Textual Definition¶

Attach to a term in a controlled vocabulary a stable natural-language statement of the conditions under which it is correctly applied, converting a privately-interpreted label into a publicly consultable boundary that serves as the appeal of last resort in annotation disputes.

Core Idea¶

A textual definition is a natural-language statement attached to a term in a controlled vocabulary, ontology, or standard that specifies the conditions under which the term is correctly applied — the term's intended extension expressed in human-readable prose — serving as the normative anchor for annotation decisions, curation disputes, and inter-annotator agreement.

The definition's structural role is to convert a named term from a label that users interpret according to their own judgment into a term with a publicly stated, consultable boundary. Without a textual definition, two annotators who disagree about whether a gene product should be annotated to "apoptotic process" have no shared resolution procedure: each applies private inference about what the term means. With a textual definition, the resolution procedure is to read the definition, decide whether the evidence meets the conditions it states, and if necessary to refine the definition rather than leaving the ambiguity to accumulate. The definition is thus the appeal of last resort in annotation practice — not a complete formal specification, but a stable prose statement that bounds the acceptable interpretation of the term.

Three commitments travel with the pattern as practised in ontology engineering. The definition must be attached to the term: not a separate document or a website comment but a metadata field that ships with the term wherever the term is distributed, so that any system receiving the term receives its definition. It should aim at genus-differentia form when possible — "a programmed cell death process which begins when a cell receives an internal or external signal and proceeds through ..." — naming the broader class (genus) and the distinguishing conditions (differentia), because that form is least ambiguous and most composable when terms are imported into larger ontologies. And it must be non-circular: if term A is defined in terms of B and B in terms of A, the definition gives no usable information and the annotation discipline collapses.

OBO Foundry Principle FP-006 codifies these commitments for biomedical ontologies: every term must have a textual definition, the definition must follow genus-differentia style, and a definition-source relation must record who wrote it and from what authority. The Gene Ontology, the Human Phenotype Ontology, ChEBI, and the Cell Ontology all carry textual definitions at this standard. The ISO 704 standard for terminology work and the ISO 25964 standard for thesauri codify structurally identical requirements for standards and library vocabularies. Legal definitions sections at the front of statutes and contracts serve the same function in legal drafting.

The pattern manages a specific scaling failure: as a controlled vocabulary grows to thousands of terms and is used across dozens of institutions, the implicit shared understanding that allowed the founding community to apply terms consistently degrades. New annotators, translators, and downstream tool-builders cannot reconstruct that understanding from term labels alone. The textual definition externalises enough of that understanding — imperfectly, in natural language, with acknowledged edge cases — that the vocabulary can function as a coordination mechanism across communities that never communicated directly.

Structural Signature¶

Sig role-phrases:

the defined term — a label in a controlled vocabulary, ontology, or standard whose conditions of correct application are at issue
the prose boundary — a natural-language statement of the term's intended extension, converting a privately-interpreted label into a publicly consultable boundary
the attachment commitment — the requirement that the definition ship as a metadata field with the term, so any receiving system inherits the boundary
the genus–differentia form — the preferred structure naming the broader class plus the distinguishing conditions, the form most composable when terms are imported into larger ontologies
the non-circularity commitment — the prohibition on A-defined-by-B-defined-by-A, so the appeal terminates in usable information
the appeal-of-last-resort status — the engineered role: the normative anchor that makes an application dispute resolvable by testing evidence against the stated conditions, refining the definition if it cannot decide
the definition-source provenance — the recorded authorship and authority by which a contested boundary's standing is adjudicated and its revisions traced
the natural-language medium constraint — the defining limitation: the prose stays ambiguous, accrues unforeseen edge cases, and must be curated and versioned, and has no counterpart outside human-readable-term substrates

