Sugihara Visa Archive

What the register gives us — and what it does not

Everything starts from Sugihara's own ledger: 2,139 numbered lines, each with a name, a nationality, a visa type, a date and a fee. That is all. There are no birth years and no birthplaces. Today 77 of those lines have a birth year attached, and only because a surviving document supplied it.

The typescript itself is damaged in a specific way: on 935 lines the leading letter of a name is missing, so Swislocki reads as wislocki and Rozenbaum as ozenbaum. 996 lines carry a repaired reading. A repair is a judgement, and it is recorded as one — the original text is never overwritten, and every repair is listed in the corrections log.

What counts as evidence

In roughly this order of weight:

• The document itself. A visa, a Curaçao endorsement, a Polish citizenship certificate, a passport page — bearing a number, a date, or a hand we can compare with others.
• A roster that names the same person. The Kobe and Shanghai refugee lists, JDC and yeshiva rolls, Japanese landing records. These were written by different people for different purposes, which is what makes agreement between them meaningful.
• Family testimony. Indispensable, and treated as evidence rather than decoration — but a family's account of a name is not independent confirmation of that name, since both may descend from the same memory.
• Published scholarship, cited to its source, including where it disagrees with us.

Matching a document to a line

A name in 1940 Kaunas could be written in Polish, Lithuanian, German, Yiddish or Russian, by a clerk hearing it for the first time, and then changed again after the war. So the archive compares names allowing for the ordinary drift between them: sz / s / sch, cz / c, rz / z, w / v, dropped leading letters, given names in either order, and post-war anglicisation — Herszberg becoming Hershberg, Klejn becoming Klein.

A name that agrees is where a match starts, not where it ends. The question that decides it is whether anything else agrees: the visa number, the date of issue, the nationality, the family travelling on the same sheet, a birthplace, a photograph.

Why a matching name is not proof

The large archives hold millions of records. Searching one of them for a common name returns a crowd: a single line of our register, searched against a European archive, returned ten men of that exact name born between 1866 and 1931. Any of them might be the right one. Nothing on our side can say which — and a coin-flip published as a fact is worse than an open question.

Where a name is genuinely rare, a match carries real weight. Where it is common, the archive records the record as a lead and says so on the page, rather than quietly turning a possibility into an identification.

The four states a record can be in

• Documented and matched — a surviving document tied to a register line. 338 lines are here.
• Documented but unmatched — a real document whose owner we cannot yet place on the register. 110 of these. They are published as their own records, because a document nobody can name is still evidence.
• A name and nothing more — the majority. Most of the 2,139 are still only a line of typing.
• An unconfirmed lead — an outside record under the same name that nothing yet ties to this person. Shown, labelled, and never counted as corroboration.

What this archive will not do

• Publish an identification that rests on a matching name alone.
• State what became of somebody on the strength of a name found in a names database. A hit there is a lead to check, not a fate to announce.
• Alter an original photograph. Where a document was photographed at an angle and has been squared up for reading, the corrected image is a generated copy and the original photograph stays one click away, because the photograph is the evidence and the correction is our reading of it.
• Publish a living person's material — a photograph, a family story, a message from a community group — without their permission.

When we get it wrong

Identifications here have been withdrawn after they were published, and will be again. When that happens the record says so rather than quietly reverting: a proposed match set aside is left visible as a question, unlinked from the register line it was claiming. Every change the archive has made — 843 recorded so far — is kept and published in the corrections log, with the reasoning.

The standard the archive tries to hold is simple: it is better to leave a document unattached than to attach it to the wrong person. An empty line can be filled later. A wrong name, once a family has read it, travels.