Authority Record¶

Fix one governed identity per entity in a curated registry — preferred label, variant set, disambiguating notes, and a persistent identifier — so every catalog entry cites it rather than re-spelling the name, and one edit propagates everywhere.

Core Idea¶

An authority record is a curated registry entry that establishes the preferred form for naming a single entity — a person, corporate body, geographic name, work title, or controlled subject — together with its known variants, disambiguating context, and cross-references to related records, so that all other catalog entries in the system can resolve to a single canonical identity by citing it.

The record's structure is precise and institutionally governed. At its core: a persistent unique identifier (the numeric key that other records cite), a preferred label (the authorized heading in a specified form, such as "Twain, Mark, 1835–1910"), and a variant-label set listing aliases, alternate transliterations, former names, pseudonyms, and common misspellings. Alongside the label set: scope notes and disambiguating context — birth and death dates, professional field, place of activity — that distinguish this entity from namesakes who might otherwise share a heading. Relation fields link the record to related, broader, or narrower authority records (an author to a corporate body they directed; a place name to its broader administrative parent). A change-history section records corrections, merges of duplicate records, and splits when one record turns out to cover two distinct entities.

What makes the authority record structurally distinctive inside the bibliographic tradition is the deferral discipline it enforces: descriptive records elsewhere in the catalog do not re-specify the entity's name; they cite the authority record's identifier. A global rename — say, a geographic name change or a posthumous scholarly consensus that two records describe the same person — is effected once, in the authority record, and propagates automatically to everything that cites it. This makes the authority record the single point of truth for naming, not a reference copy of the name.

The defining institutional implementations are: the Library of Congress Name Authority File (LCNAF), which governs personal names, corporate bodies, conferences, and works for U.S. cataloging practice; the Virtual International Authority File (VIAF), which clusters authority records from dozens of national libraries under a shared identifier; the Getty vocabularies (ULAN for artists, AAT for art and architectural terms, TGN for geographic names); ORCID for researcher identities; and ISNI for public identities across publishing and media. Each enforces one preferred form per entity within its scope, routes all variant forms through the record's identifier, and resolves ambiguity by governance rather than by convention. Compliance is specified by descriptive standards — RDA (Resource Description and Access), ISAAR(CPF) for archival authority records — and records exchange via MARC authority format or BIBFRAME.

Outside the library, archive, and museum tradition the full role-set — curated preferred label, governed variant set, scope notes, descriptive-standard compliance, exchange format, change history — does not assemble in one artifact; adjacent systems (database primary keys, canonical URLs, DNS, CAS registry numbers) carry some of these functions but not the curated controlled-vocabulary layer that makes the authority record its own thing within bibliographic control.

Structural Signature¶

Sig role-phrases:

the governed entity — a single person, corporate body, place, work, or subject whose identity the record fixes
the persistent identifier — the numeric key that other records cite in place of writing the name again
the preferred label — the one authorized heading form selected by rule (e.g. "Twain, Mark, 1835–1910")
the variant set — aliases, transliterations, former names, pseudonyms, and misspellings, all routed through the identifier
the disambiguating scope notes — dates, profession, place of activity distinguishing the entity from namesakes
the relation cross-references — links to broader, narrower, and related authority records
the change history — logged corrections, merges of duplicates, and splits when one record turns out to cover two entities
the curating authority — the institution maintaining the registry under a controlled-vocabulary policy and descriptive standards (RDA, ISAAR)
the deferral-and-propagation guarantee — descriptive records cite the identifier rather than copy the name, so one registry edit reaches exactly its citation set
the note-versus-split adjudication — ambiguity resolvable by a scope note stays in the record; ambiguity that is not signals two entities and a split
the full-role-set limitation — the controlled-vocabulary-plus-standards layer is library-tradition-bound; strip it and only a thinner identifier pattern remains

