ARCHIVES · RECORDS · EVIDENCE · MEMORY · PUBLIC KNOWLEDGE

How Archives Work

An archive is not simply a room full of old things. It is a continuity system that decides what records survive, preserves the relationships that make them intelligible, protects their integrity, describes them so they can be found, and creates controlled pathways from past action to present evidence.

Records are created because something happened. Archives exist so that, after the people and systems that created those records have moved on, we can still ask what happened, who acted, what was known, what changed, and how we know.

That sounds simple until we look at the scale of the problem. Governments create files, email, maps, databases, photographs, audiovisual recordings and transaction logs. Families create letters, photographs and messages. Companies create contracts, designs and accounts. Schools create registers, examination records and policies. Scientists create notebooks, data and instrument outputs. Publishers create manuscripts, proofs, editions and correction histories. Most of this cannot be kept forever. Some of it should not be kept forever. Some of it must be protected for decades. Some becomes essential evidence long after its original purpose has disappeared.

The archive therefore sits at an unusual junction between memory and forgetting. It is not a passive warehouse. It is a governed process of selection, custody, description, preservation and access.

The short answer

ACTIVITY
  → RECORD CREATED
  → RECORD MANAGED
  → RETENTION / APPRAISAL
  → SELECT FOR PERMANENT VALUE
  → TRANSFER / ACCESSION
  → PROVENANCE PRESERVED
  → ARRANGE & DESCRIBE
  → PRESERVE
  → CONTROL RIGHTS & ACCESS
  → DISCOVERY
  → USE AS EVIDENCE
  → REINTERPRETATION
  → CONTINUITY

The chain matters. A photograph with no date, creator or context may still be interesting, but it is weaker evidence than the same photograph embedded in a documented record series whose creator, purpose, custody and relationships are known. A spreadsheet whose formulas changed silently may contain numbers, but its evidentiary integrity is different from a versioned dataset with preserved metadata and checksums. Archives work by protecting not only content, but context.

1. What is an archive?

The word archive is used in several ways. It can mean an institution, a collection of records, a repository, a storage function, or an action such as archiving email. Professional archival practice is more precise. Archives are records preserved because of their continuing value, together with the systems and professional work needed to maintain their meaning, integrity and accessibility over time.

The International Council on Archives describes records and archives as part of the evidentiary and cultural heritage of humanity. That combination is important. Archives are not only about heritage. They are also about proof. A government decision file, land record, court record, scientific notebook or signed agreement may have historical value, but it may also establish rights, obligations, chronology and accountability.

2. Archive, library and museum are not the same thing

Libraries, archives and museums overlap, but their organising logics differ.

Institution	Typical object	Primary organising question
Library	Published books, journals, media, databases	What work is this, what is it about, and how can a reader find it?
Archive	Records created by people, organisations and activities	Who created these records, in what activity, in what relationship, and what must remain intact for them to function as evidence?
Museum	Material objects, specimens, artworks and artefacts	What is this object, what is its material and cultural context, and how should it be preserved and interpreted?

A published annual report may live in a library as a publication and in an archive as part of the record of the organisation that produced it. A camera might enter a museum as an artefact while the photographer’s negatives, contracts and correspondence enter an archive. The same world can therefore be represented through different institutional lenses.

3. Archives begin before anything becomes “old”

One of the most useful misconceptions to remove is that archives begin at the end of a record’s life. Good archival outcomes begin when records are created. If a digital system produces files with poor metadata, allows silent overwriting, loses creator identity, stores timestamps inconsistently or depends on proprietary software with no export path, an archivist years later may inherit a damaged evidentiary environment.

Records management and archival preservation therefore meet at the beginning. Naming conventions, identifiers, retention schedules, access controls, audit trails, export formats, rights metadata and versioning may look operational today, but they determine what can still be understood tomorrow.

4. Record creation: the event leaves a trace

A record is usually created as part of an activity. A meeting produces minutes. A transaction produces an invoice and ledger entry. A building project produces drawings, approvals, contracts and photographs. A scientific experiment produces protocols, instrument readings and analysis. A publication produces a manuscript, edits, proofs, metadata and a released edition.

