Colemak vs Dvorak vs QWERTY: what the evidence says

What the layouts change on paper

QWERTY dates from the 1870s and was shaped by the typewriters of its day, not by an analysis of the hands. Dvorak (patented 1936) and Colemak (released 2006) were both designed from letter frequencies: put the common letters under resting fingers, keep the pinkies light, and avoid awkward sequences.

The headline measure is home-row use. The Colemak project's own analysis says Colemak uses the home row 14% more than Dvorak and 122% more than QWERTY; in other words, QWERTY keeps well under half as much typing on the home row as Colemak does. The same analysis has fingers traveling 10% further on Dvorak than on Colemak and 102% further on QWERTY, so both alternatives roughly halve QWERTY's finger distance. Those are the designer's numbers on the designer's corpus, and other analysers weight fingers differently, but nobody disputes the direction.

Then there are same-finger bigrams, where one finger has to press two keys in a row. QWERTY has a lot ("ed", "un", "ce"), and they're slow because the finger has to lift and move before the second press. Colemak's analysis counts same-finger typing as 60% more frequent on Dvorak and 340% more frequent on QWERTY than on Colemak. Dvorak accepts more of them in exchange for strict hand alternation: vowels on the left home row, common consonants on the right.

The largest typing dataset we know of is QWERTY-only, but it still bears on the argument. Dhakal et al.'s 2018 analysis of 136 million keystrokes found that the timing of letter pairs typed with alternating hands correlated with overall speed at r = -0.72, while letter repetitions correlated at only -0.32. Dvorak maximizes alternation. Colemak favors same-hand "rolls" on adjacent fingers, like "st" or "ne", on the theory that a roll is faster still. Both camps can cite that study, and neither can settle the argument with it.

Is Dvorak faster than QWERTY?

Here the data is thin, and we'd rather say so than pretend.

There is no large, controlled study comparing experienced Dvorak or Colemak typists with experienced QWERTY typists. The 1944 Navy experiment that Dvorak's supporters cite was conducted by August Dvorak himself. The General Services Administration study that QWERTY defenders cite was run by Earle Strong in 1956 with about ten typists per group, and Strong had a public record of opposing new keyboard designs. Economists Stan Liebowitz and Stephen Margolis reviewed the history in the 1990s and concluded that the better-documented experiments suggest little or no advantage for Dvorak. Everything since is self-reports and leaderboards, and both are self-selected: the people who post Colemak speeds are the people who stuck with Colemak.

The one number from Strong's study worth remembering is the cost. According to Liebowitz and Margolis, his ten retrained typists needed "well over 25 days of four-hour-a-day training" to catch up to their old QWERTY speeds, which is more than 100 hours, and after that they improved more slowly with further Dvorak training than the matched QWERTY group did with further QWERTY training. The study is small and its author was not neutral, but the catch-up time is the part switchers still recognize.

What switchers' own accounts consistently describe is the same shape: experienced typists regain their old speed over weeks to months, land close to it, and say typing feels smoother.

We don't yet have enough layout switchers of our own to add to that, and we won't publish a comparison from a few dozen results. Every layout here records the same per-key and letter-pair statistics, so when the numbers exist they'll show where the time goes, not just the headline.

So: comfort, plausibly; speed, not really. If "it feels smoother" is worth a month to you, it's a reasonable trade.

Is Colemak worth it? The learning curve

A switch costs more than people expect.

The usual pattern in switch logs is a steep drop in the first week, fast gains over the next few, and a long tail where the last 10 WPM takes longer than the first 50. The Colemak project itself says learning time varies from person to person. Colemak is generally described as the easier switch for QWERTY typists because only 17 keys move, Z, X, C and V stay put, and so do most punctuation keys, apart from the semicolon, which moves to where QWERTY has P. Dvorak moves nearly everything, punctuation included: the apostrophe, comma and period take QWERTY's Q, W and E, and the brackets go up to the number row.

The hidden cost is every other keyboard: your laptop, a colleague's machine, the shared terminal at work. Some people keep both layouts in their fingers; many find their QWERTY degrades for a while. If you type on machines you don't control, that matters more than any home-row percentage.