What It Is Not¶

Not a complete formal specification. A textual definition is a stable prose anchor — the appeal of last resort — not a logical axiom giving necessary-and-sufficient conditions in a formal language. It aims at genus–differentia form where it can, but it acknowledges edge cases and lives in natural language; treating it as a complete formalization mistakes a human-readable boundary for a machine-decidable predicate.
Not a guarantee of unambiguous meaning. It bounds the acceptable interpretation of a term, it does not eliminate ambiguity: the medium is natural language, so definitions accrue unforeseen edge cases and must be curated and versioned. Its contribution is to make inter-annotator disagreement resolvable — testable against stated conditions, refinable when they cannot decide — not to make disagreement impossible.
Not descriptive commentary about the term. The definition is normative and attached: it ships as a metadata field with the term wherever the term is distributed, and it is the resolution authority when annotators disagree. A note on a website or a separate document is not a textual definition in this sense — what makes it one is that it travels with the term and binds its correct application, not that it merely describes it.
Not satisfied by any prose attached to a label. The pattern carries hard commitments: a definition that defines A by B and B by A is circular and conveys no usable information, and one that names no genus and differentia is less composable when imported into a larger ontology. A term with text that fails these is defective, not merely terse; the discipline is what separates a controlled vocabulary from a tag-soup.
Not the general boundary prime, and it does not travel across substrates. Its substrate-independent core — publish a stable rule of inclusion and exclusion alongside the term — is boundary, which recurs across membranes, type systems, and conceptual categories. Textual definition is the documentation-substrate instance: its defining bite is the natural-language medium, which has no counterpart outside human-readable-term substrates, so the genus–differentia style and definition-source recording are curation disciplines within one substrate, not a portable structural pattern.

Scope of Application¶

The textual definition lives within the documentation of controlled vocabularies — one broad substrate spanning several term-curation practices, every endpoint a curated human-readable specification of term meaning. Its reach is bounded by the natural-language medium that is its defining commitment: outside human-readable-term substrates the structurally adjacent role is filled by type signatures and axiomatic definitions, a different artifact. The substrate-independent core (publish a stable rule of inclusion and exclusion) is the parent boundary, of which this is the documentation-substrate instance. Within the domain the same attached, genus–differentia, non-circular discipline recurs across these practices.

Biomedical and life-science ontologies — OBO Foundry Principle FP-006 mandates a textual definition, genus–differentia style, and a definition-source relation for every term (Gene Ontology, Human Phenotype Ontology, ChEBI, Cell Ontology).
Standards bodies — ISO 704 terminology work, ISO 25964 thesauri, and IEEE definitions, with the convention that the definition is normative and surrounding prose illustrative.
Library cataloging — MeSH and LCSH scope notes and thesaurus definitions bounding the application of controlled subject headings.
API design and software contracts — docstrings, type-signature comments, and OpenAPI description fields specifying in prose what an endpoint or function accepts and returns.
Legal drafting — the "Definitions" section at the front of a statute or contract, binding defined terms to specific textual meanings for the rest of the document.
Lexicography — dictionary definitions in genus–differentia style, with the convention that words be defined by simpler words.

Clarity¶

Naming the textual definition as a distinct artifact makes legible what separates a controlled vocabulary from a tag-soup: a term with a definition has a publicly consultable boundary, while a term without one is only a label that each annotator interprets by private judgment. The distinction the concept sharpens is between a name and an interpretable boundary — between having a word for "apoptotic process" and having a stated rule for when the word correctly applies. With that distinction in hand, the question an ontology curator asks when two annotators disagree shifts from the unanswerable "what does this term really mean?" to a procedural one: does the evidence meet the conditions the definition states, and if the definition cannot decide the case, should the definition be refined? This is what gives the definition its specific role as the appeal of last resort — not a complete formal specification, but a stable prose anchor that makes inter-annotator disagreement resolvable rather than left to accumulate as silent interpretive drift.

The concept also makes a scaling failure visible that term labels alone conceal. While a founding community shares enough tacit understanding to apply its terms consistently, the absence of definitions costs nothing; the textual definition earns its keep precisely when the vocabulary grows to thousands of terms used across institutions that never communicated, where new annotators and downstream tool-builders cannot reconstruct that understanding from labels. Naming the artifact lets the engineer see the three commitments that make it function — that it must ship attached to the term (so any system receiving the term receives its boundary), aim at genus–differentia form (naming the broader class and the distinguishing conditions, the form most composable when terms are imported), and remain non-circular (since A-defined-by-B-defined-by-A conveys no usable information) — and to recognize that these are not stylistic preferences but the conditions under which externalised meaning actually coordinates work across communities, with the standing cost that natural language remains ambiguous and the definitions require curation and versioning to track an evolving domain.

