The QWERTY Story — and the Honest Debate

Look down at your keyboard. The top row of letters spells, more or less, a word that isn’t one: QWERTY. Nobody sat down and reasoned their way to that order the way you’d design a good tool. It is an accident from the 1870s that has outlived the machine it was designed for, every rival that tried to replace it, and probably will outlive you. For a certain kind of economist, that little row of keys is the closest thing to a smoking gun that history matters — proof that the market can get stuck on the wrong answer and never climb off.

That is the famous story, and we’re going to tell it properly. Then we’re going to do the thing a good thinker must do and take seriously the people who spent years arguing the famous story is largely a fable. Because the point of this lesson isn’t to hand you a cute anecdote to repeat at parties. It’s to inoculate you against your own laziness — the temptation to look at anything popular-but-imperfect and declare, triumphantly, “aha, a locked-in historical accident!” without ever checking whether that’s true.

The standard QWERTY narrative

Here’s the version you’ll find in most textbooks, most famously laid out by the economist Paul David in a 1985 paper with the irresistible title “Clio and the Economics of QWERTY” (Clio being the muse of history — David’s whole point was that history, not just present-day efficiency, shapes which technology we end up with).

The story goes like this. In the early 1870s, Christopher Latham Sholes was tinkering with one of the first commercially viable typewriters. Early machines had a mechanical problem: each key swung a metal type-bar up to strike the paper, and if you hit two nearby keys in quick succession, their bars collided and jammed. The legend is that Sholes rearranged the letters specifically to separate common letter-pairs, so the bars for frequently-adjacent letters would swing from different directions and clash less often. Whether the goal was literally “slow the typist down” or “reduce jams” is debated even in the standard story — but either way, the layout was optimized for 1870s mechanical constraints, not for human fingers.

Then came the amplifier. The layout got attached to the Remington typewriters, which sold in large numbers. Typing schools, needing a standard to teach, taught the layout that was already on the machines people owned. A widely-retold 1888 typing contest was won decisively by a fast touch-typist working on a QWERTY machine, which helped convince the world that QWERTY was the professional’s choice. Now the loop is running: more typists trained on QWERTY → offices bought QWERTY machines so they could hire trained typists → schools taught QWERTY because that’s what offices used → more typists trained on QWERTY. Each turn of the loop made the next person’s rational choice “learn QWERTY,” regardless of whether some other arrangement of keys was better in the abstract.

Decades later, in 1936, August Dvorak patented the Dvorak Simplified Keyboard, deliberately designed to put the most common letters on the home row and balance the work between hands. Dvorak and his supporters claimed large speed and comfort gains. And yet — this is the punchline — Dvorak never displaced QWERTY. Even after the type-bar jamming problem vanished entirely (electric typewriters, then computers, have no bars to jam), we kept the layout designed to prevent a mechanical failure that no longer exists. That, in the standard telling, is lock-in: an inferior standard, preserved by the accumulated weight of everyone already using it, long after its original reason expired.

Analogy. It’s like a village that built its only bridge at a bend in the river because that’s where the ford happened to be in 1700. Three centuries later the town has grown, the traffic is terrible, and everyone agrees a bridge half a mile downstream would be better — but every house, road, and shop was built assuming the bridge is here, so nobody can afford to be the first to move. The bridge’s location isn’t optimal; it’s just historical, and history has fingers.

Why it’s such a seductive example

Notice how perfect QWERTY is as a story. It’s vivid — the evidence is literally under your hands right now. It’s universal — nearly every reader has one, which makes it feel like proof rather than anecdote. And it flatters a satisfying, slightly rebellious idea: that the market, which we’re often told ruthlessly selects the best option, actually failed, and has been failing continuously for 150 years while everyone politely pretends the emperor’s keyboard is optimal.

Let’s name the reinforcing loop exactly, because the rest of the course lives inside it:

The QWERTY loop. Value of learning QWERTY ↑ as (number of QWERTY machines + number of QWERTY-trained typists) ↑ → each new typist rationally learns QWERTY → each new office rationally buys QWERTY → both totals ↑ → value of learning QWERTY ↑ again.

Every arrow is an individually sensible choice. Nobody is being irrational; each person is correctly optimizing given what everyone else already did. That’s what makes lock-in so slippery — it isn’t a story about stupidity, it’s a story about how a hundred smart local decisions can add up to a globally stuck outcome. (If you’ve done the feedback-loops and critical-mass courses, you’ll recognize this as a positive feedback loop that has already crossed its tipping point.)

The critique — and take it seriously

Now for the twist, and please resist the urge to defend the story you just learned. In 1990, economists Stan Liebowitz and Stephen Margolis published “The Fable of the Keys” — a direct, well-armed attack on the QWERTY legend. They did not deny that path dependence can happen in theory. They attacked the evidence that it happened here. Their argument, roughly, has three prongs:

1. The “Dvorak is dramatically better” evidence is thin. Many of the strongest claims trace back to studies conducted or promoted by Dvorak himself — hardly a disinterested referee — and to a famous U.S. Navy wartime study whose methodology (who ran it, how typists were selected, how “improvement” was measured) they argue was too flawed to bear the weight placed on it. Later, more controlled experiments found the Dvorak advantage to be, at most, modest — a few percent, not the revolution the legend implies. If the better mousetrap is only slightly better, then not switching isn’t a market failure; it’s arguably a sensible refusal to pay large switching costs for a small gain.

