule, track them by a persistence kind, in that order — and the transfer is secure because none of the three rules names a substrate: a database architect resolving customer records, a biologist delimiting a holobiont, a lawyer fixing corporate continuity, an auditor deciding what is one transaction, a physicist counting indistinguishable particles, and a historian periodizing a war are running the identical individuation pipeline, distinguished only by what the domain is and what content fills the three rules. What stays substrate-specific is only the content of each rule (which relation confers unity, which property fixes sameness, which category supplies persistence); the structure — three ordered, constitutive criteria — is portable without modification.

Examples

Formal/abstract

OntoClean is the prime in its purest worked form, because it makes all three individuation criteria first-class and shows them operating jointly to validate an ontology. Take a domain of categories to be organized into a subsumption hierarchy — the raw material the criteria must carve. OntoClean attaches to each category three meta-properties that are exactly the prime's three sub-criteria. Unity (+U/−U) is the composition criterion: it marks whether a category's instances are wholes unified by a single relation (an integral whole like an organism) or not (a mere amount of matter), fixing one-whole-from-parts. Identity (+I/−I) is the co-reference criterion: it marks whether a category carries a criterion for re-identifying its instances across presentations, fixing same-or-distinct. Rigidity (+R/−R) is the persistence-kind criterion: a rigid category (like person) is one its instances cannot stop belonging to without ceasing to exist, while an anti-rigid one (like student) is a role they shed — exactly the cease-to-exist test that fixes what-supplies-identity-over-time. The constitutive invariant is operative: the same domain, tagged with different meta-properties, yields a different inventory of well-formed individuals and a different set of admissible subsumptions. The ordered-dependency invariant is what gives OntoClean its bite: its formal constraints forbid, for instance, a rigid category subsuming an anti-rigid one, and an anti-rigid category supplying identity to a rigid one — constraints that are violated precisely when a modeler resolves the criteria out of order or assigns incompatible ones, smuggling a hidden individuation assumption into the hierarchy that produces inferential errors downstream. The structural payoff is that "is this ontology well-formed?" reduces to "are its unity, identity, and persistence criteria each assigned and mutually compatible?" — the three-criteria umbrella made into an executable validation.

Mapped back: OntoClean instantiates every component — a domain of categories (the raw material), a unity meta-property (composition), an identity meta-property (co-reference), a rigidity meta-property (persistence-kind), different taggings yielding different admissible inventories (the constitutive invariant), and ordering constraints among the meta-properties (the ordered-dependency invariant) — and shows the prime's core claim: an inventory of well-formed individuals is the joint product of three ordered, constitutive criteria.

Applied/industry

A health system building a longitudinal patient record runs the same three criteria in a data-engineering substrate, and discovers that getting any one wrong corrupts the count. The raw material is a stream of records arriving from admissions, the lab, the pharmacy, and outside referrals — none pre-individuated into patients. The composition criterion fixes what counts as one patient-episode versus one patient versus one care-team: a single admission, a readmission, a multi-site encounter — the unity rule that decides which records belong to one clinical whole before any matching runs. The co-reference criterion is the master-patient-index matching rule — deterministic on a national identifier where present, probabilistic on name, date of birth, and address where not — resolving which presentations are the same patient, with transitive closure as its equivalence regime (so one over-permissive pairwise match can cascade unrelated records into a phantom patient). The persistence-kind criterion fixes that the individuated entity is a person, not a patient (a role) or an active case (a phase): modeling the role as the kind is the canonical bug, where discharging a patient or deactivating a record makes the person's history vanish — the cease-to-exist test averts it. The constitutive invariant is concrete and consequential: a tax authority and a marketing department resolving the same customer database hold different unity rules (household versus individual) and different identity tests (strict identifier-anchored versus loose household-level), and each produces a different, correct-for-its-purpose count from identical data — the inventory is an artifact of the criteria, not a reading of the world. The ordered-dependency invariant bites in practice: trying to re-identify "the same patient" across two sources that individuated the patient differently (one as the person, one as the per-encounter record) is ill-posed until the unity rule is shared, and the matching keys reference incommensurable units. The prime's discipline is the fix — fix the unity rule, then the identity rule, then the persistence kind, and tune each against its own directional error cost (a false-merged allergy can kill; a fragmented record hides a contraindication; a role-as-kind loses the person).

