Hollerith Tabulator

Hollerith claims recording, verifying and tabulating as one connected art. The machine is only the third stage; the fixed hole positions are what make it possible.

Invent 20 people, each with three attributes: trade, age band, birthplace. Write them as sentences on paper.

Count how many are farmers. Time yourself. Now count farmers over forty — and notice you had to read all 20 again.

Mark 12 positions in a fixed grid on a card: 4 for trade, 4 for age band, 4 for birthplace. Position carries the meaning — nothing is written down.

Make one card per person, punching only the positions that apply. Keep the punch clean — a half-punched hole is the classic failure.

Cut a labelled template and lay it over each card to confirm the holes mean what you intended. Hollerith makes verification its own claimed stage.

Stack the cards, push a needle through the "farmer" position and lift. Farmers stay behind, everyone else rises. You just counted without reading.

Push needles through "farmer" AND "over forty" together. What drops satisfies both. This is a logical AND, done with wire.

Fix a strip of foil below the card and mount a springy wire pin above one index position, wired through a battery to an LED.

The card insulates the pin from the foil. No circuit, no light. Card stock is the switch.

The pin drops through the hole and touches the foil. The LED lights. A hole is a closed circuit — this single fact is the whole invention.

Put pins on "farmer" and "over forty" in series. The LED lights only when BOTH holes are present — the combined question is answered by the wiring, not by a person.

Now it lights for EITHER. Same cards, same pins, different question. Re-asking costs a wire, not another pass through the records.

Feed the stack through, ticking a counter each time it lights. Time it and compare with step 3.

The patent. US 395,782, "Art of compiling statistics", filed 23 September 1884 and granted 8 January 1889 to Herman Hollerith. Note the granted title: an art, not a machine. Hollerith claimed the whole method — record, verify, tabulate — and that breadth is why the patent mattered commercially.

This patent uses punched paper STRIPS, not cards. The rolled strip is what the 1889 document describes, and it has an obvious defect: a strip must be read in order, so you cannot pull one record out, re-sort the set, or replace a single entry. The independent card — which can be sorted, counted, re-sorted and filed — arrives in Hollerith's later work and is what everyone pictures. Retro-fitting the card onto this patent is a small error that hides the actual design step.

It worked, and the numbers are the argument. The 1890 census was tabulated in a fraction of the time its predecessor took, on a larger population, and produced cross-tabulations that had previously been unaffordable to ask for. Hollerith formed the Tabulating Machine Company; through a 1911 merger and a 1924 renaming it became IBM. The line from a census clerk's bottleneck to the twentieth-century computer industry is unusually direct.

Where the idea came from, and where it went. Punched-hole control was not new — the Jacquard loom had been using punched cards to control weaving since 1804, and Hollerith knew it. His change is what the hole does: in Jacquard's loom a hole permits a mechanical hook to pass, controlling one machine action; in Hollerith's it closes an electrical circuit, and electricity can be branched, combined and counted in ways levers cannot. That is why steps 12 and 13 are the heart of this build. Wiring pins in series asks AND; in parallel asks OR. Hollerith is doing Boolean logic on physical media half a century before anyone builds a computer, and the reason his machine could answer questions no clerk had time for is that the question lives in the wiring, not in the records.

The uncomfortable part, stated plainly. A technology that makes populations efficiently countable and cross-classifiable is powerful in whoever's hands hold it. Hollerith-type punched-card systems were later used by states for purposes far darker than census statistics, including by Nazi Germany. That history belongs beside the invention rather than after it: the machine is genuinely brilliant, and what it does is make it cheap to sort people into categories.