What It Is Not¶

Not the name itself. An authority record is a governed identity, not a name string typed into a heading field. "Twain, Mark" looks like text, but the record is the maintained registry entry — preferred form, variant set, disambiguating context, identifier — that fixes which entity any heading resolves to. Treating it as just the authorized spelling misses that its work is binding all variants to one canonical identity, not storing one string.
Not authentication. The record standardizes reference to an entity; it does not verify that a claimed identity is genuine. It answers "which entity does this name resolve to, in its authorized form?" — not "is this person who they say they are?" An ORCID record fixes a researcher's canonical identity for citation; it does not prove the claimant's identity at login. Conflating authoritative reference with identity verification confuses two different operations.
Not a bare identifier or primary key. The persistent identifier is one role among several, and the identifier alone is not an authority record. What makes it its own thing is the full role-set assembled in one curated artifact: a preferred form chosen by controlled-vocabulary rule, a governed variant set, scope notes, change history, and descriptive-standard compliance. Strip those and a database surrogate key or a canonical-URL tag remains — a thinner pattern the concept does not reduce to.
Not "authority" in the normative sense. The record does not exert power or issue binding decisions over anyone; the curating institution has authority over the record, but the record itself is a descriptive artifact, not a governance relation between a rule-system and the ruled. "Authority" here means authoritative reference — the single point of truth for naming — not the legitimate right to command.
Not a reference copy of the name. A descriptive record that re-specifies an entity's name has copied a value; an authority record is what such records cite, the single source they defer to. A global rename, merge, or split is effected once in the authority record and propagates to its citation set — so the record is the point of truth, not one more place the name is written down. Entries that copied the string rather than citing the identifier are precisely the ones a rename fails to reach.

Scope of Application¶

The authority record lives within bibliographic control and the wider library/archive/museum/scholarly-publishing tradition — the subfields where the full role-set (curated preferred form, governed variant set, scope notes, descriptive-standard compliance, exchange format, change history) assembles in one maintained registry artifact; its reach is within that governed-identification domain (the DNS, CAS-registry-number, biological-nomenclature, and surrogate-key cousins are co-instances of the parent primes — identification and representation — capturing only part of the role-set, not the controlled-vocabulary-plus-standards layer that makes an authority record its own thing).

Library name and subject cataloging — the home: the Library of Congress Name Authority File governs personal names, corporate bodies, conferences, and works, so searches for "Twain, Mark" and "Clemens, Samuel L." resolve to one authorized identity.
Federated and international authority clustering — the Virtual International Authority File clusters national-library records under a shared identifier, preserving authority integrity across institutions despite local naming variation.
Museum and gallery cataloging — the Getty vocabularies (ULAN for artists, AAT for materials and styles, TGN for places) govern how a collection database refers to its entities.
Researcher and public-identity registration — ORCID for researcher identities across publications, grants, and affiliations, and ISNI for public identities across publishing and media.
Bibliographic metadata exchange — cross-institution record sharing via MARC authority format and BIBFRAME under descriptive standards (RDA, ISAAR(CPF)), making "the same entity" resolvable across systems.

Clarity¶

Naming the authority record makes legible a distinction the catalog constantly blurs: the difference between a name string and a governed identity. Without the concept, "Twain, Mark" looks like text a cataloger types into a heading field, and the recurring troubles of bibliographic description — namesakes colliding under one heading, an author scattered across "Clemens," "Twain," and "Snodgrass," a corporate body that renamed itself, a posthumous merge of two records found to describe one person — read as so many unrelated nuisances to be patched record by record. Recognizing the authority record reframes all of them as one thing: a failure or success of deferral to a single point of truth. The question stops being "what name do I write here?" and becomes "which authorized identity does this resource resolve to, and does its record yet exist?"

The concept also sharpens the line between holding the name and citing it. A descriptive record that re-specifies an entity's name has copied a value; one that cites the authority identifier has subscribed to a governed source — and only the latter lets a global rename, disambiguation, merge, or split propagate cleanly instead of being chased across every catalog entry that mentioned the entity. That makes the authority file's central validity criterion stateable: is there one preferred form per entity within scope, with every variant routed through the record's identifier? A cataloger can now ask whether two registries cross-walk losslessly, whether a federated catalog preserves authority integrity under clustering, and whether a heading's ambiguity belongs in a scope note or signals that one record is silently covering two entities — questions that have no purchase until the maintained-registry-with-preferred-form discipline is named as the object being maintained.