Manages Complexity¶

Term-application disagreement in a large controlled vocabulary is, left unmanaged, an unbounded and recurring sprawl: for every term, an open class of borderline cases, each disputed afresh by annotators applying private judgment, with no shared resolution procedure and the ambiguity accumulating as silent interpretive drift across thousands of terms and dozens of institutions. The textual definition compresses that sprawl onto one artifact per term. Instead of re-arguing each case, an annotator consults a single consultable boundary — a stable prose statement of the term's conditions of correct application — and the entire "does X count?" question collapses to "does the evidence meet the conditions the definition states?", with the further move, when the definition cannot decide, of refining the definition once rather than litigating instances repeatedly. Because the artifact ships attached to the term, the resolution procedure travels with the vocabulary to every receiving system; because it aims at genus–differentia form, the boundary stays composable as terms are imported into larger ontologies; because it is non-circular, the appeal actually terminates in usable information. A curator therefore reasons over one normative text per term — the appeal of last resort — in place of an ever-growing, distributed body of case-by-case argument, which is exactly what lets the vocabulary externalise enough shared understanding to coordinate work across communities that never communicated directly, at the standing cost that the prose stays ambiguous and must be curated and versioned as the domain evolves.

Abstract Reasoning¶

Reduced to one normative prose artifact per term — the consultable boundary that ships attached, aims at genus–differentia form, and stays non-circular — the pattern licenses the inferences an ontology curator or annotator draws when applying, disputing, or maintaining terms, all turning on the difference between a name and a stated rule of correct application.

Diagnostic — resolve an application dispute by testing the evidence against the stated conditions, and detect a defective definition. The signature move is procedural: when two annotators disagree about whether an instance falls under a term — whether a gene product should be annotated to "apoptotic process" — the curator does not ask the unanswerable "what does this term really mean?" but reads the definition and infers whether the evidence meets the conditions it states. The inference runs from a stated boundary to an inclusion/exclusion verdict, and it terminates because the definition is the appeal of last resort. A complementary diagnostic reads the definition's own health: a definition that defines A by B and B by A is detected as circular and therefore information-empty, and one that fails to name a genus and differentia is flagged as less composable — the curator infers a definition is defective from its form, not merely from a contested case. The absence of any definition is itself diagnostic: a term that is only a label is recognised as having no shared resolution procedure, so disagreements about it will accumulate as silent interpretive drift.

Interventionist — refine the definition when it cannot decide, and predict the coordination effect of each commitment. When the stated conditions cannot settle a borderline case, the practitioner's move is to refine the definition once rather than litigate the instance repeatedly — predicting that a sharpened boundary resolves not just this dispute but the open class of like cases, converting recurring case-by-case argument into a single amendment. The three commitments are read as design levers with predicted consequences: attaching the definition to the term predicts that every receiving system inherits the boundary (detach it and downstream tool-builders cannot reconstruct it); aiming at genus–differentia form predicts the boundary stays composable when the term is imported into a larger ontology (abandon it and imports become ambiguous); enforcing non-circularity predicts the appeal actually terminates in usable information. Each is thus a prediction about whether externalised meaning will coordinate work — and the standing cost is predicted too: because the medium is natural language, definitions accrue unforeseen edge cases and must be curated and versioned as the domain evolves.

Boundary-drawing — decide what counts as a defined term, and when a vocabulary needs definitions at all. The pattern gates what separates a controlled vocabulary from a tag-soup: a term with a definition has a publicly consultable boundary and may anchor annotation, curation disputes, and inter-annotator agreement; a term that is only a label cannot, and the curator distinguishes the two as a difference in kind, not polish. A second, scale-sensitive boundary judgment is when the artifact earns its keep: while a founding community shares enough tacit understanding to apply its terms consistently, the absence of definitions costs nothing — the practitioner infers that definitions become load-bearing precisely when the vocabulary grows to thousands of terms used across institutions that never communicated, where labels alone no longer transmit meaning. So the decision to invest in textual definitions is itself read off a scaling threshold rather than treated as universally mandatory.