This is why archival thinking begins with function rather than nostalgia. The question is not merely, “Is this old?” It is, “What activity created this record, what did the record do, what relationships does it document, and what would be lost if it disappeared?”

5. Records cannot all be kept forever

Every large organisation produces far more information than can sensibly receive permanent archival preservation. Keeping everything sounds safe, but it creates other problems: cost, privacy exposure, discovery failure, duplicated versions, meaningless noise and preservation burden. The discipline therefore needs a principled way to distinguish temporary records from records of enduring value.

This process is called appraisal. Appraisal asks which records should be retained permanently, which should be kept for a defined period, and which may be disposed of when legal, administrative and operational obligations have been met.

6. Appraisal is an argument about future value

Appraisal is difficult because the archivist must make a decision in the present about possible uses in the future. Records may carry administrative value, legal value, evidential value, informational value, cultural value, research value or value for communities whose importance was not recognised by the record creator.

A strong appraisal process therefore avoids the fantasy that selection can be perfectly neutral. Decisions are documented, criteria are explicit, the wider body of records is considered, and specialists may be consulted where the significance cannot be judged from archival expertise alone. This matters because what is not preserved may become impossible to reconstruct later.

7. The silence of the archive

Archives are powerful partly because they preserve, and dangerous partly because preservation can be mistaken for completeness. A surviving record proves that something survived. It does not prove that all relevant voices were recorded or preserved.

Institutions tend to generate abundant records about activities they administer. People with less formal power may appear mainly through records created about them. Oral history, community archives, private papers and ephemera can sometimes widen the record, but every archive remains shaped by creation, survival, appraisal, acquisition, law, resources and historical accident.

An archive is evidence. It is never the whole past.

8. Transfer and accession: custody becomes explicit

When records move into an archival repository, the transfer should be documented. The archive needs to know what arrived, from whom, under what authority, with what restrictions, in what physical or digital condition, and with what rights.

Accessioning is the formal process of taking records into custody and registering them. It creates an initial control point: identity, source, date, extent, conditions, ownership or custody status, and the basic relationship to a collection or record group.

9. Provenance: keep the creator visible

One of the defining principles of archival practice is provenance. Records from one creator or source should not simply be mixed with records from another as though they were interchangeable subject fragments. The creator and originating context matter because they help explain why the records exist and what they mean.

The Society of American Archivists describes provenance in relation to keeping records of the same origin together and understanding the context of the activities and functions in which they were created. This is fundamentally different from taking thousands of documents and reorganising every item by topic.

10. Original order: relationships can be evidence

A second major principle is original order: where meaningful and recoverable, the arrangement created by the record maker is preserved or intellectually represented because that arrangement can reveal relationships, workflows and priorities.

Imagine a project folder organised as proposal → approval → contract → field reports → invoices → completion certificate. Rearranging every page alphabetically by subject might make some words easier to browse, but it could destroy the sequence that shows how the project moved from authorisation to completion.

Original order is not worship of disorder. Records may arrive scrambled, damaged or without a coherent system. The professional task is to identify which relationships are evidentially meaningful, document interventions, and avoid inventing a false historical order.

11. Arrangement: archives are usually described collectively

Large archives cannot always catalogue every sheet individually. Archival description often works hierarchically: collection or fonds → series → subseries → file → item. The appropriate depth depends on the records, resources, risk and research value.

This collective logic is efficient, but it also reflects archival meaning. A single memo may make much more sense when seen inside the series of records produced by the office that created it.

12. Description: the archive builds a map

Preservation without discovery creates a dark archive: material may survive but remain practically invisible. Description creates the map. Titles, dates, creators, scope notes, administrative histories, relationships, access conditions, identifiers, subjects and container information help researchers move from a question to relevant records.

A finding aid is therefore not the archive itself. It is a representation of the archive. It can be excellent, partial, outdated or biased by the description practices of its time. Researchers should read it as a guide, not as a perfect mirror of the collection.