Mapped back: The patient-record system runs the prime end-to-end — an uncarved record stream (raw material), an episode/patient/team unity rule (composition), an MPI matching rule with transitive closure (co-reference), person-as-the-identity-providing-kind (persistence), tax-versus-marketing yielding different correct counts (the constitutive invariant), and ill-posed cross-source matching until unity is shared (the ordered-dependency invariant) — and demonstrates the transfer: an ontology engineer, a database architect, and a clinician are running the same three-criteria individuation pipeline, distinguished only by what fills the three rules.

Structural Tensions

T1 — Criteria-Constituted versus Criteria-Independent Inventory (the Constitutive Tension). The prime's foundational tension is between treating the inventory of individuals as made by the criteria and treating it as a fact read off the world. The failure mode is inventory realism — disputing "how many organisms / customers / events are there?" as if it had a criterion-independent answer, so two parties argue endlessly about the world when they hold different unity, identity, or persistence rules. Diagnostic: ask whether the count would change under different individuation criteria; if it would, the dispute is about which criteria are in force, not about the world, and is settled by naming and agreeing the rules — not by gathering more data about the domain.

T2 — Ordered Pipeline versus Entangled Criteria (the Sequencing Tension). The three sub-criteria run in a fixed order — composition, then co-reference, then persistence — but in practice they are entangled and applied simultaneously, so a downstream criterion is resolved over an upstream one that was never fixed. The failure mode is out-of-order individuation: re-identifying "the same entity" across sources whose unity boundaries differ, or assessing persistence over an entity whose identity was never settled, producing an ill-posed question that no amount of effort on the downstream criterion can answer. Diagnostic: confirm the composition criterion is fixed and shared before applying the identity criterion, and the identity criterion before the persistence criterion; where an upstream criterion is unsettled, the downstream question is not hard but ill-posed.

T3 — Discovered versus Stipulated Criteria (the Realism Tension). Some individuation criteria seem discovered (nature makes organism the unit for a living thing) and others are plainly stipulated (a legal regime decrees what counts as one corporation; a schema elects an identity key). The tension is between reading the criteria as a fact about the domain and reading them as a design choice. The failure mode is misjudged necessity: treating a stipulated convention as a discovered necessity (so a defensible alternative individuation looks like an error) or treating a discovered natural criterion as freely stipulable (so a real persistence fact is overridden by convention). Diagnostic: ask whether anything outside the system's conventions fixes the criterion; where personhood or entity-continuity is conferred by rule, the criteria are designed and disputes about them are about what to stipulate, whereas a biological persistence criterion holds whether or not anyone tracks it.

T4 — Single Regime versus Layered Individuations (the Scopal Tension). A system often admits multiple legitimate individuation regimes over the same domain — a body is one organism biologically and body-plus-microbiome is one organism metabolically; a firm is one entity legally, another economically, another by brand. The tension is between fixing a single regime and tolerating coexisting ones. The failure mode is silent layer-mixing: counting under one regime, re-identifying under a second, and holding accountable under a third, so the system contradicts itself about how many individuals exist without noticing. Diagnostic: where regimes can be layered, require every operation (counting, matching, persistence, accountability) to declare which regime's criteria it uses; the danger is not having layers but using different ones without flagging it.

