Every researcher has had this experience: a published family history states a fact with total confidence — a birth date, a birthplace, a maiden name — and it turns out to be wrong. Not maliciously wrong. Just wrong, in the ordinary way that happens when one researcher copies another researcher’s conclusion, who copied it from someone else, who misread a document decades earlier and never went back to check. The fact travels for a century, gaining authority with every retelling and losing accuracy the whole time. Understanding why this happens — and how to guard against it — starts with a distinction genealogists learn early and then have to keep relearning: the difference between an original source and a compiled one.

Three Kinds of Sources, Not Two

The earlier discussion of Evidence Explained in this series covered the distinction between original sources (the first recording of information) and derivative sources (copies, transcriptions, abstracts, and indexes built from an original). Genealogist Thomas W. Jones adds a useful third category to that framework: narratives.^[1] A narrative is an authored work — a published genealogy, a family history, a journal article, a lineage-society application, a research report — that synthesizes and interprets other sources rather than simply reproducing them.

This third category matters because narratives behave differently from either original or derivative sources. A derivative source, done carefully, aims to preserve the original information faithfully — a transcription error is a mistake, not a design feature. A narrative, by contrast, is built out of a researcher’s judgment calls: which sources to trust, how to resolve a conflict, which relationship to infer from indirect evidence. Those judgment calls can be excellent, and often are. But they are still judgment calls, made by a fallible person, and every published genealogy inherits whatever those judgment calls got wrong.

Primary, Secondary, and Beyond: A Quick Definition

Source type (original, derivative, narrative) answers one question: what kind of object is this? A separate question asks something different: how directly does the information inside it connect to the event it describes? That’s the primary/secondary/tertiary distinction, and it cuts across source type rather than lining up neatly with it.^[2]

Primary information comes from someone with firsthand, contemporary knowledge of the event — ideally recorded at or near the time it happened. A physician’s note on a birth certificate, written the day of the delivery he attended, is primary information about that birth. A soldier’s own diary entry describing a battle he fought in that afternoon is primary information about the battle. A photograph taken at the moment of a wedding is primary information about who attended.

Secondary information comes from someone without firsthand knowledge, or is recorded well after the fact, typically relying on memory, hearsay, or someone else’s earlier account. A birth date recorded on a death certificate eighty years later, supplied by a grieving adult child working from memory, is secondary information about that birth — even though the certificate itself, as a source, is an original record. A family Bible entry made years after a child was born, or an obituary’s account of where the deceased was born, are both secondary information for the same reason: distance in time and lack of direct observation, not the format of the record.

Beyond that lies tertiary information — material compiled or summarized from secondary (and sometimes primary) sources, often several steps removed from anyone with direct knowledge of the underlying event. Genealogical encyclopedias, general historical reference works, published indexes of other researchers’ genealogies, and the automated “hint” systems that surface other users’ trees on genealogy websites are all tertiary in this sense. They’re genuinely useful for orientation, context, and leads — a tertiary source pointing toward a promising record is worth following up. They are rarely usable as evidence for a specific fact on their own, precisely because so many unseen judgment calls sit between them and anyone who actually witnessed the event.

The same document routinely contains more than one level at once. A death certificate carries primary information about the death itself (the attending physician was there) and secondary information about the deceased’s birth date and parents’ names (supplied by a relative working from memory) — a point discussed in more detail in the companion piece on Evidence Explained. Recognizing this is often more useful than trying to label an entire document as simply “primary” or “secondary” — the real question is always which specific fact you’re trying to establish, and who was actually in a position to know it.

Why Published Genealogies Are Almost Always Secondary

For nearly every event a published family history narrates — a birth, a marriage, an immigration — the family history itself is a secondary source, and often a tertiary one. The author almost certainly wasn’t there. What the author had was other sources: some original, some derivative, some other people’s narratives, woven together into a coherent story.^[3] That synthesis step is exactly where things go wrong, and it goes wrong in three predictable ways: a transcription slip carried forward from an earlier source, a plausible-sounding inference presented as settled fact, or a prior researcher’s error accepted uncritically because it fit the story being told.

A published genealogy that cites original sources for each claim effectively strengthens those claims. One without citations offers secondary information with no way to trace it back to an informant who was actually there.

That single distinction — cited versus uncited — is the most useful filter a researcher has when evaluating any compiled genealogy, including a century-old county history, a self-published family book from the 1970s, or a stranger’s public tree online. A citation doesn’t guarantee the conclusion is right. It does mean the conclusion can be checked.