Provenance — trace and adjudicate by the definition's source authority. Because the pattern carries a definition-source relation recording who wrote a definition and from what authority, the curator can reason about a boundary's standing: a contested definition is adjudicated partly by appeal to its source, and a definition's revisions trace the term's evolving intended extension over time — provenance the practitioner consults when deciding whether to trust, cite, or amend the stated boundary.

Knowledge Transfer¶

Within the documentation of controlled vocabularies — the home substrate — the textual-definition pattern transfers as mechanism, because the cargo is one artifact with three commitments: a normative prose boundary that ships attached to the term, aims at genus–differentia form, and stays non-circular, serving as the appeal of last resort. It carries across a whole family of term-curation practices with the same shape: biomedical and life-science ontologies (OBO Foundry FP-006 mandating definitions, genus–differentia style, and a definition-source relation; instantiated by GO, HPO, ChEBI, Cell Ontology), standards bodies (ISO 704 terminology work, ISO 25964 thesauri, IEEE definitions, with the convention that the definition is normative and surrounding prose illustrative), library cataloging (MeSH and LCSH scope notes), API design and software contracts (docstrings, type-signature comments, OpenAPI description fields), legal drafting (the "Definitions" section binding terms for the rest of a statute or contract), and lexicography (dictionary definitions in genus–differentia style). Across all of these the apparatus carries without translation — the resolve-the-dispute-by-testing-evidence-against-stated-conditions move, the refine-the-definition-once intervention, the circularity and genus–differentia health checks, the scaling-threshold judgment about when definitions earn their keep, and definition-source provenance for adjudicating a boundary's standing. But this breadth, while real, stays within one broad substrate: every endpoint is a curated human-readable specification of term meaning, and the porting is between natural-language-prose practices, not across structurally different substrates.

Beyond curated-term documentation the pattern does not recur with the same shape, and the limiting fact is its defining commitment to the natural-language medium. In formal systems the structurally adjacent role is filled by type signatures and axiomatic definitions, which are a different artifact; there is little (A) metaphor temptation, because "a prose rule of correct application attached to a term" simply has no counterpart outside substrates where natural-language descriptions are read by humans or by NLP systems consuming them. The honest reading is the (B) one: what is genuinely substrate-independent about the candidate is already carried by its parents — boundary (the demarcation move of publishing a stable rule of inclusion and exclusion), controlled_vocabulary / index (the curated-vocabulary infrastructure), and representation (the medium choice) — and the textual specificity adds convention, not new structure. boundary is the substrate-independent pattern that transfers across physical membranes, type systems, ecological niches, and conceptual categories; textual definition is its documentation-substrate instance, the genus–differentia style, definition-source recording, and inter-annotator protocols being curation disciplines within one substrate rather than a portable structural pattern that opens new analytical moves in unrelated domains. The cross-domain lesson therefore belongs to boundary (with semantic_narrowing_widening as the diachronic counterpart whose changes a definition's revisions trace, and symbolic_boundaries as the social-category counterpart); textual definition keeps the ontology-engineering cargo (the attached normative prose, the appeal-of-last-resort status, the natural-language medium constraint) that stays home. See Structural Core vs. Domain Accent.

Examples¶

Canonical¶

The Gene Ontology term "apoptotic process" (GO:0006915) shows the pattern in its designed form. Rather than leaving the label to each curator's judgment, GO attaches a normative prose definition in genus–differentia style: the genus is "a programmed cell death process," and the differentia specify that it begins when a cell receives an internal or external signal and proceeds through a characteristic cascade of biochemical events ending in the cell's death — the conditions that separate apoptosis from necrosis or other death processes. The term also carries a definition-source relation recording the authority the wording was drawn from. When two annotators dispute whether an experimental result should be annotated to GO:0006915, they do not argue about what the label "really means"; they read the definition and test whether the evidence meets its stated conditions, refining the definition once if it cannot decide.