Manages Complexity¶

A catalog of any size accumulates a combinatorial sprawl of naming trouble: every entity can surface under many strings (aliases, transliterations, pseudonyms, pre- and post-rename forms, misspellings), every string can collide with a namesake, and every resource that mentions an entity is a site where that trouble can recur. Faced record by record, the cataloger must re-adjudicate identity at each occurrence — chasing a geographic rename across thousands of headings, deciding afresh whether this "J. Smith" is that one, patching each merge or split wherever the entity was ever named. The authority record collapses that field of independent decisions to a single governed object per entity: the disambiguation, the variant set, the preferred form, and the change history are all resolved once and stored once, and every descriptive record reduces to a citation of one identifier. The cataloger no longer tracks N occurrences of a name but one authorized identity; a global rename, disambiguation, merge, or split is a single edit whose effect is read off automatically wherever the identifier is cited. The space of "what could go wrong with this name, everywhere it appears" compresses to the small, checkable question of whether one preferred form per entity exists in scope with every variant routed through its identifier — the qualitative health of the catalog's naming reads off that single registry rather than from auditing each entry in turn.

Abstract Reasoning¶

Holding "one governed identity per entity, cited not copied" as the object of analysis licenses a handful of concrete cataloging inferences.

Diagnostic — read a hidden identity fault off a surface symptom. When the same authorized heading retrieves resources that biographical or contextual evidence says belong to two different people, infer that one record is silently conflating two entities and is due for a split; the surface signature (a heading whose scope note cannot consistently disambiguate its own attached resources) points to a structural fault in the registry, not in any one descriptive record. The mirror inference runs the other way: when searches for several variant strings — "Clemens," "Twain," "Snodgrass" — fail to converge on one author, infer either a missing authority record or unlinked duplicate records, i.e. variants never routed through a single identifier. And when a global rename has been entered yet stale forms persist in some descriptive entries, infer those entries copied the name string rather than citing the identifier — they subscribed to nothing, so nothing propagated to them. The reasoning is: from a retrieval anomaly (over-collapsed, under-collapsed, or partially-updated headings) to the specific deferral failure (conflation, missing record, duplicate records, or copied-not-cited reference) that produced it.

Interventionist — predict what one registry edit will and will not move. Because descriptive records resolve through the identifier, the analyst reasons forward from a single authority-file edit to its catalog-wide effect: change the preferred form once and every citing entry displays the new form automatically; merge two records found to describe one entity and their formerly-split resource sets unify under the surviving identifier; split one record found to cover two entities and the previously-conflated resources must be re-pointed — the one edit that does not propagate silently, because each affected resource now needs reassignment to the correct identifier. The predicted effect is precise in scope: it reaches exactly the entries that cite the edited record and no others, so the intervention's blast radius is the citation set, which is itself enumerable from the identifier. Conversely, to predict that a contemplated change will fail to propagate is to predict that the affected entries hold copied name strings — the same fact the diagnostic move detects after the fact.

Boundary-drawing — decide where ambiguity belongs and whether the discipline even applies. Faced with a heading that retrieves a confusing mix, the cataloger decides between two branches: if a scope note (dates, profession, place) can stably separate the cases under one genuine identity, the ambiguity belongs in the record as disambiguating context; if no scope note can, the case has crossed the line into two entities and demands a split. That branch decision — note versus split — is the recurring adjudication the concept makes crisp. At a coarser grain, the concept draws the boundary of its own applicability: a system qualifies as authority-controlled only if it maintains a curated preferred form with every variant routed through a persistent identifier under governance; a bare primary key, a canonical-URL tag, or an ad hoc nickname list that lacks the controlled-vocabulary and change-history layers falls outside the regime, so the inferences above (clean propagation, scope-note-versus-split adjudication, lossless cross-walk between registries) cannot be assumed to hold of it.

Order-of-resolution — when identity must be settled relative to description. The deferral discipline implies a sequencing constraint: the authorized identity must exist (or be created) before a descriptive record can correctly cite it, so a resource catalogued for an entity with no authority record yet is reasoning ahead of its registry and will either spawn an uncontrolled heading or have to be re-pointed once the record is made. Reasoning from this order tells the cataloger that the health of the catalog's naming is settled at the registry first and reads out into the descriptive layer second — never the reverse.