T5 — Tuning One Criterion versus Coupled Error Costs (the Coupling Tension). Each sub-criterion has a directional error cost — spurious-versus-fragmented wholes, false-merge-versus-false-split, role-versus-kind — and tuning one criterion can shift the burden onto another rather than reducing total error. The failure mode is local optimization: tightening the identity test to cut false merges, which fragments entities and pushes the cost onto the persistence criterion (now tracking phantoms), or loosening the unity rule to capture more wholes, which floods the identity test with near-duplicates. Diagnostic: price the three criteria's errors jointly, not one at a time; because the criteria compose into one pipeline, an operating-point change on any leg propagates downstream, and the binding constraint may be a criterion you did not touch.

T6 — Numerical Individuation versus Qualitative Description (the Boundary Tension). The prime governs the specifically numerical questions — oneness, sameness, persistence — and is constantly confused with the qualitative questions of description and classification that presuppose the individuals already exist. The tension is between fixing what the individuals are (individuation) and saying what they are like or which kind they fall under (description, classification). The failure mode is category drift: treating a classification disagreement ("are these records persons or households?") as an identity disagreement, and tuning the matching rule when the dispute is upstream at which kind is being individuated. Diagnostic: ask whether the parties have fixed the same individuals before comparing their properties or kinds; a kind or description mismatch masquerades as an individuation-criterion disagreement and cannot be resolved by adjusting the unity, identity, or persistence rule.

Structural–Framed Character

Criteria of individuation sit at the pure structural end of the structural–framed spectrum, with a frontmatter aggregate of 0.0 — every diagnostic reads zero, and the prime is a canonical structural prime: a bare mereological-and-identity skeleton (composition, co-reference, persistence-kind) carrying no normative or institutional content.

The carve-a-domain-into-countable-individuals rule-set is medium-neutral and demonstrably recurs across substrates. The pattern carries no home vocabulary that must travel (vocab_travels 0.0): the identical three-criteria question appears as unity/identity/rigidity meta-properties in ontology engineering, biological individuality for holobionts and modular organisms, particle indistinguishability in physics, consolidation and personhood in law, primary keys and entity resolution in databases, what-counts-as-one-transaction in accounting, extensionality and the identity of indiscernibles in logic, and periodization in history — each told in its own field's words, which is why an ontology engineer, a biologist, a physicist, a lawyer, a database architect, an auditor, and a historian are running the same individuation pipeline. It carries no evaluative weight (evaluative_weight 0.0): an individuation criterion is neither good nor bad — different unity or identity rules are each correct for their purpose, and the prime frames the choice as a value-neutral operating point with directional error costs, not a judgment. Its origin is formal-relational (institutional_origin 0.0), a bare metaphysical-and-mereological question with no institutional load, even though some of its applied cases are legal. It is not human-practice-bound (human_practice_bound 0.0): metabolic integration individuates a symbiosis and the persistence criteria of a biological kind hold with no human in the loop; the symmetrization rules that count indistinguishable particles hold in a star's interior with no observer present; and a BEGIN/COMMIT boundary confers unity on a transaction with no human practice required. And invoking it recognizes rather than imports (import_vs_recognize 0.0): to apply individuation criteria is to make explicit a counting discipline already implicit in any system that maintains an inventory of countable individuals, adding no interpretive frame.

The contrast with the prime's nearest neighbor underscores the structural read: where essentialism names necessary defining properties in general, criteria_of_individuation isolates the specifically numerical sub-questions — oneness, sameness, persistence — and equips each with an operational test (unity threshold, sameness criterion, cease-to-exist test) that bare essentialism lacks. The 0.0 aggregate is correct — a paradigm structural prime, recognized rather than translated wherever a domain is carved into countable individuals.