Mapped back: GO:0006915 is the defined term; its prose statement is the prose boundary, shipped in the term's metadata as the attachment commitment. "Programmed cell death process" plus the signal-initiated cascade is the genus–differentia form, and the recorded source is the definition-source provenance. The read-and-test-the-conditions resolution is the appeal-of-last-resort status, mandated for GO by OBO Foundry FP-006.

Applied / In Practice¶

Statutory drafting deploys the same artifact in the "Definitions" section that opens many laws. The U.S. Copyright Act (17 U.S.C. § 101) begins with a long definitions section that binds terms for the entire statute: it specifies, for instance, exactly what counts as a "work made for hire," "publication," or a "derivative work." When a later dispute turns on whether a freelancer's contribution is a "work made for hire," courts do not consult ordinary usage of the phrase; they apply the statutory definition's stated conditions, and the definition governs the rest of the Act wherever the term appears. The definition ships attached to the statute, is normative rather than illustrative, and functions as the appeal of last resort — the same discipline as an ontology term, in a legal substrate.

Mapped back: "Work made for hire" is the defined term; the § 101 entry is the prose boundary, part of the statute itself (the attachment commitment), and it is the appeal-of-last-resort status courts test evidence against. Its enactment authority is the definition-source provenance, while the persistent litigability of edge cases (e.g., contractor versus employee) is the natural-language medium constraint that keeps such definitions under amendment.

Structural Tensions¶

T1: Bounds interpretation versus eliminates ambiguity (the medium that is both asset and limit). A textual definition makes inter-annotator disagreement resolvable — testable against stated conditions, refinable when they cannot decide — but it does not make disagreement impossible, because the medium is natural language, which accrues unforeseen edge cases and must be curated and versioned as the domain evolves. The tension is that the artifact's central value and its permanent limitation are the same choice of medium: prose is what lets a boundary be human-consultable and travel across communities that never communicated, and prose is exactly what keeps the boundary from ever being fully decidable. Demanding that a definition eliminate ambiguity mistakes its job; treating its residual ambiguity as failure discards the only medium that could coordinate work at all. Diagnostic: Is the definition being asked to make disagreement resolvable, or impossibly to make it vanish — and is its residual ambiguity being read as a defect or as the cost of the natural-language medium?

T2: Appeal of last resort versus complete formal specification (a stable anchor, not an axiom). The definition is the normative anchor an annotation dispute terminates in — read it, test the evidence against its conditions — but it is a stable prose statement that acknowledges edge cases, not a logical axiom giving necessary-and-sufficient conditions in a formal language. The tension is that it must be authoritative enough to end an appeal yet informal enough to live in human-readable prose, and those pull against each other: a fully formalized predicate would be machine-decidable but lose the accessibility that lets non-specialists and downstream tool-builders consult it, while a merely descriptive note would be readable but could not terminate a dispute. Treating a textual definition as a complete formalization mistakes a human-readable boundary for a decidable one. Diagnostic: Is the definition being relied on as the terminating authority for disputes (its actual role), or as a machine-decidable specification (which its prose medium cannot be)?

T3: Refine-once versus litigate-instances (the fix that seeds future disputes). When the stated conditions cannot settle a borderline case, the disciplined move is to refine the definition once — resolving not just this dispute but the open class of like cases — rather than re-arguing instances forever. But each refinement is itself prose, and prose accrues new edge cases, so the amendment that closes one class of disputes opens the ground for the next. The tension is that the very mechanism that converts recurring case-by-case argument into a single durable amendment also guarantees the definition is never finished: refinement is both the resolution and the renewal of ambiguity. A vocabulary that never refines lets drift accumulate; one that refines endlessly churns its own boundaries. Diagnostic: When a case cannot be decided, is the response a one-time refinement of the shared boundary, or a re-litigation of the instance that leaves the definition (and the next dispute) untouched?