Knowledge Transfer¶

Within bibliographic control the concept transfers as mechanism, with its full role-set and diagnostics intact. The authority-record discipline — a curated registry of one preferred form per entity, a governed variant set routed through a persistent identifier, scope-note disambiguation, change history, and the deferral rule that descriptive records cite the identifier rather than copy the name — carries across the library/archive/museum/scholarly-publishing family without translation: the Library of Congress Name Authority File, the Virtual International Authority File clustering national-library records, the Getty vocabularies (ULAN, AAT, TGN), ORCID for researcher identities, ISNI for public identities. A cataloger trained on LCNAF reads the same diagnostics (a heading conflating two people is due for a split; non-converging variant searches signal a missing or duplicate record; stale forms after a global rename mean entries copied the name instead of citing it) and the same interventions (one registry edit propagates to the citation set; merges unify resource sets, splits require re-pointing) in any of them. These are co-instances of one governed-identification discipline, not analogies — each enforces one preferred form per entity, routes variants through the identifier, and resolves ambiguity by governance, exchanged via shared formats (MARC authority, BIBFRAME) under shared standards (RDA, ISAAR(CPF)).

Beyond the descriptive-cataloging tradition the honest reading is case (B): the structural insight that generalizes — one canonical identifier per entity, with variants and cross-references resolving to it — is a more-general mechanism already carried by the parent primes identification (the general process of distinguishing entities) and representation (the model–referent relation), and that is what recurs across genuinely distinct domains. Close cousins really do exist as co-instances of the canonical-identifier-with-variants pattern: DNS (a canonical name with alias records), database surrogate keys with natural-key lookups, scientific nomenclature (the International Code of Zoological Nomenclature governing preferred binomials with synonyms), chemistry's CAS registry numbers, and package-manager namespaces. But the tell that the named authority record does not travel is that each cousin captures only some of the role-set — an identifier, sometimes a variant set — while dropping the rest. What does not travel out is the home-bound cargo that makes an authority record its own thing: the four library-tradition-bound commitments assembled in one artifact — a curated registry under active institutional maintenance, a controlled-vocabulary policy selecting one preferred form by rule, descriptive-standard compliance (RDA, AACR2, ISAAR), and exchange formats (MARC authority, BIBFRAME) — plus the scope-note-versus-split adjudication and the curated preferred-label discipline. Strip any of those and you no longer have an authority record but a thinner pattern (a primary key, a canonical-URL tag, a nickname list) that the parent primes already cover. So when the cross-domain lesson is needed — make identity resolvable to one canonical referent, route variants through it — carry the identification/representation parent, not the bibliographic "authority record," whose controlled-vocabulary-plus-descriptive-standard layer is a governed domain implementation that does not survive extraction. Calling a DNS canonical name or a CAS number an "authority record" borrows the resolve-to-one-identity shape while dropping the curation and standards machinery that give the original its force; that is the analogy boundary, and should be marked so (see Structural Core vs. Domain Accent).

Examples¶

Canonical¶

The Library of Congress Name Authority File record for the author Mark Twain is a defining specimen. It carries a persistent identifier (an LC control number that catalog records cite), an authorized preferred heading — "Twain, Mark, 1835-1910" — and a governed set of variant forms routed to that one identity, including his legal name "Clemens, Samuel Langhorne, 1835-1910" and pseudonyms he published under. Birth and death dates in the heading act as scope notes distinguishing him from any namesake. Because U.S. catalog records for his books cite this record rather than independently spelling out a name, a search for "Clemens, Samuel L." and a search for "Twain, Mark" both resolve to the same authorized identity, and every holding attaches to one governed entity rather than scattering across variant strings.

Mapped back: The author is the governed entity; the LC control number is the persistent identifier and "Twain, Mark, 1835-1910" is the preferred label. "Clemens" and the pseudonyms form the variant set routed through the identifier; the dates are the disambiguating scope notes. The Library of Congress under RDA is the curating authority, and catalog records citing the number rather than re-spelling the name is the deferral-and-propagation guarantee.