Substrate Independence

Criteria of individuation are about as substrate-independent as a prime can be — composite 5 / 5 on the substrate-independence scale. The signature — three ordered, constitutive rules (composition, co-reference, persistence-kind) by which a domain is carved into countable individuals — is stated in pure mereological-and-identity terms with no commitment to any medium, so it is recognized rather than translated when it surfaces in a new field, which earns structural abstraction a full 5. And it demonstrably recurs across substrates with the identical structure: the unity/identity/rigidity triad of OntoClean in ontology engineering; the delimitation of organisms, holobionts, and modular individuals in biology; the indistinguishability of identical particles and the symmetrization rules of quantum statistics in fundamental physics; consolidation, party-matching, and corporate continuity in law; primary keys, transaction boundaries, entity resolution, and identity-providing kinds in databases and software; what-counts-as-one-transaction and reconciliation in accounting; extensionality and the identity of indiscernibles in logic and set theory; and event periodization, source-matching, and continuation criteria in historiography — a domain breadth (5) spanning formal, physical, biological, legal, computational, financial, and historical substrates, about as wide as the catalog admits. The transfer is exact and recurs case after case (the same three-stage diagnosis carries from a database to a taxonomy to a particle ensemble to a legal register), and is well documented under the three specific sub-questions — OntoClean's hidden-individuation-assumption result, the competing-unity-criteria analysis in biology, the entity-resolution discipline in data systems, the haecceity debate in the metaphysics of physics. What holds transfer evidence at 4 rather than 5 is that the unity is more often catalogued under the three separate sub-questions (unity test, identity test, identity-providing kind) than under a single umbrella name, so the cross-substrate identity of the combined tripod is recognized rather than already documented as one pattern. Maximal abstraction and maximal breadth with strong (not maximal) single-name transfer put this among the catalog's canonical high structural primes.

  • Composite substrate independence — 5 / 5
  • Domain breadth — 5 / 5
  • Structural abstraction — 5 / 5
  • Transfer evidence — 4 / 5

Relationships to Other Primes

One-hop neighborhood: parents above, mutual partners to the right, children below.Criteria ofIndividuationsubsumption: Identity-Providing KindIdentity-Provid…subsumption: Identity TestIdentity Testsubsumption: Unity TestUnity Test

Foundational — no parent edges in the catalog.

Children (3) — more specific cases that build on this

  • Identity-Providing Kind is a kind of Criteria of Individuation

    child of emergent criteria_of_individuation

  • Identity Test is a kind of Criteria of Individuation

    child of emergent criteria_of_individuation

  • Unity Test is a kind of Criteria of Individuation

    child of emergent criteria_of_individuation

Neighborhood in Abstraction Space

Criteria of Individuation sits in a moderately populated region (55th percentile for distinctiveness): it has near-neighbors but no dense thicket of synonyms.

Family — Inference & Evidence (26 primes)

Nearest neighbors

Computed from structural-signature embeddings · 2026-06-14

Not to Be Confused With

The most important confusions are with the prime's own three children, because each is part of criteria_of_individuation rather than a synonym for it, and mistaking a part for the whole is the commonest error. Each child is recoverable as a special case of the genus by restricting attention to one sub-criterion and adding the machinery that sub-criterion alone needs — and naming that added machinery is the cleanest way to see why the child is narrower. The unity_test is the composition sub-criterion — the rule fixing one-whole-from-parts — and on top of the genus it adds a parts-and-unifying-relation apparatus: a candidate whole, the constituents, the relation they must bear, a threshold on that relation, and the whole-level predicates the resulting oneness licenses. The genus does not carry that apparatus; the unity_test specializes the genus to "the composition leg, equipped to confer wholeness on a heap." Answering it settles only how many wholes there are, not whether two presentations are the same whole or what a whole must remain to persist. The identity_test is the co-reference sub-criterion — the rule resolving same-or-distinct over already-individuated entities — and on top of the genus it adds an equivalence-regime-and-error-profile apparatus: a sameness criterion, the question of whether it is reflexive/symmetric/transitive, and measurable false-merge and false-split rates. It presupposes that composition has fixed what counts as one entity and says nothing about persistence; it specializes the genus to "the co-reference leg, equipped with a tunable matching rule." The identity_providing_kind is the persistence sub-criterion — the rule fixing which category an instance cannot survive losing — and on top of the genus it adds the cease-to-exist test and the single-bearer invariant: among the many categories an instance satisfies, exactly one supplies persistence, and the rest are roles or phases. It presupposes the entity has been individuated and re-identified; it specializes the genus to "the persistence leg, equipped to sort kinds from roles." Each child is one leg of a tripod with its own extra equipment; criteria_of_individuation is the tripod, carrying none of the three apparatuses but the composition and ordering of all three. A practitioner who reaches for only one child solves only one third of the individuation problem: name unity alone and you can count wholes but cannot re-identify or persist them; name identity alone and you re-identify entities whose unity is unsettled (ill-posed); name the persistence kind alone and you track the survival of an entity that was never carved or matched. The umbrella's distinctive content is exactly the composition of the three plus the ordered dependency among them, which no single child carries.

