MODULE 2 · PROVENANCE · 35–50 MINUTES

Provenance: source the source

Provenance means tracing where something came from and the context in which it was created, preserved and transmitted. Senior research becomes much stronger when you stop asking only “What does this say?” and start asking “How did this evidence get here?”

WHO CREATED IT?Who wrote, spoke, photographed, measured, recorded or otherwise produced the original evidence?
WHEN?When was it created? Is that different from the date it was uploaded, transcribed, republished or interviewed?
WHERE / IN WHAT CONTEXT?What place, event, institution or situation shaped its creation?
WHAT FUNCTION OR PURPOSE?What was the source doing — recording, persuading, administering, remembering, selling, protesting? State purpose only where the source/context supports it; do not mind-read.
FOR WHOM?What audience was addressed or expected? Public, officials, whānau, voters, researchers, customers — or is this unknown?
HOW DID IT REACH ME?Original item, archive scan, transcription, quotation, screenshot, repost, search result, summary or AI-mediated output?

Creator, custodian and mediator are different jobs

A historical petition displayed by the National Library was not created by the modern web page that hosts it. A newspaper article found through Papers Past was created by a historical publication, preserved and digitised by institutions, then delivered to you through a modern interface.

For difficult sources, name three roles separately:

Sometimes one organisation fills more than one role. Sometimes a role is unknown. Do not collapse them just because one logo appears at the top of the page.

Transmission is part of provenance

A claim can travel through many layers before you see it. Every layer can select, crop, summarise, translate, rank or rephrase.

Original record / testimony / measurement
↓ preserved, archived, transcribed or digitised
Scholarship or journalism
↓ summarised, quoted or indexed
Search / platform / social feed
↓ selected or synthesised
Generative AI output
↓ copied into notes
Your argument

The more mediated the claim becomes, the more important it is to reconstruct the chain before treating it as evidence. Mediation does not automatically make a source poor; hidden mediation makes the limits harder to inspect.

Practice 1: a screenshot is not a provenance chain

Imagine someone sends you a screenshot containing a dramatic quotation attributed to a politician. The screenshot has no visible date or link.

  1. List every provenance question you cannot answer from the screenshot alone.
  2. What search terms could help you locate the original speech, interview, transcript or post?
  3. If you find ten reposts but no original, what has your confidence actually gained?
  4. What would count as the strongest accessible carrier?

Practice 2: creator or carrier?

Case: you discover an 1893 newspaper editorial through Papers Past after following a link from an NZHistory page returned by Google.

Identify separately:

  1. the historical creator/publication;
  2. the modern custodian/digitisation layer;
  3. the orientation source that sent you there;
  4. the search layer that made the orientation page visible;
  5. which of those carriers you would cite for a claim about what the 1893 editorial actually said.

Practice 3: the same words, different provenance

Version A: a sentence copied from an archive scan with creator, date, collection reference and page number.

Version B: the same sentence appears in an image macro on social media with no link.

Version C: the same sentence appears in an AI answer that says “a contemporary observer wrote…” but gives no source.

The wording might be identical. The evidentiary status is not, because your ability to inspect context and authenticity differs.

Your task: explain what you would need to do before using each version in formal senior work.

Do not invent purpose

“This source was made to deceive people” may be a hypothesis, not a demonstrated fact. Prefer evidence-based wording such as: “The source advocates for X, addresses Y audience and selects evidence that supports Z.” If the creator's private intention is unknown, say so.

The same rule will matter later with AI: record the user's prompt/task and the observable output rather than pretending a model has a human-like private motive you can source.

Practice 4: map a real provenance chain

Take one claim from a current assignment, article, video or AI conversation. Use the Trace the Claim Home tool and move upstream until you reach the strongest accessible evidence.

Do not stop merely because you find a citation. Open it. Check whether the source actually supports the claim being made.

Year 13 extension: provenance and power

Provenance is not only technical metadata. Institutions decide what is recorded, preserved, described, digitised and discoverable. A beautifully catalogued state record may have strong documentary provenance while still representing the viewpoint and administrative priorities of the state that produced it.

Extension: explain why “well-preserved” and “complete representation of reality” are different claims.

Mastery check

Before moving on, take one source and answer all six provenance prompts without guessing. Then identify its creator, custodian and any mediator you used. If one answer is unknown, write unknown and state how you would try to find out. Honest uncertainty is stronger than invented context.