Applied / In Practice¶

ORCID does the same governed-identity work for living researchers, doing real work against the acute problem of name collisions in science (many researchers share a name like "Wei Wang" or "J. Smith," and people change names or affiliations). Each researcher is assigned a persistent 16-digit ORCID iD; journals, funders, and repositories increasingly require authors to cite that iD on submissions, grants, and datasets rather than relying on the name string. A publication is then bound to the person's canonical identity, not to a spelling. When a researcher changes surname or institution, the update is made once in the ORCID record and the linkage holds — every work already tied to the iD still resolves correctly — so the scholarly record disambiguates two same-named authors and survives a name change without re-editing every citing document.

Mapped back: The researcher is the governed entity and the 16-digit iD is the persistent identifier; recorded name variants and former names are the variant set. Publishers and funders citing the iD rather than the name enact the deferral-and-propagation guarantee — one record edit (a name or affiliation change) reaches the citation set — and ORCID as institution is the curating authority resolving common-name ambiguity by governance rather than convention.

Structural Tensions¶

T1: Single point of truth versus single point of error. The deferral discipline's payoff is that one registry edit propagates automatically to every citing record — a global rename is effected once. But propagation is symmetric: one wrong edit propagates just as cleanly, so a mistaken merge, a bad preferred-form choice, or an erroneous disambiguation instantly corrupts the entire citation set, and the authority becomes a chokepoint whose errors have catalog-wide blast radius. Copied name strings, for all their redundancy, at least fail independently; a governed identity fails collectively. The tension is that the centralization which makes correction efficient makes error efficient too, so the same architecture that lets one edit fix everything lets one edit break everything, and the value of the single point of truth is inseparable from the concentrated risk of the single point of failure. Diagnostic: Does routing everything through this one record make corrections cheap enough to justify that a single bad edit would propagate to the whole citation set?

T2: Governance resolves ambiguity versus governance imposes a contested canonical name. Enforcing one preferred form per entity resolves collisions and scatter — the discipline's core service. But "one preferred form by rule" is a choice, and choosing it exercises real power over how an identity is represented: which transliteration, a person's chosen name versus legal versus former name, a contested or indigenous-versus-colonial place name, an author's deadname. The concept insists it means "authoritative reference," not normative authority — yet selecting the single canonical heading is itself a normative act with stakes for the named. The tension is that the disambiguation which makes the catalog tractable is also an imposition the catalog cannot avoid making, so the registry that merely "resolves ambiguity" is simultaneously deciding, for everyone downstream, whose version of a name is authorized. Diagnostic: Is the preferred form here an uncontroversial technical choice, or a contested representation the governance is imposing under the cover of neutral disambiguation?

T3: Cite-not-copy propagation versus the standalone legibility indirection sacrifices. Citing an identifier rather than copying the name is what lets renames, merges, and splits propagate cleanly — the deferral rule that makes the record a point of truth. But the indirection couples every descriptive record to the authority infrastructure: the name resolves only if the registry is available, maintained, and the link intact, so a citation becomes a dangling pointer if the authority dies, the identifier is reassigned, or the network to it is lost. A copied string, by contrast, is legible standalone forever, needing nothing. The tension is that citation buys clean propagation at the cost of self-sufficiency — the governed record is more maintainable but less durable in isolation, betting the legibility of the whole catalog on the continued life of the registry it defers to. Diagnostic: Is the propagation benefit of citing the identifier worth making every record depend on the authority remaining alive, reachable, and stable to be resolvable at all?

T4: The note-versus-split branch as crisp adjudication versus fuzzy, shifting identity. The concept makes the recurring adjudication crisp: if a scope note can stably disambiguate, ambiguity stays in the record; if not, it signals two entities and a split. Decidable in principle. But whether one entity or two is often a genuinely hard, contestable, and evolving question — a pseudonym that grows into a separate authorial persona, a corporate body that merged then split, identity that is a matter of degree rather than a binary. The registry forces a call that reality resists, and merges and splits get redone as scholarship shifts. The tension is that the clean branch the concept supplies presupposes that entity-hood is determinate, while the cases that actually trigger the adjudication are precisely the ones where it is not, so the crisp rule is applied where its input is irreducibly fuzzy. Diagnostic: Is "one entity or two" here a settled fact a scope note or split can capture, or a contested, drifting identity the binary adjudication will have to redo?