A second genuine confusion is with classification. Both produce groupings and both feel like "sorting the domain," but they operate at different levels and in a fixed order. Classification answers "what kind is this entity?" — it assigns already-individuated entities to types. Criteria of individuation answers the prior questions of "what is one entity, which two are the same entity, and what must it remain?" — the rules that produce the entities classification then sorts. The dependency is strict: you must have carved a countable individual (composition), be able to re-identify it (co-reference), and know what it must stay (persistence) before you can ask what kind it is. Conflating them hides where a disagreement lives: two parties whose counts differ may agree on every kind-assignment but have individuated differently (one treating records as persons, another as households), so the dispute is upstream at individuation, not in classification — and no amount of refining the type hierarchy resolves an individuation mismatch.

A third confusion — the one the brief flags as the nearest non-child neighbor, and the sharpest to draw — is with essentialism and the broader ontology it lives in. Essentialism holds that categories have necessary defining properties that fix what kind of thing something is; criteria of individuation isolates the specifically numerical sub-set of those questions — oneness, sameness, persistence-over-time — and, crucially, supplies operational tests for each (the unity threshold, the sameness criterion, the cease-to-exist test) that bare essentialism lacks, while also covering composition and co-reference questions essentialism does not reach. The gap is cleanest where the two come apart: a fully specified essence leaves the count open (knowing the essence of water settles nothing about how many lakes there are; knowing the essence of person tells a registry nothing about which two records co-refer), and conversely a working individuation criterion need carry no essence at all (a primary key individuates rows by fiat, with no claim about their natures). The limiting case is physics, where essentialism is silent but individuation is forced: two electrons share every intrinsic property — same mass, charge, spin — so there is no essence that distinguishes them, yet the formalism must still settle how many there are, which it does through a symmetrization rule rather than an essence. That is exactly the identity of indiscernibles pressed to its breaking point: where essentialism would say "same properties, therefore same thing," individuation must decide whether indiscernibles are nonetheless numerically two, and it answers with a stipulated rule, not a discovered nature. Not every essential property is individuating, and not every individuation criterion is an essence. Likewise an ontology is the whole specification of a domain's categories and relations, while criteria of individuation is one load-bearing discipline within it — the rules by which the ontology counts individuals (the unity, identity, and rigidity meta-properties operationalized by frameworks like OntoClean). Confusing the prime with essentialism-in-general or ontology-in-general lets a modeler think that naming a domain's categories and their essential properties has settled how the domain is counted, when the constitutive, ordered individuation question — what makes something one, the same, and persisting — is still untouched.

For a practitioner these distinctions decide where to intervene and in what order. Mistake the umbrella for one of its children and you solve a third of the problem and leave the inventory inconsistent. Mistake it for classification and you tune a type hierarchy when the disagreement was about which individuals exist. Mistake it for essentialism or ontology and you assume a complete category specification has fixed the count, when the numerical rules were never set. The unifying discipline is the prime's three-stage, order-sensitive check: fix the composition criterion (what is one whole), then the co-reference criterion (which presentations are the same whole), then the persistence-kind criterion (what the whole must remain) — and treat the resulting inventory as the artifact of those criteria, correctable by changing the rules rather than by gathering more facts about a world the rules were always going to carve.

Solution Archetypes

No catalogued solution archetypes reference this prime yet.