Guides
Part of People by heritage in one sourced place in 2027
5 points on reading people by heritage data
People by heritage data is a chain of documented links, each a record naming both parties, with negative evidence kept and no percentage at the end.
A documented descent is a chain of separate links, each one a record that names two people and states the relationship between them. Break any link and the chain below it is unsupported, however good the rest looks.
This page is about building and storing that chain: what counts as a link, what to do when two records disagree, and why the output is never a percentage.
What to take away
- One link is one record naming both parties and stating the relationship, not a name matched across two documents.
- Negative evidence has to be recorded, or the same fruitless search gets repeated for years.
- Conflicting records are kept, both of them, with a note on which is preferred and why.
- A chain of links cannot be converted into a fraction of anything.
What makes a link, and what does not
| Counts as a link | Does not count |
|---|---|
| A record naming a child and a parent together | Two records with the same name in the same county |
| A will naming a relationship in its own words | A shared surname and a plausible date gap |
| A record of a life event that states a relationship | An index entry with no image and no citation |
| A record naming a party and their origin place | A family story about where the family came from |
The distinction is whether the document itself asserts the relationship. Matching two names across two documents is your inference, not the record's, and it belongs in a notes field until a document supports it.
Storing the chain
Each link needs six values, and they fit on one row.
- The two parties, by internal identifier rather than by name.
- The relationship as the document words it.
- The record type and series.
- The repository and reference, precise enough for someone else to order it.
- The date of the record, with precision.
- Your confidence, as established, probable, or unresolved.
Store links, not people. A person is then a query over links, which means a corrected link updates every conclusion that rested on it instead of leaving a stale summary somewhere. Storing links instead of people is exactly the design that makes a people by heritage list have four different owners rather than one.
Negative evidence is evidence
A search that found nothing is a result worth keeping: which series, which years, which index, which spellings tried, and the date searched. Without it, the same ground gets covered repeatedly and an absence never accumulates into anything.
It also protects you from the wrong conclusion. A gap can mean the event was not registered, that registration had not started in that jurisdiction, that the record was lost, or that access is restricted. National agencies document those causes series by series, including in their guidance on naturalization records and on how immigrant records were created in the first place. Read the scope note before turning a gap into a claim.
When two records disagree
Keep both. The instinct to pick the tidier one destroys the only signal you have that something is wrong.
A workable resolution note is three sentences: what each record says, which is closer to the event and to a party who knew the facts, and what would settle it. Prefer the record created nearest the event by someone with direct knowledge, and say that is what you did. Where nothing settles it, the field stays unresolved, which is a real state and not a failure.
Why there is no percentage
Each generation back doubles the number of ancestral slots, and past a few generations the slots outrun any plausible population, so lines fold back on each other. That folding is pedigree collapse, and it is the reason a fraction has no stable meaning: the answer depends entirely on how many generations you counted and which lines happened to leave papers.
A chain of links supports a different, stronger sentence: this person is connected to that person by these six documented links. That sentence names its evidence and survives scrutiny. "Part something" does neither.
Nothing here produces a claim about a living person's ancestry, ethnicity, nationality, or religion. Those are covered on the heritage pillar, which sets out why this site publishes none of them. For how the place fields in these records behave, see the country pillar and the US state pillar; for the dating problems, the historical figures pillar.
Common questions
Is a well-sourced family tree on a genealogy site a link?
No. It is a lead, sometimes an excellent one. Follow it to the records it cites and record those. A tree without images or citations is somebody's conclusion.
How do I handle an adoption in the chain?
As two different relationships, legal and biological, stored separately, and neither published about a living person. The legal record is a record of parentage; it is not research material about origins.
What if the only evidence is a transcript, not the original?
Record it as a transcript, name the transcriber, and mark the confidence accordingly. Transcripts introduce errors at predictable points, especially in dates and in names that crossed a script.
Can I publish the chain itself?
For people long dead and documented, yes, with citations. For anyone living, no. The people in the chain did not choose to be research subjects, and their appearance in your file is not consent.