T5: Curated governance as the source of quality versus its cost, latency, and unscalability. What makes an authority record more than a bare primary key is the curated, standards-compliant governance layer — controlled vocabulary, scope notes, change history, human adjudication. That curation is the distinctive value. But it is slow, expensive, and does not scale: the order-of-resolution constraint means an entity with no record yet spawns an uncontrolled heading, and the curation backlog leaves vast swaths of the real world ungoverned, so the discipline's quality is gated by a maintenance capacity that cannot keep pace with the entities needing governance. The tension is that the human curation which gives the authority record its force is exactly the bottleneck that limits its coverage, so the more rigorous the governance the fewer entities it can actually reach — quality and scale trade directly against each other. Diagnostic: Does the entity at issue actually have a curated record yet, or is the governance capacity so far behind the need that most naming here is happening outside the authority discipline entirely?

T6: Autonomy versus reduction (a bibliographic artifact or the identification/representation parents). The authority record is a fully specified bibliographic construct — curated preferred label, governed variant set, scope notes, descriptive-standard compliance (RDA, ISAAR), exchange format (MARC, BIBFRAME), change history — and across the library/archive/museum/scholarly-publishing family (LCNAF, VIAF, Getty, ORCID, ISNI) it transfers intact as mechanism. But the structural insight that generalizes — one canonical identifier per entity, with variants and cross-references resolving to it — is carried by the parents identification and representation, and the cousins (DNS canonical names, CAS registry numbers, database surrogate keys, biological nomenclature) each capture only part of the role-set while dropping the curated controlled-vocabulary-plus-standards layer. The tension is that the cross-domain lesson (make identity resolvable to one canonical referent) belongs to the parents while the governed-curation apparatus stays home. Diagnostic: Resolve toward identification/representation when carrying the resolve-to-one-canonical-identity lesson across domains; toward the authority record when a curated registry governs one preferred form per entity under descriptive standards in situ.

Structural–Framed Character¶

The authority record sits at the framed-leaning position — an evaluatively neutral construct that is nonetheless deeply an artifact of a maintaining institution, which keeps it well off the structural side and clearly framed of a mechanism like isostasy that runs observer-free. The five criteria split with a lopsided tilt. On evaluative weight it reads structural: an authority record renders no verdict and praises or blames nothing — it is descriptive infrastructure, a governed identity, not a normative judgment, and the entry is careful that "authority" here means authoritative reference, not the legitimate right to command. But on human-practice-bound it reads firmly framed: the record is constituted by the practice of bibliographic control — curated by a maintaining institution, adjudicated by human catalogers, kept alive by ongoing governance — and it dissolves the instant that practice is removed, leaving at most a bare string; nothing about it runs in a world without the cataloging tradition, unlike a rebounding lithosphere. On institutional origin it is as framed as an entry gets: the authority record is literally an artifact of named institutions (LCNAF, VIAF, Getty, ORCID, ISNI) operating under descriptive standards (RDA, ISAAR) and exchange formats (MARC authority, BIBFRAME) — an invention of a tradition, not a fact of nature anyone discovered. On vocab-travels it reads framed: the full role-set vocabulary — preferred label, controlled-vocabulary policy, scope notes, descriptive-standard compliance, change history — is pinned to the library/archive/museum substrate, and stripping it leaves only a thinner identifier pattern. On import-vs-recognize the profile is bimodal in the entry's own terms: within bibliographic control the discipline is genuinely recognised as the same mechanism across LCNAF, VIAF, Getty, and ORCID, but beyond it only the resolve-to-one-identity shape travels by analogy, while cousins like DNS and CAS numbers capture part of the role-set and the portable content is carried by the parents.

