ARCHIVAL DESCRIPTION · METADATA · FINDING AIDS · PROVENANCE · RECORDS IN CONTEXTS · DISCOVERY

How Archival Description and Metadata Work

An archive can preserve a million records and still fail a researcher if nobody can tell what those records are, where they came from, how they relate, what they contain, what restrictions apply, or how to find them.

Preservation keeps the record alive. Description teaches the future how to find and understand it.

Archival description is the system by which records are represented through titles, dates, creators, histories, scopes, relationships, identifiers, conditions of access and other metadata. It creates intellectual control over material that may be physically distributed across boxes, repositories, servers and formats.

Unlike ordinary cataloguing, archival description often begins from relationships. Records are produced by people and organisations performing activities. They arrive in groups. Their order may carry evidence. A single item may make little sense outside the series, office or process that created it.

The short answer

RECORDS
  → IDENTIFY CREATOR
  → PRESERVE PROVENANCE
  → UNDERSTAND FUNCTION / ACTIVITY
  → ARRANGE OR REPRESENT RELATIONSHIPS
  → ASSIGN IDENTIFIERS
  → DESCRIBE COLLECTION / SERIES / FILE / ITEM
  → RECORD DATES / EXTENT / SCOPE / RIGHTS / ACCESS
  → CREATE AUTHORITY CONTEXT
  → PUBLISH FINDING AID / METADATA
  → INDEX / LINK
  → SEARCH / BROWSE / GRAPH
  → USER RETURNS TO RECORD

The description is therefore not the record. It is a structured map of the record world.

1. Why archives need description

Large collections are impossible to navigate by memory. Staff change. Buildings change. Box locations change. Digital systems change. Description externalises knowledge that would otherwise disappear with the people who processed the collection.

Good description gives both humans and machines durable anchors.

2. Description is not transcription

Transcription reproduces the text of a record. Description represents the record and its context.

A handwritten letter might be fully transcribed yet badly described if the archive does not record its creator, date, recipient, collection, provenance and relationship to surrounding correspondence.

3. Description is not summary alone

A summary can tell us what a document discusses. Archival description must often tell us more: why it exists, who created it, where it belongs, what dates it spans, whether access is restricted, and how it relates to other records.

Subject is only one dimension of archival meaning.

4. Provenance is the first organising intelligence

Provenance keeps the origin of records visible. Records produced by one creator are understood in relation to that creator and the activities through which they arose.

This prevents a misleading reorganisation in which every record is detached from its administrative or personal history and reduced to topical fragments.

5. Original order preserves process

Where a meaningful original order exists, preserving or representing it can reveal workflow. A sequence of application → review → approval → contract → implementation may carry evidentiary meaning beyond the contents of any one file.

Original order is not blind obedience to disorder. Archivists document when order is missing, damaged, reconstructed or multiple.

6. Archival description is often hierarchical

Archives commonly describe records from broader groups to narrower units:

FONDS / RECORD GROUP / COLLECTION
  → SERIES
  → SUBSERIES
  → FILE
  → ITEM

Not every archive uses every level. The hierarchy expresses intellectual relationships and allows broad contextual description to be inherited by more specific levels.

7. Why not catalogue every item?

Item-level description can be valuable, but it is expensive. A collection may contain millions of pages or files. Describing every item before making anything accessible can leave the entire collection closed for years.

Archival description therefore balances depth against access. More detailed description is added where value, risk, demand or complexity justifies it.

8. A finding aid is a map, not the territory

A finding aid describes a body of records and helps users navigate it. It may include collection title, creator history, dates, extent, scope, arrangement, series descriptions, container lists and access conditions.

The finding aid is itself an interpretive object. It reflects descriptive decisions, terminology and knowledge available at the time it was written.

9. Titles should identify without inventing

Some records have formal titles. Others need supplied titles. A supplied archival title should distinguish the record clearly without pretending the creator used words that were invented later by the archivist.

Good description makes intervention visible rather than disguising it as original evidence.

10. Dates need precision and uncertainty

Archival dates can refer to creation, accumulation, publication, capture, digitisation or other events. The description should make the date type clear.

When dates are estimated, uncertainty should be represented. False precision is not better metadata.

11. Extent tells users the scale

Extent may be measured in boxes, folders, linear metres, volumes, items, files, bytes or duration. It helps researchers and archivists understand how large the record body is.

Extent is operational metadata as well as descriptive metadata: it affects storage, transfer, digitisation planning and research expectations.

12. Scope and content describe what the records cover

A scope-and-content note explains the activities, subjects, record forms, people, places and chronological coverage represented.

It should help the researcher decide whether the collection is worth deeper investigation without collapsing the entire record body into a simplistic abstract.