The Telephone Game

The mechanism by which a single error becomes an accepted “fact” is depressingly simple, and every experienced genealogist has watched it happen. A local history published in the 1890s misreads a faded church register and records a birth year one digit off. A family genealogist in the 1950s cites that county history rather than the original register, not unreasonably — it was the more accessible source, and no one had reason to doubt it. That 1950s book gets digitized decades later, indexed, and uploaded to an online tree-building platform. Thousands of users, each doing their own honest research, find the tree, see a birth year with no obvious red flags, and add it to their own trees with a click. The error is now attached to thousands of family trees, most of which cite each other rather than the one faded, still-existing church register that actually answers the question — and would show a different year, if anyone went back to look.

None of the people in that chain did anything careless by the standards of their own moment. The 1950s genealogist trusted a published local history, which is a completely reasonable thing to do when you don’t know it contains an error. The online tree users trusted a genealogy that, by the time they found it, looked like an established fact rather than one person’s conclusion. The error compounds not because anyone was lazy, but because compiled sources look exactly as authoritative whether or not they’re correct, and nothing about a confidently stated fact reveals how many uncritical copies stand between it and the original record.

Treating the DFA’s Own Family Group Sheets as What They Are

It would be inconsistent to lay all of this out and exempt the DFA’s own work from it. A Family Group Sheet is a narrative source in exactly Jones’s sense: a compiled synthesis, built from original records where they survive, from derivative sources like online indexes where originals aren’t accessible, and inevitably from some judgment calls about which conflicting piece of information to trust. That makes every Family Group Sheet in this database a compiled source pointing toward the original records behind it — not a replacement for them, and not a final answer simply because it has an official-looking group number.

This is the same reasoning behind conventions already in place across this series: transcribing source typos verbatim rather than silently correcting them, so a future researcher checking the original record isn’t misled by a well-intentioned fix; flagging genuine data conflicts for human review rather than quietly resolving them one way; and treating every statistical finding in the “What the Numbers Tell Us” series as provisional, open to revision as the underlying dataset grows or a closer read turns up an oversight. A compiled source that behaves this way — showing its work, flagging its own uncertainty, and pointing back toward the records it was built from — is doing the only thing a narrative source can honestly do.

For Dillman Researchers

• Ask whether a genealogy cites its sources before you trust its conclusions. An uncited claim in a published family history is someone’s best guess, however confidently it reads.
• When a Family Group Sheet cites a source, follow it back to the original if you can. The sheet is there to point you toward the record, not to substitute for reading it yourself.
• Treat agreement between multiple online trees with suspicion, not confidence. A hundred trees that all trace back to the same one uncited source are one data point, not a hundred.
• When you find a conflict between a compiled genealogy and an original record, trust the original — and note the conflict rather than silently picking a side, in case a later source resolves it differently than either one.
• If you find an error in one of the DFA’s own Family Group Sheets, that’s not a failure of the project — it’s exactly the kind of correction a compiled source is supposed to invite. Bring it to the DFA so the sheet can be revised.

A compiled genealogy, done well, is one of the most useful tools available to a family historian — it saves enormous time, and it often gets things right. The discipline isn’t to distrust it. It’s to remember, every time, that it’s a claim about the past built by a fallible person out of other sources, and that the only way to know how much to trust it is to be able to trace it back to where it came from.

Sources

1. Thomas W. Jones, “Assessing Genealogical Sources,” in Professional Genealogy: Preparation, Practice & Standards, ed. Elizabeth Shown Mills, excerpted at Genealogical.com (accessed July 2026).
2. “Primary vs. Secondary Sources in Genealogy Research,” Genealogy Authority (accessed July 2026).
3. “A Checklist of Compiled Sources and Where to Find Them,” FamilySearch Research Wiki (accessed July 2026).
4. “How to Evaluate Sources in Genealogy Research: Original or Derivative?” Family Tree Magazine (accessed July 2026).

AI Disclosure: This article was researched and written with the assistance of Claude, an AI assistant developed by Anthropic. General claims about source classification are drawn from the sources cited above; DFA-specific claims are drawn from the association’s own prior published articles and Family Group Sheet Database.

Notes

1. References to the DFA’s Family Group Sheets database and its published conventions are drawn from the Dillman Family Association, Family Group Sheet Database, compiled 2026, and from the DFA’s own published articles.