T4: Scaling threshold (when definitions earn their keep). While a founding community shares enough tacit understanding to apply its terms consistently, the absence of definitions costs nothing — the artifact earns its keep only when the vocabulary grows to thousands of terms used across institutions that never communicated, where labels alone no longer transmit meaning. The tension is that the investment is worthless before the threshold and load-bearing after it, and the threshold is crossed silently: the community applying terms fluently from shared context has no signal that the tacit understanding is degrading until new annotators, translators, and tool-builders can no longer reconstruct it. Mandating definitions universally over-invests in small vocabularies; withholding them past the threshold lets interpretive drift accumulate invisibly. Diagnostic: Has the vocabulary outgrown the shared tacit context of its founding community — and is the decision to invest in definitions keyed to that scaling threshold or treated as a blanket rule?

T5: Attachment versus description (normative-and-shipped, not commentary). What makes prose a textual definition is not that it describes the term but that it is normative and attached — it ships as a metadata field with the term wherever the term is distributed, so any receiving system inherits the boundary, and it is the resolution authority when annotators disagree. A note on a website or a separate document, however accurate, is not a textual definition in this sense. The tension is that the commitment which makes the artifact function (attachment) is exactly what distinguishes it from mere commentary, and prose that fails it is defective rather than merely informal: detach the definition and downstream tool-builders cannot reconstruct the boundary, no matter how good the prose. Description is abundant and free; a binding, traveling boundary is the scarce thing. Diagnostic: Does the prose ship attached to the term and bind its correct application, or does it merely describe the term from a separate location?

T6: Autonomy versus reduction (a curation artifact or the documentation-substrate instance of the boundary parent). The textual-definition pattern has unusually broad within-substrate reach — biomedical ontologies (OBO Foundry FP-006), ISO 704 and 25964, MeSH scope notes, docstrings, statutory definitions, dictionary entries — yet every one of those endpoints is a curated human-readable specification of term meaning, one broad natural-language-prose substrate rather than structurally different domains. What is genuinely substrate-independent is already carried by the parents: boundary (publish a stable rule of inclusion and exclusion), controlled_vocabulary / index (the curated infrastructure), and representation (the medium choice). boundary travels across membranes, type systems, ecological niches, and conceptual categories; textual definition is its documentation-substrate instance, and the genus–differentia style, definition-source recording, and non-circularity commitment are curation disciplines within that substrate, adding convention rather than new structure. Off human-readable-term substrates the adjacent role is filled by type signatures and axioms — a different artifact. Diagnostic: Resolve toward the parent boundary when asking what recurs across substrates; toward the textual definition when the question is the natural-language-medium curation discipline that keeps a term's boundary consultable and attached.

Structural–Framed Character¶

Textual definition sits at the framed-leaning position on the structural–framed spectrum: not a normative verdict, but a curation artifact constituted by the practice of documenting controlled vocabularies and codified by standards bodies, whose distinctive content is inseparable from that practice. The criteria line up almost entirely on the framed side. Its evaluative weight is low and mostly points structural — a definition is a boundary, not a conviction — though it carries a mild normative edge, since a definition that is circular or names no genus and differentia is judged defective, a quality standard rather than a neutral description. Every other criterion points framed. It is strongly human-practice-bound: a textual definition is not something a term has in nature but something curators attach and consult — it exists to serve as the appeal of last resort in annotation disputes, ships as a metadata field, and is refined when it cannot decide a case, so strip away the community of annotators, curators, and downstream tool-builders and there is no textual definition, only prose next to a label with nothing to adjudicate. Its institutional origin is pronounced and explicit: the concept is codified by OBO Foundry Principle FP-006, ISO 704 and 25964, IEEE conventions, and statutory drafting practice — its three commitments (attachment, genus–differentia form, non-circularity) and its definition-source provenance are disciplines legislated by standards authorities, artifacts of terminology-work traditions, not facts nature marks. On vocab_travels it is domain-pinned by its very defining commitment — the natural-language medium — which has no counterpart off human-readable-term substrates, where the structurally adjacent role is filled by type signatures and axioms, a different artifact. And on import_vs_recognize it patterns as recognition within curated-term documentation and, beyond it, not even as analogy but as the parent boundary doing the work.