13. Biographical and administrative history restore context

Records often outlive the institutions and people that created them. A concise biographical or administrative history can explain who the creator was, what functions they performed, how offices changed and why the records exist.

This is especially important when organisational names change across decades.

14. Authority records separate people from spellings

One person may appear under initials, full name, married name, title, pseudonym or transliteration variants. Authority control links these name forms to a stable identity.

This lets a search system recognise that variant strings may refer to the same person while preserving historically accurate name forms inside the records.

15. Organisations also need authority control

Departments merge, ministries are renamed, companies restructure and committees dissolve. If every historical name is treated as an unrelated keyword, institutional continuity becomes hard to see.

Authority metadata can record predecessor, successor, parent and related bodies.

16. Identifiers make records addressable

A stable identifier lets staff and researchers refer to the same collection or record without relying on title text alone.

Identifiers are especially important when titles change, records move physically, descriptions are revised or systems exchange data.

17. Location should not be identity

A shelf number, box location or server path tells us where an object currently sits. It should not be the only identity of the record.

Locations change. Stable identifiers allow the record to move without becoming a different record.

18. Access conditions belong in metadata

A researcher needs to know whether records are open, restricted, partly redacted, awaiting review, limited by donor conditions or subject to other permissions.

Restriction metadata should record reason and review state where appropriate. “Closed” without explanation can become permanent through neglect.

19. Rights metadata is different from access metadata

A record may be open for viewing but still protected by copyright. Another may be in the public domain but restricted because it contains sensitive personal data.

Access asks who may see the record. Rights asks what may lawfully be done with it.

20. Condition and preservation information can affect access

A fragile physical record may require a surrogate. A corrupted digital file may have limited renderability. A nitrate negative may need specialised storage and handling.

Preservation metadata therefore influences the access route.

21. Descriptive metadata has layers

Metadata layer	Typical job
Descriptive	Helps identify and discover the record.
Administrative	Supports management, ownership and access.
Technical	Describes formats, equipment and digital characteristics.
Preservation	Records fixity, transformations, events and preservation history.
Structural	Explains relationships among components.
Rights	Records copyright, permissions and restrictions.

Different schemas may divide these layers differently. The important principle is that long-term interpretation depends on more than subject keywords.

22. Metadata is evidence about evidence

Metadata may record who described the object, when, under what standard, what source supplied a date, and what changes were later made.

That makes metadata itself part of the provenance system. A catalogue correction should not become an invisible rewrite when the change materially affects interpretation.

23. Archival standards create interoperability

Standards allow institutions to exchange and understand description more consistently. They provide common concepts and structures while still allowing local implementation choices.

Historically, international archival description used standards such as ISAD(G), ISAAR(CPF), ISDF and ISDIAH. The International Council on Archives is now developing and implementing a more integrated framework called Records in Contexts.

24. Records in Contexts changes the shape of description

The International Council on Archives describes Records in Contexts, or RiC, as a standard designed to support more contextual, flexible, accurate, dynamic and connected archival description.

Instead of relying only on a single hierarchy, RiC models records alongside people, groups, positions, activities, places, dates and relationships.

25. RiC has complementary parts

The ICA framework includes RiC-FAD for foundations, RiC-CM for the conceptual model, RiC-O for the ontology, and RiC-AG for application guidance.

As of 2026, RiC-FAD and RiC-CM 1.0 are stable releases; the ICA lists RiC-O 1.1 as the latest official ontology release, while the current RiC Application Guidelines are published as version 0.1 draft guidance for practitioner feedback.

26. Why a graph can express archival reality better than one tree

A traditional hierarchy is powerful for representing collection structure. But real archival relationships cross branches. One person can serve several organisations. One function can move between departments. One event can generate records across multiple creators.

A graph model can represent these intersecting relationships without forcing every fact into one tree.

PERSON
  → HELD POSITION
  → IN ORGANISATION
  → PERFORMED ACTIVITY
  → CREATED RECORD SET
  → DOCUMENTED EVENT
  → RELATED TO PLACE
  → CONNECTED TO OTHER RECORDS

27. Linked data makes relationships computable

RiC-O expresses archival description as an ontology that can support RDF and linked-data implementations. This lets systems represent entities and relationships in ways machines can query and connect.

Linked data does not remove the archivist. It gives archival judgement a richer machine-readable form.

28. Machine-readable does not mean machine-decided

Metadata extraction can be automated. Dates can be parsed. Names can be suggested. File formats can be detected. OCR can generate text.

But archival description includes interpretation: which creator matters, what function produced a record, whether two identities are the same, how uncertain dates should be expressed, and what relationships deserve representation.