The hidden benefit, and the one we'd put weight on, is that relearning forces proper fingering. Plenty of QWERTY typists at 50 WPM have habits (index finger on "c", ring finger on "p") that cap them. A new layout means learning every key fresh with the correct finger, and some "Colemak made me faster" reports are really "learning properly made me faster." You can get that on QWERTY for less: our practice drills target the keys your results show you miss.

The variants

Colemak-DH moves D and H out of the center columns, where the index fingers stretch inward, and onto the bottom row where they're reached by curling. It was developed by the Colemak community, comes in ANSI and ISO versions, and has become a popular starting point for new learners. We ship both versions, because the extra key on ISO keyboards changes where the moved keys land.

Workman (2010) follows Colemak's approach and adds a penalty for lateral index-finger stretches. Its own figures put 6% of keystrokes on the two middle columns, against 12% for Colemak. It changes 21 characters from QWERTY, and the semicolon is one of them: it moves from the home row to where QWERTY has P. We have Workman; few people use it.

BÉPO is the French layout in Dvorak's spirit, built for French letter frequencies and included in the AFNOR keyboard standard of 2019. AZERTY and QWERTZ are not alternative layouts in this sense; they're regional relatives of QWERTY, and emulating them is mostly about putting the accented letters and punctuation where a French or German typist expects them.

How layout emulation works here

The browser reports the physical key you pressed as a key code (KeyJ, for example) regardless of what your operating system thinks that key means. Emulation maps that code to the character the chosen layout puts there, so you can practice Colemak on a QWERTY keyboard with no OS change and no stickers. The result records which layout was emulated, the share card shows it, and the keymap under the test redraws in that layout so you can glance at it instead of the keycaps.

It's also why personal bests and history filters are kept per layout. A Dvorak day and a QWERTY day are different data, and a weak-key model that mixed them would drill pairs that are only slow on the layout you're not using.

One limit to be plain about: the keystroke log can't tell an emulated Colemak run from a QWERTY run typed by someone whose operating system is already set to Colemak. Both arrive as the same stream of key codes. The leaderboards don't separate layouts yet, so treat a layout label on a shared result as self-reported.

Two more limits. There is no emulation for Hindi or other Indic layouts, so for the Hindi typing tests you set Inscript or Unicode Remington GAIL in your operating system and we read what it produces. Kruti Dev, which some exam notices mention, is a legacy non-Unicode font typed with a Remington-style mapping; its output isn't Unicode Devanagari, so it won't match our test text. And emulation maps single keys only, so a character your layout reaches through a dead-key combination needs the layout set in your OS.

What's the best keyboard layout for typing?

If your wrists or fingers hurt on QWERTY: try Colemak-DH for a month with emulation, on the Colemak-DH test, before changing anything in your OS. The reduced travel is real, and you'll know within a couple of weeks whether your hands prefer it.

If you want to be faster: stay on QWERTY and spend the month on the practice plan. Fix the fingering and drill the slow pairs. The gain is bigger and it transfers to every keyboard you touch.

If you're a programmer: layouts move symbols too, and Dvorak relocates the brackets and much of the punctuation you use most. Touch typing for programmers covers that; symbol placement matters more for code than letter placement.

If you're learning from scratch: it's a genuine choice, because you have nothing to relearn. Colemak-DH is the one we'd pick for an English typist starting today. We'd also say, with some regret, that QWERTY on every borrowed keyboard for the rest of your life is a stronger argument than it sounds.

What we'd ask you not to do is switch because a leaderboard screenshot showed a Colemak typist at 180 WPM. That typist would very likely be fast on QWERTY too.

Frequently asked questions

Can I try Colemak or Dvorak without changing my keyboard settings?

Yes. The Colemak and Dvorak tests remap your keys in the browser, so nothing changes in your operating system.

How long does it take to switch from QWERTY to Colemak?

Most accounts describe weeks to reach a usable speed and one to a few months to get back to the old one, with a lot of variation between people.

Does Colemak keep the QWERTY shortcuts?

Mostly. Z, X, C and V stay in place, so undo, cut, copy and paste don't move. Dvorak moves them.