The one structural-looking feature is the portable skeleton the entry itself isolates: publish a stable rule of inclusion and exclusion alongside the thing it bounds — boundary, the demarcation move that recurs across membranes, type systems, ecological niches, and conceptual categories, supported by controlled_vocabulary/index (the curated infrastructure) and representation (the medium choice). That boundary skeleton is genuinely substrate-portable, which is what tempts a structural reading. But it does not lift textual definition off the framed side, because publishing an inclusion/exclusion rule is precisely what textual definition instantiates from the boundary umbrella, not what makes "textual definition" itself travel: the cross-substrate reach belongs to boundary, while the entry's distinctive content — the attached normative prose, the genus–differentia curation style, the definition-source recording, the appeal-of-last-resort status, and above all the natural-language-medium constraint — is domain accent that stays home as curation discipline within one documentation substrate, adding convention rather than new structure. Its character: an evaluatively light but practice-constituted, standards-codified documentation artifact, structural only in the inclusion/exclusion boundary it instantiates and dresses in the natural-language curation machinery of ontology engineering.

Structural Core vs. Domain Accent¶

This section decides why textual definition is a domain-specific abstraction and not a prime, and it carries the case for its domain-specificity in the same move.

What is skeletal (could lift toward a cross-domain prime). Strip the ontology engineering and a thin relational structure survives: publish, alongside the thing it bounds, a stable rule of inclusion and exclusion, so that whether a case falls inside is decided by testing it against the stated rule rather than by private judgment. The portable pieces are abstract — a bounded thing, a stable demarcating rule, and a resolution procedure that tests candidates against the rule. That skeleton is boundary (the demarcation move of publishing an inclusion/exclusion rule), supported by controlled_vocabulary/index (the curated infrastructure the rule attaches to) and representation (the choice of medium the rule is stated in). It is genuinely substrate-portable — boundary recurs across physical membranes, type systems, ecological niches, and conceptual categories — which is exactly why it is the core textual definition shares with those parents, and not what makes the entry the particular thing it is.

What is domain-bound. Almost everything that makes the concept textual definition in particular is curated-term-documentation furniture, and its single most defining commitment — the natural-language medium — is precisely the part that does not survive extraction. The requirement that the definition ship attached as a metadata field so any receiving system inherits the boundary; the preference for genus–differentia form so terms stay composable when imported into larger ontologies; the non-circularity prohibition; the definition-source provenance recording who wrote it and by what authority; and the appeal-of-last-resort status in annotation disputes are the worked disciplines of a specific practice, codified by OBO Foundry FP-006, ISO 704 and 25964, IEEE conventions, and statutory drafting. The decisive test is the medium itself: off human-readable-term substrates the structurally adjacent role is filled by type signatures and axiomatic definitions — a genuinely different artifact, machine-decidable rather than human-consulted — so "a prose rule of correct application attached to a term" has no counterpart at all where no natural-language description is read by a human (or an NLP system standing in for one). The artifact is also constitutively practice-bound: strip away the community of annotators, curators, and downstream tool-builders and there is no textual definition, only prose next to a label with nothing to adjudicate.

Why this does not clear the prime bar. A prime is a relational structure whose vocabulary travels and whose cross-domain transfer is recognition of the same mechanism, not analogy. Textual definition's transfer is unusually broad within one substrate and stops there. Within curated-term documentation it travels as mechanism — biomedical ontologies (GO, HPO, ChEBI, Cell Ontology under FP-006), standards bodies (ISO 704/25964, IEEE), library cataloging (MeSH/LCSH scope notes), API docstrings and OpenAPI fields, statutory definitions sections, and dictionary entries all carry the identical apparatus (resolve-by-testing-evidence, refine-once, the circularity and genus–differentia health checks, the scaling-threshold judgment, definition-source provenance) — but every one of those endpoints is a curated human-readable specification of term meaning, one broad natural-language-prose substrate, so this is within-substrate recognition, not cross-domain reach. Beyond that substrate the pattern does not recur with the same shape — and notably not even as tempting analogy, since the parent boundary simply does the work directly. When the bare structural lesson is wanted across substrates, it is already carried, in more general form, by the parents textual definition instantiates: boundary for the inclusion/exclusion rule (with semantic_narrowing_widening as the diachronic counterpart a definition's revisions trace, and symbolic_boundaries as the social-category counterpart), controlled_vocabulary/index for the curated infrastructure, representation for the medium choice. The cross-substrate reach belongs to boundary; "textual definition," as named, is its documentation-substrate instance, adding curation convention — attachment, genus–differentia style, definition-source recording, the natural-language medium — rather than new portable structure, and that convention should stay home.

