People by US city search: how to find and verify records

Most searches for a person in a city fail for a mechanical reason rather than a research one. The record exists, it is online, and the query cannot reach it because of how the index was built.

Knowing which of the two search systems you are querying, and what each one is physically incapable of finding, saves more time than any amount of persistence with a name box.

What to take away

• Catalog search and full-text search fail in opposite ways. One cannot see inside a document; the other cannot see documents that were never transcribed.
• Search by address, employer or associate rather than by name. Names are the least stable field in every source.
• Record what you searched and found nothing in. An undocumented negative result is worth nothing and will be repeated.

Two systems, two blind spots

Catalog and finding aid search queries descriptions written by an archivist: collection titles, box and folder headings, scope notes. Archival material is usually described at collection level, not item level, so a person can sit in a folder inside a described collection while their name appears nowhere in the description. No keyword search will surface them. You get there by identifying the collection through its subject and then reading the finding aid.

Full-text search queries machine transcription of scanned pages. It reaches inside documents, and it only sees what the transcription produced. Where the scan is poor, the type worn, the paper stained or the layout complex, the transcription is wrong and the page is invisible to a correct query.

The practical rule: use catalog search to find where the material is, and full-text search to find lines inside material you already know exists. Using either for the other's job is where hours disappear.

What machine transcription does to a name

Predictable failures, all worth building into a query:

• Similar letter shapes are interchanged, so a surname can be transcribed with the wrong initial letter and drop out of every alphabetical and phonetic search.
• Older typography, including archaic letterforms and tight ligatures, transcribes badly, and a name that crossed a script carries several defensible forms of romanization besides.
• Newspaper columns bleed into each other, so words from two stories merge into one token.
• A name hyphenated across a line break becomes two fragments, neither of which matches.
• Dense small type, the standard setting for directories and classified listings, produces the worst results of all.

So run a name several ways: with truncation, with a wildcard in the middle, with the first name alone plus a street name, and with an initial rather than a full given name. If the interface offers phonetic matching, remember what it does. Common phonetic keys are anchored on the first letter, so a variant that changes the first letter defeats them, and they cannot help at all with a name recorded in a different alphabet and transliterated inconsistently. The broader problem of unstable spelling is covered on the historical records page.

Search the place, not the person

The strongest city searches barely use the name field.

• Work an address. Street sections of directories, tax and assessment rolls, permits and plat books are organized by property. An address search finds people whose names you could never have guessed correctly, and the directory page explains why that section is the powerful one.
• Work an employer or institution. Payroll lists, membership rolls, school registers and union records place a group of people at once.
• Work the neighbors. Households next door in one source tend to recur in others, and a neighbor's unambiguous name is a way into a page where your subject is misspelled.
• Work an event. A property transfer, a court case, a license application or a funeral notice pins a date, and a dated event lets you search a narrow run of a newspaper page by page instead of by keyword.

Reading a short run of pages directly is often faster than a further hour of query variations, and it shows you context that a search result never does.

An index is not the record

Every online result is an index entry created by somebody after the fact. It may be a transcription, an abstract, or a machine reading, and each of those introduces its own errors on top of whatever the original contained. Library guidance on searching a digitized newspaper collection sets out what its own index does and does not reach.

Always follow the entry to the image of the document, and where no image exists, to the repository holding the original. Then record which you saw. A finding based on an index entry alone is a finding about the index. That distinction is the same one that makes the difference between a citation you can defend and one you cannot, and it belongs in the source column of any dataset built from these searches.

Log the negatives

The result most researchers throw away is the one worth keeping. "Not found" is only useful if it says what was actually looked at.

Field	Record
Source	Title, repository, and the run of years searched
Query	Exact terms, spellings and wildcards used
Coverage	Which years or volumes are missing from the source itself
Interface	Which index or platform, since the same source is searchable differently on different platforms
Date	When you searched, because collections grow
Result	Found, not found, or unable to search

With that in hand, a negative result becomes a piece of evidence: this name is absent from this source across these years, which is a real finding when the source is one where the person should appear. Without it, you will search the same thing again in six months and learn nothing new. Coverage gaps and custodian questions are set out on the city facts page.

Knowing when to stop

Set the stopping rule before you begin, because this kind of search has no natural end. A workable one: stop when you have checked every source class that would record the event you are after, in the years the event could have happened, in the jurisdiction that would have held it, and logged the gaps. What remains after that is a documented unknown, which is a legitimate thing to publish.

The alternative, continuing until something plausible turns up, selects for weak evidence by construction.

Living people

None of the above applies to searching for a private individual, and this site does not do it. We do not compile addresses, family details or movements for living people, we do not use commercial people-search services, and we do not publish anyone's birthplace or residence.

Where a living person has spoken publicly about where they are from, the search is a search for that statement: the interview, the outlet, the date. What gets published is the statement with its attribution, never a location assembled from records. The city pillar treats this as a safety rule rather than an editorial preference.

Bottom line

Pick the right search system for the question, expect transcription to have mangled the name, search the place and the event instead, follow every hit to the document itself, and write down what you failed to find. That sequence turns searching from a lottery into a method with an end.

Common questions

A database says the record is there but the image will not load. Now what?

Note the citation exactly as the database gives it and request the item from the holding repository. Databases license images and lose access to them; the underlying record does not move.

Two platforms index the same collection and return different results. Which is right?

Both are incomplete, differently. Different transcription passes and different search behavior produce different misses. Where a source matters to your conclusion, search it on more than one platform and say which you used.