The portable structural skeleton is one canonical identifier per entity, with variants and cross-references resolving to it — a governed model-referent binding that lets identity be cited rather than re-copied. That skeleton is genuinely substrate-spanning, but it is exactly what the authority record instantiates from its umbrella primes identification (distinguishing entities) and representation (the model-referent relation), not what lets "authority record" itself travel: the cross-domain reach to DNS canonical names, surrogate keys, and biological nomenclature belongs to those parents, while the domain-accented specifics — the curated controlled-vocabulary layer, the descriptive-standard compliance, the exchange formats, the note-versus-split adjudication — stay home and cross only as analogy. Its character: an evaluatively neutral but thoroughly institution-constituted cataloging artifact whose only substrate-spanning content is already carried by its umbrella primes identification and representation, wrapped in a curated-registry-plus-standards apparatus that pins it to bibliographic control — framed-leaning, structural only in the canonical-identifier skeleton it borrows from its parents.

Structural Core vs. Domain Accent¶

This section decides why the authority record is a domain-specific abstraction and not a prime, and it carries the case for its domain-specificity along with it.

What is skeletal (could lift toward a cross-domain prime). Strip bibliographic control and a thin relational structure survives: one canonical referent per entity, with variant forms and cross-references resolving to it, cited rather than re-copied, so that a single edit at the referent reaches exactly its citation set. The portable pieces are abstract — a set of distinguishable entities, a designated canonical stand-in for each, a family of variant surface forms bound to that stand-in, and a deferral rule under which downstream records point at the referent instead of holding their own copy. That skeleton is genuinely substrate-portable, which is precisely why it recurs beyond libraries as DNS canonical names with alias records, database surrogate keys with natural-key lookups, CAS registry numbers, and the binomials of biological nomenclature — and why the entry reads it as an instance of the parents identification (distinguishing one entity from all others) and representation (the governed model–referent binding). But it is the core the authority record shares, not what makes it its own thing.

What is domain-bound (cannot peel away without becoming a looser thing). Almost everything that makes the concept an authority record in particular is library-tradition furniture, and none of it survives extraction. Four commitments assemble in a single maintained artifact: a curated registry kept alive by active institutional maintenance, a controlled-vocabulary policy that selects one preferred form by rule, descriptive-standard compliance (RDA, AACR2, ISAAR(CPF)), and exchange formats (MARC authority, BIBFRAME). Around them sit the scope-note disambiguation, the change history logging corrections, and the note-versus-split adjudication that decides whether an ambiguity stays inside one record or signals two entities and a split. The decisive test is the entry's own: strip the curated controlled-vocabulary-plus-standards layer and what remains is a database primary key, a canonical-URL tag, or an ad hoc nickname list — a thinner identifier pattern the concept does not reduce to. The cousins named above each keep some of the role-set and drop the rest; only the full assembly, governed and human-curated, is an authority record.

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. The authority record's transfer is bimodal. Within the library/archive/museum/scholarly-publishing family it moves as mechanism, its full role-set and diagnostics intact — LCNAF, VIAF, the Getty vocabularies (ULAN/AAT/TGN), ORCID, and ISNI are co-instances of one governed-identification discipline, and a cataloger trained on one reads the same merges, splits, and propagation rules in any other. Beyond that family it travels only by analogy: calling a DNS canonical name or a CAS number an "authority record" borrows the resolve-to-one-identity shape while dropping the curation and standards machinery that give the original its force. And when the bare cross-domain lesson is needed — make identity resolvable to one canonical referent, route variants through it, edit once — it is already supplied in more general form by the parents the record instantiates: identification and representation. The substrate-spanning reach belongs to those parents; "authority record," as named, carries bibliographic baggage that does not and should not travel.

Relationships to Other Abstractions¶

Current abstraction Authority Record Domain-specific

Parents (7) — more general patterns this builds on

Authority Record is a kind of Representation Prime

An Authority Record is a Representation whose target is one entity and whose governed record medium preserves identity-relevant labels, notes, and relations.
Authority Record is part of, typical Alternative label Domain-specific

Authority records typically contain governed Alternative Labels for aliases, transliterations, former names, pseudonyms, and misspellings.
Authority Record is part of Authorized access point Domain-specific

The sanctioned heading string is a strict internal constituent of the full Authority Record, which additionally stores variants, notes, identifiers, and history.
Authority Record is part of, typical Cross-reference Relation Domain-specific

Cross-reference Relation is typically an internal constituent of Authority Record; removing that role breaks part of the child's defining mechanism.
Authority Record is part of Persistent Identifier Prime

A Persistent Identifier is the record's strict machine-stable constituent, cited by descriptive records so name edits propagate without rewriting citations.
Authority Record is part of Versioning Prime

Versioning is an internal constituent of the maintained record: corrections, preferred-form changes, merges, splits, and retirements are logged over time.
Authority Record is a decomposition of Alias-to-Authority Mapping Prime

Strip bibliographic standards from an Authority Record and one maintained identity with many variant routes remains—the Alias-to-Authority Mapping core.