13. Preservation: survival is engineered

Physical records face heat, humidity, light, pests, mould, pollutants, acidic materials, mechanical damage, water, fire and handling. Preservation reduces risk through suitable enclosures, storage conditions, environmental monitoring, disaster planning, conservation treatment and careful use.

Preservation does not mean freezing an object into perfect permanence. Nothing material is permanent. The goal is to extend usable life, slow deterioration, stabilise vulnerable items, reduce avoidable risk and preserve enough physical and contextual integrity for future use.

14. Digitisation is not the same as digital preservation

Scanning a document can improve access and reduce handling of a fragile original. It does not automatically solve preservation. Digital files have their own risks: bit corruption, storage failure, format obsolescence, broken links, lost metadata, software dependency, expired credentials and undocumented transformations.

Digital preservation treats a file as an object with identity and history. Common controls include checksums or other fixity information, redundant storage, monitored integrity, documented formats, metadata preservation, controlled migration, emulation where appropriate, access copies separated from preservation masters, and logs of significant actions.

15. Born-digital records change the scale

A born-digital archive may contain millions of files, email threads, databases, websites, social media captures, source code and multimedia. The record may not be a single self-contained document. It may depend on relationships between tables, software, permissions, external services and time-sensitive links.

This changes archival work from “keep the file” to “preserve enough of the environment, structure, metadata and behaviour to retain meaning.” Sometimes that means normalising formats. Sometimes it means preserving original binaries as well as transformed copies. Sometimes it means documenting what cannot be preserved exactly.

16. Authenticity, integrity and reliability are different questions

A record can be authentic but wrong. A genuine memorandum may contain an incorrect claim. A photograph may be original yet misleadingly framed. A database export may accurately preserve corrupted source data. Archives protect authenticity and integrity; researchers still have to evaluate truth and interpretation.

Authenticity: is this what it purports to be?
Integrity: has it remained complete and unaltered in relevant ways, or are transformations documented?
Reliability: was it created through a process dependable enough for the purpose for which we are using it?
Accuracy: are the facts or measurements correct?
Context: can we understand who created it, why, when and within what activity?

17. Access is part of preservation’s purpose

An archive that can never be used may preserve material but fail part of its public purpose. Access turns preservation into research, accountability, education, memory and cultural return.

Yet access is not always immediate or unlimited. Records may contain personal data, confidential information, national security material, culturally sensitive content, copyright restrictions or donor conditions. Archives therefore balance openness with law, ethics, rights and legitimate restrictions.

18. Restriction should have a reason and a state

Good access control is not simply “open” or “closed”. It records why a restriction exists, who may authorise access, whether redaction can enable partial access, when review should occur, and what event changes the status.

OPEN
OPEN_WITH_CONDITIONS
RESTRICTED_PRIVACY
RESTRICTED_RIGHTS
RESTRICTED_SECURITY
RESTRICTED_DONOR_TERMS
CLOSED_FOR_PROCESSING
REVIEW_DUE
RELEASED_AFTER_REVIEW

This is particularly important in digital systems, where a vague restriction can become permanent simply because nobody knows why it was imposed.

19. Search is not the archive

Modern researchers often meet archives through a search box. Search is useful, but it can flatten hierarchy and privilege records with better metadata or machine-readable text. A keyword may miss handwritten material, variant spellings, historical terminology, undigitised files or records described only at series level.

The expert researcher therefore alternates between keyword search and structural navigation: finding aids, creator histories, record groups, series descriptions, accession information and adjacent collections.

20. OCR creates access, not certainty

Optical character recognition can make scanned text searchable, but OCR output is a derived representation. Old typography, damaged paper, complex layouts, handwriting and multilingual material can produce errors. A search hit based on OCR should therefore return to the image or original record when the exact wording matters.

The same principle applies to automated transcription and AI-generated summaries. Derived text can accelerate discovery. It must not silently replace the archival object as the evidentiary source.

