Cistercian numerals

How it works

One vertical staff. Four quadrants around it, one per decimal place: units top right, tens top left, hundreds bottom right, thousands bottom left. A few strokes in each quadrant name that digit.

A zero is simply an empty quadrant. There is no sign for zero, and a bare staff with nothing on it was never defined as a number, so the system covers 1 to 9999 and stops.

Only nine shapes exist. The other three quadrants reuse the same nine, mirrored — across the staff for tens, across the staff's midline for hundreds, both ways for thousands.

The shapes build on each other: 5 is 1 plus 4, 7 is 1 plus 6, 9 is 1 plus 2 plus 6. Where the figure looks closed, the staff is the missing side.

Where it comes from

The four-quadrant system appears in the first half of the thirteenth century in Cistercian houses on the Hainaut border, in what is now the Belgian–French frontier country — probably Aulne-sur-Sambre or Vaucelles. It extends an earlier English shorthand, credited to John of Basingstoke, that could only write 1 to 99.

The motive was practical, not mystical. The Cistercians were early builders of indexes and concordances, and needed short reference numbers for cramped margins. In Roman numerals, column 884 is dcccixxxiiii.

Two families existed. The horizontal-staff form came first and was the more common one in medieval manuscripts; the vertical form spread from northern France in the late thirteenth century. Every modern revival, this one included, uses the vertical.

What it was actually used for

Numbers, written down. Not arithmetic — the system has no zero, no columns, no way to carry, so you cannot add two of these glyphs. Where real calculation appears in the same manuscripts, it is done in Hindu-Arabic numerals.

• Indexes and concordances. The oldest known use: a concordance at Aulne-sur-Sambre, first half of the thirteenth century, giving the column number for each keyword across 884 columns.
• Page numbering. A Bodleian manuscript from the Low Countries carries Roman numerals in black on each recto and Cistercian ciphers in red in the margin of each verso, folios 10 to 297.
• One astrolabe. A brass astrolabe made in Picardy around 1375 has every number on it — degree scales, latitudes, calendar — in these ciphers. Exactly one such instrument survives, and it is in private hands.
• Wine gauging around Bruges and Damme. Flemish manuscripts of the fifteenth and sixteenth centuries list barrel volumes in ciphers, for the trade of measuring wine casks with a marked rod. No rod and no marked barrel survives; the evidence is three manuscripts.
• Year dates. A set of astronomical tables of 1473 writes only the year column in ciphers, while every other number on the page is Hindu-Arabic — apparently to make that column stand out.

The corpus is about two dozen manuscripts across three centuries. Set against the tens of thousands of medieval manuscripts that survive, this was never a widespread notation, and there is no evidence the order adopted it institutionally.

They died out for concrete reasons: you cannot calculate with them, the ceiling is 9999, they were never standardised — which corner meant what varied by region — and they set terribly in movable type, where every one of 9,999 combinations would need its own sort. Printers who tried got them wrong. Hindu-Arabic numerals arrived in northwestern Europe at almost exactly the moment the ciphers were invented and did everything they did, plus arithmetic. The Cistercians switched too.

One footnote, because it is usually told wrong: these were not a secret code. They were used openly, in indexes and on instruments. The single documented case of anyone using them to hide something is a guild of Bruges wine-gaugers around 1720, five centuries later, keeping their trade craft from customers.

This version

The glyphs this generator draws are not facsimiles. I lettered them by hand, giving the strokes a calligraphic weight the plain historical forms do not have.

The small diamonds are mine too. Historically a zero is nothing at all — an empty quadrant. But a numeral like 1000 then leaves three quarters of the field empty and the composition falls to one side. The dots mark the corners a quadrant's strokes never reach, so every numeral keeps its balance and 78 reads as a considered 0078 rather than a fragment. They carry no numeric meaning and you can switch them off in the generator.

The schematic diagrams above are drawn separately, in plain geometry, so the historical system stays legible underneath my version of it.