METHODS & CAUTIONS
The List is built from one public dataset: the U.S. Social Security Administration's national baby-names file (names.zip, years of birth 1880 through 2025), one file per year, each row a name, a sex, and a count. Every figure on this site is computed from that file and nothing else. The honesty of the whole enterprise lives in the following cautions, which the management asks the visitor to read before believing anything.
To protect privacy, the SSA publishes a name only when it received at least five births for that sex in that year. A name absent from a year's file had zero, one, two, three, or four recorded births that year — and no one outside the agency can ever say which. Accordingly, this site never says a name is extinct or dead. It says the name fell below the federal reporting line in a given year and has not returned. "Critically low" means five to nine births: within one bad year of the line. A "return" means crossing back above the line, nothing more.
The counts are tabulated from Social Security card applications, keyed to the year of birth on the application — not from birth certificates at the time of birth. Consequences:
1. The early decades under-register, severely. Social Security numbers did not exist until 1936. People born 1880–1936 appear only if they later applied for a card, so the earliest files reflect the subset of each cohort who survived into the SSN era and applied. Counts before roughly 1937 understate true births — worst for the earliest years, for people who died young, and for groups less connected to the formal economy. The apparent explosion of names in the 1910s is partly this coverage effect, not only fashion.
2. Universal coverage is recent. Only when enumeration at birth became routine in the late 1980s do the files approximate all U.S. births. Comparisons that cross the 1937 and late-1980s boundaries mix coverage changes with fashion changes.
3. Spelling variants are separate names. Blanch and Blanche, Bulah and Beulah, Myrle and Merle are independent rows in the file and independent entries here. A spelling can vanish while its sound thrives. Every claim on this site is about a spelling, and should be read that way.
4. Sex-split artifacts. Nearly every common name appears under both sexes in the file, and some of that — especially midcentury — is widely believed to include recording error. Small opposite-sex counts of strongly sexed names deserve suspicion, and the List treats them with it.
5. A returned string is not always a returned name. When a string reappears after eighty years, it is often a modern coinage colliding with an old spelling — a surname-name or an import, not a revival. The Returned room enjoys the coincidence; it does not claim continuity of fashion.
6. Names are as-entered at SSA. Hyphens, spaces, and diacritics are stripped or normalized; very short and very long names are excluded or truncated by the agency's processing.
7. The latest year is provisional in spirit. 2025 counts reflect applications processed by the file's cut; late applications trickle in, and recent-year counts can rise slightly in future editions.
8. Censoring at both ends. A name last seen in 1985 has had forty years to return and has not; a name last seen in 2005 has had only twenty. Decade-by-decade comparisons of "die-offs" must carry this caveat, and no room on this site should be read as a prophecy.
| THE VANISHED | Last reported 40+ years ago (last year above the line ≤ 1985) |
|---|---|
| LONG MISSING | Last reported 20–39 years ago (1986–2005) |
| CRITICALLY LOW | Reported in 2025 with 5–9 births |
| THE RETURNED | Spent 10+ consecutive years below the line, then reappeared |
| STURDY GHOSTS | First reported by 1900 and present in every year since |
The rooms overlap where the data does: a returned name may now be critically low; a sturdy ghost may be standing at seven births. A name appears in every room it qualifies for, because that is where it is.
The raw SSA files are processed by a deterministic script into a table of one row per name and sex — first and last year, totals, peak, gaps, returns — and every page of this site is generated from that table. Flagship claims were additionally spot-checked against the raw year files, byte by byte. Charts start their y-axis at zero, and years below the reporting line are drawn as gaps, never interpolated.
Source: U.S. Social Security Administration national names file (1880–2025). Names below 5 occurrences in a year are not reported.
Somewhere, a grandmother is unsurprised.
A collection of THE MUSEUM OF HOPEFUL MACHINES