21. Oral history is an archive of remembered experience

Oral history shows why archival evidence is richer than paper. An interview records lived experience, memory, interpretation, voice and perspective. It can preserve knowledge that never entered formal institutional records.

But memory is not a mechanical recording device. Time, later knowledge, trauma, social context and interview framing influence recollection. Oral history is therefore powerful evidence of experience and memory that should be contextualised and, where relevant, triangulated with other sources.

22. Photographs are records, not windows

A photograph can feel immediate, which makes it easy to forget mediation. Someone chose where to stand, what to include, when to release the shutter, which image to keep and how to caption it. Cropping, sequencing, retouching and later metadata can alter interpretation.

Archival description can preserve photographer, date, location, original caption, collection context, negative or file identity and related images. Those relationships turn a picture from a floating illustration into stronger evidence.

23. Maps and plans are arguments about space

Maps, cadastral records and building plans often become crucial long after their operational use. They may document boundaries, infrastructure, design intent, demolished places or changes in land use. Their scale, projection, edition, annotations and creator matter. A map is not the territory; it is a purpose-built representation of the territory.

24. The archive needs a disaster plan

Fire, flood, burst pipes, mould outbreaks, ransomware, hardware failure and human error can turn preservation into emergency response. A resilient archive plans before the incident: risk registers, priorities, contact trees, backup locations, salvage supplies, vendor arrangements, recovery procedures and regular tests.

For digital archives, a backup is necessary but not sufficient. A backup that cannot be restored, whose credentials are lost, or whose corruption was copied faithfully into every replica is not resilience. Recovery testing is part of preservation.

25. Archives need succession, not just storage

A collection may outlive the software, institution, staff and funding structure that first preserved it. Long-term stewardship therefore needs succession planning. Who owns the records? Who can transfer custody? What happens if a repository closes? Are formats exportable? Are identifiers portable? Is the preservation metadata understandable outside the current system?

This is the deeper reason archives are continuity systems. The archive must survive not only physical decay, but organisational change.

26. Singapore: archives as national memory and evidence

In Singapore, the National Archives of Singapore collects, preserves and manages public and private archival records relating to Singapore. Its holdings include government records, photographs, maps and building plans, oral histories, audiovisual materials, speeches, press releases and private records. Selected material is discoverable through Archives Online, while access to other records may require the Archives Reading Room or specific permissions.

This is an important distinction for learners: the national archive is not simply a website of old photographs. It is an institution that manages custody, preservation, description, access and continuity across multiple record forms.

27. Publishing creates another archive

Publishing is usually thought of as producing books, journals, websites and articles. But every publishing system also creates records: commissions, contracts, drafts, editorial correspondence, copy-edits, proofs, design files, rights clearances, metadata, release files, corrections and new editions.

When those records survive, future researchers can see not only what was published, but how the publication came into being. The archive can reveal choices that the finished edition hides: what was cut, challenged, corrected, reframed or rejected.

28. Legal deposit connects publishing to national memory

Legal deposit is one of the clearest bridges between publishing and archival continuity. In Singapore, the National Library Board collects and preserves Singapore publications as part of the country’s published heritage. The Legal Deposit system extends the preservation responsibility beyond what individual publishers happen to keep.

The principle is larger than compliance. Publication is an event in the present; legal deposit turns part of that event into a durable cultural record for future readers.

29. Archives and AI: retrieval becomes powerful, provenance becomes essential

AI can help transcribe, classify, extract entities, suggest descriptions, detect duplicates, translate, cluster large collections and answer questions across records. That is useful precisely because archives are large and unevenly described.

But AI also creates a new risk: an answer can become detached from its record. If a system synthesises five documents into one paragraph without preserving citations, dates, creator context and uncertainty, it may produce fluent knowledge while weakening archival evidence.

A good archival AI therefore needs a return path:

QUESTION
  → SEARCH
  → RECORD SET
  → PROVENANCE CHECK
  → ACCESS / RIGHTS CHECK
  → EXTRACT
  → SYNTHESISE
  → CITE RECORDS
  → SHOW UNCERTAINTY
  → RETURN TO ORIGINAL CONTEXT

The machine can help traverse the archive. It should not erase the archive.

