Museum methodology

How We Know

Great museums put their methods on display — conservation labs behind glass, provenance notes on the placards. This page is our glass wall. Every verdict in these galleries is built the same way: trace the evidence, rank it, state the confidence honestly, and show the work.

The museum's standard is simple to say and demanding to meet: a quote is not false because the internet says so, and not true because everyone repeats it. Each artifact is authenticated by tracing an evidence trail — earliest known appearance, claimed source, verified source, textual comparison — never by repeating internet consensus.

The evidence ladder

Not all sources are equal. When our researchers authenticate an artifact, each piece of evidence is weighed by where it sits on this ladder:

1. Primary sources — the original film, book, speech, letter, manuscript, broadcast, or official transcript. The words themselves, where they were first fixed.
2. Contemporary documentation — newspaper reports, reviews, and records from the time the words were said.
3. Institutional sources — national archives, library collections, publisher and estate archives.
4. Scholarly sources — peer-reviewed research, university press books, academic biographies.
5. Reputable secondary sources — established journalism, serious biographies, citation-backed quotation references.
6. Derivative sources — later summaries and explainers; useful leads, weak proof.
7. Unsourced quote sites — aggregators, meme pages, quote graphics. These can show a misquote spreading, but they are never accepted as evidence of what was actually said.

Seen through the glass: everyone remembers “Houston, we have a problem” from Apollo 13. The primary source — NASA's official air-to-ground voice transcription of April 13, 1970 — records something slightly different: “Okay, Houston, we've had a problem here” (Jack Swigert), then “Ah, Houston, we've had a problem” (Jim Lovell). The present-tense version was popularized by the 1995 film, where the tense was changed for dramatic effect. A transcript outranks a blockbuster, no matter how many times the film line is repeated — that's the ladder doing its job.

What the confidence levels mean

Every artifact carries a confidence level stating how strongly the available evidence supports the museum's verdict. We would rather tell you the honest strength of the record than pretend to certainty history doesn't offer — preserving uncertainty is a founding principle here, not a hedge.

• High — the verdict is strongly supported by primary, institutional, or excellent secondary evidence. The corrected wording can be checked directly against a verified source, and the trail is strong enough for another researcher to reproduce the finding.
• Likely — reliable sources agree and nothing serious contradicts them, but one part of the origin trail remains incomplete. We say so.
• Medium — the verdict is plausible but incomplete, disputed, or built on indirect evidence. These exhibits carry clear uncertainty notes and say “current evidence suggests,” because that is exactly what we mean.
• Low — the quote is suspicious, but the evidence is weak or fragmentary. Artifacts like this stay in research rather than on display, unless the exhibit is explicitly about an unresolved mystery.
• Unknown — the museum cannot responsibly classify the artifact yet. No verdict is published.

Before any exhibit opens, it passes a Museum Standards review — quote accuracy, source quality, and confidence language are checked line by line, and a human curator signs off. Nothing goes on display without that gate.

When we get something wrong

A museum that never expects to issue a correction isn't being rigorous — it's being optimistic. Our corrections policy:

• Every exhibit shows its work. Each artifact has a public provenance record listing the evidence and sources behind it, so you can check our reasoning against the record yourself.
• Confirmed errors get fixed at the source first. When new evidence changes a verdict, the exhibit itself is corrected with a visible correction note, then every other surface that repeated it is updated.
• A human approves every correction. Corrections are never automated — a curator reviews the new evidence and signs off, permanently.
• Nothing is silently overwritten. Revisions are recorded in the artifact's version history: the old verdict, the new one, and why it changed.
• Verdicts have review dates. Published artifacts are re-reviewed on a schedule, because the evidence record grows even when nobody is looking.

Think we got one wrong?

Good — that's how the collection improves. Check the provenance record on the exhibit, and if the evidence points somewhere we didn't, write to the registrar. And if you know a famous quote that deserves authentication, the acquisitions desk is open: suggest an artifact.