Is your brick wall a missing record or a missing search?
A brick wall is the point where the searches that used to work stop returning your ancestor. There are only two causes underneath it: the record you want was never created or no longer survives, or it survives and your search cannot reach it. Most walls are the second kind.
The two need opposite responses, which is why the diagnosis comes first. A records problem means you stop hunting that document and start hunting whatever captured the same fact. A research-design problem means the document is sitting in an archive, or in a database you already pay for, filed under a spelling, a county, or a name you have not tried. Guess wrong and you can spend a year re-reading one index.
Work the design problems first. They cost an afternoon each to test, and they account for most stalled lines.
What do you actually know, and where did each fact come from?
List every fact you hold about the person, the source for each one, and how directly that source supports it. Walls hide inside unsourced facts inherited from someone else's tree. A surprising number of them turn out to belong to a different person, which is exactly why the search keeps failing.
The Board for Certification of Genealogists sets out five parts of a sound conclusion: reasonably exhaustive research, complete and accurate citations, analysis and correlation of the evidence, resolution of conflicting evidence, and a written conclusion that follows from all of it (see the board's ethics and standards). Reasonably exhaustive does not mean every record ever made. It means enough that you would notice if something important were missing.
The audit often breaks the wall on its own. A birth year that came from a cousin's tree, a birthplace that came from a transcription of a transcription: mark each fact with where it came from, and the unsourced ones become your research questions instead of your foundation.
Why does a search that finds nothing still matter?
Record every search that returns nothing: the database, the exact terms, and the years and places it covered. A search you cannot remember running is a search you will run again. Written negative results also become evidence once you can show the record set was complete for those years.
Genealogists separate two things that sound identical. Negative search results mean you looked and found nothing. Negative evidence is an inference drawn from a silence that should not be there. A man who appears on the county tax list every year for twelve years and then does not, in years where the list survives complete, has told you something: he died, he left, or he stopped owning what was taxed.
That second kind only works if the record set is intact. Which years survive, which pages are missing, whether the index covers the whole run: the boring questions are the ones that turn an empty search into an argument.
A usable log entry has five things in it: the date you searched, the database or archive and the specific collection, the exact terms and spellings, the years and places the collection actually covers, and the result. Five lines per search. The payoff arrives about six months later, when a new index goes online and you can tell in a minute which of your old searches is now worth repeating.
Could the record be filed under a spelling you have not tried?
Indexes are transcriptions of handwriting, so a record can be present and unfindable at the same time. Try phonetic and wildcard search, search a first name plus a place with no surname at all, and search for a spouse or child instead. Exact-match surname search is the most common reason a search fails.
Names drift for reasons that have nothing to do with your ancestor: a clerk wrote what he heard, a transliteration went one of several valid ways, a nickname stood in for a formal name. Our guide to ancestor name variants covers why, and the free nickname finder turns Polly back into Mary in either direction.
- Drop the surname entirely and search a given name plus a place and a year range.
- Search for a sibling, a spouse, or a child whose name indexed more cleanly.
- Browse the images for the right district page by page when the index is thin.
- Try initials only, and try the surname with the first letter wrong (a capital L read as an S, an H read as a K).
Did the county line move instead of your family?
Jurisdiction changes file records in a courthouse your ancestor never visited. US counties were created, split, renamed, and re-bordered constantly, so an 1830 event can sit in a parent county that no longer touches the town. Always check what the place was called on the day of the event.
The scale is easy to underestimate. The Newberry Library's Atlas of Historical County Boundaries maps roughly 17,700 versions of the 4,000-odd counties and county equivalents that have existed in the United States, because most counties have existed in more than one shape.
Practically: find which county held the ground on the event date, then find where that county's records went. Older records usually stay with the parent county rather than following the land, and state-level copies of tax and land records often survive when local ones do not. Border changes can cross state lines too, so a family that never moved can appear in two states (West Virginia separated from Virginia in 1863). If the terminology in the deeds and wills is the obstacle rather than the filing, our records glossary translates it.
