How do two ancestors get merged into one person?
Two men of the same name in one county rarely get merged by a single dramatic error. It happens through a run of small acceptances: a hint that matched on a name and a state, a tree merge, a census household that looked close enough. The result is one person assembled out of two lives.
The symptoms are recognisable once you know them. A man who marries at nineteen and again at seventy-one, two states apart. Land bought in two counties in the same week. Children spread across thirty-five years. A death date followed by more records. Any of those means you are probably holding two people, not one unusually eventful ancestor.
The fix is to split first and re-attach afterwards. Create both candidates, give each a label you will recognise, then move every record onto one of them for a stated reason. What refuses to attach to either is the interesting part.
What did Senior and Junior mean to a clerk?
In tax lists and deeds, Senior and Junior usually distinguished the elder and the younger man of one name in one district, with no relationship implied at all. An uncle and nephew, or two unrelated men, could carry the labels. Reading Jr. as "son of" has built entire branches that never existed.
The labels also move. A man is Junior while the elder lives and Senior afterwards, and a third arrival can push both down a place. That makes a suffix useful for dating a record and useless for defining a family. The same applies to "the elder" and "the younger" in English parish records.
How do you build a timeline that separates two men?
Build a separate timeline per candidate, one line per record, with date, place, and source. Identity questions are usually settled by physical impossibility rather than by resemblance: one man cannot be enumerated in two households on one day, or father a child after his own estate was settled.
- Plot every dated record, including the ones you think belong to the other man.
- Treat these as disqualifying: two places on one date, an event before birth or after death, an age that moves backwards across censuses, a wife who appears while a previous wife is still recorded as living.
- Do not treat these as qualifying: the name matches, the state matches, the birth year matches.
That last point deserves emphasis. If you searched on a birth year, then records agreeing on that birth year tell you nothing about identity; they only tell you the search worked. Corroboration has to come from something the search did not specify.
Can geography tell two same-name ancestors apart?
Plot every residence on a map with dates attached. Same-name men usually differ in where they were rather than in what they were called. A record whose place contradicts every place your person is known for, and which names none of the relatives you already have, belongs to someone else until something ties it in.
Judge distance in period terms. Thirty miles in 1790 was a different decision than it is now, and a family that appears in two counties is more often two families than one that commuted. Deeds and land records give the most precise locations available, down to the adjoining owners, which is why they resolve identity questions that census districts cannot.
One caution before you split on geography: check whether the county line moved rather than the family. Boundary changes can make one farm look like two places in two jurisdictions, a trap covered in our brick walls guide.
Why do the witnesses settle identity questions?
The other names on a record are the strongest discriminator available: bondsmen, marriage witnesses, baptism sponsors, executors, appraisers, adjoining landowners, and the households either side in a census. Mills' FAN principle (friends, associates, and neighbors) treats identity as something the surrounding cluster proves.
In practice one candidate turns out to move through life with the same handful of people: the neighbor who witnesses three deeds across twenty years, the brother-in-law who appraises the estate, the same two families standing as sponsors at every baptism. The other candidate has a different cluster. Once you can see both clusters, most of the disputed records assign themselves. Mills' QuickLesson on identity problems works a case of exactly this shape.
Which records prove that two people existed?
Some record sets enumerate a population exactly once: one household per census, one probate file per estate, one entry per taxable head in a district per year. When both candidates appear in the same list in the same year, you have proof of two men. When only one does, you have a question.
Annual tax lists are the workhorse here. They are name-by-name, they repeat every year, and the year a name arrives or disappears dates a migration, a coming of age, or a death. They also let you separate two men by district in years when nothing else does. When a name stops appearing in a list that survives complete, that absence is evidence in its own right rather than a failed search, which is the distinction our brick walls guide sets out.
How does AncestorIQ separate namesakes?
Deep Research runs an identity check on every source before its people can reach your tree, deciding between your person and a same-name stranger. A source is only dropped when that judgment is highly confident, because a wrong drop costs you a real record and a wrong keep corrupts a line.
- Individually impossible facts go first, with no model in the loop: a death for someone still living, an event dated before the person was born, a birth outside the known window.
- A record whose state contradicts every place your person is known for, and which names none of the relatives already in your tree, keeps its facts but loses its people, so a stranger's spouse and children are never proposed as your ancestor's.
- Agreement on the birth year never counts as corroboration, because the search was anchored on that year to begin with.
- When an attempt to expand a family produces relatives whose dates cannot fit the person's lifespan, the whole expansion is rejected rather than half-attached.
- Every dropped source is recorded with its reason, so you can look at what was excluded and overrule it.
We built that because namesake handling is where automated research fails hardest, and the failure is invisible: a confident tool hands you a coherent, sourced, wrong person. See how Deep Research works for the pipeline, and our AI genealogy guide for what to check before you accept anything an AI tool proposes.
How do you record the split once you have made it?
Once you separate the men, record the separation: label each one, list which records belong to which and why, and state what conflicting evidence you resolved. The Board for Certification of Genealogists treats resolution of conflicting evidence as a required part of a sound conclusion, not an optional footnote.
Label by something distinguishing rather than by number: Thomas Reid of Chester County and Thomas Reid the weaver will still mean something to you in a year, where Thomas Reid (1) will not. Keep the rejected records attached to the rejected candidate instead of deleting them, so the next person to look (including you) can see the case rather than just the verdict.