Hierarchy paths (42) — routes to 12 parentless roots

Authority Record → Representation → Abstraction

Show alternative paths (41)

Not to Be Confused With¶

Descriptive (bibliographic) record. The catalog record that describes a resource — a book, an article, a dataset — with its title, imprint, physical description, and subject headings. It is the entry that cites an authority record rather than being one: where the descriptive record says "by this author," the authority record fixes who that author canonically is. The relation is deferral, one-directional — a descriptive record subscribes to authority records, and a rename edited once in the authority propagates to it. Tell: does the record describe a resource you could hold or read (descriptive record), or fix the governed identity of an entity that many resources point to (authority record)?
Authority file / controlled vocabulary. The whole governed system — the maintained set of authority records under a controlled-vocabulary policy (LCNAF as a file, a subject thesaurus as a vocabulary). An authority record is a single unit within such a file; the file is the collection plus its rules of preferred-term selection and broader/narrower relations. The part-whole relation is explicit: the record is one governed identity, the file is the registry of all of them. Tell: are you naming one entity's governed entry (authority record), or the entire maintained registry-plus-policy it belongs to (authority file / controlled vocabulary)?
Persistent-identifier systems (DOI, Handle, ARK, URN). Infrastructures that mint and resolve stable identifiers, typically for documents, objects, or locations, guaranteeing a link keeps working. They supply the persistent-key role but not the curated controlled-vocabulary layer: a DOI resolves to a resource without a governed preferred label, variant set, scope notes, or note-versus-split adjudication. An authority record may use a persistent identifier, but the identifier alone is not the record. Tell: does the thing merely guarantee a stable pointer to a resource (PID system), or curate one preferred form per entity with governed variants and disambiguation (authority record)?
Descriptive standards and exchange formats (RDA, AACR2, ISAAR(CPF), MARC authority, BIBFRAME). These are the rulebooks and encodings that specify how an authority record is to be constructed and transmitted — how to form a preferred heading, which fields to use, how to serialize the record for exchange. They govern the artifact but are not the artifact: the standard is the recipe, the authority record is the dish. Compliance with RDA is a property an authority record has, not a synonym for it. Tell: is the thing a specification or serialization format that many records conform to (standard / exchange format), or the individual governed registry entry that conforms to it (authority record)?
A disambiguation page (e.g. Wikipedia's). A page that lists the several distinct entities sharing a name and routes the reader to choose among them. It is almost the inverse of an authority record: the disambiguation page presents ambiguity for human resolution and makes no canonical choice, whereas the authority record resolves the ambiguity by governance — fixing one preferred form per entity and, where two entities lurk under one heading, splitting them into separate records. Tell: does the artifact enumerate namesakes and leave the reader to pick (disambiguation page), or bind one entity to one authorized identity under curation (authority record)?
The DNS / surrogate-key / CAS-number cousins and the identification / representation parents it instances. These are the cross-domain co-instances of the thin skeleton — a canonical referent per entity with variants resolving to it — and the umbrella primes that carry it, not confusable peers. Each cousin keeps part of the role-set (an identifier, sometimes an alias set) while dropping the curated controlled-vocabulary-plus-standards layer; the portable "resolve to one canonical identity" lesson belongs to identification and representation, not to "authority record." Tell: strip the curated registry, controlled vocabulary, descriptive standards, and change history and what remains is a bare canonical-identifier pattern — at which point you are using identification / representation (or a cousin like DNS or a surrogate key), not an authority record. (Treated fully in a later section.)

Neighborhood in Abstraction Space¶

Authority Record sits in a moderately populated region (52^nd percentile for distinctiveness): it has near-neighbors but no dense thicket of look-alikes.

Family — Authority Control & Identity (11 abstractions)

Nearest neighbors

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