What do you search when the obvious record does not exist?
Substitution is the core brick-wall skill: stop hunting the document and start hunting the fact. A birth date lives in a baptism, a census age, a marriage bond, a death certificate, a pension file, or a family Bible. Several independent weak sources that agree can carry a conclusion.
| The fact you need | Records that also carry it |
|---|---|
| A birth date and parents | Baptism entry, census ages, marriage record, death certificate, draft registration, pension file, delayed birth certificate |
| A wife's maiden name | Marriage bond and licence, children's birth and death records, her own death certificate, her father's will, baptism sponsors |
| Which of two same-name men is yours | Annual tax lists, deeds and their witnesses, timeline overlap, one-slot record sets like a single census household |
| Anything in a county that burned | Re-recorded deeds, state-level tax copies, church registers, newspapers, published abstracts made before the loss |
Each row has its own guide: can't find a birth record, how to find a maiden name, same-name ancestors, and researching around destroyed records.
Two rules keep substitution honest. The sources have to be independent: a county history, an old published genealogy, and an online tree that all say 1847 are usually one source repeated three times. And the closer a record sits to the event, in time and in the knowledge of whoever reported it, the more weight it carries. A baptism entered by the minister that month outranks an age given fifty years later by a son-in-law, even though both are just numbers on a page.
How do the people around your ancestor break the wall?
When someone leaves too few records, research their circle. Elizabeth Shown Mills' FAN principle (friends, associates, and neighbors) treats identity as something the surrounding cluster proves. Witnesses, bondsmen, baptism sponsors, executors, and the households either side in a census state relationships no record about your ancestor states outright.
The habit to build is reading the whole page instead of the line you searched for. Who signed the bond. Who appraised the estate. Who owned the adjoining land, year after year. Mills' QuickLesson on identity problems makes the case that clues proving nothing on their own are what point you at the records that do.
Cluster research is also the standard fix for the hardest version of this problem, which is two people sharing one name in one place. That gets its own treatment in same-name ancestors.
What if the records really were destroyed?
If the loss is documented (a courthouse fire, the 1890 US census, the 1922 destruction in Dublin), stop searching the missing series and start on its substitutes. Knowing exactly what burned, which years, and which duplicates sat elsewhere turns an open-ended search into a short list.
Losses are almost never total, and archives publish their own gap lists. The 1890 census exists in fragments; burned Virginia counties have state-level tax copies; Irish research has land-based name lists that pre-date and post-date the returns that burned. Our guide to destroyed records goes loss by loss with what to search instead.
Which next step matches the wall you have?
Most stalled lines fit one of a handful of shapes. Name the shape, then take the matching step instead of repeating the search that already failed. When two shapes fit, work the cheaper one first: a spelling test costs an afternoon, ordering a file from an archive can cost weeks.
| What you are seeing | Probable cause | Next move |
|---|---|---|
| Exact-name searches return nothing anywhere | Indexing or spelling | Phonetic and wildcard search, then search a relative instead |
| The person appears, then vanishes after one record | Migration, death, or a name change | Search by age, birthplace, and known associates rather than by name |
| Records exist for the town but not for your family | Wrong jurisdiction | Identify the county on the event date and follow where its records went |
| One person seems to have lived two lives | Two people conflated | Split the candidates and give each one its own timeline |
| The series you need is documented as lost | Records problem | Switch to substitutes for that place and those years |
| Everything agrees but nothing is sourced | Inherited error | Re-source each fact from the original record before going further |
Doing all of that by hand is a weekend per candidate, which is why most walls stay up. AncestorIQ's Deep Research runs the wide search for you: name variants, record sets across several archives, and an identity check on every source before anything reaches your tree. What AI is genuinely good and genuinely bad at in this work is set out in our guide to AI genealogy.
