A record-correction · the keyboard under your hands
The Keys Were Spread, Not Slowed
Everyone repeats it: QWERTY was designed to slow typists down. It gets the cause exactly backwards. A jamming typebar stops you cold — so a layout that jams less lets you go faster, not slower. Type anything below and measure the fingerprint that's actually there.
In 1868 the typewriter was a basket of metal arms. Press a key, an arm swings up and strikes the paper through the ribbon, then falls back. Press two keys whose arms sit near each other too fast, and the second arm catches the first on its way down. The bars lock together at the platen. The machine halts; you reach in and unstick them. A clash isn't a slowdown — it's a full stop.
So the engineering problem Christopher Latham Sholes spent five years tinkering at was the opposite of the legend: how do you keep the busiest letter-pairs from clattering into each other? Reduce the clashes and the typist speeds up. "Designed to slow you down" isn't just unproven — it points the wrong way down the causal arrow.
Type something. Watch the hands take turns.
The clearest fingerprint of an anti-clash layout isn't where the keys sit — it's whether your two hands take turns. When the next letter is on the other hand, that hand is already moving while this one strikes: faster, and on a typebar machine, bars coming down from opposite sides instead of hammering neighbours.
The clash bench
Letters are coloured by hand: left / right. A red outline marks a same-hand pair — the slow, clash-prone keystroke a good layout tries to avoid.
Last keystrokes (most recent on the right):
| For your text | Alpha | QWERTY | Dvorak |
|---|---|---|---|
| Hands alternate | – | – | – |
| Same-hand pairs (clash-prone) | – | – | – |
| Presses on home row | – | – | – |
Type a long enough sample and a stable gap opens up. Across a whole book (the verified figures below), QWERTY's hands alternate on 51.7% of letter-pairs — more than 70.9% of random letter arrangements. A plain alphabetical board manages only 49.0%. The nudge is real, if modest: five years of moving letters around left a board that keeps the hands taking turns a little more than chance would — exactly what you'd expect from fighting clashes, and the precise opposite of a brake.
Then the machine changed — and the goal flipped.
Once typewriters went electric, the typebars were gone and nothing could jam. The thing worth minimising was no longer clashes — it was travel. In 1936 August Dvorak built a layout for that opposite world: pile the most-used letters onto the home row so the fingers barely move. Toggle the two boards and watch where the work lands.
Where the typing lands
Darker = more presses, from the same public-domain corpus. The home row is underlined in green. Same letters, same English — a third of the work on QWERTY's home row, two-thirds on Dvorak's.
This is the quiet punchline. QWERTY and Dvorak don't disagree about how to type well — they were solving different machines. QWERTY answered "don't let the bars clash"; Dvorak answered "don't make the fingers walk." The first goal vanished the day the hardware did, which is most of why we never switched: there was never a jam left to fix.
One honest catch: the obvious metric is the wrong one.
You might guess QWERTY simply throws common pairs far apart on the board. We measured it — and it's false. The average straight-line distance between the two keys of a letter-pair, weighted by how often English uses it, is 3.20 key-widths on QWERTY: actually a hair below the 3.29 average of random layouts. Raw distance carries no signal.
Stranger still, Dvorak has the largest raw distance of the three (4.32) — because alternating hands so aggressively flings each next key onto the far hand, across the whole board. So "the keys are spread out" is the wrong picture twice over. The signal that survives measurement is hand-alternation, not key-distance — which is why the bench above counts turns, not gaps. We'd have led with the distance story if the numbers had let us; they didn't, so we didn't.
The check
Every figure on this page is recomputed by research/qwerty-myth/verify.mjs from 563,924 letters of a public-domain corpus (Austen, Pride and Prejudice, Project Gutenberg #1342), under a stated modern-keyboard model (row offsets 0 / 0.25 / 0.75 key-widths; left hand = columns 0–4). 8/8 checks green.
- Home-row load — QWERTY 33.3%, Dvorak 70.4% (close to the long-cited ~32% / ~70%), Alphabetical 35.8%.
- Hand-alternation — QWERTY 51.7%, Dvorak 71.2%, Alphabetical 49.0%; QWERTY beats 70.9% of 5,000 random layouts (random mean 49.1% ± 4.9%).
- Mean key-distance — QWERTY 3.20 (below the random mean of 3.29), Dvorak 4.32: the naive "spread far apart" metric fails, and inverts for Dvorak.
- What's measured vs. what's history. The numbers are properties of the layouts and real text — not a proof of what Sholes intended. Intent lives in the documentary record below, where the genuine uncertainties are named, not smoothed.
What's settled, and what isn't
The mechanism above is sturdy. The history has real open seams — and honesty means flying them, not hiding them.
The narrow thing this page settles is narrow on purpose: not "here is the true origin of QWERTY" — historians still argue that — but "the one explanation everyone repeats, that it was built to slow you down, is the one we can rule out." A clash stops you. Stopping less is going faster. The arrow only points one way.