The listing says 1885. The brass plaque by the door says 1885. The neighbor, who has lived on the street for forty years, says 1885.
Here is the uncomfortable part: none of those is a record. The date on a real estate listing usually comes from the tax roll, and tax rolls copy from each other. Plaques are voted on by committees working from local lore. The neighbor is repeating the plaque. The single most common way a house gets a wrong birth year is that somebody, a century ago, wrote down the date the land was first sold, and every later document inherited the mistake. Deeds transfer dirt. They say nothing about what stands on it, or when.
So how do you actually date an old house? The same way you date a manuscript: provenance. You build a chain of independent records and see where two of them agree.
1. Walk the chain of title. The county recorder (or registry of deeds, or clerk, depending on the state) keeps a grantor-grantee index: who sold, who bought, when, for how much. Start with the current owner and walk backwards. This gives you the land's story, the lots, the subdivisions, the speculators. It will not give you the house. What it gives you is a bracket: if your lot sat empty farmland in 1860, no part of the house predates 1860, whatever the plaque says.
2. Read the fire insurance maps. The Sanborn Fire Insurance Maps are the quiet hero of this whole business. Mapmakers surveyed American towns block by block from 1867 onward, drawing every building's footprint, labeling it DWG (dwelling) or store or barn, so insurers could price fire risk. The Library of Congress put its enormous Sanborn collection online, free, no login. Find your town, find your block, and step through the years. The first map year your house's footprint appears on the paper is one of the hardest brackets you will ever get. (Coverage in the free set runs to about 1922; after that, local collections take over.)
3. Pull the newspapers. Chronicling America, another free Library of Congress project, has digitized millions of pages of local newspapers. Nineteenth-century local papers covered real estate the way modern ones cover sports. "New dwelling, 14 Maple Street, for Mr. Brown, to be completed by October" is a sentence that settles arguments. Also look for auction notices, fire reports (a house that burned in 1891 and was rebuilt is a different house), and "just finished" advertisements by builders.
4. Find the people. Federal census records (every ten years) and city directories (every year, in most cities) tell you who lived at your address and when. Two traps here. First, street numbers change: many cities renumbered whole streets, so the house called 14 Maple in 1910 may be 38 Maple today. Directories often note the renumbering. Second, the census is a sieve, enumerators missed households, but misses are themselves information.
5. Ask the photographers. Assessors photographed houses to tax them. New York City's Municipal Archives tax photos, a full portrait of every building in the city, are the famous set, and other cities kept versions. Historic American Buildings Survey (HABS), at the Library of Congress, documented thousands of structures with measured drawings and written histories. And if your house has any chance of being on the National Register, read the actual nomination at the Register's NPGallery database, not the plaque. Nominations cite their sources, which means you can check them, which matters, because nominations also carry errors.
6. Then, and only then, the style. "It looks Victorian, so it's 1880s" is the weakest evidence in the room. Styles linger for decades, builders ordered trim and doors from catalogs well after a style's heyday, and the Sears mail-order house claim, beloved of thousands of owners, is notoriously overdiagnosed, as the people who authenticate these houses keep documenting. Style brackets. Records close the bracket.
The rule that makes it work Do not accept any single record. Deeds get miscopied. Maps get updated late. Newspapers print optimism. A date is only solid when two independent records, records that did not copy from each other, agree. Until then your answer is a bracket: "built between 1886 and 1890," which sounds less satisfying than "1888" and is worth more.
That is not bureaucratic fussiness. It is the same discipline a medievalist applies to a manuscript: a weak link in the provenance chain can undo the whole claim of what a thing is and where it came from. Houses are not different from books in this. The difference is that the manuscript's chain was kept by monks and the house's chain was kept by clerks who were also paid to be quick.
Circa is not a hedge. Circa is what honesty looks like written down.
Sources named: Library of Congress, Sanborn Maps collection (loc.gov/collections/sanborn-maps); Chronicling America (chroniclingamerica.loc.gov); National Register NPGallery; HABS/HAER at the Library of Congress; NYC Municipal Archives tax photos; Erik Kwakkel, "Perfect imperfection," medievalbooks.nl, November 2024, for the provenance-chain framing.