29. Controlled vocabularies reduce accidental fragmentation

If one catalogue uses “Second World War”, another “World War II” and another “WW2” without relationships between terms, retrieval fragments.

Controlled vocabularies and authority systems help connect variants while preserving historically appropriate wording in the record itself.

30. Historical language should not be silently modernised

Archival records may contain outdated, offensive or discriminatory terminology. Description has to balance historical fidelity, discoverability, community impact and contemporary professional standards.

A responsible system can preserve original titles or quotations where evidentially necessary while adding contextual notes and modern access terms rather than erasing the historical record.

31. Description should be corrigible

Archives learn. A person in a photograph may be identified decades later. A date may be corrected. A newly discovered accession may change provenance. Community knowledge may expose a misdescription.

Description should therefore support revision while preserving enough history to understand material changes.

32. Citizen description can extend knowledge

Public transcription, tagging and identification projects can add information that staff alone could not provide at scale.

Community contributions still need provenance: who suggested the description, whether it was reviewed, and how confidence is represented.

33. OCR should be a discovery layer, not the archival truth layer

OCR makes scanned records searchable, but its errors can be severe in handwriting, old typography, damaged pages and multilingual documents.

Search indexes can use OCR aggressively while interfaces preserve a route back to the scanned image or original record.

34. AI-generated description needs provenance

AI can draft summaries, extract entities and suggest keywords. These outputs should be marked as generated or reviewed when that distinction matters, with confidence and source routes retained.

A fluent generated description should never become more authoritative than the archival evidence from which it was derived.

35. AI-ready archives need record-level return paths

A machine should be able to traverse from question to description to record and back:

QUESTION
  → ENTITY / TOPIC / DATE
  → COLLECTION
  → SERIES
  → RECORD ID
  → CREATOR / PROVENANCE
  → ACCESS STATE
  → DIGITAL OBJECT
  → CITABLE EVIDENCE

The archival graph should make retrieval stronger without hiding the original context.

36. Singapore: description makes national records discoverable

The National Archives of Singapore Archives Online provides discovery routes across archival records including government files, photographs, maps, building plans, oral histories and audiovisual materials.

The public interface illustrates an important point: users may search individual records, but archival access remains connected to custody, permissions, reference numbers and collection context.

37. Description also manages access expectations

NAS guidance distinguishes records available online from records that require requests, clearances or reading-room access. Description therefore performs a service function: it tells the researcher not only what exists, but what can be done next.

This turns metadata into routing infrastructure.

38. Failure modes

Failure	What breaks
Describe only by subject	Creator, function and record relationships disappear.
Item catalogue everything first	Backlogs prevent broader access.
Location used as identity	Records lose continuity when moved.
No authority control	Name variants fragment discovery.
Restriction with no reason	Closed status may become indefinite.
Metadata rewritten silently	Cataloguing history becomes invisible.
OCR treated as exact text	Machine errors become false quotations.
AI description without provenance	Synthetic interpretation becomes indistinguishable from archival description.
One rigid hierarchy	Cross-cutting relationships remain hidden.
Graph without archival context	Linked data becomes a network of decontextualised facts.

39. A practical description checklist

Identify the record creator or source.
Understand the activity or function that produced the records.
Preserve meaningful provenance and order.
Assign stable identifiers.
Describe from broad context to useful detail.
Record dates with uncertainty honestly.
State extent and material or digital form.
Explain scope and content.
Record access and rights separately.
Use authority control for people and organisations.
Preserve relationships between collections, series, files and items.
Make material corrections traceable.
Support both search and structural browsing.
Keep machine-generated metadata tied to source records and review states.

40. The deeper model: metadata is a navigation system through time

Metadata is sometimes dismissed as administrative detail. In an archive it is much more. It is the mechanism that allows a record to remain findable after offices close, people die, technologies change and physical locations move.

The record survives materially. Metadata preserves the routes around it.

A record without context may survive as an object. A record with good description survives as evidence.

41. Why this matters to civilisation

Civilisations create more records than any individual can read. Their memory therefore depends on representations: catalogues, finding aids, indexes, authority files, graphs and metadata.

When those representations are weak, collective memory becomes inaccessible. When they are strong, future generations can traverse the evidence instead of merely knowing that boxes or files exist somewhere.

Source and authority routes

Continue the Archives and Publishing series

Publication control: Wintour House · eduKate Publishing · evidence, metadata, edition, correction and archive gates.

World Return: When you preserve a record, preserve its routes too: who created it, what activity produced it, where it belongs, which restrictions apply, what changed, and how a future reader can travel from a question back to the evidence.

How Archival Description and Metadata Work | From Provenance and Finding Aids to Records in Contexts and Machine Discovery