2. Markets do switch when the prize is big enough. Liebowitz and Margolis pointed out that the same economy supposedly frozen on QWERTY has, over and over, abandoned entrenched standards the moment a genuinely superior alternative appeared and the gain justified the cost. Their claim: real, large improvements get adopted; only tiny ones get stranded, and stranding a tiny improvement is not a scandal.

3. The conclusion was asserted, not proven. Their sharpest point is almost philosophical. “QWERTY is both locked-in and inferior” is two claims welded together, and the second one — that a clearly better option exists — was mostly assumed because the story is so appealing, not established with clean evidence.

You are allowed to find this deflating. That’s the point. A model is only as trustworthy as the evidence for its headline example, and if the headline example is shaky, an honest thinker has to notice. (Note: the Dvorak-vs-QWERTY empirical question is still debated — some researchers dispute Liebowitz and Margolis right back. The lesson is not “the critics won.” It’s “the case is contested, so stop treating it as settled proof.”)

What the debate actually teaches

Here’s the real payoff, and it’s more valuable than either side’s verdict on keyboards. The QWERTY debate forces you to split one lazy claim into two claims that must each be earned:

• Claim A — the mechanism is real. Path dependence, driven by increasing returns to adoption, is a genuine force. Early accidents can be amplified into durable standards. This part is not seriously disputed by anyone, including the critics. You should keep it in your latticework permanently.
• Claim B — this particular thing is a harmful lock-in. To assert that about any specific case, you owe two pieces of evidence: (1) that a meaningfully better alternative actually exists (not just a different one, not just a slightly better one), and (2) that what’s keeping us on the current option is switching costs — the sunk investment in the old standard — rather than the current option simply being good enough that the alternative isn’t worth the change.

The failure mode is collapsing B into A: seeing the mechanism could apply and concluding it does. That’s the just-so story. The discipline is to treat “we’re locked on the worse option” as a hypothesis that needs a counterfactual, not a conclusion you get for free from noticing that something is old and imperfect.

Worked example — running the two-claim test. Suppose a friend says, “The reason we all use [some old file format] is pure lock-in — there’s a much better format, but we’re trapped.” Run the checklist. Claim A: is there a plausible increasing-returns loop (everyone’s files, tools, and colleagues use it)? Usually yes — so the mechanism is available. Claim B(1): is the alternative meaningfully better, or just different / marginally better? If it’s a 3% improvement, stranding it isn’t a scandal. Claim B(2): if it is much better, is the blocker really switching costs — or is it that the old format quietly does something the new one doesn’t? Only if all three boxes are ticked have you earned the word “lock-in.” Most casual accusations fail at B(1).

A cleaner example to anchor the mechanism

Because QWERTY is contested, it’s a bad proof even though it’s a great illustration. So let’s pin the mechanism to a case where the increasing-returns loop is less disputable: the standard-gauge railway (rails set 1,435 mm — about 4 feet 8.5 inches — apart).

That oddly specific width has a folk history involving cart ruts and even Roman chariots; ignore the more colorful legends. What matters is the loop, which is textbook-clean:

The result: standard gauge spread across much of the world and became extremely hard to dislodge, even in places where engineers argued a different width would have been marginally better for local conditions. Here the two claims are easier to inspect than with keyboards. Claim A (increasing returns) is obvious — a rail network’s whole point is connecting to other track, so compatibility is the value. Claim B is more honest too: converting a national rail gauge is a famously enormous, sometimes literally overnight-army-of-workers undertaking, so the switching cost is not hypothetical — it’s measured in fortunes and years. Whether any particular country is stuck on a worse gauge is still an evidence question, but the lock-in force is undeniable and quantifiable in a way QWERTY’s inferiority never quite was.

Pitfall. Even here, don’t overclaim. “Standard gauge locked in” (mechanism, well-supported) is a different sentence from “standard gauge is the worst option and we’re all suffering” (harm claim, needs its own evidence). The mechanism can be rock-solid while the harm is small or zero — a standard can lock in and also happen to be perfectly fine. Lock-in is not the same as regret.

The honest verdict

So where does that leave the little row of keys under your hands? Exactly where it should: as a superb illustration of the mechanism and a lousy proof that markets routinely lock onto bad options. Use QWERTY to explain how increasing returns can amplify an early accident — that’s what it’s genuinely good for. Do not use it to prove that the market is broken, because that specific empirical claim is contested and you’d be leaning on a fable.

Compress the whole debate into one durable distinction:

• The mechanism of path dependence is certain. Early accidents plus increasing returns can, and demonstrably do, produce durable standards. Keep it.
• The frequency and severity of harmful lock-in — how often we’re actually stuck on a worse option, and how much it costs us — is genuinely debated. Don’t assume it; investigate it, case by case, with a counterfactual in hand.

That distinction is the adult version of this model. The child’s version (“QWERTY proves the market is dumb”) is fun at parties and wrong in arguments. Lesson 6 will come back to this line — where the model lies — and map exactly when a path-dependence explanation is doing real work versus dressing up a hunch. For now, carry the two-claim test with you: mechanism is free, harm must be earned.

A last check

Where this goes next

You now hold both halves: the mechanism is real, the famous proof is contested, and the fix is to demand a counterfactual before shouting “lock-in.” But we’ve been treating “increasing returns to adoption” as a single black box — the more people use it, the more valuable it gets. That box actually has distinct machinery inside it. In Lesson 3 — the four engines of increasing returns, we crack it open and separate the specific forces that amplify an early accident (learning effects, coordination, complementary goods, and expectations) so you can spot which engine is running in any real case — and how strongly.