Relationships to Other Abstractions¶

Current abstraction Textual Definition Domain-specific

Parents (1) — more general patterns this builds on

Textual Definition is a kind of Boundary Prime

A textual definition is a boundary specialized to term application and expressed in stable natural-language documentation.

Hierarchy path (1) — routes to 1 parentless root

Textual Definition → Boundary

Not to Be Confused With¶

Formal / axiomatic definition (OWL logical definition, type signature). A machine-decidable specification giving necessary-and-sufficient conditions in a formal language — an ontology's logical axioms, a function's type signature. A textual definition is the human-consulted prose anchor that acknowledges edge cases and lives in natural language; the two often coexist on one term but do different work. Off human-readable-term substrates, the formal artifact is what fills the adjacent role. Tell: is the boundary stated as a formal predicate a reasoner can evaluate (axiomatic definition), or as prose a curator reads and tests evidence against, refining it when it cannot decide (textual definition)?
The bare term or label (a name without a boundary). A word in a vocabulary with no stated conditions of correct application — "apoptotic process" as a label each annotator interprets by private judgment. This is precisely the contrast case the concept exists against: the label has no publicly consultable boundary and no shared resolution procedure, so disputes about it accumulate as silent interpretive drift. Tell: is there a stated rule you can test a candidate against (textual definition present), or only a name whose meaning each user reconstructs privately (bare label)?
Descriptive commentary / non-normative note. Accurate prose about a term that sits in a website comment, a wiki, or a separate document. Unlike a textual definition it is neither normative nor attached: it does not ship as a metadata field with the term and it is not the resolution authority in disputes. Detach the boundary and downstream tool-builders cannot reconstruct it, however good the prose. Tell: does the prose travel attached to the term and bind its correct application (textual definition), or merely describe it from a separate location with no adjudicating force (commentary)?
Data dictionary (data modeling). A catalog of a dataset's or schema's fields with their meanings, types, and constraints. It overlaps textual definition in spirit — documenting what a name means — but its home is data-field/column semantics and it typically mixes prose with type and constraint metadata rather than serving as the genus–differentia appeal-of-last-resort for a controlled-vocabulary term. A sibling in the same family, different endpoint. Tell: is the artifact bounding the correct application of a vocabulary term under curation discipline (textual definition), or documenting the fields of a data structure with their types and constraints (data dictionary)?
Boundary (the parent prime it instantiates). The substrate-neutral demarcation move — publish a stable rule of inclusion and exclusion alongside the thing it bounds — that recurs across membranes, type systems, ecological niches, and conceptual categories. Not a confusable peer but the umbrella; textual definition is its documentation-substrate instance, adding the natural-language curation conventions. Tell: when asking what recurs across substrates, the portable content is boundary — treated more fully elsewhere — while the attached normative prose, genus–differentia style, and natural-language medium are textual definition's home-bound accent.
Semantic narrowing / widening. The diachronic process by which a term's meaning drifts — contracts or expands — over time through usage. Textual definition is the synchronic artifact that states a boundary at a point in time; its successive revisions are what trace such narrowing or widening. One is the moving meaning, the other the versioned record that pins and tracks it. Tell: are you naming the historical drift of what a term covers (semantic narrowing/widening) or the attached prose rule whose amendments document that drift (textual definition)?

Neighborhood in Abstraction Space¶

Textual Definition sits in a sparse region of the domain-specific corpus (78^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Trust Infrastructure for Shared Knowledge (6 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-07-12