30. What an archive can prove — and what it cannot

An archive can show that a particular record exists.
It can often show where the record came from and how it relates to other records.
It can preserve evidence of decisions, transactions, communications and observations.
It can reveal change across time through series of records.
It can support accountability and rights.
It cannot guarantee that every relevant record was created.
It cannot guarantee that every created record survived.
It cannot guarantee that preserved records are factually correct.
It cannot remove the need for interpretation, comparison and source criticism.

31. How to read an archive well

When you encounter an archival record, ask more than “What does it say?” Ask:

Who created it?
What activity produced it?
When was it created, and is that date certain?
For whom was it created?
What did the creator expect the record to do?
What series or collection does it belong to?
What came before and after it?
How did it reach this archive?
Has it been digitised, transcribed, redacted or otherwise transformed?
What restrictions or rights affect use?
What other sources could confirm, challenge or complicate it?
Whose perspective is missing?

32. The deeper model: an archive is a memory machine with brakes

It is tempting to think of an archive as a memory machine. That is only half right. A good archive also has brakes: appraisal prevents indiscriminate accumulation, access controls protect legitimate interests, provenance prevents false recombination, description records uncertainty, preservation logs transformations, and correction prevents cataloguing errors from becoming invisible facts.

The archive is therefore better understood as a governed memory system. It preserves enough of the past to make future questioning possible while recording the boundaries of what has survived.

33. Failure modes

Failure	What goes wrong
Keep everything	Noise, privacy exposure, cost and discovery collapse.
Keep only famous things	Everyday systems and marginal voices disappear.
Rearrange by modern subject	Creator relationships and original context can be destroyed.
Scan and discard context	Images survive while provenance, order and metadata disappear.
Trust OCR as text	Derived transcription errors become mistaken for the record.
Silent digital migration	Future users cannot distinguish original characteristics from transformations.
Permanent restriction	Closed records never reach scheduled review or release.
Search-only discovery	Undescribed, undigitised or poorly indexed records become invisible.
AI without citations	Fluent answers detach from evidentiary sources.

34. The archival loop

Archives are often drawn as a one-way pipeline from record creation to preservation. In reality, use creates new knowledge that can improve description, reveal misidentification, expose missing context and change appraisal priorities. Archives learn from researchers, communities and new evidence.

CREATE
  → MANAGE
  → APPRAISE
  → ACCESSION
  → PRESERVE
  → DESCRIBE
  → ACCESS
  → RESEARCH
  → NEW CONTEXT
  → DESCRIPTION / POLICY / COLLECTION IMPROVEMENT

35. Why archives matter to civilisation

Large societies depend on memory that outlives individual people. Law depends on records. Property depends on records. Public administration depends on records. Science depends on records of method and observation. Families depend on memory. Publishing depends on manuscripts and edition histories. Accountability depends on being able to reconstruct who decided what and when.

Without archives, every generation would inherit more stories than evidence. With archives, the past is still incomplete—but it remains interrogable.

The purpose of an archive is not to make the past silent and safe. It is to preserve enough evidence that the future can still ask difficult questions.

Source and authority routes

International Council on Archives — international archival profession, records and archival heritage.
Society of American Archivists Dictionary: Provenance — provenance and original-order terminology.
Describing Archives: A Content Standard — Statement of Principles — archival description and original-order principles.
National Archives of Singapore: Our Roles — Singapore archival functions, preservation and access.
National Archives of Singapore: Archives Online — public discovery route.
National Library Board: Legal Deposit — preservation of Singapore’s published heritage.

Continue through eduKateSingapore

World Return: If a record matters, do not ask only whether it has been saved. Ask whether its identity, context, rights, relationships and integrity have been preserved well enough for someone in the future to understand what it was and why it mattered.

How Archives Work | From Record Creation to Appraisal, Preservation, Access and Memory